Embed Size (px)
Citation preview
75�)������
:,/0$5��:3����'HOLYHUDEOH�'����6\VWHP�VWDELOLW\�DQDO\VLV�
�
1RYHPEHU������
7(&+1,&$/�5(3257�SUBJECT/TASK (title)
:,/0$5�:3��±�'HOLYHUDEOH�'����6\VWHP�6WDELOLW\�DQDO\VLV� �CONTRIBUTOR(S)
Ian Norheim (SINTEF), Elin Lindgren (KTH), Sanna Uski (VTT), Poul Sørensen (Risø), Clemens Jauch (Risø) CLIENTS(S)
SINTEF Energy Research Address: NO-7465 Trondheim, NORWAY Reception: Sem Sælands vei 11 Telephone: +47 73 59 72 00 Telefax: +47 73 59 72 50 www.energy.sintef.no Enterprise No.: NO 939 350 675 MVA
TR NO. DATE CLIENT’S REF. PROJECT NO.
TR F6212 2005-03-09 Attention 12x267 ELECTRONIC FILE CODE RESPONSIBLE (NAME, SIGN.) CLASSIFICATION
050309153251 Kjetil Uhlen Public ISBN N0. REPORT TYPE RESEARCH DIRECTOR (NAME, SIGN) COPIES PAGES
82-594-2924-1
Deliverable Petter Støa 100 DIVISION LOCATION LOCAL FAX
Energy Systems Sem Sælands vei 11 +47 73597250 RESULT (summary)
This report is a deliverable of the Work Package 5 (WP5) in the EU supported project Wind Power Integration in Liberalized Electricity Markets (WILMAR). The main objective of the WP5 work in the WILMAR project has been to: 1. Identify and quantify potential system stability problems in particular related to large-scale integration of intermittent renewable energy generation into the power system 2. Identify and evaluate various solutions to eliminate/reduce the problems The report describes and analyse results from different stability studies related to the main objectives of WP5. Three major areas of system stability has been the focus for the analysis. These were considered as most critical from a system operation point of view:
1) Frequency stability The minute to minute impact of variations in wind power, load, and production on the system frequency in the Nordic Nordel system has been simulated and analysed based on output results from the Work Package 6 (WP6) Joint Market Model (JMM) in the WILMAR project. Also the main power flow constraints in the Nordel system and in a small part of the European UCTE system consisting of West Denmark and North-West Germany have been assessed. The results illustrate system frequency and power flow problems that can not be analysed in detail with the JMM. Solutions to potential frequency stability and balancing control problems have been suggested.
2) Small signal stability Large-scale integration of wind power in Norway influences the damping of the inter-area oscillation modes in the Nordel system. The impact of different wind power technologies on these modes has been. It is shown that modern variable speed wind turbines with power electronic interface to the network may have a de-stabilising effect. However, there exist low cost control methods to solve this problem.
3) Transient stability It is demonstrated how a new offshore wind park in Eastern-Denmark may improve the transient stability in the Nordel system when including a grid frequency controller in the wind farm.
.(<:25'6�
Large-scale wind power integration Frequency stability SELECTED BY AUTHOR(S)
Small signal stability Transient stability
2
12x267 TR F6212
TABLE OF CONTENTS Page
1 INTRODUCTION ...................................................................................................................4 1.1 OBJECTIVES OF THE WORK IN WORK PACKAGE 5 .........................................4 1.2 REPORT CONTENT ...................................................................................................6
2 SYSTEM STABILITY AND OPERATIONAL CONSTRAINT...........................................7 2.1 OPERATIONAL CONSTRAINTS..............................................................................7
2.1.1 Frequency bias ..................................................................................................7 2.1.2 Active reserves..................................................................................................8 2.1.3 Transmission capacity limits...........................................................................11
2.2 TYPES OF SYSTEM STABILITY PROBLEM........................................................11
3 TOOLS AND METHODS ....................................................................................................14 3.1 PSS/E ..........................................................................................................................14 3.2 POWER FACTORY FROM DIGSILENT.................................................................15 3.3 STEPWISE POWER FLOW......................................................................................15
3.3.1 Stepwise Power Flow loop..............................................................................15 3.3.2 Simulating the Stepwise Power Flow loop with two synchronous systems
connected through HVDC-links......................................................................25 3.4 LP OPTIMIZATION WITH DC POWER FLOW.....................................................28
3.4.1 Optimization Problem Formulation ................................................................28
4 MODELS AND DATAFLOW..............................................................................................34 4.1 THE 23 GENERATOR MODEL OF THE NORTHERN EUROPEAN
SYSTEM.....................................................................................................................34 4.2 DATA CONVERSION FROM WP6 OUTPUT TO WP5 INPUT ............................37
4.2.1 Conversion program........................................................................................37 4.2.2 Nodes and regions...........................................................................................37 4.2.3 Loads...............................................................................................................41 4.2.4 Generator production ......................................................................................41 4.2.5 Transmission ...................................................................................................42 4.2.6 Export / import ................................................................................................43 4.2.7 Generator rating (PSS/E Mbase).....................................................................44 4.2.8 Balancing power list........................................................................................45 4.2.9 Wind power forecast errors.............................................................................47
5 SECONDARY CONTROL...................................................................................................48 5.1 CASE DESCRIPTION ...............................................................................................48
5.1.1 2001 case.........................................................................................................48 5.1.2 2010 cases .......................................................................................................52
5.2 RESULTS FROM STEPWISE POWER FLOW CALCULATIONS........................59 5.2.1 2001 case.........................................................................................................59 5.2.2 2010 cases .......................................................................................................62
5.3 RESULTS FROM LP OPTIMISATION CALCULATIONS....................................68 5.3.1 2001 case.........................................................................................................68 5.3.2 2010 case.........................................................................................................69
5.4 RESULTS FROM EXTENDED TERM DYNAMIC SIMULATIONS ....................70 5.4.1 2001 case.........................................................................................................70
3
12x267 TR F6212
6 SMALL SIGNAL STABILITY ANALYSIS........................................................................73 6.1 SMALL SIGNAL STABILITY STUDIES ................................................................73
6.1.1 Case description ..............................................................................................73 6.1.2 Wind power modelling and integration ..........................................................74 6.1.3 Results.............................................................................................................77
6.2 ANALYSIS.................................................................................................................79
7 TRANSIENT STABILITY ANALYSIS...............................................................................81 7.1 THE NORDIC POWER SYSTEM MODEL .............................................................81 7.2 MODEL OF WIND POWER INSTALLATIONS IN THE NORDIC POWER
SYSTEM.....................................................................................................................82 7.2.1 Topology of the Grid in Eastern Denmark......................................................82 7.2.2 Model of Wind Farm Feeders .........................................................................83 7.2.3 The Wind Farm Models ..................................................................................85
7.3 SIMULATIONS, RESULTS AND DISCUSSIONS..................................................87 7.3.1 Case 1..............................................................................................................88 7.3.2 Case 2.0...........................................................................................................90 7.3.3 Case 2.1...........................................................................................................91 7.3.4 Comparison of the Grid Frequency Response of Case 1, Case 2.0 and
Case 2.1...........................................................................................................92
8 CONCLUSIONS ...................................................................................................................94 8.1 FREQUENCY STABILITY.......................................................................................94 8.2 SMALL SIGNAL STABILITY..................................................................................96 8.3 TRANSIENT STABILITY.........................................................................................96
9 REFERENCES ......................................................................................................................98
4
12x267 TR F6212
�� ,1752'8&7,21� ���� 2EMHFWLYHV�RI�WKH�ZRUN�LQ�:RUN�3DFNDJH��� The present report is a deliverable of the Work Package 5 (WP5) in the EU supported project Wind Power Integration in Liberalized Electricity Markets (WILMAR). The main objective of the WILMAR project can be stated as in the following:
• Development of a planning tool for analysing the consequences of introducing substantial amounts of wind power in an electricity system covering the Nordic countries and Germany in 2010, in terms of:
- What are the technical impacts? - What is the impact on the spot and balancing power markets? - What does it cost to integrate a certain amount of wind power? - How is the performance of different integration measures? - How should the market be designed from the viewpoint of integration of wind
power? WP5 focuses on the technical impacts, but its results may have influence on the other issues mentioned in the above list. The main objective of the WP5 work in the WILMAR project has been to:
1. Identify and quantify potential system stability problems in particular related to large-scale integration of intermittent renewable energy generation into the power system.
2. Identify and evaluate various solutions to eliminate/reduce the problems. The purpose of WP5 is to perform different types of stability studies with large scale wind power in the Northern European system. This is important in order for the different stakeholders to make the correct decisions concerning wind power integration. For a transmission system operator stability studies will be useful in order to assess:
� How much wind power can be integrated from an operational security point of view?
� How much operating reserves are needed? � Congestion management. � Secondary control and protection issues. � System stability (Voltage stability, Angle stability).
From the generation owner / wind farm developer point of view the system stability studies are useful in order to make assessments regarding:
� Choice of wind farm technology (control systems, protections, etc.) � Additional costs (or reduced income) due to network constraints
5
12x267 TR F6212
Some of the studies described in this report, like the more classical stability studies, are made independent of the analyses performed with the Joint Market Model (JMM) in Work Package 6 (WP6), while the stability studies concerning frequency and secondary control has been performed based on WP6 data. The issues relevant for analysis by WP6 models and WP5 simulations are indicated in Figure 1.1. For chosen consecutive hours simulated in WP6, the effects that may occur within these hours due to the dynamic changes in wind, load, and production are simulated in WP5. The idea is to get an indication if the strategy chosen for the balancing power market in WP6 is able to deal with the continuous changes in load, wind and other power production.
� Long-term planning� Future demand and system expansion
� Future prices� Optimisation of resources
� Network planning� How much wind energy can be integrated
� Secondary control and protection
� Congestion managemenet, reserves, etc.
� System stability and primary control� Voltage stability
� Angle stability� Transient stability
� Damping
WP6
WP5
)LJXUH�����6XEMHFWV�RI�UHOHYDQFH�IRU�:3��DQG�:3�� The main approach in the WP5 studies is to use a simplified or reduced model [1] of the Northern European power system to analyse different stability problems. The reduced model has been established at SINTEF with the purpose of demonstrating some main dynamic properties of the Nordic power transmission system. Additionally, the equivalent network data (lines and transformer impedances) are adjusted to fairly well reflect the flow of power in the different corridors between the countries. The chosen model is utilised in different computer programs to simulate different types of stability problems. The different simulation tools and methods are described in Chapter 3. Some of the case studies have been performed jointly with WP6 to investigate potential problems on secondary control that are not possible to investigate with the JMM. WP6 simulate future scenarios like year 2010. With a large increase of installed wind power capacity in the Northern-European system, it is expected that there will be increased difficulties to keep the system frequency in the Nordel system within acceptable limits. It is therefore important to analyse the
6
12x267 TR F6212
balancing power market in the context of keeping the system frequency within the requirements also for 2010 cases. Power flow on main transmission corridors will also be influenced by large scale wind power integration and acceptance of balancing power market bids. These are operational challenges that have been assessed as part of the case studies based on WP6 simulations. A more general approach has been taken in the case study concerning small signal stability or power oscillation damping. In this context the most important issue was to investigate the general influence on the system damping with large-scale wind power integration in the system. In the transient stability analysis one case study has been performed on how a new wind farm in East Denmark would influence the Nordel transmission system after a transient fault close to the new wind farm. The impact of the transient fault on the wind farm itself was also studied. Voltage stability is regarded as a more regional/local network problem, and therefore this has not been a subject for analysis in the WP5 work. ���� 5HSRUW�FRQWHQW� In Chapter 2 the typical operational constraints in the Nordel system that are relevant for system stability analysis are identified and described. An overview is made on the types of stability problems that are relevant for the WP5 analysis, and it is described which case studies and stability problems that were chosen for analysis in WP5. Chapter 3 contains an overview and description of the tools and methods that have been utilised in the stability issues. Chapter 4 describes the Northern European model equivalent and the data flow between WP6 and WP5. The data flow within the WP5 is also described. In Chapter 5 the case studies on secondary control and frequency stability are described and the results of the simulations are presented and commented. In Chapter 6 the case study on small signal stability are described and the results of the simulations are presented and commented. In Chapter 7 the case study on transient stability are described and the results of the simulations are presented and commented. Chapter 8 contains the main conclusions.
7
12x267 TR F6212
�� 6<67(0�67$%,/,7<�$1'�23(5$7,21$/�&21675$,17� In the present chapter it is focused on the different system stability problems that are analysed in WP5. In [2] a more thorough introduction to the different stability problems can be studied. The operational constraints define some limits that help the TSOs to operate the system according to normal security standards. An introduction to some of these is given in Section 2.1. Much of the description about the operational constraints in this present section is based on [3]. ���� 2SHUDWLRQDO�FRQVWUDLQWV� The operational constraints are set to ensure safe technical operation of the power system. In the NORDEL system several ancillary services are defined and must be available for safe operation of the system. A list of the ancillary services is shown here: - )UHTXHQF\�$FWLYDWHG�2SHUDWLQJ�5HVHUYH��LQFO��)UHTXHQF\�%LDV��- )UHTXHQF\�$FWLYDWHG�&RQWLQJHQF\�5HVHUYH�- )DVW�&RQWLQJHQF\�5HVHUYH�- 6ORZ�&RQWLQJHQF\�5HVHUYH - Reactive Reserve - Voltage Activated Contingency Reserve (used for HVDC connections) - Load Following - System Protection - Ramping (currently not used) - Black Start - Secondary control of HVDC connection (West Denmark only) - Automatic Load Shedding - Manual Load Shedding - Fast Forecast Reserve (bundled with Fast Contingency Reserve) - Fast Counter-trading Reserve - Peak Load Reserve (under preparation) In the present report it is the four first ancillary services that are considered as these are the most relevant for the daily operation of the power system and for the types of stability studies that were seen as most relevant for the WILMAR WP5 work. A description of them is found below in Sections 2.1.1 and 2.1.2. ������ )UHTXHQF\�ELDV� The basis for controlling the short term balance between demand and supply is the system frequency. The control responsibility is distributed among the partners according to frequency bias, but Norway and Sweden have the main responsibility to control frequency and time deviation [4].
8
12x267 TR F6212
In the Nordel system the frequency is allowed to float within ±0.1 Hz in normal operation. The stationary frequency deviation following a contingency must not exceed 0.5 Hz, but there is normally no requirement for the WUDQVLHQW�frequency deviation. Only once have we found a remark where ”system frequency following a Dimensioning Fault1 should not go below 49.0 Hz” [5]. The total frequency bias (system bias) of the Nordel system must be at least 6000 MW/Hz ZLWKLQ�WKH�IUHTXHQF\�UDQJH�����������+] [4]. This amount is annually allocated among the partners according to last year’s energy consumption, as shown in Table 2.1. A formal basis for the number 6000 MW/Hz has not been found, but this value had been experienced with satisfactory system performance over time when the requirements were designed. During the last years the frequency quality in the Nordic system has declined. This is partly due to periods with large import and low system bias, combined with large and rapid load changes on international interconnections. The largest deviations especially occur during morning and evening load changes and around midnight. The official target is to keep the accumulated periods of system frequency outside the normal range of 49.9-50.1 Hz below �����PLQXWHV�SHU�\HDU [6]. The actual operation is still far from reaching this target, as shown in Figure 2.1.
0
1000
2000
3000
4000
5000
6000
7000
2001 2002 2003 pr. Aug-04
<HDU
0LQX
WHV Registered
Target
)LJXUH������$FFXPXODWHG�WLPH�RI�RSHUDWLRQ�RXWVLGH�QRUPDO�IUHTXHQF\�UDQJH�RI���������+]�� ������ $FWLYH�UHVHUYHV� The active power reserves in the Nordel system are divided into $XWRPDWLF�5HVHUYH (Primary Reserve), )DVW�5HVHUYH (Secondary Reserve) and 6ORZ�5HVHUYH (Tertiary reserve) [4].
1 See Section 2.1.2 under subheadline automatic reserve for a definition of “Dimensioning Fault”
9
12x267 TR F6212
$XWRPDWLF�UHVHUYH��The Automatic Active Reserve, or Primary Reserve, is further divided into )UHTXHQF\�$FWLYDWHG�2SHUDWLQJ�5HVHUYH (FAOR) and )UHTXHQF\�$FWLYDWHG�&RQWLQJHQF\�5HVHUYH (FACR)2. In normal operation, the system frequency is allowed to float between 49.9-50.1 Hz. With a minimum system bias of 6000 MW/Hz, the minimum )$25 is thus 600 MW, fully activated at 49.9 Hz. Similar to frequency bias, this amount is annually allocated among the partners according to last year’s energy consumption. At least 2/3 of the reserve (responsibility) must be SK\VLFDOO\ allocated within the respective partners own system in case of grid splitting or islanding [4]. Thus, a certain amount of trading of FAOR and frequency bias between the partners is allowed in the Nordel system. The required amount of )UHTXHQF\�$FWLYDWHG�&RQWLQJHQF\�5HVHUYH (FACR) is defined as 'LPHQVLRQLQJ�)DXOW minus self-regulation of load. Dimensioning Fault is defined as the single contingency that can happen with a probability of once every third year that leads to the largest loss of generating capacity [7]. The Dimensioning Fault is normally in the range of 1200 MW, corresponding to a nuclear power plant in Sweden or the largest hydro power plant in Norway. The self-regulation of the load was estimated to 200 MW already in 1982 [8]. This number also includes the voltage dependence of the load [5]. Note that the self-regulation only includes the load response, while the total V\VWHP�ORDG�IUHTXHQF\�FKDUDFWHULVWLF also includes response from generating units. The total load in the Nordel system has increased from around 35 GW to 65 GW since 1982 [9], but the value of the self-regulation has not been updated. The total FACR requirement is thus 1000 MW, which is allocated among the partners in proportion to the GLPHQVLRQLQJ�IDXOW in each partner’s system [4]. This allocation should be updated weekly. There is, however, no systematic change in FACR requirement depending on system load level like in the England/Wales system [10]. The activation of FACR will start at 49.9 Hz, and the reserve will be fully activated at 49.5 Hz. The reserve response should be nearly linear in the frequency range 49.9-49.5 Hz. Furthermore, at an instantaneous frequency drop to 49.5 Hz, 50% of the FACR must be activated within 5 seconds, 100% within 30 seconds. This response requirement is similar to the previous requirement in the UCTE system [10]. Figure 2.2 shows the correlation between FAOR and FACR.
2 These English abbreviations are used in this report for the benefit of the reader. They are not generally approved or
used in the Nordel system.
10
12x267 TR F6212
49.4 49.5 49.6 49.7 49.8 49.9 50
Frequency (Hz)
Reserve (MW)
2SHUDWLQJ�5HVHUYH������0:
&RQWLQJHQF\5HVHUYH�������0:
6000MW/Hz
2500 MW/Hz
)LJXUH������)UHTXHQF\�DFWLYDWHG�UHVHUYHV�LQ�1RUGHO�� $XWRPDWLF�ORDG�VKHGGLQJ can also be included in the FACR, provided the load fulfils the response requirements stated above. �)DVW�UHVHUYH��A required amount of )DVW�&RQWLQJHQF\�5HVHUYH�(FCR), or Secondary Reserve, is not specified in the current system agreement [4]. Each TSO is given the responsibility to ensure that their fast reserves are sufficient to:
1. re-establish FAOR and FACR after a contingency or load forecast deviation, and 2. re-establish scheduled power flows after a contingency or load forecast deviation, thus 3. return system to normal operation within 15 minutes after a contingency
This is the same principle as in the European UCTE system: The primary reserves (FACR) are allocated among each country/partner. Thus, everybody will contribute to stabilize the system following a contingency. When the primary reserves are activated, the system is in a stable state, but the frequency deviates from nominal value, the power interchanges deviate from the scheduled values, and primary reserves are “spent”. Secondary reserves then have to be activated such that the country/partner responsible for the deviation covers the imbalance, and the system is brought back to nominal frequency and scheduled interchanges [10]. The FCR must be available within 15 minutes, and be located such that the system will be returned to normal operation following a contingency. The FCR can be shared among the partners, provided there is no potential congestion in the transmission system that might prevent the activation of the reserves.
11
12x267 TR F6212
The following table summarizes the current Nordel requirements of active reserves [4]: 7DEOH������6XPPDU\�RI�FXUUHQW�DFWLYH�UHVHUYH�UHTXLUHPHQWV�LQ�1RUGHO��� &RQVXPSWLRQ�
������7:K��)$25��0:��
)$&5��0:��
)&5��DSSUR[����0:��
)UHT��%LDV��0:�+]��
(DVW�'HQPDUN� 14 24 90 600 240
:HVW�'HQPDUN� - - 75 620
)LQODQG� 85 141 205 1 000 1 410
1RUZD\� 115 192 313 1 600 1 920
6ZHGHQ� 145 243 303 1 200 2 430
727$/� ���� ���� ������ ������ ������ �6ORZ�UHVHUYH��The Slow Active Contingency Reserve (SCR) comprises capacity available after 15 minutes. If necessary, these are activated to re-establish fast reserves that are “spent” after a contingency. The forecast part of the slow reserve should be seen as an integrated part of the generation scheduling. These three levels of reserves (Automatic, Fast, Slow) are also generally termed 3ULPDU\, 6HFRQGDU\ and 7HUWLDU\ reserves to emphasize the hierarchical structure where slower and less expensive reserves are activated to re-establish the faster and expensive ones. ������� 7UDQVPLVVLRQ�FDSDFLW\�OLPLWV� The trading on the elspot market and the balancing power market is limited through the transmission capacity limits in the system. The transmission capacity limits are set by the TSOs according to angle and voltage stability simulations based on the n-1 criteria. The ability of the system to maintain in synchronism after chosen critical contingencies (for instance loss of a line, a generator, or a short circuit) or the amount of power that can be transferred on lines between regions without causing voltage instability are typical cases for determining the transmission capacity. In some cases, it might be the thermal capacity that determines the transmission capacity (cases where the stability studies indicate a higher transmission capacity than the thermal capacity of the line). ���� 7\SHV�RI�V\VWHP�VWDELOLW\�SUREOHP� Stability analysis of the power system is a large area that covers many different topics. In the context of large scale integration of wind power in the Northern European system all areas of power system stability are influenced. It is convenient to describe power system stability in three categories:
1) Angle stability, which is related to the ability of the electrical machines in the system to remain in synchronism after a large or small disturbance. The problem might be local or
12
12x267 TR F6212
on a system level. It is common to distinguish between transient stability analysis which considers large system disturbances, and small signal analysis which considers small system disturbances.
2) Voltage stability, which is related to the ability of the power system to maintain acceptable voltages at all buses in the system under normal operating conditions and after being subjected to a disturbance. It is usual to consider voltage instability as a local problem, while a voltage collapse is considered to involve voltages in a larger part of the system.
3) Frequency stability. Stability related to dynamics which influence the system frequency in the range from typically 10 seconds to 10s of minutes. In this context frequency stability can be analysed as a Mid-Term or Long-Term stability problem. Physically, frequency stability can be related to issues like lack of active reserves, slow control actions, poor coordination of protection and inadequacies in system equipment.
Within the WP5 activity in the WILMAR project the main focus has been on Long-Term stability, and specifically on the issue frequency deviation due to imbalance between production and load within simulated hours. A large scale integration of wind power in the Northern European system causes increased variation and increased unpredictability of the production compared to the present day situation. For this situation it is of importance to investigate if the power system is able to maintain a satisfying balance between production and load. This is measured by the frequency deviation from 50 Hz. In the NORDEL system it is required that the system frequency during normal operation should not deviate more than ± 0.1 Hz from 50 Hz. To prevent this from happen the transmission system operators (TSOs) trade power on the balancing power market (which is trading of Secondary reserves). In the Nordel system, power on the balancing power market should be activated within 15 minutes. Few tools exist to analyse secondary control problems. A new model called Stepwise Power Flow (SPF) has partly been developed through the WILMAR project. This model makes it possible to analyse the frequency deviations within a simulated hour when some key information about how the system changes is available as input data, i.e. the load change, wind power change, scheduled production change, and the bids on the balancing power market. In case of large frequency deviations balancing power is activated according to the bid list. This way it is possible to examine if there is a sufficient amount of balancing power available on the bid list. The main input data, such as the bid list, production plans, variation in load and wind power are available from WP6. Thus, the cases selected and simulated with the SPF routine will indicate if there are potential operating problems within the hour problems that are not possible to observe with simulation of the WP6 JMM model. For selection of cases it has been chosen to consider some hours with different interesting properties. This way, it was expected to get a general impression (without an exaggerated amount of simulations) on if the output results from the JMM model are reasonable with respect to Long-Term stability. The SPF routine is explained more thoroughly in Chapter 3. Alternative approaches to the SPF routine were also performed in order to have a stronger validation of the results and methods. These approaches are long term simulation in the power system simulation tool PSS/E and LP optimisation with DC power flow.
13
12x267 TR F6212
Voltage stability issues are mostly local, and do not affect the system as a whole. Due to this it is not a specific Northern European power system problem because of large scale integration of wind power production, and it has been left out of the WP5 activity. Integrating a large amount of wind power in the Nordel system will influence the oscillation modes in the system. In the context of considering the whole system it is the inter-area oscillation modes that were seen as most interesting for the WP5 work. Simulations on how different technologies influenced the damping of the inter-area oscillation modes were performed in the PSS/E. A new 198 MW offshore wind power park has been planned in Eastern Denmark. This has been the basis for the transient stability studies in WP5. For this purpose the power system simulation tool Power Factory from DIgSILENT was used as this was the tool available for the project partner performing this task. A substantial work was put into converting the 23 generator model used as basis for all simulations in WP5 (see Section 4.1) from PSS/E format to Power Factory format.
14
12x267 TR F6212
�� 722/6�$1'�0(7+2'6� The present chapter describes the tools and methods used for the analysis in the WP5 work. Two of the simulation tools, PSS/E and Power Factory from DIgSILENT, are commercial products. These are well suited for traditional stability studies as small signal stability and transient stability. PSS/E also has a module that could be used to study long-term stability. The other simulation tools used in WP5 are designed especially for simulating stability with respect to secondary control. These are Stepwise Power Flow developed at SINTEF, and LP optimisation with DC power flow developed at KTH. ���� 366�(� PSS/E is short for Power System Simulator for Engineering and is manufactured through the Siemens Power Transmission & Distribution, Inc., Power Technologies International (Siemens PTI) [11]. SINTEF and VTT have extensive experience in using PSS/E for simulating power system stability. The most important modules in PSS/E for the use required by the WP5 work are:
1) Power flow In the power flow simulations stationary power flow for a given system is calculated. This power flow may then be used as the initial state of the simulated grid in a dynamic simulation.
2) Dynamic simulations In this modular the dynamic simulations are performed. One set control systems and other parameters that influence the dynamic behaviour of the system, for instance after a disturbance. User programmed models may be integrated in the dynamic simulation. The results of a dynamic simulation are given in the time domain. The user must choose a fixed time step that ensures that the dynamic properties of the simulated system are preserved. This module is suited to run large power systems in the time range 10 seconds to 1 minute.
3) Extended term dynamic simulation PSS/E extended term simulation allows users to study longer term effects, such as frequency deviations as affected by prime mover response and voltage changes caused by protective equipment, and yet minimize computer time by providing possibility of changing simulation time step during simulation. The integration method depends on the size of the time step, and therefore smaller time step can be used during and right after large changes/faults, and larger time step at other times. Thus, PSS/E extended term module simulates the mid- to long-term dynamic response of the system.
15
12x267 TR F6212
���� 3RZHU�)DFWRU\�IURP�',J6,/(17� The Power Factory software is developed by DIgSILENT GmbH in Germany. Power factory is an integrated power system analysis tool covering an extensive range of standard and more specialized applications, including models of modern wind power installations. Risø has extensive experience in using Power Factory for simulating power system stability. The most important modules in Power factory for the use required by the WP5 work is:
1. Power flow. In the power flow simulations stationary power flow for a given system is calculated. This power flow may then be used as the initial state of the simulated grid in a dynamic simulation.
2. Dynamic RMS (or stability) simulations. In this module, dynamic simulations similar to PSS/E simulations can be performed. Besides the positive sequence simulation offered in PSS/E, Power Factory also offers simulation of the unsymmetrical sequences. RMS simuleations are normally used in stability studies.
3. EMT dynamic simulation. For more detailed studies of the transient behaviour, Power Factory also offers simulations including the electromagnetic transients. The EMT simulation mode corresponds to simulations with EMTDC, and requires much smaller time steps than the RMS simulations.
���� 6WHSZLVH�3RZHU�)ORZ� In the case of power and frequency control, typically Fast reserves, Secondary control and AGC (Automatic Generation Control) are difficult to handle in standard tools since these operate in the 5-15 minutes range, sometimes longer. Adding the coordination with generation scheduling and load forecast errors the time span is even longer. To enable such studies of the power system including Secondary control SINTEF Energy Research has developed a method, called 6WHSZLVH�3RZHU�)ORZ��63)��[3]. The basic methodology of Stepwise Power Flow was developed in the Statnett project ”Operational Security and Control” [12], using the MatPower toolbox [13] for MATLAB [14]. In the context of the WILMAR project the Stepwise Power Flow routine has been further developed to take into account wind power variations, and to in one simulation, simulate the two synchronous systems in the Northern-European system connected through HVDC-links. Due to the new concept that the SPF routine stands for and the developments of the routine that has been performed in the WILMAR project a thoroughly explanation of the routine as used in the WILMAR project is given in Sections 3.3.1 - 3.3.2. ������ 6WHSZLVH�3RZHU�)ORZ�ORRS� The main idea with the Stepwise Power Flow method is to use a sequence of stationary power flow analyses to capture slow system dynamics in the minutes range from stationary droop response through fast reserves (Balancing Power from Balancing Market) to scheduled generation changes, while transient dynamics are neglected. Regular power flow calculations assume balance
16
12x267 TR F6212
between scheduled generation and actual load, but this is formally correct only once or a few times during the hour in a real power system where the load is dynamically changing. With two power flow cases imported from PSS/E (one case is the power flow in the beginning of the simulated hour and the other is the power flow at the end of the simulated hour), a sequence of modified power flow analyses are executed in a defined number of steps between the basic cases. In the present version the system load is assumed to change linearly through the hour (WP6 assumes no stochastic load changes), while scheduled generation is changed only at the change of hour (the instant of change of the hour is user defined). Wind power production may be changed as the user defines it every time step (for instance every fifth minute) of the simulation. Figure 3.1 illustrates how it within an hour typically will be a deviation between production and load due to the fact that the scheduled generation does not follow the ramping of the load. The system operator observes the deviation between production and load as a deviation in frequency from 50 Hz.
8 000
10 000
12 000
14 000
16 000
18 000
20 000
22 000
6:00 7:00 8:00 9:00 10:00
Time (h)
Load
(M
W)
GenerationLoad + Exch.
Period of analysis
8 000
10 000
12 000
14 000
16 000
18 000
20 000
22 000
6:00 7:00 8:00 9:00 10:00
Time (h)
Load
(M
W)
GenerationLoad + Exch.
Period of analysis
)LJXUH�����3HULRG�RI�DQDO\VLV�IRU�IUHTXHQF\�DQG�SRZHU�FRQWURO� In regular power flow analyses the 6ZLQJ�%XV generator compensates the imbalance between generation, load and losses. In SPF, this imbalance is for each step shared between all synchronised units according to their droop. The droop of wind power is assumed to be zero which mean that wind power does not contribute to frequency control. This is likely to be a correct assumption due to the unpredictability of wind (the TSO is not likely to accept wind power production for the balancing power market, and the bidder is not likely to reserve wind power for secondary control). The Stepwise Power Flow routine perform a quasi-stationary simulation of
- system frequency - active power generation and consumption - spinning reserves and system bias (MW/Hz) - critical line flows
The inputs to the algorithm are
- power flow and topology data from PSS/E for two consecutive hours - scheduled generation incl. deviations from schedule
17
12x267 TR F6212
- primary control/unit droop - wind power production series (modeled as negative load series that are added to the load) - HVDC ramping - volume and price of available Balancing Power - outages (generation, load or lines) - generator tripping and load shedding
The SPF flowchart is shown in Figure 3.2 for a simulation with changing load and wind power. The flowchart could have contained outages and generator tripping also, but this has not been the subject for the analyses performed in WP5 and has therefore been left out. The different functions are described below.
∆P = ∆Pwind(n) - ∆Pload(n)
∆f = ∆P / B(δ)∆Pgen = f(δ, ∆f)
8QLW�ORDG GURRSV
∆f < ∆flim => %3
3RZHU IORZ
Q� �Q���Pload(T), Pgen(T)Pload(T+1), Pgen(T+1)8QLW�ORDG GURRSV��δi )UHTXHQF\ %LDV� B(δ)%DODQFLQJ SRZHU �%3�Sum of unit droops, δι ÆFrequency bias, B [MW/Hz]
∆P = ∆Pwind(n) - ∆Pload(n)
∆f = ∆P / B(δ)∆Pgen = f(δ, ∆f)
8QLW�ORDG GURRSV
∆f = ∆P / B(δ)∆Pgen = f(δ, ∆f)
8QLW�ORDG GURRSV
∆f < ∆flim => %3∆f < ∆flim => %3
3RZHU IORZ3RZHU IORZ
Q� �Q���Q� �Q���Pload(T), Pgen(T)Pload(T+1), Pgen(T+1)8QLW�ORDG GURRSV��δi )UHTXHQF\ %LDV� B(δ)%DODQFLQJ SRZHU �%3�Sum of unit droops, δι ÆFrequency bias, B [MW/Hz]
)LJXUH������)ORZ�GLDJUDP�IRU�6WHSZLVH�3RZHU�)ORZ��63)�� During initialisation, power flow data for start �Q� ��� and end �Q� �1� cases are read. In the current version the demand is assumed to increase linearly for each node through the whole period: L1
Q31Q33 ORDGLORDGLORDGL ∀
=−==∆
)0()( (MW)
3.1
After an initial power flow calculation, the model is running in a loop of 1 steps with incremental demand in each node. The demand 3GL in node with index L�at step Q�� is thus given as:
18
12x267 TR F6212
L3Q3Q3 ORDGLORDGLORDGL ∀∆+=+ )()1( (MW) 3.2
The total demand change between time step Q�� and Q becomes:
∑=
∆+−=∆ORDG1
LORDGLORDGORDG 3Q3Q3
1
)1()( (MW)
Where 1ORDG is the number of loads in the system
3.3
The same calculation is made also for reactive demand. The wind power production at node L�is given as a negative load, and can be defined by the user in two different ways:
1. A user defined wind power production series, which describes the amount of production on the node for each time step of the simulation
2. Based on a measured wind series, the power curve of the technology used, the range of the region the node represents, and statistical data from the wind series a wind power production series can be generated. The method is described in [15].
In time step Q, the wind power production at the node with index N is given by:
NQ3Q3 ZLQGNZLQGN ∀= )()( (MW) 3.4
Thus, the wind power production at any time step at node N, is independent of the wind power production in all other time steps. The total wind power production in a synchronous system becomes:
NQ3Q3ZLQG1
NZLQGNZLQGQ ∀= ∑
=1
)()( (MW)
Where 1ZLQG is the number of nodes with wind power production.
3.5
The total change in load (included wind power) at time step n is given by:
)()()( Q3Q3Q3 ORDGZLQGQWRW ∆−∆=∆ (MW)
Where, )()0()( Q33Q3 ZLQGQZLQGZLQG −=∆
3.6
The change in demand and wind power causes a power imbalance at each time step resulting in an HVWLPDWHG stationary frequency deviation given by:
)(
)()( Q%
Q3QI WRW∆−=∆ (Hz) 3.7
19
12x267 TR F6212
�
∑=
⋅⋅=
JHQ1
M MM1M 63Q%
1
2)(
δ (MW/Hz)
3.8
� and 1JHQ� - no. of generators 31M� - rated output of generator M 6M� - connection status of generator M δM - droop of generator M
The Swing Bus will still compensate the change in network losses, since these are omitted from ���. The estimated frequency deviation in ��� is used to calculate the new output of each generator M according to its droop settings: M3QIQ3Q3
M
1MJMJM ∀⋅∆⋅−−=
δ)(2)1()( (MW) 3.9
Equations ���- ��� assume a linear system frequency bias. Normally, this would be a sufficient assumption, but if the new output of one or more generators exceeds the rated output, the calculation is modified to take the missing capacity and reduced system bias into account:
1MJM
M
1MJM
1MJMJM3Q3ZKHUHM
63Q3
3Q33≥∀
=
=
−=∆
)(
0
)(
)(
’
’
’
3.10
The total missing capacity is calculated, resulting in reduced system bias %¶�Q� and increased frequency deviation:
∑=
∆=∆JHQ1
MJMGHI 3Q3
1
’)( (MW) 3.11
∑=
⋅⋅=
JHQ1
M MM1M 63Q%
1
’’ 2
)(δ
(MW/Hz) 3.12
=> )(
)()(
’’
Q%Q3QI GHI∆
−=∆ (Hz) 3.13
=> )()()( ’ QIQIQI ∆+∆=∆ (Hz) 3.14
The algorithm is looping though Equations ���-���� until 0)( =∆ Q3GHI . This will create a non-
linear response in system frequency as illustrated in Figure 3.3. After the total frequency deviation and corresponding generator outputs are calculated, a regular power flow is run to find the line flows, voltages etc. No such adjustments are currently made for reactive power.
20
12x267 TR F6212
B1B1
B2B2
f1
P1
f1
P1
f2
P2
f2
P2
f3
P3
f3
P3 )LJXUH�����1RQ�OLQHDU�IUHTXHQF\�UHVSRQVH�GXH�WR�FDSDFLW\�OLPLWV� �����([DPSOH�RQ�WKH�XVH�RI�WKH�63)�DOJRULWKP��In a case where the SPF algorithm is used on a peak load hour in the 18-bus Nordel model [1] the frequency response looks like in Figure 3.4 (in this example no wind power and HVDC-links were modeled). The simulation is here made in 5 minutes time steps between two sequential PSS/E power flow situations where the scheduled generation is assumed to come online after 25 minutes. The resulting “saw-tooth” frequency response is very similar to the measured response in Figure 3.5.
49.85
49.90
49.95
50.00
50.05
50.10
50.15
0 10 20 30 40 50 60
7LPH��PLQ�
)UHT
XHQF
\��+]
�
)LJXUH�����6LPXODWHG�IUHTXHQF\�GXH�WR�ORDG�LQFUHDVH�
21
12x267 TR F6212
50.1
50.0
49.905:00 05:30 06:00 06:30 07:00 07:30 08:00
Frequency (Hz)
Time
50.1
50.0
49.905:00 05:30 06:00 06:30 07:00 07:30 08:00
Frequency (Hz)
Time )LJXUH�����0HDVXUHG�IUHTXHQF\�GXULQJ�PRUQLQJ�ORDG�LQFUHDVH�LQ�1RUGHO�V\VWHP� The capacity situation in Figure 3.4 is deliberately reduced to create a stressed system. Shortly before the change of hour the slope of the droop response increases due to generators reaching rated capacity. This will cause a dramatic drop in system bias (see Eq. ����), as shown in Figure 3.6. Shortly before new generation is brought online, the system bias is reduced to less than 6000 MW/Hz, while the System Operator has in his hourly schedule an estimated bias of 11.500 MW/Hz! If an outage should occur in this period the consequences could be dramatic. A similar but inverted droop response would result if the simulated hour had load GHFUHDVH instead of increase (tested by switching the initial and final power flow cases). Note also that the algorithm described above is decoupled in time. Each time step is calculated incremental to the previous without further knowledge about the history of the system. This makes the algorithm much simpler and easier to adjust or expand.
0
2000
4000
6000
8000
10000
12000
14000
16000
0 5 10 15 20 25 30 35 40 45 50 55 60
7LPH��PLQ�
6\VWH
P�ELD
V��0:
�+]�
)LJXUH�����5HGXFHG�V\VWHP�ELDV�GXH�WR�JHQHUDWRUV�UHDFKLQJ�UDWHG�FDSDFLW\� �����(QG�RI�H[DPSOH���
22
12x267 TR F6212
$FWLYDWLRQ�RI�EDODQFLQJ�SRZHU�� In Figure 3.4 the frequency drops to just above the minimum requirement of 49.9 Hz before scheduled generation is coming online. If the scheduled generation was delayed another 5 minutes the frequency would drop below this limit and the System Operator would need to act to prevent this. The secondary control in the Nordel system is handled manually by the Balancing Market (BM), and such functions are very difficult to model in a concise way. Four major aspects should be considered:
- The current model has no memory or foresight, and does not “anticipate” a frequency violation before it actually occurs.
- A human operator, on the other hand, will generally observe the trend of the frequency for some time and anticipate the frequency violation before it actually occurs.
- The required response time of regulating power is 15 minutes. No or very little response will be available within the same 5-minute time step as the frequency violation occurs, unless the capacity was called at a previous step.
- The required Balancing Power will be delivered partly from units already online, partly by committing new units.
Trying to incorporate all these aspects in “hard code” the following heuristic methodology is chosen:
1. When a frequency violation occurs due to gradually increasing demand, the algorithm assumes that this was anticipated by the System Operator. The amount of Balancing Power to activate in the present step is therefore set to KDOI of the total amount needed in the SUHYLRXV step:
2/)1()1()( %QIQ53 ⋅−∆−=∆ (MW) �����
2. The required amount is loaded on the cheapest units specified in the BM list. If the new output exceeds the specified rated capacity, the unit rating is increased accordingly. This is done because the current version uses aggregated generator models, not single unit representation.
3. With the new generation added, the droop response according to Eq. ���-���� is re-calculated.
�����([DPSOH�RQ�VLPXODWLRQ�RI�DFWLYDWLRQ�RI�EDODQFLQJ�SRZHU� The same case as previous in the present section is used. The only difference is a 5 minute delay in the generation schedule. The Balancing Power yields the droop response as shown in Figure 3.7. The methodology can be summarized as follows:
- The model is using droop response only, until a frequency violation is observed (dashed curve, W = 30 min).
23
12x267 TR F6212
- When a frequency violation is observed the model takes one step back, activates Balancing Power and re-calculates the next time step (solid curve).
49.85
49.90
49.95
50.00
50.05
50.10
50.15
0 10 20 30 40 50 60
7LPH��PLQ�
)UHT
XHQF
\��+]
�
)LJXUH�����6LPXODWHG�IUHTXHQF\�ZLWK�%DODQFLQJ�3RZHU� The added Balancing Power input to the system will reduce the need for droop control from other units, thus re-establishing some of the system bias as shown in Figure 3.8.
0
2000
4000
6000
8000
10000
12000
14000
16000
0 5 10 15 20 25 30 35 40 45 50 55 60
7LPH��PLQ�
6\VWH
P�ELD
V��0:
�+]�
)LJXUH�����6\VWHP�ELDV�ZLWK�DFWLYDWLRQ�RI�%DODQFLQJ�3RZHU� �����(QG�RI�([DPSOH� In the current version only activation of Balancing Power when the frequency exceeds the normal operating limit of +/- 0.10 Hz is implemented. However, the methodology is prepared to handle several levels of warnings and actions in a flexible way. At present, the following 6 rules are specified. Below an example on how these rules can be specified is shown:
24
12x267 TR F6212
**** Defining Balancing Power Rules and Actions ***** Rule 1: Bias(n) <= Bias(n-1) Action 1 (Warning): Print warning Rule 2: |dF| >= 0.08 (Warning level) Action 2 (Warning): Print warning Rule 3: Bias <= 8000 (Warning level) Action 3 (Warning): Print warning Rule 4: |dF| >= 0.10 (Critical level) Action 4 (Critical): Activate BP Rule 5: Bias <= 6000 (Critical level) Action 5 (Critical): Activate BP Rule 6: Branch overload Action 6: List the N_BR highest loaded branches >BM_rules = [1; 1; 1; 1; 0; 1];
The user can set the warning- and critical levels he wants, and then specify in the last command line which of these rules are active. Typical messages during simulation can be:
> WARNING (1): System bias is reduced > WARNING (3): System bias below limit 1 > CRITICAL (4): Frequency below limit 2
The following BM list has been used in the simulation example above:
Balancing Power list: bus Price [NOK/MW] Vol [MW] 5300 135.00 200.00 6000 140.00 200.00 5600 145.00 250.00 6100 160.00 100.00 6700 165.00 200.00 6500 170.00 100.00 5400 175.00 350.00 5500 176.00 100.00 5100 180.00 100.00
In the current version of the algorithm the exact amount of calculated Balancing Power is activated from the BM list. In reality, BP would be activated in blocks of 25 or 50 MW. After the activation of BP as shown in Figure 3.7, the BM list is updated as follows:
BP activated = 338.43 MW Clearing Price = 140.00 NOK New Balancing Power list: bus Price [NOK/MW] Vol [MW] 5300 135.00 0.00 6000 140.00 61.57 5600 145.00 250.00 6100 160.00 100.00
25
12x267 TR F6212
6700 165.00 200.00 6500 170.00 100.00 5400 175.00 350.00 5500 176.00 100.00 5100 180.00 100.00
Note also the last “rule” which is only a request to list the highest loaded branches (lines or transformers) for each time step. The output from this request will look like:
> Checking for branch overload ...
Highest branch loads: Fbus Tbus Flow [MVA] Flow [%] 5600 5601 1153.64 114 5500 5501 524.50 104 5101 5501 1153.39 67 5101 5100 611.09 61 5400 6100 512.24 60 5300 5301 570.75 57 5603 5602 567.84 57 5102 5100 965.83 48 5601 6001 1152.94 46 5500 5603 635.90 43
������ 6LPXODWLQJ�WKH�6WHSZLVH�3RZHU�)ORZ�ORRS�ZLWK�WZR�V\QFKURQRXV�V\VWHPV�FRQQHFWHG�WKURXJK�+9'&�OLQNV�
The 23 generator model of the Northern European system used in WP5 consists of two synchronous systems connected through HVDC-links. These are the Nordic Nordel system and the North-West Germany plus the West-Denmark part of the European UCTE-system. As these systems are not synchronous the frequency, bias, reserves etc. will be different in them. However, they are not without influence on each other as they are connected through HVDC-links. In order to achieve proper load flow calculations, bias calculations, frequency calculation etc. it is necessary to run the SPF routine as a function for the two synchronous areas for each time step. See Figure 3.9
26
12x267 TR F6212
)LJXUH�����)ORDW�GLDJUDP�IRU�VLPXODWLQJ�WZR�+9'&�FRQQHFWHG�V\QFKURQRXV�DUHDV�ZLWK�WKH�63)�URXWLQH� If a synchronous area is importing power from a HVDC-link the HVDC-link in this area is modelled as a negative active load and a positive reactive load with size 0.4 times the absolute value of the active load the HVDC-link represents. For a synchronous area that is exporting power through a HVDC-link the HVDC-link in this area is modelled as a positive active load and a positive reactive load with size 0.4 times the absolute value of the active load the HVDC-link represents. The SPF-routine will then be modified and have the form as shown in Figure 3.10. The change in ∆3KYGF�W� (see Figure 3.10) if the line between Jutland and Germany exceed its transmission limit is explained later in the present section. �
SPF- calculations
Area=Area+1 t = t+1
27
12x267 TR F6212
)LJXUH������63)�URXWLQH�ZKHQ�LQFOXGLQJ�LPSRUW�H[SRUW�IURP�+9'&�OLQNV�
$FWLYDWLRQ�RI�EDODQFLQJ�SRZHU�EHWZHHQ�WKH�V\QFKURQRXV�DUHDV��The HVDC-links can be utilized to make balancing power in one synchronous area available in the other synchronous area. In the Northern European system studied in the WILMAR project it is unlikely that balancing power is traded from UCTE to the Nordel system, as the bids for the balancing power market normally will be substantially lower in the hydro dominant Nordel system. It is also unlikely that the frequency in the UCTE system will deviate more than ± 0.1 Hz due to the strength of the UCTE system. However, if for instance the import of power to Jutland from Germany exceed the transmission limit on the line between Jutland and Germany (see Figure 4.1), a change in the power transmitted on the HVDC-connections between Nordel and Jutland is interesting in order to avoid or decrease the overload. Importing more power from Nordel in such a situation will cause the flow of power from Germany to Jutland to decrease, and the line will eventually be loaded according to its stability limit. However, this will increase the load in Nordel and balancing power may have to be activated in order to avoid the frequency to fall below 49.9 Hz.
∆P = ∆Pwind(t) - ∆Pload(t)-∆Phvdc(t)
∆f = ∆P / B(δ) ∆Pgen = f(δ, ∆f)
8QLW�ORDG�GURRSV��
∆f < ∆flim => %3�
�3RZHU�IORZ��
�W� �W����Pload(T), Pgen(T) Pload(T+1), Pgen(T+1) 8QLW�ORDG�GURRSV��δi
)UHTXHQF\�%LDV� B(δ) %DODQFLQJ�SRZHU��%3��:LQG��+9'&�UDPSLQJ�Sum of unit droops, δι Æ Frequency bias, B [MW/Hz]
PGER-JUT>PLIM⇒ change ∆Phvdc(t) In the next time step
Valid when looping the Jutland-Germany system
28
12x267 TR F6212
In the SPF routine implemented in the WILMAR project, the flow on the HVDC-link between Nordel and Jutland is changed if the flow in the line Germany-Jutland exceeds the transmission limits. The flow on the HVDC-links is changed with 10 % of the transmission limit on the line between Germany and Jutland plus the amount of power that this line was overloaded with. If there is not enough available power on the HVDC-links to activate all this, only the amount that not causes the HVDC-links to be overloaded are activated. ���� /3�RSWLPL]DWLRQ�ZLWK�'&�SRZHU�IORZ� A different approach to model secondary control is developed by KTH in the PhD project “Minimizing Costs for Reserve Power in Integrated Power Systems with Large Amounts of Wind Power”. The basic idea is to use an optimization problem formulation to minimize costs for balancing power, and the model is implemented in GAMS [21] and MATLAB [14]. An earlier version of the model was presented in e.g. [22], but some modifications have been made for the WILMAR project to make comparisons with SPF possible, and the optimization problem formulation will be presented in the following subsections. ������ 2SWLPL]DWLRQ�3UREOHP�)RUPXODWLRQ� The following formulation is to model the power system for one hour, where the each half-hour has its system bias, scheduled production and list of balancing power. A useful feature of optimization is that it is simple to add or exclude properties just by adding or excluding constraints. One property excluded here, which was included in another version, is the possibility for the TSO to move the scheduled production up to 15 minutes, instead of accepting bids close to a change of hour, but this has also a cost. ����������1RPHQFODWXUH� Sets
{1 }, «L M«,= , regions {1 }1 «P Q«1= , nodes
L1 nodes located in region L , subset of 1
{1 }/ « «/= A region lines ( )L M, , L M<
/∗ region lines ( )L M, , L M>
HVDC/ HVDC lines, subset of /�{1 }7 «W«7= time steps {1 }+ «K«+= hours
K7 time steps in hour K , subset of 7
{1 }% «E«%= bids
K% bids available in hour K , subset of %�Q% bids located in node Q , subset of %
29
12x267 TR F6212
Parameters
QW' load change in node Q at time W planQW3 scheduled change in production in node Q at time W windQW3 forecast error of wind power production in node Q at time W
0)A total transmission in line A at time 0W =
E9 available volume of bid E
EF price per volume for bid E actEW activation time for bid E [no. of time steps]
QW5 bias in node Q at time W minXA min transmission limit between region L and M maxXA max transmission limit between region L and M rampW) scheduled change of transmission in HVDC lines at time W�finalK7 the last time step in hour K
0I nominal frequency
devI maximal deviation from nominal frequency
maxG maximal time deviation
stepW length of each time step [seconds]
Variables�
QW3 sum of forecast errors, scheduled production changes, activated bids and change in load in node Q at time W , i.e. the change of net generation in node Q at time W except changes of primary reserves
QW* change in activated primary reserve in node Q at time W WI frequency at time W W)A change in transmission in line A at time W totW)A total transmission in line A at time W HVDCW) total transmission on HVDC lines at time W maxWHA exceeding of max transmission limit in line A at time W
minWHA exceeding of min transmission limit in line A at time W startEW\ activated volume of bid E at time W stopEW\ stopped volume of bid E at time W WG time deviation at time W QµA transmission coefficients to distribute the net generation of node Q to
line A , calculated from DC power flow ] objective function
����������2EMHFWLYH��
30
12x267 TR F6212
The objective of the optimization is to minimize the cost of all bids of balancing power for the TSO and to minimize overload on transmission lines between regions. The objective function is then start stop final act min maxmin ( ) ( ) ( )
K
E E EW EW K E OW OWE % O /
W 7W 7K +
] F 9 \ \ 7 W W H H∈ ∈
∈∈∈
= − − − + +∑ ∑ 3.16
����������&RQVWUDLQWV�RQ�JHQHUDWLRQ� The change of net generation QW3 in node Q at time W is defined as the sum of scheduled
production changes, wind power forecast errors, accepted bids and load change. A bid is assumed to be accepted at time W but the activation time of the power gives that the power is available at
actEW W+ . The net generation is then
act act
plan wind start stop( )E E
Q
QW QW QW E QWE W W E W WE %
3 3 3 9 \ \ ' Q 1 W 7− −
∈
= + + − − ∀ ∈ , ∈∑ 3.17
The change in generation to or from the primary reserve QW* in node Q at time W changes in
proportion to the biasQ5 in node Q as
PWP 1
QW QP
P 1
** 5 Q 1 W 75
∈
∈
= ∀ ∈ , ∈∑∑
3.18
Here no limitations on the primary reserves are included. ����������&RQVWUDLQWV�RQ�IORZ��The flow in one direction is defined as being equal to the negative flow in the other direction: ( )LMW MLW) ) L M / W 7= − ∀ , ∈ , ∈ 3.19
The sum of the changes in generation in node Q and the change in transmission to and from node Q at time W must be equal to zero, as all losses are neglected: ( )
L
LMW QW QWQ 1M / /
) 3 * L , W 7∗ ∈∈ ∪
= + ∀ ∈ , ∈∑ ∑ 3.20
31
12x267 TR F6212
To obtain flows corresponding to a DC load flow, the transmission coefficients QµA is used. The
change in the flow in line A at time W is given by ( )W Q QW QW
Q 1
) 3 * / W 7µ∈
= + ∀ ∈ , ∈∑A A A 3.21
and the total flow in line A at time W is given by tot
0W VV 7 V W
) ) )∈ : ≤
= + ∑A A A 3.22
To be able to determine the impact on frequency from changes on the HVDC lines between Nordel and UCTE, the total transmission HVDC
W) is calculated as
HVDC
+9'&
W W/
) )∈
= ∑ AA
3.23
with the constraints that the HVDC transmission is as scheduled except when bids are accepted from UCTE:
act act
HVDC ramp start stop( )E E
Q
W W E E W W E W WE %Q 8&7(
) ) 9 \ \− −
∈∈
= − −∑ 3.24
Previously, the transmission limits of region lines have been fix, which forced the solution to handle overload by accepting up-regulation and down-regulation in different areas, to lower the flow on critical lines. As will be shown in the case studies, no down-regulation bids are available here, and therefore this feature has been modified, and the overload is instead minimized. Thus, the constraints for transmission limits between regions is min max tot max max
W W WX H ) X H / W 7− − ≤ ≤ + ∀ ∈ , ∈A A A A A A 3.25
with min max, 0W WH H / W 7≥ ∀ ∈ , ∈A A A 3.26
����������&RQVWUDLQWV�RQ�IUHTXHQF\� The frequency at time W differs from the previous frequency at time 1W − with the total change in net generation at time W divided by the total system bias, and is also affected by changes in transmission on the HVDC lines between Nordel and UCTE. The contraints for the Nordel frequency is then
32
12x267 TR F6212
HVDC
1
QW WQ 1
W WQ
Q 1
3 )I I W 75
∈−
∈
−= + ∀ ∈
∑∑
3.27
The frequency must stay within its limits:
0 dev 0 devWI I I I I W 7− ≤ ≤ + ∀ ∈
3.28
In the case studies this will correspond to the limits49.9 Hz 50.1 HzWI≤ ≤ , but as will be shown
the upper constraint has to be removed to find a feasible solution since no down-regulation is available. The time deviation is defined by an integral and is here calculated as
step0
0
( )W VV 7 V W
WG I II ∈ : ≤
= −∑
3.27
The time deviation must stay within its limits:
max maxWG G G W 7− ≤ ≤ ∀ ∈
3.28
����������&RQVWUDLQWV�RQ�ELGV� The bid acceptance is determined from positive variables between 0 and 1. For example, start
EW\
equals 0.5 if 50 % of bid E is accepted at time W and stopEW\ equals 0.3 if 30 % of bid E is stopped
at time W , which gives start stop0 , 1EW EW\ \ E % W 7≤ ≤ ∀ ∈ , ∈
3.29
A bid can only be activated up to 100% of the available volume, and the same is valid for de-activation: start 1EW
W 7
\ E %∈
≤ ∀ ∈∑
stop 1EW
W 7
\ E %∈
≤ ∀ ∈∑
3.30
All activated bids must also be stopped, at change of hour or earlier, but not before they have been accepted:
33
12x267 TR F6212
start stopEV EV
V 7 V W V 7 V W
\ \ E % W 7 W 7∈ : ≤ ∈ : ≤
≥ ∀ ∈ , ∈ : <∑ ∑
start stopEW EW
W 7 W 7
\ \ E %∈ ∈
= ∀ ∈∑ ∑
3.31
A bid can only be accepted the hour it is available and a bid cannot be accepted closer to the end of an hour than the activation time since it is impossible to activate the bid within the hour. Bids and time steps not available are set to zero: start 0EW K K\ E % W 7 K += ∀ ∈ , ∉ , ∈
start final act0EW K K E\ E % W 7 W 7 W K += ∀ ∈ , ∈ : > − , ∈ start final act
10EW K K E\ E % W 7 W 7 W K +−= ∀ ∈ , ∈ : < − , ∈
3.32
Corresponding constraints are necessary for stopping of bids: stop 0EW K K\ E % W 7 K += ∀ ∈ , ∉ , ∈
stop final act0EW K K E\ E % W 7 W 7 W K += ∀ ∈ , ∈ : ≥ − , ∈ stop final act
10EW K K E\ E % W 7 W 7 W K +−= ∀ ∈ , ∈ : ≤ − , ∈
3.33
34
12x267 TR F6212
�� 02'(/6�$1'�'$7$)/2:� In the present chapter the model used for the stability studies performed in WP5 is presented. The flow of data from WP6 to WP5 input simulation files are also described. ���� 7KH����JHQHUDWRU�PRGHO�RI�WKH�1RUWKHUQ�(XURSHDQ�V\VWHP� The basis for all simulations in WP5 is the 23 generator model of the Northern European system. The 23 generator model determines the topology of the grid, where in the grid conventional production is located and where in the grid the loads are located. For the secondary control cases the output from WP6 determines the amount of production and on which units, the amount and location of wind power, the bid list for the balancing power market, and the HVDC ramping. In Section 4.2 it is described how the data flow from the WP6 output to the WP5 input has been performed. For the cases of analysing damping and transient stability selected data sets has been chosen out of the purpose of the simulations. In these cases the wind power were modelled according to the type of technology that were considered. It should also be mentioned that a substantial work was put into converting a base case of the 23 generator model from PSS/E format to DIgSILENT format. The possibility to examine if the power flow and dynamic responses was preserved in this process and the fact that the project partner Risø needed the DIgSILENT format in order to perform the transient stability studies were the motivation for performing the conversion. It turned out to be a not straight forward process, but eventually a good correspondence between the PSS/E model and the DIgSILENT model was obtained. The 23 generator model has been developed in PSS/E at SINTEF Energy Research. The model has been developed through several steps, starting with a 16 generator model of the Nordel system (Jutland and Germany were excluded from this model). How this model was developed is described in [1]. The main motivation for developing the simple model of the Nordel system was: - It was desirable to study slow dynamic phenomena as control of frequency and active power.
Thus, a simpler model reduced the simulation time to acceptable levels. In the context of the WILMAR project, the 23 generator model is suitable as it have a significant correspondence in power flow and in dynamic responses if comparing with a full scale model of the Nordel system. For analysis connected to secondary control the reduced size and still significant accuracy makes the 23 generator model favourable. In addition it is no restriction in distributing it among the different project partners (available full-scale models are restricted information and can not be distributed). Thus, the different partners could work on the same model as input for their simulation tools and the methods especially for simulation of long term stability could be compared. In addition, the data from WP6 to WP5 did not have to be converted to several models. In the 23 generator model of the Northern European the lines and generators are located and adjusted in such a way that they to a significant degree reflect the real production and the most
35
12x267 TR F6212
interesting bottlenecks in the Nordel system. The impedances are adjusted in such a way that the power flow to a significant degree will correspond to a full-scale model. Finally the dynamic models of the aggregates have been adjusted in such a way that the major dynamic properties of the 23 generator model reflect the major dynamic properties in the full-scale model. In Figure 4.1 the locations of the different generators equivalents in the 23 generator model are shown. The node number of the different generators is also shown. In Figure 4.2 the one-line circuit diagram of the 23 generator model is shown.
�)LJXUH�����/RFDWLRQ�RI�JHQHUDWRUV�LQ�WKH����JHQHUDWRU�PRGHO�
G
G
G
G
G
G
G GG
G
G
G
G
G
G
G
G
G
G
G
G
G
G
G
G
G
G GG
G
G
G
G
G
G
G
G
G
G
G
G
GG
����� �����
�����
���������������
�����
�����
�����
���������� �����
�����
����������
����������
���������������
�����
�����
�����
36
12x267 TR F6212
)LJXUH�����1RGHV��JHQHUDWRUV��DQG�ORDGV�LQ�WKH����JHQHUDWRU�PRGHO�RI�WKH�1RUWKHUQ�(XURSHDQ�V\VWHP�
8001 8002
5603
5602 5600
5601
6000
6100 5300
5301
5501 5401
5402
6001 5102
5103
5100
5101
6500
6700 6701
3701
3244
7100 3115
3249 3000
7000
3245
3100 3200
3300
3360
8003 8004
8005
8500
9000
9001
5500
5400 3359
12B6�
'.B:(67 �
*HUPDQ\ �
12B0 �
), � 6( �
'.B($67�
37
12x267 TR F6212
���� 'DWD�FRQYHUVLRQ�IURP�:3��RXWSXW�WR�:3��LQSXW� This section describes how the data from the market analysis in WP6 are converted to data for the stability studies in WP5. The main task of the data conversion is to generate PSS/E files to WP5 with load flow input data to the 23 generator model (section 4.1) for selected hours. The load flow files include data for generation, load, DC transmission and export/import, which are consistent with simulations performed with the Joint Market Model (JMM) in WP6. Besides, a “balancing power list” according to section 3.3.1 with bus number, price and volume of the secondary reserves according to JMM simulations is generated for the selected hours. Finally, files with wind power forecast errors for each of the selected hours are generated. ������ &RQYHUVLRQ�SURJUDP� A conversion program “WILMARJMM2PPPE” has been developed to support automatic conversion of the data. The conversion program uses input from:
• The Input Data Base (IDB), which also provides input data to the JMM simulations. • Simulation results with JMM. • A PSS/E load flow file based on the 23 generator model.
The input from the IDB is:
• A list specifying the JMM region of each PSS/E node. • A list specifying the JMM region of each JMM unit group. • A list specifying the unit group and capacities of each JMM unit.
The JMM simulation results input
• Production • Online capacity (secondary reserve) • Primary reserves (positive and negative) • Loads • Marginal production costs
The output from the data conversion program is:
• PSS/E files for every JMM hour. • Balancing power lists for every JMM hour.
In the present version, the conversion program generates the PSS/E files with consistent data for load and generation. The data for transmission and export/import has to be written manually into the PSS/E file. ������ 1RGHV�DQG�UHJLRQV�
38
12x267 TR F6212
In WP2 of the WILMAR project, the different regions have been defined as shown on the map in Figure 4.3. These regions are used by the JMM in WP6. With reference to Figure 4.2 the nodes that are located within the different WP 6 regions are as described below. 7KH�IROORZLQJ�QRGHV�DUH�ORFDWHG�LQ�6ZHGHQ�� a) Sweden south of corridor 4 (SE_S) 3300 SV-SW 3301 SE-SW-WI b) Sweden north of corridor 4 and south of corridor 2 (SE_M) 3000 STCKH 3200 KRLSK 3359 SV-W 3360 KONT_135 c) Sweden north of corridor 2 (SE_N) 3115 SV-NI 3100 STRFN 3244 SV-M2 3245 SV-M1 3249 SV-N2 3701 SV-N2B 7KH�IROORZLQJ�QRGHV�DUH�ORFDWHG�LQ�1RUZD\�� a) Norway South (NO_S) 5100 EAST3 5101 EAST4-A 5102 EAST4-B 5103 EAST4-C 5300 HDAL3 5301 HDAL4 5400 CNTR3-A 5401 CNTR4-A 5402 CNTR4-B 5500 CNTR3-B 5501 CNTR4-C 5600 SOUTH3A 5601 SOUTH4A 5602 SOUTH4B 5603 SOUTH3B 6000 WEST300 6001 WEST400 6100 NWEST3
39
12x267 TR F6212
b) Norway Mid (NO_M) 6500 MID300 c) Norway North (NO_N) 6700 NOR300 6701 NOR400 7KH�IROORZLQJ�QRGHV�DUH�ORFDWHG�LQ�)LQODQG�� 7000 SO-FIN 7100 NO-FIN 7KH�IROORZLQJ�QRGHV�DUH�ORFDWHG�RQ�-XWODQG��:HVW�'HQPDUN��� 8001 SKAG_150 8002 SKAG_400 8003 KONT_150 8004 KONT_400 8005 JYLLAND 7KH�IROORZLQJ�QRGH�LV�ORFDWHG�RQ�=HDODQG��(DVW�'HQPDUN��� 8500 SJOLLAND 7KH�IROORZLQJ�QRGH�LV�ORFDWHG�LQ�*HUPDQ\�� 9000 TYSKLAN 9001 WIND-TY
40
12x267 TR F6212
)LJXUH�����1DPH�DQG�ORFDWLRQ�RI�WKH�:,/0$5�SURMHFW�GHILQHG�UHJLRQV�LQ�WKH�1RUWKHUQ�(XURSHDQ�V\VWHP��
GER POL
RUS
NO_S
NO_M
SE_M
DK1
SE_N
SE_S
FI
DK2 DC
DC
DC DC
DC
DC
NO_N
41
12x267 TR F6212
������ /RDGV� From the JMM simulations, the loads are given in each JMM region in each hour. The load flow files of the 23 generators model require the loads in nodes. Since each node belongs to a specific region, it is straightforward to distribute the JMM simulated loads in a certain region on the nodes in that region. Formally, the loads ( )K3 (366
ORDGL/ in PSS/E node # L in hour K are given from the loads ( )K3 -00
ORDG5 in
JMM region 5 at hour K according to: ( ) ( )
( ) ( ) 5LK3EDVH3EDVH3K3 -00
ORDG5
5N
(366ORDGN
(366ORDGL(366
ORDGL ∈⋅=∑∈
/
// (MW) 4.1
Here, ( )EDVH3 (366
ORDGL/ is the value of the PSS/E node L in the PSS/E base load flow file.
The conversion of loads is performed automatically by the conversion program “WILMARJMM2PPPE” ������ *HQHUDWRU�SURGXFWLRQ� From the JMM simulations, the generator production is given for each XQLW�JURXS in each hour. A unit group is a collection of generation XQLWV with the same market related characteristics, i.e. same production costs, startup times etc. Units in the same unit group also belong to the same region. The units are specified in the ,QSXW�'DWD�%DVH (IDB), where they are also attached to a certain unit group. The idea has been first to specify the actual units in the system, and subsequently to group them into a reduced number of unit groups in order to make the JMM simulations faster. This way, it should be easier to understand and maintain the input database. A simple way to convert the JMM simulation results to PSS/E input would be to distribute the JMM simulated generator production in a certain region proportionally to the generators in the PSS/E base file in that region, i.e. similar to the conversion of loads. However, since each JMM unit is in principle specified geographically, each unit can in principle be attached to a specific PSS/E generator. The PSS/E generator of a unit is identified in the IDB by adding two fields, specifying a PSS/E node number and generator id respectively to each unit record. Generator id “1” is used for conventional generators, while generator id “2” is used for wind power generators. This selection makes it possible to convert the wind generators to negative loads in the stepwise power flow studies. Although in principle, the units are thus attached to a specific PSS/E generator, the units specified in some regions represent groups of generators, and therefore the PSS/E node field cannot be specified. In that case, the JMM simulated generator production is distributed proportionally to the ratings of the generators in the PSS/E base file.
42
12x267 TR F6212
Formally, the procedure can be expressed as follows: First, the contribution of installed generation capacity JQ*L3 of unit group * to PSS/E generator L in region 5 is summed up according to
∑
∈
⋅=*N
,'%JQNNLJQ*L 33 α (MW) 4.2
Here, ,'%
JQN3 is the installed capacity of unit N according to the IDB, and αNL is a fraction determined
as
( )( )
)id( )id( if 0
)id( )id( and specifiednot is )node( if
)id( )id( and )node( )node( if 1
)()(
/
/
LN
LNNEDVH3EDVH3
LNLN
NL
NLGKLG5K
(366JQK
(366JQN
NL
NL
≠=
==
===
∑=∈
α
α
α
4.3
Note that if node(N) is not specified in the IDB, only the PSS/E generators with the same generator id share the installed capacity. This ensures that conventional generation units without specified PSS/E generator are not converted to wind power production by the data conversion. Together with the contribution of installed capacity to a specific PSS/E generator L, the total installed capacity 3JQ* of unit group * is calculated as ∑
∈
=*N
,'%JQNJQ* 33 (MW) 4.4
Now, for each hour K, the generator production ( )K3 (366
JL/ in each PSS/E generator L is calculated as
the sum of contributions from the productions ( )K3 -00J* of unit groups * as simulated with JMM,
according to ( ) ( )∑
∈
⋅=5*
-00J*
JQ*
JQ*L(366JL K33
3K3 / (MW) 4.5
This conversion of generator production is performed automatically by the conversion program “WILMARJMM2PPPE” ������ 7UDQVPLVVLRQ� JMM calculates the transmission between all the connected regions in WP6 for every hour K. The JMM simulations optimize the operation with specified transmission capacity limits between each region. The applied transmission capacities are mainly based on public available information on www.nordel.org.
43
12x267 TR F6212
When data is converted to WP5 PSS/E load flow files, it is only possible to specify the load flow transmissions in the DC links. The AC links will distribute the flow depending on the impedances in the grid. In the present state, the conversion of transmission in the DC links from WP6 JMM simulation to WP5 PSS/E format is performed manually, based on the JMM simulated transmission. The active power 3'& transmitted in the link is obtained directly from the JMM results, and it is assumed that the converter stations in the DC links consume reactive power 4'& determined as '&'& 34 ⋅= 4.0 4.6
For future work, automatic conversion could be implemented the conversion program “WILMARJMM2PPPE”. As stated above, the flow in the AC links is distributed depending on the impedances in the grid. The present version of the JMM optimizes the power system operation without considering the actual distribution of the flow due to the impedances in the grid. In principle, this can lead to an unrealistic JMM simulation of the transmission, which can cause overload when the load and production is converted to a load flow. Bjørndal and Jørnsten [16] have shown how the influence of the grid reactance can be easily implemented in a market optimization model, using a linear restriction to the flow in loops of the system. In transmission systems, it is generally accepted that the resistance can be neglected, and in that case the restriction for the flow in a loop / can be expressed linearly as 0=⋅∑
∈ /LMLMLM 3; 4.7
Here, ;LM is the reactance in branch between node L�and node M and 3LM is the power flowing in the branch. The equation ensures that the sum of phase angles over the branches in the loop is zero, which is actually as fundamental a restriction as to require that the sum of the flow into a node is zero. One problem is, however, that the “nodes” here represents a region, so some approximations have to be done to determine the reactances between the regions. However, it should be possible to provide applicable values for the reactances, based on the grid in the 23 generators model. It is considered to implement this in future versions of JMM. ������ ([SRUW���LPSRUW� Export to and import from 3rd countries is also simulated in JMM. The export / import is represented in the PSS/E raw file as loads, which are obviously positive in case of export.
44
12x267 TR F6212
The JMM simulations are done with all the regions in the Nordic system and with 3 regions representing Germany. On the other hand, the 23 generators model only has one node to represent Germany. To keep consistency, the one node in the 23 generators model is assumed to represent the North-West region in Germany: DE_NW. Thus, transmission simulated with JMM to other regions is treated as export / import. For export / import on AC lines, the reactive power flow is assumed to be zero while export / import on DC links, ��� is applied. ������ *HQHUDWRU�UDWLQJ��366�(�0EDVH�� For the 23 generators model, the rated power ( )K3 (366
QL/ of each generator L in hour K must be set in
addition to the production ( )K3 (366JL
/ found in section 4.2.4. In PSS/E terms, the rated power in p.u.
is denoted the Mbase. The rated power is calculated based on the same fractions as the generator production in section 4.2.4. JMM assumes primary reserves according to requirements in Nordel and UCTE codes, and for the purpose of transferring data to WP5, JMM generates an output file with the resulting positive primary reserve ( )K3 -00
L*Pr of each unit group *. This primary reserve is used to set the rated power
according to ( ) ( ) ( )( )∑
∈
+⋅=5*
-00L*
-00J*
JQ*
JQ*L(366QL K3K33
3K3 Pr/ (MW) 4.8
This conversion of generator rating is performed automatically by the conversion program “WILMARJMM2PPPE” During the WP5 work it was decided to change the generator ratings resulting from the JMM simulations, so that the generator ratings are set to be 10 % above the actual production in each case. This way one ensured more realistic primary reserves, which are different at low load and high load cases. This correction of the generator ratings was done manually. More specifically, the primary reserves are set to 10 % of the production for generators which primarily represent hydro units, and to 0 % for generators which primarily represent thermal units. This is a simplification, which does not take into account that the thermal units contribute to primary control to meet requirements to the national contributions according to table Table 2.1 and the Nordic Grid Code [4]. However, the actual contribution from thermal units to the primary reserves is quite small, and will therefore have only marginal influence on the frequency bias.
45
12x267 TR F6212
������ %DODQFLQJ�SRZHU�OLVW� The balancing power list is used in the frequency stability studies, and it is a list of secondary reserves, specifying the bus, price and volume as shown in section 3.3.1. The secondary reserves are calculated from JMM simulations based on the same fractions as the generator production and rating in section 4.2.4 and 4.2.7 respectively. The calculation of the secondary reserve is based on the JMM simulated “Capacity On Line” (COL) ( )K3 -00
&2/* . This term should not be understood as the installed generator capacity, but in the
JMM it is “instantaneously” available power understood as available from one hour to the next. In the JMM this capacity can not be higher than the installed capacity times the availability of the capacity. Often the capacity online will be significantly lower than the available capacity in the JMM, because it costs money to keep surplus capacity online. An important exemption to this is hydropower, where the capacity online is put equal to the available capacity, because it is assumed in the JMM that the costs connected to start up and partload operation of hydropower is negligible. COL includes spinning reserve which is not classified as primary reserve and can be ramped up according to the requirements in Nordic Grid Code, but the main contribution to COL is hydro, which has high ramp rates and is able to start production from standstill. Formally, for each hour K, the contribution to the secondary reserve ( )K3 (366
6HF*L/ in each PSS/E
generator L from generators in JMM unit group * is calculated according to ( ) ( ) ( ) ( )K3K3K33
3K3 -003LU*
-00J*
-00&2/*
JQ*
JQ*L(3666HF*L −−⋅=/ (MW) 4.9
Note that the contributions from all unit groups is not summed as it was with production in equation ��� and rated power in equation ���. This is because the price depends on which unit group that provides the reserve. Therefore, a single PSS/E generator will have multiple reserves with different prices. Concerning the price of the reserve power, JMM generates a file with marginal costs of each unit group every hour. This marginal costs in this file is used as the price of the reserve. Equation ��� is the general formula used to determine the secondary reserve. However, the hydro power in a region in JMM is modeled as a plant with a time varying run-of-river generation that occupies differing amounts of the total hydropower capacity and a stepwise supply curve with 10 increasing steps of production costs. This ensures that the optimisation routine in JMM dispatches the hydro power realistically between the regions, and it ensures that the reduction in available controllable hydropower when run-of-river production is large is taken into account. More specifically it is done by representing hydropower in a region in JMM as 11 joined unit groups. The first unit group represents run-of-river (ROR) hydro but not the capacity of ROR, and the
46
12x267 TR F6212
remaining 10 unit groups (named S01, S02, .. S10) represent 10 steps of fully controllable hydro power with increasing production costs for increasing step number. For hydro, the startup costs are set to zero, and therefore the COL ( )K3 -00
&2/* of a hydro unit group G is determined by the constant installed capacity -00
,QVW*3 and the time varying availability ( )K*α
according to ( ) ( ) -00
,QVW**-00&2/* 3KK3 ⋅= α (MW) �����
The total installed capacity ( )K3 -00
J,QVWRe of such a joint of hydro power unit groups in a region is the
constant sum according to ∑
=
+=10
01
Re
LL
-00,QVW6LL
-00,QVW525
-00J,QVW 333 (MW) �����
The amount of ROR hydro generation ( )K3 -00
J525 varies in time, depending on the season. The run-
of-river generation occupies hydropower capacity such that the amount of installed capacity ( )K3 -00
,QVW6LL and corresponding capacity on line ( )K3 -00&2/6LL is reduced with ( )K3 -00
J525 . The reduction is
done such that it is the most expensive hydropower unit groups that get their capacity reduced by run-of-river generation.Therefore, Equation ��� will calculate negative reserves ( )K3 (366
6HF525L/ for the
ROR unit group and corresponding too high reserves ( )K3 (3666HF6LLL
/ for the controllable unit groups.
The conversion of secondary reserves is performed automatically by the conversion program “WILMARJMM2PPPE”. The version of WILMARJMM2PPPE which was used to convert the data applied in the secondary control studies presented in chapter 5 did the mistake to discard the negative reserves ( )K3 (366
6HF525L/ assuming that negative reserves were meaningless, but still to keep the
too high reserves ( )K3 (3666HF6LLL
/ for the unit groups with reservoirs. This mistake has been corrected in
the latest version of WILMARJMM2PPPE, but the secondary control studies have not been redone with the more correct reserve volume, which is reasonable because there was plentiful of other secondary reserves available according to the secondary control studies, and therefore the too high reserves were never allocated and did not influence the result of the studies. The conversion program only include the unit groups with more than 100 MW secondary reserve capacity in the balancing power list. This reduction somehow accounts for that smaller units which are not producing already would require manning of the plants if they bid. Still, the resulting balancing power list may in many cases specify more reserve volume than what is actually available, because the availability factors for hydropower in the JMM are not (yet) properly calibrated. For instance, much of the installed hydro will be out for scheduled maintenance during the summer season, but the JMM has used an availability factor of 1 for hydropower also in the summer. However, the volume of the secondary reserve is not a limiting factor for the secondary control, since the main issue is if the frequency bias is sufficient to cope with the changes in load and generation.
47
12x267 TR F6212
������ :LQG�SRZHU�IRUHFDVW�HUURUV� Wind power forecast errors are calculated for JMM simulations in WP6 using a Scenario Tree Tool developed by the project partner IER in the University of Stuttgart. The basis for this tool is a “Wind Speed Forecast Error Module”, which generates forecasts in the market regions taking into account the correlation between the wind power productions in the regions. This Scenario Tree Tool is extended to provide data for the frequency stability simulations in WP5 as well. For WP6 applications, a typical forecast length is 36 hours corresponding to the NORDEL spot market. To handle this forecast length, 3-6 hours steps has been applied, and a scenario reduction tool has been developed. For the WP5 application, only one hour is simulated, and therefore only one hour forecast length is relevant. Therefore, the scenario reduction is not needed for the WP5 forecast errors. Instead, a “Wind power forecast errors for SPF” module has been developed. The idea has been to develop a tool, which provides a “normal case” and a “worst case” scenario of the one hour forecast error. The “Wind power forecast errors for SPF” module performs the following:
�� For each hour, one hour wind power forecast scenarios for each region considered are generated (using the same modules as for WP6).�
2. For each hour and each scenario, the total system wind power forecast error is determined. Then the wind power forecast errors are summed of each region using error sign (e.g. 10 MW for Region1 and -5 MW for Region2 gives 5 MW for the total system).
3. For each hour, the 1st, 6th, 84th and 99th percentiles of the total system error are determined.
4. For each hour, the values of these percentiles are compared to the scenarios of the total system error and for each percentile the scenario with the most similar total system error is selected individually.
5. For each hour and percentile, the wind power forecast errors within each individual region of these selected scenarios are written to a file, which is used in WP5.
The 16th and 84th percentiles are normal cases whereas the 1st and 99th percentiles are worst cases with less and more wind power production than predicted.
48
12x267 TR F6212
�� 6(&21'$5<�&21752/� The present chapter describes the results from simulations with different tools described in Chapter 3 to analyse secondary control within the hour for selected cases simulated in WP6. The tools used for simulations are the Stepwise Power Flow routine, LP optimization, and PSS/E. The data for the selected cases simulated in WP6 are described in Section 5.1.1-5.1.2. In all cases the total wind decreases during the simulated hour and the total load increases during the simulated hour. It is important to keep in mind that the results build on a simplified model of the Northern-European system. Therefore they must be considered as indicative rather than results that should be a basis for decisions and conclusions without any further investigations and analysis. ���� &DVH�GHVFULSWLRQ� ������ �����FDVH� The 2001 case serves as a test case, which is applied to all the tools in the secondary control studies. We have looked for a case where the load increases and the wind power decreases. The case is based on one of the first successful simulations with JMM including all countries. At that time, the JMM simulations were very time consuming, and only one week was simulated. Figure 5.1 shows the demands (or loads) in each of the JMM regions during the simulated weeks. Only the regions in the Nordic synchronous system are shown, because this is the system analysed in the secondary control studies. 1 January is a Monday, and daily as well as weekly variations are clearly observed. It is clear that if we look for one hour where the load increases, we must select a morning hour.
49
12x267 TR F6212
)LJXUH�����'HPDQGV�LQ�-00�UHJLRQV�IRU�WKH�ILUVW�ZHHN�LQ������� Figure 5.2 shows the simultaneous wind power variations. From the graph it can be assessed that the largest decrease in wind power in morning hours is on 6 January. However, this is a Saturday where the load does not increase as fast as on the working days. Therefore, 5 January from 7 to 8 in the morning has been selected as case. During that hour, the load in the synchronous system increases 1793 MW while the wind power decreases by 7 MW (from 232 MW to 225 MW).
0
2000
4000
6000
8000
10000
12000
14000
16000
01-Jan 02-Jan 03-Jan 04-Jan 05-Jan 06-Jan 07-Jan
Load
[MW
]
DK_EFI_RNO_MNO_NNO_SSE_MSE_NSE_S
50
12x267 TR F6212
)LJXUH�����:LQG�SRZHU�SURGXFWLRQV�LQ�UHOHYDQW�-00�UHJLRQV�IRU�WKH�ILUVW�ZHHN�LQ������� Table 5.1 shows the production as it has been distributed on nodes according to section 4.2.4. The primary reserves have been set to 10 % of the production for generators mainly representing hydro power, while other primary reserves are set to 0%. The secondary reserves are calculated automatically according to Section 4.2.8 in the first place, but more than half of the calculated secondary reserves have been removed manually afterwards, because the volume and some of the prices provided by this early version of JMM were not realistic. Finally, the installed capacities have been calculated, based on the installed capacities given in the common Input Data Base.
0
100
200
300
400
500
600
01-Jan 02-Jan 03-Jan 04-Jan 05-Jan 06-Jan 07-Jan
Win
d po
wer
[MW
]
NordicSyncDK_EFI_RSE_MSE_S
51
12x267 TR F6212
7DEOH������3URGXFWLRQ��SULPDU\�UHVHUYHV��VHFRQGDU\�UHVHUYHV�DQG�LQVWDOOHG�JHQHUDWLRQ�FDSDFLW\�DVVXPHG�LQ�WKH������FDVH��Node� Production
[MW] �Prim. reserve
[MW] �Sec. reserve
[MW] �Installed [MW] �
����� 6450 0*) 0 9769
����� 1471 147 0 2391
����� 2343 234 100 3808
����� 98 10 0 293
����� 873 87 0 1419
����� 3229 323 0 5247
����� 1308 0*) 0 3017
����� 4513 0*) 40 6479
����� 1554 155 140 2591
����� 2374 237 210 3959
����� 1640 164 140 2735
����� 863 86 80 1440
����� 2374 237 210 3959
����� 86 9 0 144
����� 1209 121 110 2015
����� 2503 250 220 4175
����� 3107 311 120 5410
����� 866 87 150 1714
����� 6417 0*) 0 12432
����� 2341 234 40 10785
����� 2739 0*) 230 4376
727$/� ������ ����� ���� �� 88158�*) According to section 4.2.7, the primary reserves are set to 10 % of the production for generators which primarily represent hydro units, and to 0
% for generators which primarily represent thermal units. This is a simplification, which does not take into account that the thermal units contribute
to primary control to meet requirements to the national contributions according to table Table 2.1 and the Nordic Grid Code [4].
**) The secondary reserves are calculated automatically according to Section 4.2.8 in the first place, but more than half of the calculated secondary
reserves have been removed manually afterwards, because the volume and some of the prices provided by this early version of JMM were not
realistic.
The wind power forecast error is calculated with the Scenario Tree Tool according to Section 4.2.9. The resulting 1% (“worst case”) is shown in Table 5.2. A negative forecast error means that the wind production is less than predicted.
52
12x267 TR F6212
7DEOH������)RUHFDVW�HUURUV�XVHG�LQ�WKH������FDVH��5HJLRQ� )RUHFDVW�HUURU�
>0:@�'.B(� -37
),B5�� 1
12B0�� -1
6(B0�� -4
6(B1�� 2
6(B6�� -3
727$/�1RUGLF�6\QF� ���� The forecast HUURUV�LQ�7DEOH�����DUH�TXLWe small. This is partly because the installed wind power is relatively small in the Nordic Synchronous system. Most of the Danish wind power is installed in DK_W, which does not belong to the Nordic Synchronous system. In the selected hour, the wind power production is only 232 MW. Another reason why the forecast error is relatively small is that the “worst case” 1 % percentile is found for the total wind power forecast error including West Denmark and Germany. For future studies of secondary control in the Nordic system, it would be more correct to identify the “worst case” based on the total forecast error in the Nordic synchronous system only. ������ ������FDVHV� Two 2010 scenarios have been defined in WP6. The scenarios involve commissioning of new power plants, decommissioning of existing plants, prognosis for load development, and reinforcement of the transmission system. The general rule in the selection of new power plant investments has been to only select plants that are under construction or far in the planning process. The main transmission system reinforcement is the DC connections: Fennoscan 2, Storebælt and Norway-Nederland. Figure 5.3 shows the assumed development in electricity consumption from the 2004 data to the 2010 scenario. Generally, it is observed that a moderate increase in consumption is assumed.
53
12x267 TR F6212
A more significant change is assumed on the development in installed wind power as seen in Figure 5.4. In the present (2004) situation, wind power is mainly installed in Denmark, and most of the Danish wind power is installed in West. Therefore, the present wind power capacity in the Nordic synchronous system is only 1443 MW of which more than half is installed in Denmark East. The 2010 base wind scenario assumes a total of 4100 MW wind power in the Nordic synchronous system, whereas the 10% high wind scenario assumes 18652 MW. In both 2010 scenarios, the wind power is distributed more evenly on the countries than in 2004 as seen in Figure 5.4.
0
20
40
60
80
100
120
140
160
180
Denmark(East)
Finland Sweden Norway
Country
Ele
ctric
ity c
onsu
mpt
ion
[MW
]
20042010
)LJXUH�����(OHFWULFLW\�GHPDQG�GHYHORSPHQW�LQ�WKH������VFHQDULR��
0
1000
2000
3000
4000
5000
DK_E FI_R NO_M NO_N NO_S SE_M SE_N SE_S
5HJLRQ
,QVWDO
OHG�Z
LQG�SR
ZHU�>0:
@
End 20042010 base wind2010 10% high wind
)LJXUH�����,QVWDOOHG�ZLQG�SRZHU�LQ������EDVH�ZLQG�VFHQDULR�DQG������KLJK�ZLQG�VFHQDULRV��)RU�UHIHUHQFH��WKH�DFWXDO�LQVWDOOHG�ZLQG�SRZHU�LQ�HQG������LV�DOVR�JLYHQ��
54
12x267 TR F6212
����������%DVH�ZLQG�VFHQDULR� Two weeks have been simulated with the 2010 base wind scenario. The wind power production during those two weeks is shown in Figure 5.5.
Again, a morning hour with load increase and wind power decrease has been identified: 22 Feb from 6 to 7. The electricity consumption increases with 3592 MW in that hour (from 56663 MW), and the wind power production in the Nordic synchronous system decreases with 57 MW (from 2093 MW). Table 5.3 shows the production as it has been distributed on nodes according to section 4.2.4. The primary reserves have been set to 10 % of the production for generators mainly representing hydro power, while other primary reserves are set to 0%. The secondary reserves are calculated automatically according to Section 4.2.8, i.e. only the unit groups with more than 100 MW secondary reserve capacity have been included. Finally, the installed capacities have been calculated, based on the installed capacities given in the common Input Data Base.
0
500
1000
1500
2000
2500
3000
3500
12 F
eb
13 F
eb
14 F
eb
15 F
eb
16 F
eb
17 F
eb
18 F
eb
19 F
eb
20 F
eb
21 F
eb
22 F
eb
23 F
eb
24 F
eb
25 F
eb
:LQG
�SRZH
U�>0:
@ NordicSyncDK_EFI_RNO_MNO_SSE_MSE_NSE_S
)LJXUH�����:LQG�SRZHU�SURGXFWLRQV�LQ�1RUGLF�V\QFKURQRXV�V\VWHP�IRU�WZR�ZHHNV�LQ�)HEUXDU\�������
55
12x267 TR F6212
7DEOH����. Production, primary reserves, secondary reserves and installed generation capacity assumed in the 2010 base wind case.
Node� Production [MW] �
Prim. reserve [MW] �
Sec. reserve [MW] �
Installed [MW] �
����� 7396 0*) 300 10473
����� 1388 139 709 2391
����� 2211 221 1128 3808
����� 146 15 0 348
����� 824 82 421 1419
����� 3047 305 1555 5247
����� 832 0*) 0 2603
����� 5248 0*) 133 7277
����� 1442 144 867 2694
����� 2203 220 1324 4116
����� 1522 152 915 2844
����� 801 80 397 1497
����� 2203 220 1324 4116
����� 80 8 0 150
����� 1122 112 674 2095
����� 2323 232 1397 4340
����� 3141 314 1584 5510
����� 940 94 501 1814
����� 9144 0*) 359 13713
����� 2709 271 515 5268
����� 2533 0*) 335 3866
727$/� 51255� 2610� 14438**)� 85589�*) The primary reserves are set to 10 % of the production for generators which primarily represent hydro units, and to 0 % for generators which
primarily represent thermal units. This is a simplification, which does not take into account that the thermal units contribute to primary control to
meet requirements to the national contributions according to table Table 2.1 and the Nordic Grid Code [4].
**) The secondary reserves are calculated automatically according to Section 4.2.8, i.e. the resulting balancing power list may specify more reserve
volume than what is realistically available, because the availability of hydropower in the JMM run has been set to 1, which is a too large value.
The error in the conversion program because of full water reservoirs does not influence this case, because no reservoirs were full in this February
case..� The wind power forecast error is calculated with the Scenario Tree Tool according to Section 4.2.9. The resulting 1% (“worst case”) is shown in Table 5.4. A negative forecast error means that the wind production is less than predicted.
56
12x267 TR F6212
7DEOH������)RUHFDVW�HUURUV�XVHG�LQ�WKH������EDVH�ZLQG�FDVH��5HJLRQ� )RUHFDVW�HUURU�
>0:@�'.B(� -12.99
),B5�� 59.78
12B0�� 6.98
12B6�� 10.30
6(B0�� 27.18
6(B1�� -18.76
6(B6�� -27.62
727$/�1RUGLF�6\QF� 44.87� Note that the total forecast error in the Nordic Synchronous system in Table 5.4 is positive, corresponding to higher actual wind power production than forecasted. This is because the 1% percentile is identified for the total wind power forecast error including West Denmark and Germany as mentioned in Section 5.1.1. This confirms that for future studies of secondary control in the Nordic system, the “worst case” should be identified based on the total forecast error in the Nordic synchronous system only.�������������+LJK�ZLQG�VFHQDULR� The high wind scenario case has been selected based on 10 % high wind scenario data for all 2010. Figure 5.6 shows demand and wind power in the Nordic synchronous system for all 2010.
57
12x267 TR F6212
The idea of selecting the case based on a whole year of data is to come closer to a worst-case. Again, we are looking for a morning hour with load increase and wind power decrease, but on top of that, we are looking for a low-load (low-demand) hour. With these criteria, 5 July from 5 to 6 has been selected. Table 5.5 shows the production in the high wind case as it has been distributed on nodes according to section 4.2.4. The primary reserves have been set to 10 % of the production for generators mainly representing hydro power, while other primary reserves are set to 0%. The secondary reserves are calculated automatically according to Section 4.2.8 in the first place, but only the unit groups with more than 100 MW secondary reserve capacity have been included. Finally, the installed capacities have been calculated, based on the installed capacities given in the common Input Data Base.
0
10000
20000
30000
40000
50000
60000
70000
80000
Jan-2010 Apr-2010 Jul-2010 Oct-2010 Jan-2011
>0:@ Demand
Wind
)LJXUH�����:LQG�SRZHU�SURGXFWLRQV�DQG�GHPDQG�LQ�1RUGLF�V\QFKURQRXV�V\VWHP�IRU�������
58
12x267 TR F6212
7DEOH����. 3URGXFWLRQ��SULPDU\�UHVHUYHV��VHFRQGDU\�UHVHUYHV�DQG�LQVWDOOHG�JHQHUDWLRQ�FDSDFLW\�DVVXPHG�LQ�WKH������KLJK�ZLQG�FDVH��Node� Production
[MW] �Prim. reserve
[MW] �Sec. reserve
[MW] �Installed [MW] �
����� 3516 0*) 348 10473
����� 1210 121 2097 2391
����� 1927 193 3339 3808
����� 62 6 0 348
����� 718 72 1245 1419
����� 2656 266 4601 5247
����� 387 0*) 0 2603
����� 2562 0*) 266 7277
����� 1284 128 2303 2694
����� 1962 196 3518 4116
����� 1356 136 2431 2844
����� 713 71 1279 1497
����� 1962 196 3518 4116
����� 71 7 0 150
����� 999 100 1791 2095
����� 2069 207 3710 4340
����� 1449 145 4806 5510
����� 829 83 468 1814
����� 4363 0*) 0 13713
����� 1263 126 964 5268
����� 803 0*) 176 3866
727$/� 32162� 2053� 36860**)� 85589�*) The primary reserves are set to 10 % of the production for generators which primarily represent hydro units, and to 0 % for generators which
primarily represent thermal units. This is a simplification, which does not take into account that the thermal units contribute to primary control to
meet requirements to the national contributions according to table Table 2.1 and the Nordic Grid Code [4].
**) The secondary reserves are calculated automatically according to Section 4.2.8, i.e. the resulting balancing power list probably specify more
reserve volume than what is realistically available. With the corrected conversion program taking into account capacity reserved for ROR
generation (see Section 4.2.8), the total secondary reserve was calculated to 20784 MW, i.e. much less than the 36860 MW used in the studies. The
error in the version of the conversion program used to calculate the shown secondary reserves also explains why the sum of production and
reserves in some nodes above is greater than the installed capacity! However, this did not influence the result as stated in Section 4.2.8.
�The wind power forecast error is calculated with the Scenario Tree Tool according to Section 4.2.9. The resulting 1% (“worst case”) is shown in Table 5.6. A negative forecast error means that the wind production is less than predicted.
59
12x267 TR F6212
7DEOH������)RUHFDVW�HUURUV�XVHG�LQ�WKH������KLJK�ZLQG�FDVH� 5HJLRQ� )RUHFDVW�HUURU�
>0:@�'.B(� 84.78
),B5�� 24.32
12B0�� -36.18
12B1�� -59.34
12B6�� 104.61
6(B0�� 0.12
6(B1�� -16.37
6(B6�� 28.46
727$/�1RUGLF�6\QF� 130.4� Again, the total forecast error in the Nordic Synchronous system in Table 5.6 is positive, corresponding to higher actual wind power production than forecasted. ���� 5HVXOWV�IURP�6WHSZLVH�3RZHU�)ORZ�FDOFXODWLRQV� ������ �����FDVH� Only for the Nordel system the response in the system frequency as load and wind power production is considered. Jutland and Northern Germany belongs to the UCTE system which is considered so strong that the simulated changes in load/production in Jutland and Northern Germany are not significantly influencing the frequency in the UCTE system. Only the flow on the AC-line between Jutland and Germany is considered. As it is hydro power that is mainly used for up and down regulation of production (thermal power and especially nuclear power tends to have less changes in production for up and down regulation) it is assumed that the scheduled production is put online within 5 minutes and that the ramping of the scheduled production is starting after 30 minutes of the simulated hour. There is a considerable change in flow on the Skagerrak HVDC-links (DK1-NO_S) and Kontiskan HVDC-links (DK1-SE_M) at the beginning of the hour and at the end of the hour. This change is assumed to be ramped in 15 minutes and the ramp starts after 25 minutes of the simulated hour. The ramping of these links does not exceed the Nordel maximum ramp limit of 30 MW/min [4]. The droop of the generators in the Nordel system is set to 6 % for hydro generators and 0 % for all thermal and nuclear generators. The primary reserves are set to be 10 % of the online capacity. This gives a system bias equal to 9917.7 MW/Hz in the beginning of the hour and a system bias equal to 10795.6 MW/Hz in the end of the hour for the Nordel system.
60
12x267 TR F6212
The response in frequency in the Nordel system is shown in Figure 5.7. The response is for the case without any forecast errors. However, these are so small for the 2001 case that they have been neglected (i.e. they are so small that they are not assumed to significantly influence the results that are simulated without any forecast errors). The frequency drops below 49.9 Hz at the time step representing 30 minutes of the simulated hour. This is seen by following the blue line. However, the red line shows that the system operator would avoid this by activating balancing power causing the frequency to be 49.925 Hz after 30 minutes. The amount of balancing power activated was 370.7 MW, and it was activated in Northern Finland (40 MW), Zealand (230 MW) and in Southern Norway (100.7 MW). The generators activated were the cheapest units on the bid list for balancing power. The change in the slope of the frequency fall between 25 and 30 minutes is due to the ramping of the HVDC lines and not a change in the system bias. Thus, it is correct that also for the blue bars the system bias remains unchanged (these correspond to the bias if following the blue line for the system frequency, and likewise the red bars correspond to the bias if following the red curve for the system frequency.). In Table 5.7 the flow on the corridors between the regions in the JJM model is compared to what is expected to be the maximum transmission limits between these regions. The table shows that most of the corridors are not overloaded. The only transmission that clearly is overloaded is the Fennoskan link (SE_M-DK1). However, this is caused by the (wrong) assumption in the WP6 simulations, that the Fennoskan link has a transmission limit of 730 MW. The limit for power flowing from GER to DK1 is also set too high compared to its real limit, but it was convenient on a certain point to take a decision on the transmission capacity for all WP5 partners and therefore the limits shown in Table 5.7 were set as reference. Later it was found in [20, p.17] that this limit is 800 MW (The limit for power flowing the other way from DK1 to GER is 1200 MW as in Table 5.7). In the simulated case a maximum of 825.9 MVA (780.8 MW) is flowing into the Danish (DK1) region, and this corresponds to a 852.1 MVA (836.6 MW) injected on the transmission GER-DK1 on the German side. However, this is not a heavy overload (if at all considering that the flow into the Danish system is within the 800 MW limit), and it is expected that for a short period like in the simulation it could be accepted.
61
12x267 TR F6212
0 10 20 30 40 50 6049.85
49.9
49.95
50
50.05
50.1
Time [min]
Fre
quen
cy [
Hz]
0 5 10 15 20 25 30 35 40 45 50 55 600
5000
10000
15000
Time [min]
Bia
s [M
W/H
z]
)LJXUH�����5HVSRQVH�LQ�V\VWHP�IUHTXHQF\�WR�FKDQJH�LQ�ORDG�DQG�ZLQG�SRZHU�SURGXFWLRQ��6FKHGXOHG�SURGXFWLRQ�LV�SXW�RQOLQH�DIWHU����PLQXWHV��� 7DEOH�����$FWXDO�DQG�PD[LPXP�DOORZHG�IORZ�LQ�WKH�FRUULGRUV�EHWZHHQ�WKH�:3��UHJLRQV� &RUULGRU� 0D[LPXP�)ORZ�
�09$��0D[LPXP�DOORZHG�IORZ��0:)
5HJLRQ�$� 5HJLRQ�%� $�±�%� $��%� %�±�$�NO_N NO_M 331.7 600 600 NO_N SE_N 590.2 600 700 NO_M SE_N 288.1 600 500 NO_M NO_S 88.7 300 300 NO_S SE_M 185.3 2050 1850 NO_S DK1 1008.7 1040 1040 SE_N FI 934.4 1600 1200 SE_N SE_M 3253.9 7000 7000 SE_M FI 0 550 550 SE_M DK1 734.8 670 630 SE_M SE_S 824.5 4000 4000 SE_S GER 371.3 600 600 SE_S DK2 1169.1 1775 1700 DK2 DK1 0 600 600 DK1 GER 825.9 1200 1200
62
12x267 TR F6212
������ �����FDVHV� Four cases have been simulated. Case 1 and case 2 are simulation of the base wind scenario described in Section 5.1.2.1 As these are case with relative high Nordel system load they are referred to as high load cases in the present section. Case 3 and case 4 are simulation of the high wind scenario (relative high wind power production compared to the system load) described in Section 5.1.2.2 As these are cases with relative low load in the Nordel system they are referred as low load cases in the present section. The same conditions for droop, ramping of scheduled production and the UCTE system as for the 2001 case are valid here. However, in the simulations for case 1 and case 2 of the 2010 scenario the ramping of the HVDC connections were done slower as it is much more power that needs to be ramped than in the 2001 case. Thus, one ramp the HVDC-links from 15 minutes of the simulation time until 45 minutes of the simulation time. This way one manage to fulfil the NORDEL requirement on the DK1-NO_S and DK1-SE_ HVDC-links which is a maximum planned ramping of 30 MW/min (see [4] Bilaga 7.7). This limit is also assumed for the other modelled HVDC-links. Case 1 and Case 2 was also simulated with LP optimisation calculations, the results from these calculations are described in Section 5.3.2. In the simulations in case 3 and case 4 of the 2010 scenario the ramping of the HVDC-link flows were done equally as the ramping of the load. This means a linear ramping with the same ramp rate throughout the whole simulated hour. As changes in transfer on the HVDC-links were relatively small from the beginning of the hour and to the end of the hour, the practical impact of this on system frequency and flow is not expected to be large compared to a ramping only lasting for instance 15 minutes. &DVH����+LJK�ORDG�ZLWK�IRUHFDVW�HUURUV�IRU�ZLQG�SRZHU�SURGXFWLRQ�H[FOXGHG��The response in frequency for the high load case without any forecast errors is shown in Figure 5.8. Several times the system operator here needs to activate balancing power. After 20 minutes the system operator has activated 466.4 MW balancing power, after 25 minutes 743.2 MW more has been activated, and finally between 25 minutes and 30 minutes 786.9 MW of balancing power is activated. This gives a total of 1996.5 MW of balancing power activated before hour shift, which is well within the balancing power requirement in the existing Nordel requirements for secondary reserve (see Table 2.1) which totals to approximately 5000 MW (4400 MW if disregarding Jutland) and within the available secondary reserves simulated with JMM. However, the simulation shows that despite of this the system frequency is not kept above 49.9 Hz. This is due to the simulated time it takes for the system operator to activate balancing power (see Eq. 3.15). The rule for simulation of activation of balancing power might be discussed, but it is not considered to be unreasonable that 2000 MW is put online within 15 minutes. The simulation show the challenges the system operators might encounter with high change in load, HVDC-ramping and change in total wind power production at the same time. A lot of trading on the balancing power market is necessary to avoid the frequency to drop below 49.9 Hz.
63
12x267 TR F6212
After the scheduled production has been put online, the frequency exceeds 50.1 Hz. Currently the Stepwise Power Flow routine does not take any action for over frequency. Thus, this has not been a focus. One could argue that it is enough to study under-frequency because as with over frequency it is a result of change in load, HVDC-ramping, and change in wind power production. To deal with over over-frequency in the context of secondary control one would have to include a bid list for disconnecting production/connecting load. Table 5.8 shows that there are no significant problems with overloads on lines between regions during the simulated hour. The flow of active power on the connection NO_N – SE_N is from SE_N to NO_N, thus the maximum flow on this connection is within the allowed limits. As mentioned earlier the flow of power limit on the HVDC-connection between SE_M and DK1 was set to 730 MW in the WP6 simulation and therefore this connection is overloaded. There is also a small overload on the line between DK1 and GER. The flow of active power is from DK1 to GER. In MW the maximum import to GER is 1130.3 MW, and therefore it is within the MW limit if one measures the flow on the German side. If the flow is measured on the Danish side the flow in MW is 1250.6 MW, which is a 4.2 % above the maximum flow limit. In some cases the system operator may have chosen to trade on the balancing power market in order to avoid this. Nevertheless, such a small flow above the capacity may be acceptable for a limited period.
0 10 20 30 40 50 6049.8
49.9
50
50.1
50.2
Time [min]
Fre
quen
cy [
Hz]
0 5 10 15 20 25 30 35 40 45 50 55 600
5000
10000
15000
Time [min]
Bia
s [M
W/H
z]
)LJXUH�����5HVSRQVH�LQ�V\VWHP�IUHTXHQF\�WR�FKDQJH�LQ�ORDG�DQG�ZLQG�SRZHU�SURGXFWLRQ��6FKHGXOHG�SURGXFWLRQ�LV�SXW�RQOLQH�DIWHU����PLQXWHV��
64
12x267 TR F6212
7DEOH�����$FWXDO�DQG�PD[LPXP�DOORZHG�IORZ�LQ�WKH�FRUULGRUV�EHWZHHQ�WKH�:3��UHJLRQV� &RUULGRU� 0D[LPXP�)ORZ�
�09$��0D[LPXP�DOORZHG�IORZ��0:)
5HJLRQ�$� 5HJLRQ�%� $�±�%� $���%� %�±�$�NO_N NO_M 344.3 600 600 NO_N SE_N 635.1 (SE_N-NO_N) 600 700 NO_M SE_N 527.9 600 500 NO_M NO_S 156.7 300 300 NO_S SE_M 1234.9 2050 1850 NO_S DK1 1008.7 1040 1040 SE_N FI 384.87 1600 1200 SE_N SE_M 2477 7000 7000 SE_M FI 488.9 550 550 SE_M DK1 734.8 670 630 SE_M SE_S 1320.1 4000 4000 SE_S GER 371.3 600 600 SE_S DK2 1606.9 1775 1700 DK2 DK1 228.7 600 600 DK1 GER 1252.9 (DK1-GER) 1200 1200 �&DVH����+LJK�ORDG�ZLWK�IRUHFDVW�HUURUV�IRU�ZLQG�SRZHU�SURGXFWLRQ�LQFOXGHG��The forecast errors for wind power production are also an input from WP6. They are calculated as one-hour ahead worse case forecast-errors for the wind power production in each defined region in the Northern European system. This means that the given forecast errors show how much the predicted wind in worse case could be wrong in the end of the simulated hour. For the case described in the present section the forecast errors became (see Table 5.4): NO_N 0 MW NO_M 6.98 MW NO_S 10.30 MW SE_N -18.76 MW SE_M 27.18 MW SE_S -27.62 MW FI 59.78 MW DK2 -12.99 MW It is seen that the forecast errors in the Nordel system (all regions except DK1) are quite small and to a large degree they cancel each other. The total forecast error in Nordel is only 44.87 MW. This forecast error is in addition positive and indicates an improved frequency response compared to the case without forecast errors. Due to the small forecast errors, it is not considered as necessary to show detailed results of the simulated line flows with forecast errors included. From Figure 5.9 one can observe that the response in system frequency is when including forecast errors in the
65
12x267 TR F6212
wind is almost identical to the case where the forecast errors in the wind was excluded (see Figure 5.8).
0 10 20 30 40 50 6049.8
49.9
50
50.1
50.2
Time [min]
Fre
quen
cy [
Hz]
0 5 10 15 20 25 30 35 40 45 50 55 600
5000
10000
15000
Time [min]
Bia
s [M
W/H
z]
)LJXUH�����5HVSRQVH�LQ�V\VWHP�IUHTXHQF\�WR�FKDQJH�LQ�ORDG�DQG�ZLQG�SRZHU�SURGXFWLRQ��6FKHGXOHG�SURGXFWLRQ�LV�SXW�RQOLQH�DIWHU����PLQXWHV��� &DVH����/RZ�ORDG�FDVH�ZLWKRXW�IRUHFDVW�HUURUV�LQ�ZLQG�SRZHU�SURGXFWLRQ�LQFOXGHG��The simulated frequency response from Stepwise Power Flow calculations is shown in Figure 5.10. It is seen that the system frequency stays above the 49.9 Hz limit. After 15 minutes 289.8 MW of balancing power has been activated. The location of this power is in Mid-Norway. After 20 minutes 434.7 MW more balancing power has been activated. 244.2 MW of this was activated in Mid-Norway and the rest of this in Northern-Sweden. Then no further balancing power is observed activated until 30 minutes of the simulation. The activated balancing power that then comes online is 359.2 MW and it is located in Northern-Sweden. The total amount of activated balancing power before hour shift was therefore 1038.1 MW, and it is well within the available balancing power in Nordel for the simulated hour. For this case there are at least two unacceptable overloads on the transmission lines between the regions. These are the transmission between Mid-Norway and Northern-Sweden and the transmission between Northern-Sweden and Finland.
66
12x267 TR F6212
Table 5.9 shows that the overload is more than 100 % on the transmission between Mid-Norway and Sweden. The maximum overload occurs before hour shift and is enhanced by all the activation of balancing power in Mid-Norway. If for instance removing the bids in Mid-Norway from the balancing power bid list, the maximum overload before hour shift would only have become 918.1 MVA (and 918.1 MW) on the line between Mid-Norway and Northern-Sweden. Comparing this with Table 5.9 gives an approximately 50 % reduction in overload. This shows that location of balancing power is also an important factor when choosing the bids and not only the price of balancing power. The case also indicates that an upgrade of the line Mid-Norway – Northern-Sweden is necessary for the large scale integration of wind power in Mid-Norway (in the case more than 1600 MW wind power is online in Mid-Norway at the start of the simulated hour). The Norwegian TSO is aware of the capacity problem between Mid-Norway and Northern-Sweden and an upgrade of the transmission capacity on this line is planned in 2008/2009 [23]. Taking this into account would maybe provide the needed transmission capacity between these regions in the simulated case, but if not one would have to trade balancing power (for instance increasing production in Sweden and decreasing production in Mid-Norway). The Stepwise Power Flow in the current version do not activate balancing power due to overload of AC-lines (except Germany-Denmark), thus one could not prevent this overload by trading balancing power during the simulation. However, the planning of the system operation should normally prevent the need to trade balancing power due to overloads. Finally, the case may also show the importance of including impedances in the reserve analysis. Not only to reveal overloads, but also to show how the power is flowing between regions. For instance does Mid-Norway in this case import power from Northern-Norway. One could expect that in sum there is enough transmission capacity out of Mid-Norway, but due to the impedances in the grid, some of the power in Northern-Norway flows into Mid-Norway and are worsening the transmission problem between Mid-Norway and Northern-Sweden. A plain planning tool without any grid modelling would not reveal this problem. The transmission capacity between Northern-Sweden and Finland is also overloaded. In this case it is simply because there is a too large deviation between the load and production in Finland. Thus, one would have to trade balancing power to prevent this.
67
12x267 TR F6212
0 10 20 30 40 50 6049.85
49.9
49.95
50
50.05
50.1
50.15
Time [min]
Fre
quen
cy [
Hz]
0 5 10 15 20 25 30 35 40 45 50 55 600
2000
4000
6000
8000
10000
Time [min]
Bia
s [M
W/H
z]
)LJXUH������5HVSRQVH�LQ�V\VWHP�IUHTXHQF\�WR�FKDQJH�LQ�ORDG�DQG�ZLQG�SRZHU�SURGXFWLRQ��6FKHGXOHG�SURGXFWLRQ�LV�SXW�RQOLQH�DIWHU����PLQXWHV�DQG�IRUHFDVW�HUURUV�DUH�QRW�LQFOXGHG�� 7DEOH�����$FWXDO�DQG�PD[LPXP�DOORZHG�IORZ�LQ�WKH�FRUULGRUV�EHWZHHQ�WKH�:3��UHJLRQV� &RUULGRU� 0D[LPXP�)ORZ�
�09$��0D[LPXP�DOORZHG�IORZ��0:)
5HJLRQ�$� 5HJLRQ�%� $�±�%� $���%� %�±�$�NO_N NO_M 256.8 600 600 NO_N SE_N 351.6 600 700 NO_M SE_N 1248.1 (1242.1 MW,
NO_M-SE_N) 600 500
NO_M NO_S 113 300 300 NO_S SE_M 2147.6 (2093.6 MW
NO_S-SE_M) 2050 1850
NO_S DK1 958.2 1040 1040 SE_N FI 2174 (2137.3 MW,
SE_N-FI) 1600 1200
SE_N SE_M 4611.4 7000 7000 SE_M FI 470.8 550 550 SE_M DK1 573.2 670 630 SE_M SE_S 3126.4 4000 4000 SE_S GER 415.8 600 600 SE_S DK2 1607.4 1775 1700 DK2 DK1 442.6 600 600 DK1 GER 1147.9 1200 1200
68
12x267 TR F6212
�&DVH����/RZ�ORDG�FDVH�ZLWK�IRUHFDVW�HUURUV�LQ�ZLQG�SRZHU�SURGXFWLRQ�LQFOXGHG��Also for this case the forecast errors are quite small. Thus, the frequency response looks quite similar as in Case 3. The line flow of power between the regions will only be moderately changed. For instance is the maximum flow between Mid-Norway and Northern-Sweden now 1233.5 MVA (1232.3 MW). This is only a 1.2 % difference with the case without forecast errors. All in all the forecast errors will not significantly change the conclusions made out of the results of Case 3. � ���� 5HVXOWV�IURP�/3�RSWLPLVDWLRQ�FDOFXODWLRQV� In the present section results from LP optimisation calculations of the 2001 case and case 1 plus case 2 of the 2010 case are presented and compared with the corresponding results from SPF. These simulations were performed for comparison of results and verification of the methods. ������ �����FDVH� The same case for 2001 is optimised as simulated with SPF. The response in frequency in the Nordel system is shown in Figure 5.11. The blue line shows the situation without activation of balancing power and the red line shows the result after activation of 78 MW after 30 minutes. The total amount of balancing power is 292.7 MW lower than in the case with the SPF-method. This is partly due to the fact that the frequency is lifted to 49.9 Hz at 30 minutes (hour shift) while in the SPF-simulation it was lifted to 49.925 Hz which requires 247.9 MW in addition of balancing power. Other causes for the rest of the 292.7 MW difference between LP optimization and SPF could be differences in losses, and the way the balancing power is activated. For instance, in SPF it is the rating of the generators that is increased according to the amount of balancing power needed and not the actual production. While in LP optimization calculations it is the actual production that is increased according to the bid-list. The transmission between the regions differs only from the results shown in Table 5.7 with a few percent.
0 10 20 30 40 50 6049.85
49.9
49.95
50
50.05
50.1
Time [min]
Fre
quen
cy [
Hz]
)LJXUH������5HVSRQVH�LQ�V\VWHP�IUHTXHQF\�WR�FKDQJH�LQ�ORDG�DQG�SURGXFWLRQ��
69
12x267 TR F6212
������ �����FDVH� For the 2010 case without forecast errors in wind power production, the response in frequency in the Nordel system is shown in Figure 5.12. After 20 minutes is 431 MW secondary control activated, after 25 minutes is 526 MW more activated and after 30 minutes is 526 MW more activated, which gives a total amount of 1482 MW activated power. The blue line shows again the situation without balancing power. The total amount of balancing power is 514.5 MW lower than in the case with the SPF-method. This is partly due to the fact that the frequency is lifted to 49.9 Hz at 30 minutes (hour shift) while in the SPF-simulation it was lifted to 49.9373 Hz which requires 358.9 MW in addition of balancing power. Other causes for the rest of the 514.5 MW difference between LP optimisation and SPF could be differences in losses, could be the same as described for the comparison of the 2001 case. As mentioned before, it was here necessary to skip the upper limit of the frequency, since no down-regulation is available.
0 10 20 30 40 50 6049.7
49.8
49.9
50
50.1
50.2
Time [min]
Fre
quen
cy [
Hz]
)LJXUH������5HVSRQVH�LQ�V\VWHP�IUHTXHQF\�WR�FKDQJH�LQ�ORDG�DQG�SURGXFWLRQ�ZLWKRXW�ZLQG�SRZHU�� As with SPF, the results when including forecast errors of wind power has a small impact on the frequency, see Figure 5.13. The frequency is a little higher, because of the positive error, but less balancing power is necessary instead.
0 10 20 30 40 50 6049.7
49.8
49.9
50
50.1
50.2
Time [min]
Fre
quen
cy [
Hz]
)LJXUH������5HVSRQVH�LQ�V\VWHP�IUHTXHQF\�WR�FKDQJH�LQ�ORDG�DQG�SURGXFWLRQ�ZLWK�ZLQG�SRZHU��
70
12x267 TR F6212
���� 5HVXOWV�IURP�([WHQGHG�WHUP�G\QDPLF�VLPXODWLRQV� Due to lack of time and different unexpected problems with PSS/E due to use of the long-term dynamic simulations, only the 2001 case was simulated. The case was performed for comparison of results and verification of the different methods used in the present work for simulating frequency stability. ������ �����FDVH� In Section 3.1, the different simulation modules in PSS/E were explained. Due to challenges that were encountered during the PSS/E simulations of the 2001 case it is in addition convenient in the present section, to describe the PSS/E the simulation approach for this specific case. ����������6LPXODWLRQ�DSSURDFK� The long-term dynamic simulations were performed with the regular dynamic simulation tool of PSS/E in this study. The advantage to use extended-term dynamics option of PSS/E is basically the ability to change the simulation time-step size during the simulation, and therefore shorten the simulation time. Another, even bigger advantage of the larger step size is better ability to avoid numerical instability and multiplying calculation errors, which numerous calculation cycles might cause in a simulation of long period of time. All the PSS/E standard models, however, are not compatible with extended term simulation. For example, the HVDC-model ‘CDC1T’ can not be used in extended simulations. Because the Nordic 23 generator model consists of quite few components in number, the number of equations is not very large, and therefore the simulation time with regular PSS/E dynamic simulation does not get long even for simulation of a one hour time span. In addition, the regular dynamic simulation results seem to agree well with the extended simulation results on a skeleton model without HVDC-models (which on the other hand means that no power flows through the HVDC-links and therefore the models are needed). There were the power flow data, and dynamics data available for the Nordic 23-generator system. The dynamics data, however, did not include data for generators in nodes 3200 and 5603. In the long-term simulation, wind power generation is being as a first approach modeled as negative load. For slow dynamic responses this is not likely to cause any large deviation from the case with detailed dynamic models of the wind turbines. Besides wind power production, also scheduled production is modeled as negative load in the long-term dynamic simulation. This is reasoned with the easiness of modeling the ramping of the scheduled production as negative load instead of increase in generator production. In addition, this is more conservative approach, as the new production online does not contribute to system inertia. The amount of scheduled production is relatively small, being about 5 % of the initial power production in the Nordel system, and therefore this should not interfere the PSS/E ability to run the simulation.
71
12x267 TR F6212
User defined dynamic load models were made to ramp the loads, and wind power as negative loads in the simulations. The load ramping model needs as inputs the real and reactive power at the beginning and at the end of simulation period, as well as the time at which the linear ramping will be started and finished. The loads remain in constant beginning and end values in pre- and post-ramping stages. Over ramping period, the load values are being extrapolated on every time-step of simulation between their initial and final values. The ramping of HVDC-lines was performed by setting the IMIN of the HVDC-model (‘CDC1T’) which turned out to be the most practical way to influence the HVDC-transmission as there was no accurate means found to model the HVDC-transmission. As there was no dynamic model made to do the ramping, a routine (with IPLAN, which is a programming language provided with PSS/E) was made to run the simulation, in such a way, that the IMIN value for each ramping HVDC-link was calculated and set at every 10 seconds over the HVDC ramping period. 10 seconds was chosen to be suitable time step to do this, as the changes would remain reasonably small not to cause hardly any system response to step-changes. The IMIN values of the links were calculated roughly based on the desired real power flow in the HVDC-line, and the HVDC-line nominal voltage
'&1
'&0,1 8
3,,
= ����
It is possible to model the HVDC-link and ramp it in a more accurate way, but it requires better parameterization (knowledge of the HVDC-link parameters) of the HVDC-link dynamic model and/or deeper dedication to modeling the HVDC-link or HVDC-ramping models. ����������6LPXODWLRQ�UHVXOWV� There have been used a slightly different allocation of generation on node 3359. In the original Nordic 23-generator model, there were two equivalent generators of different types in this particular node. The generator representing hydro power production, was allocated 76 % of the node production, and to other power production (thermal power), 24 % of the total production in this node. In the long-term simulations with PSS/E, the node 3359 generation (other than wind power) was modelled as 100 % hydro production with 10 % primary reserves of the power produced in the initial situation (compare Table 5.1). Whereas in SPF and LP optimisation calculations, production in this node was modelled 100 % thermal power, with 0 % primary reserve. The frequency response of the case is shown in Figure 5.14.
72
12x267 TR F6212
)LJXUH������5HVSRQVH�LQ�V\VWHP�IUHTXHQF\�WR�FKDQJH�LQ�ORDG�DQG�ZLQG�SRZHU�SURGXFWLRQ�� As seen in Figure 5.14 the frequency remains within the allowed limits throughout the simulation period, despite the fact that the original frequency is a bit below 50 Hz. There is also a notch in the frequency during the very fist seconds of the simulations. This is due to the modelling difficulties of the HVDC-links in dynamic simulation. The HVDC-links were modelled such (see Section.5.4.1.1 that the HVDC-link model is given the current as user input, which was roughly calculated. Considering the influence of different contribution (415 MW) of one equivalent generator on frequency deviation, and the modelling aspects of the HVDC-links, the long-term dynamic simulation results seemed to agree reasonably well with SPF-simulation results. In the long-term dynamic simulation results there can be seen two additional features to the SPF-simulation results. There is a slightly faster increase in frequency as the scheduled production starts to come on-line at 30 minutes after the beginning of the simulation. There also appears a spike at 35 minutes after the beginning of the simulation, as the scheduled production is all on-line after ramping it up. Also, in this simulation response, ramping of the HVDC-links (ramping at 25 to 40 minutes) is seen clearly. �
0 10 20 30 40 5049.9
49.95
50
50.05
Time [min]
Fre
quen
cy [
Hz]
73
12x267 TR F6212
�� 60$//�6,*1$/�67$%,/,7<�$1$/<6,6� The transient and small signal stability analyses have been performed independent of specific WP6 cases and are of more principle character. However, one still uses the simplified model of the Nordel system as basis for the studies analysed. ���� 6PDOO�6LJQDO�6WDELOLW\�VWXGLHV� The work described in the present section has been published both on the Nordic Wind Power Conference 2004 [17] and with an improved version in the Wind Energy Journal July-September 2005 [18]. The impact on the inter-area mode oscillations with wind power integrated into the grid is not yet well explored. It is therefore interesting to simulate the Nordic grid with large scale wind power integration, and investigate how the wind power will influence the inter-area mode in the Nordic grid. Inter-area mode oscillation can be defined as the swinging of many machines in one part of the system against machines in other parts. They are caused by two or more groups of closely coupled machines being interconnected by weak ties. The natural frequency of these oscillations is typically in the range of 0.1 to 1 Hz. In the Nordel system the two most influencing inter-area mode is the 0.30 Hz oscillation mode that is well observable in the power flowing in the AC-connections between Sweden and Finland and the 0.58 Hz mode that is observable in the power flowing in the lines between Southern-Norway and Sweden in the so called Hasle-corridor. The main objective with the dynamic PSS/E simulations that were the basis for the results presented in the present chapter was to investigate the impact of wind power generation on the damping of inter-area mode oscillations in the Nordic grid. The oscillations were excited trough different disturbances. The 23 generator model described in Chapter 4 was used to model the Nordel grid plus Germany and Jutland. The initial power flow for all dynamic simulation was a typical cold fall day situation (16300 MW production in Norway and 50400 MW production in the Nordel system). Two places are chosen to observe the inter-area mode oscillations. One is the Hasle-corridor between southern Norway and Sweden, the other is the AC-connection between Finland and Sweden in the north. ������ &DVH�GHVFULSWLRQ� Only cases with wind power integration in Norway have been studied. It is however expected that this made the results easier to interpret and that it was a good approach for making principle studies on how large-scale wind power integration may affect the Nordel system.
74
12x267 TR F6212
&DVH��� Case 1 is the reference model, and is without wind power production in Norway. For this model the production in Norway is 16.300 MW and 50.400 MW in the Nordic interconnection. This production is a medium production situation, i.e. it corresponds to the production of a cold autumn day. &DVH����In Case 2, 1.000 MW of wind power is implemented in the middle of Norway. This power replaces 1.000 MW of hydropower from the same area. Accordingly, the model is equal to the one in Case 1, but wind power is implemented into the Norwegian grid. Three different generator types are simulated, squirrel cage induction machine, SCIM, doubly fed induction machine, DFIG, and direct drive synchronous generator, DDSG. There is no wind model implemented into the wind power plant, i.e. the generators are simulated with constant mechanical torque. &DVH����In Case 3, 5.000 MW of wind power is implemented into the grid. 1.800 MW is implemented at the western coast of southern-Norway, 1.000 MW is implemented in the middle of Norway and 2.200 MW is implemented in the northern part of Norway. The wind power replaces hydropower at each node in the same manner as described in Case 2. As in Case 2, three different generator types are simulated, SCIM, DFIG and DDSG. Also as in Case 2, there is no wind model implemented into the wind power plant, i.e. the generators are simulated with constant mechanical torque. ������� :LQG�SRZHU�PRGHOOLQJ�DQG�LQWHJUDWLRQ� 7KH�ZLQG�SRZHU�UDGLDOV� Each wind power plant consists only of one radial with one generator, i.e. each wind power plant is aggregated into one generator. The generator operates at 0.69 kV, and the voltage is transformed successively to 22 kV, 132 kV and 300 kV, which is the main transmission grid. In Figure 6.1, the wind power radial that is used to implement the wind power plants in the grid is shown.
)LJXUH����7KH�ZLQG�SRZHU�UDGLDO�WKDW�LV�XVHG�WR�LPSOHPHQW�WKH�ZLQG�SRZHU�SODQWV�LQ�WKH�JULG��
75
12x267 TR F6212
6TXLUUHO�FDJH�LQGXFWLRQ�PDFKLQH��The squirrel cage induction machine, SCIM, also known as fixed speed wind turbine or constant speed wind turbine, is the oldest one that is used by the commercial wind industry. A principal model of the SCIM is shown in Figure 6.2.
�)LJXUH������$�SULQFLSDO�PRGHO�RI�D�VTXLUUHO�FDJH�LQGXFWLRQ�PDFKLQH�� For wind farm modelling CIMTR3 is recommended by PSS/E™ as a SCIM model. This model takes into account the rotor flux dynamic but neglect the stator G
GWΨ term or the so-called
transformer term in the stator voltage equations or the stator transients. Because of that, the DC offset in the stator current and associated braking torque is also ignored. This assumption would not cause excessive errors in analysis of induction machine for on-line operation when the ac network frequency stays close to nominal and the rotor slip is small enough. The CIMTR3 model is a standard PSS/E™ model in version 28.0. Since the wind fluctuations are neglected, and the generator works on constant mechanical torque, the wind turbine and the gearbox is not included in the model. However, the mass from the rotor and shaft is included into the inertia to generator model. 'RXEO\�IHG�LQGXFWLRQ�JHQHUDWRU��The doubly fed induction generator, DFIG, is a variable speed system with a constant-frequency induction generator. To allow variable wind speed operation, the mechanical rotor speed and the electrical frequency of the grid must be decoupled, [19]. The stator is directly connected to the grid while the rotor winding is connected via slip rings to a back-to-back voltage source converter that feeds the three-phase rotor winding. In this way, the mechanical and electrical rotor frequencies are decoupled and the electrical stator and rotor frequencies can be matched, independently of the mechanical rotor speed. Hence, with this generator type it will be possible to control the speed (or torque) and also the reactive power on the stator side of the induction generator. A principal model of a doubly fed induction generator is shown in Figure 6.3.
76
12x267 TR F6212
)LJXUH�����$�SULQFLSDO�PRGHO�RI�D�GRXEO\�IHG�LQGXFWLRQ�JHQHUDWRU� The PSS/E™ version 28.0 has not any model for the DFIG in the library. Hence, Shaw Power Technologies, Inc™ has made a wind turbine package based on General Electrics 1.5 MW wind turbine. This package has been made available for SINTEF. The General Electrics wind turbine model is not only a generator model as CIMTR3, it can also have wind information, pitch control, over and under voltage protection and over and under frequency protection as input to the model. The generator in the GE 1.5 MW wind turbine is a three-phase doubly fed induction generator. The DFIG is, as the SCIM, simulated without influence from wind. The wind turbine works on constant mechanical torque. Further, the pitch control, over and under voltage protection and over and under frequency protection are neglected. 'LUHFW�GULYH�V\QFKURQRXV�JHQHUDWRU��The direct drive synchronous generator, DDSG, is as the DFIG a variable speed turbine, and the mechanical rotor speed and the electrical frequency are also here decoupled. As distinct from the DFIG, the generator in DDSG is completely decoupled from the grid by a power electronics converter. The generator is excited using either an excitation winding or permanent magnets, [19]. A principal model of a direct drive synchronous generator is shown in Figure 6.4.
)LJXUH�����$�SULQFLSDO�PRGHO�RI�D�GLUHFW�GULYH�V\QFKURQRXV�JHQHUDWRU�� The PSS/E™ version 28.0 has not any model for the DDSG. Hence, the DDSG model is simplified dramatically. It is supposed here that the converter, which decouples the synchronous
77
12x267 TR F6212
generator to the grid, is much quicker than the oscillations that occur in the grid. The DDSG is therefore modelled as a negative load, i.e. a load that supplies power into the grid instead of extracting power from the grid. ������ 5HVXOWV� This section presents the main results of this work. The cases that are referred to, is the same as the cases that are described in Section 6.1.2. Case 1, which is the reference model, is without any wind power integrated in Norwegian grid. Case 2 is with 1.000 MW wind power integrated in Norwegian grid. Case 3 is with 5.000 MW wind power integrated in Norwegian grid. The results will be presented in the following manner: First, the results from the simulation of the 0.58 Hz inter-area mode between Norway and Sweden will be presented. Case 1 (reference case) is compared to Case 2 (1.000 MW wind power) and Case 3 (5.000 MW wind power). Secondly, the results from the simulation of the 0.30 Hz inter-area mode between Finland and Sweden will be presented. Case 1 (reference case) is also here compared to Case 2 (1.000 MW wind power) and Case 3 (5.000 MW wind power). ,QWHU�DUHD�PRGH�RVFLOODWLRQ�EHWZHHQ�1RUZD\�DQG�6ZHGHQ��It is more difficult to analyse the 0.58 Hz inter-area mode between Norway and Sweden, than it is to observe the 0.30 Hz inter-area mode between Sweden and Finland. The reason for this will be commented later in the analyses. Disconnection and re-closing of a line close to the Hasle corridor is used to excite the inter-area mode oscillation. The disconnection time of the line was optimized with respect to excite the 0.58 Hz oscillation mode [17]. The results in Figure 6.5 show that the damping of the inter-area mode oscillation is not influenced to any great extent with 1.000 MW wind power in Norway. This is not surprising, since only 1.000 MW of wind power is installed in the grid compared to the total production of 16.300 MW in Norway. The damping of the inter-area mode oscillation decreases with different types of generators. The SCIM has the least influence on the damping, while the DDSG model has the greatest influence on the damping.
78
12x267 TR F6212
�)LJXUH�����&RPSDULVRQ�RI�WKH�GDPSLQJ�>�@�RI������+]�LQ�&DVH����UHIHUHQFH�PRGHO��DQG�&DVH����������0:�ZLQG�SRZHU��ZLWK�GLIIHUHQW�JHQHUDWRU�WHFKQRORJLHV��� Figure 6.6 shows the results from the simulations with 5.000 MW of wind power integrated into the Norwegian grid. It is now possible to observe how the inter-area mode oscillation change with large scale wind power integrated in Norway. The SCIM improves the damping significantly, while the DFIG model and the DDSG model impair the damping of the inter-area mode oscillation. The DDSG is the worst case for the grid.
)LJXUH�����&RPSDULVRQ�RI�WKH�GDPSLQJ�>�@�RI������+]�LQ�&DVH����QR�ZLQG�SRZHU��DQG�&DVH����������0:�ZLQG�SRZHU��ZLWK�GLIIHUHQW�JHQHUDWRU�WHFKQRORJLHV��� The greater the damping [%] in Figure 6.5 and Figure 6.6 the faster the inter-area mode oscillations will die out, and visa versa, the less the damping [%] in Figure 6.5 and Figure 6.6 the slower the inter-area mode oscillations will die out. ,QWHU�DUHD�PRGH�RVFLOODWLRQ�EHWZHHQ�)LQODQG�DQG�6ZHGHQ��The inter-area mode oscillation between Finland and Sweden is relatively easy to excite. In the model, the inter-area mode oscillation of 0.30 Hz is best observable between Finland and Sweden. A three-phase short circuit is chosen to excite the power system. The three-phase short circuit is performed in Sweden, close to the ac-connection between Sweden and Finland.
79
12x267 TR F6212
Figure 6.7 shows the damping of the 0.30 Hz between Finland and Sweden with 1.000 MW wind power in Norway. It can be seen from Figure 6.7 that 1.000 MW of wind power does not influence the inter-area mode oscillation to any extent.
)LJXUH�����&RPSDULVRQ�RI�WKH�GDPSLQJ�>�@�RI������+]�LQ�&DVH����UHIHUHQFH�PRGHO��DQG�&DVH����������0:�ZLQG�SRZHU��ZLWK�GLIIHUHQW�JHQHUDWRU�WHFKQRORJLHV��� Figure 6.8 shows the results from the simulations with 5.000 MW of wind power integrated in Norway. The results from Figure 6.8 give the same results as Figure 6.7. The damping of the inter-area mode oscillation between Finland and Sweden is not influenced by the integration of wind power in Norway.
)LJXUH�����&RPSDULVRQ�RI�WKH�GDPSLQJ�>�@�RI������+]�LQ�&DVH����UHIHUHQFH�PRGHO��DQG�&DVH����������0:�ZLQG�SRZHU��ZLWK�GLIIHUHQW�JHQHUDWRU�WHFKQRORJLHV�� ���� $QDO\VLV�� First, the results from the inter-area mode oscillation between Norway and Sweden are treated, then the results from the inter-area mode oscillation between Finland and Sweden. It is worth to notify the following before the analysis and discussion begins:
80
12x267 TR F6212
1) The simulation results from the Nordic equivalent of the Nordic interconnection are not
directly transferable to reality. First, the Nordic equivalent is strongly simplified. Secondly, the wind power plants are also strongly simplified.
2) The calculations of the simulation results are burdened with a large margin of errors.
When a simulation is finished, a representative part of the plot for the small signal analyses is calculated. The calculations from the simulations that are presented here are calculated for different parts of the simulation to secure the results that are presented.
Yet, the results from the simulation can be used to observe how the damping of the inter-area mode oscillations of the Nordic grid is influenced by wind power in Norway. The results show that large-scale wind power integration in Norway with the squirrel cage induction machine will make the inter-area mode oscillation between Norway and Sweden better since the damping of the 0.58 Hz inter-area mode is increased. One explanation why the squirrel cage induction generator makes the damping of the inter-area mode oscillation better can be that the squirrel cage induction machine is better damped than the ordinary synchronous generators that are used to hydropower. Also, new generator types with new time constants will change the inter-area mode oscillations. When a new generator with new time constants is introduced into a set of generators that are swinging together against another set of generators, the damping of the inter-area mode oscillations may be changed, either improved or reduced, depending on the time constants from the new machines. The doubly fed induction generator does not affect the damping of the inter-area mode oscillation in any great extent. The damping of the inter-area mode oscillation only decreases a little bit. The reason why the doubly fed induction generator model does not have the same positive reaction on the damping of the inter-area mode oscillation, despite the fact that it is an induction generator, may be that it is partly decoupled from the grid. The doubly fed induction generator model is then able to partly follow the power oscillations that appear in the grid when an excitation occurs, and the damping of inter-area mode oscillation between Norway and Sweden may then be reduced. The negative load that is used as the direct drive synchronous generator decreases the damping of the inter-area mode oscillation between Norway and Sweden. It is supposed that the converter between the generator and the grid is much quicker than the oscillations that occur in the grid. It is not surprising that the damping of the inter-area mode oscillation between Norway and Sweden decreases with such a model. The damping of the inter-area mode oscillation between Finland and Sweden is not disturbed in any way when wind power is integrated in Norway. This is not very surprising since the electrical machines in Finland swings against the electrical machines in Sweden. The electrical machines in Norway do not swing against the machines in Finland, and hence, there are no reaction of the inter-area mode oscillation between Finland and Sweden.
81
12x267 TR F6212
�� 75$16,(17�67$%,/,7<�$1$/<6,6� This chapter considers the mutual effects of wind power in power systems under transient fault situations, with the Nordic power system as study case. It is analyzed (i) how the wind turbines behave in the system when it experiences a transient fault and (ii) what impact the wind turbines have on the dynamic behaviour of the system after a fault. Characteristic for the Nordic power system is that it is geographically large, but at the same time it is of comparably small capacity, due to Norway, Sweden, Denmark and Finland being only sparsely populated countries. This makes it more vulnerable to high levels of wind power penetration if the installed turbines are uncontrolled distributed generators. Until recently wind turbines connected to the Nordic power system were not engaged in the control and support of the system. If transient faults in the system lead to considerable excursions in voltage and/or frequency the wind turbines were to disconnect and to reconnect only once the system has returned to stable operation. Increasing wind power penetration leads to the problem that considerable amount of generation might disconnect in case of a transient fault in the system, causing the system to become unstable from an otherwise harmless fault situation. To prevent such situations newly installed wind turbines in Denmark have to comply with grid connection requirements that demand wind turbines to ride through transient faults [24], [25]. The model of the Nordic power system used in these investigations is the 23 generator model described in section 4.1. For the transient stability study, the dynamic version of the 23 generators model is applied, including speed governors and voltage control of the generators. SINTEF originally developed the 23 generators model in the power system simulation software PSS/E. For the present transient stability study, the model has been converted to the power system simulation tool Power Factory from DIgSILENT [26]. At Risø National Laboratory a model of the wind power connected to the Nordic power system in eastern Denmark has been added to SINTEF’s Nordic power system model. This additional model has been developed in cooperation with the Danish transmission system operator. ���� 7KH�1RUGLF�3RZHU�6\VWHP�0RGHO� The Nordic power system stretches the countries Norway, Sweden, Denmark and Finland, and has a nominal system frequency of 50 Hz. It is divided into two synchronous areas. The biggest part of the system, comprising Norway, Sweden, Finland and the eastern part of Denmark are one interconnected, synchronous AC system. The small rest of the Nordic power system, i.e. western Denmark, is AC connected to the big UCTE system, which is the interconnected AC system of central Europe. Several HVDC links connect the Nordic synchronous system with the central European system. There are, among others, HVDC links between Norway and western Denmark, between eastern Denmark and Germany and between southern Sweden and Germany.
82
12x267 TR F6212
Since in transient fault situations the HVDC links can be considered uncontrolled voltage dependent sources and sinks, the central European part of the system is of no relevance for transient fault simulations. Hence for simplicity the term “Nordic power system” will in the following refer to the Nordic synchronous system. The model of the Nordic power system is an aggregation, which means that the generators, lines and loads in the model are lumped representations of several generators lines and loads in reality. It comprises 35 nodes and 20 synchronous generators. It is a model of the transmission system only; comprising the voltage levels 420, 300, 150 and 135 kV. In the location of eastern Denmark, the model is extended with a simplified grid to represent the connection of wind power. This simplified grid comprises all voltage levels from transmission system voltage down to generator terminal voltage. This extension is described in the following section. ���� 0RGHO�RI�:LQG�3RZHU�,QVWDOODWLRQV�LQ�WKH�1RUGLF�3RZHU�6\VWHP� In order to study the impact between the Nordic synchronous system and large wind power installations, wind power in the East Danish system has been selected. The 23 generators model has been extended considering the transmission system and parts of the distribution system in the southern part of East Denmark. The transmission system to the south of East Denmark is relatively weak, which is important because already most of the wind power in East Denmark is installed in the south, and future large offshore installations are considered to be installed there. Denmark is the country in the Nordic power system that has by fare the highest level of wind power penetration. Therefore it is only natural to model the wind power connected there. It is assumed that wind power in Eastern Denmark is the only wind power that has to be considered in the Nordic system. This is a sound assumption, as only faults in eastern Denmark will be simulated. Any wind power that is installed in another part of the system would hardly be affected by faults in eastern Denmark. Inherent for wind power is that its resources are far away from load centers and hence almost invariable far away from strong transmission systems. This is even more applicable for offshore wind farms. The largest part of wind power that will be installed in eastern Denmark in future will be offshore. ������ 7RSRORJ\�RI�WKH�*ULG�LQ�(DVWHUQ�'HQPDUN� In the original model as developed by SINTEF eastern Denmark is represented by a single busbar with a synchronous generator and a load. This busbar is called Zealand, as can be seen in Figure 7.1, which shows the topology of the power system extension used in this chapter. The synchronous generator at busbar Zealand (SG Zealand) is rated 2000 MVA and represents all the conventional power plants in eastern Denmark. The load connected to Zealand represents the load in the northern part of eastern Denmark.
83
12x267 TR F6212
In the extension, three wind farms in the south of Eastern Denmark are considered. One wind farm represents all the land-based wind turbines, which are distributed over the southern islands of East Denmark. These turbines are aggregated to a single induction generator. Another wind farm represents the Nysted offshore wind farm. This is one of the world’s largest offshore wind farms and has been connected to the system since 2003 [27]. The third wind farm is an offshore wind farm that is likely to be installed in future. The connection between this 420 kV busbar and the wind farms is modeled in a more detailed manner than the rest of the Nordic system. It considers all voltage levels from 420 kV down to generator terminal voltages. ������ 0RGHO�RI�:LQG�)DUP�)HHGHUV� The wind farms in the south of East Denmark are connected to the strong 400 kV transmission system in Spangager in the north through a relatively weak 132 kV transmission grid. The weak 132 kV transmission system limits the stability of the grid in the southern part, and is therefore represented in a simplified version in the model. The distance between Spanager and Radsted is approximately 100 km. The representation in the model is a number of equal parallel lines, whereas the real transmission grid is more complex, including segments with overhead lines and cables. The number of parallel lines is varied depending on the level of wind power penetration simulated. (Different cases are simulated as will be seen in section 7.3.)
84
12x267 TR F6212
Connected to Radsted is a load that represents the load in the south of eastern Denmark. The wind farms are connected to Radsted through 132 kV feeders and medium voltage cables, representing the cable network in the wind farms (33 kV). The 132 kV feeder connecting the distributed land-based wind turbines is assumed to be 25 km long. This length is an approximate value found by considering the distance from Radsted to a central location between all the land-based turbines [28]. From this central location 24 parallel 11 kV cables, of 20 km length, represents the medium voltage cable network to the turbines. The length of the 11 kV cable is an average distance from the turbines to the central location mentioned above. Since the distributed land-based turbines are all connected to the distribution system, three transformers are chosen to step the voltage down from the transmission system voltage to the generator terminal voltage. The connection from Radsted to the Nysted offshore wind farm is modeled with a 29 km long 132 kV line. The cable network inside the wind farm is represented by three parallel, 3.2 km long, 33 kV cables. The distance of 3.2 km is the average distance from the turbines to the transformer platform [29].
�� ����
�� ���� 33 kV 33 kV 11 kV
0.96 kV
IG
Nysted offshore WF
0.96 kV
IG
Future offshore WF
0.69 kV
IG
land-based WT
Radsted on Lolland (DK)
�� ����50 kV
��
Southern Sweden
Zealand (DK) 420 kV ������
��
Spanager on Zealand (DK) 132 kV �� ��
��
��
����
• • •
SG
������
SG
��
��
���� ��
�� �� ��
Nordic Power System
��
= P&Q measurement point
�)LJXUH�����*ULG�GLDJUDP�RI�WKH�H[WHQVLRQ�RI�WKH�1RUGLF�V\VWHP�PRGHO�LQWURGXFLQJ�WKH�ZLQG�SRZHU�LQVWDOOHG�LQ�HDVWHUQ�'
85
12x267 TR F6212
The 132 kV feeder connecting the future offshore wind farm with Radsted is assumed to be 30 km long. Just like in the case of the Nysted offshore wind farm, also here the internal farm cable network is represented by three parallel 3.2 km long, 33 kV cables. With these feeders the generators of the wind turbines are connected to the Nordic power system. The generators and their prime movers, i.e. the wind turbines are described in the following section. ������ 7KH�:LQG�)DUP�0RGHOV� ����������:LQG�0RGHO� In this article only transient fault simulations are considered. The simulated events last up to a few seconds, therefore natural wind variations need not be taken into account. Rotating wind speed variations like 3 p (e.g. the tower shadow effect) [30] can be neglected as well, because the wind power plants considered are aggregations of many single wind turbines. If many turbines are connected together their rotating wind speed variations cancel each other out. The wind speed is set to a constant 18 m/s, which is a wind speed that allows all turbines to produce rated power. A rated power operating point is chosen, as this is most burdening for the power system. ����������0RGHO�RI�WKH�'LVWULEXWHG�/DQG�%DVHG�:LQG�7XUELQHV� Substantial amount of wind power is distributed over the islands in the south of eastern Denmark. These distributed land-based wind turbines are aggregated and modeled by one squirrel cage induction generator. The prime mover is modeled by a constant mechanical torque that acts on a two masses spring and friction model, which then drives the generator (Figure 7.2). Aerodynamics and control schemes of these turbines are neglected. They are not relevant for transient fault studies as the grid connection requirements that were applicable when these turbines were installed demand them to disconnect in case of a grid fault. The capacity of the induction generator representing the aggregation of all the distributed land-based wind turbines is 235 MW. This is a value that can be worked out from the wind turbine data register of the Danish Energy Agency [28].
86
12x267 TR F6212
The protection system that disconnects the land-based turbines in case of a fault is implemented in the form of undervoltage, overspeed and overcurrent protection. The protection scheme implemented in this model disconnects the generator and its compensation unit, when
• The voltage at the generator terminals drops below 0.85 pu for 100 ms • The speed of the generator exceeds 104% of its rated speed. (When the generator cannot
export as much power as is imported through the wind, it accelerates.) • The current exceeds 200% rated current for 100 ms.
It is assumed that the reactive power compensation is implemented in such a way that only the no load reactive power demand of the generator is compensated. This is in accordance with the applicable grid connection requirements. ����������0RGHO�RI�1\VWHG�2IIVKRUH�:LQG�)DUP� The Nysted offshore wind farm consists of 72 identical active-stall wind turbines, each rated 2.3 MW [27]. Therefore this wind farm is modeled with 72 parallel 2.3 MW induction generators, driven by a wind turbine model with full mechanical (Figure 7.2) and aerodynamic representation [31]. A similar wind turbine model has been verified in an islanding experiment of a real multi-megawatt active-stall turbine. This experiment proved that the model represents the behavior of the real turbine well [32]. Figure 7.3 shows the topology of the wind turbine model.
turbine rotor main shaft gearbox generatoraerodynamic
7DH
θ:75
7OVV
θOVVNPV
FPV
��I
,:75
,JHQ
7JHQ
θJHQ
7DJ
)LJXUH���� 'ULYH�WUDLQ�PRGHO�RI�ZLQG�WXUELQH�
Z�
8�
wind speed
aero-dynamics
drive train
generator & compensation
power system
control system
7DH�
ω:75�
7DJ�
ωJHQ�
,�
Q&�
3��4�
SLWFK�
)LJXUH���� :LQG�WXUELQH�PRGHO�
87
12x267 TR F6212
An active-stall controller that finds the right pitch angle during normal, fault-free operation is implemented. It optimizes the active power production at wind speeds below rated wind speed. Above rated wind speeds it limits the active power output of the turbine to its rated value [33]. The Nysted offshore wind farm has to fulfill grid connection requirements that require certain fault ride through capabilities [25]. Therefore also a transient fault controller is implemented that allows the turbines to ride through transient faults without experiencing damaging speed excursions [34]. In accordance with the applicable grid connection requirements, a reactive power compensation unit has to keep the wind farm neutral in reactive power demand at the grid connection point. A shunt capacitor bank is implemented at the generator busbar, and a controller controls the number of connected capacitors, Q&, such that the steady state power factor is one at the high voltage side of the 132/33 kV transformer. ����������0RGHO�RI�)XWXUH�2IIVKRUH�:LQG�)DUP� The future offshore wind farm considered is also made of active-stall wind turbines. It consists of 99 identical 2 MW turbines. Hence the wind farm is modelled with 99 parallel induction generators driven by a wind turbine model similar to the one used in the Nysted wind farm (Figure 7.3). The control of the wind turbines in the future offshore wind farm is more sophisticated as these turbines are not only required to ride through transient faults, but also to contribute to the damping of grid frequency oscillations in the wake of transient faults [35]. This is not demanded in current grid connection requirements, but if wind power penetration increases this could become a requirement. Also here it is assumed that the wind farm controls its steady state reactive power production such that it is neutral in reactive power demand at the high voltage side of the 132/33 kV transformer. For that purpose also here a shunt capacitor bank is implemented. ���� 6LPXODWLRQV��5HVXOWV�DQG�'LVFXVVLRQV� Different scenarios are simulated to assess the impact of wind power in the current and the future situation. The faults simulated are 100 ms, zero impedance, 3-phase short circuits on one of the lines between Spanager and Radsted, close to Radsted (see Figure 7.1). The fault is cleared by permanent disconnection of the faulted line. This is a fault situation described in Elkraft’s grid connection requirements for wind farms connected to the transmission system [25].
88
12x267 TR F6212
������ &DVH��� The current situation is simulated, i.e. only the land-based turbines and the Nysted offshore wind farm are connected. The feeder for the future offshore wind farm, as shown in Figure 7.1 is not existent in this simulation.
0
0.2
0.4
0.6
0.8
1
1.2
0 1 2 3 4 5 6
time [s]
volta
ge [
pu]
land-based turbinesNystedRadstedZealand
)LJXUH���� &DVH����9ROWDJH�DW�GLIIHUHQW�ORFDWLRQV�LQ�WKH�V\VWHP
Figure 7.4 shows that the voltage at busbar Radsted drops to zero, as Radsted is closest to the fault location. Zealand is hardly affected by this fault as the relatively weak connection between Radsted and Zealand causes a substantial voltage drop. The voltage at the terminals of the land-based turbines gets suppressed in the beginning of the fault and after a few ms it drops to zero as the protection system of these turbines disconnects them. The voltage at Radsted recovers quickly after the clearance of the fault because the land-based turbines have disconnected, which means that they do not consume reactive power any more. In addition to that, the now unloaded cables in the feeder to the land-based turbines act as capacitors generating noticeable amount of reactive power (Figure 7.7). The voltage at Nysted recovers also relatively quickly because of the reduced reactive power demand of the Nysted generator in the first seconds. The fault excites the inherently flexible drive train of the Nysted wind turbines (Figure 7.2) to oscillations, which in the first instances after the clearance of the fault leads to a strongly reduced active power production [34]. At the same time the compensation capacitors stay connected helping the voltage to recover.
0.98
1
1.02
1.04
1.06
1.08
0 1 2 3 4 5 6time [s]
Nys
ted
gene
rato
r sp
eed
[pu]
)LJXUH�����&DVH ���6SHHG�RI�WKH�1\VWHG�ZLQG�IDUP�JHQHUDWRU�
89
12x267 TR F6212
Figure 7.5 shows the speed of the generator of the Nysted offshore wind farm. During the fault the speed of the generator accelerates steeply. Just after the clearance of the fault it exhibits oscillations with the resonance frequency of the small inertia of the generator rotor (Figure 7.2). While these oscillations subside within the first two seconds after the fault, the underlying low resonance frequency of the turbine rotor with its large inertia becomes dominant. At simulation time 3 seconds these low frequency oscillations cause a noticeable increase in generator speed, which in turn means that the generator requires more reactive power. Increasing speed in a squirrel cage induction generator means increasing slip, V.
Xr Xs Rs
Us Xm
��
��Rr s
)LJXUH�����(TXLYDOHQW�FLUFXLW�GLDJUDP�RI�D�VTXLUUHO�FDJH�LQGXFWLRQ�JHQHUDWRU�
From Figure 7.6, which shows the equivalent circuit diagram of such a squirrel cage induction generator, it can be seen that with increasing slip, V, the overall rotor resistance decreases
V5
5 UURWRU = . Lower resistance leads to more current flowing through the mainly inductive circuit of
the generator, causing higher reactive power demand. For these considerations it is sufficient to use the steady state equivalent circuit of the generator as shown in Figure 7.6. The wind turbine drive train, which causes the speed oscillations, has considerably larger time constants than the electrical circuit of the generator. Figure 7.7 shows that this reactive power demand has to be covered by SG Zealand, which produces reactive power to be transferred to Sweden (line DK-S). This reactive power surge causes a voltage drop between Zealand and Radsted, which can be seen in Figure 7.4 around the simulation time 3 seconds. Eventually the voltages in eastern Denmark settle to a slightly higher value because of the extra reactive power being generated in the cables of the land-based turbines feeder.
-350
-300
-250
-200
-150
-100
-50
0
50
100
150
0 1 2 3 4 5 6time [s]
Q [
MV
Ar]
land-based turbines
Nysted
line DK-S
)LJXUH���� &DVH����5HDFWLYH�SRZHU�DW�GLIIHUHQW�ORFDWLRQV�LQ�WKH�V\VWHP�DV�VKRZQ�LQ�)LJXUH�����
90
12x267 TR F6212
������ &DVH����� In this case, it is simulated that the future offshore wind farm has been connected and that the turbines in this wind farm have the same control capabilities as the Nysted turbines, i.e. only fault ride through as demanded by the current grid connection requirements. The rating of the 420/132 kV transformer and the number of lines between Spanager and Radstad are increased to suit the extra power installed in the new wind farm. In addition, the dispatched power of SG Zealand is reduced by 200 MW and its rated power is reduced by 200 MVA to reflect the situation that future installed wind power will substitute conventional power plants. In Figure 7.8 the voltage of the land-based turbines is not shown as this drops to zero like in the previous case. From Figure 7.8 it can be seen that the voltage at Radsted, and consequently at the terminals of the wind farms, is under considerably more strain than in the previous case. The drive trains of the turbines in the two offshore wind farms oscillate similarly causing the speed of the generators to oscillate (Figure 7.2), which causes the reactive power demand to oscillate (Figure 7.6), and this in turn causes the voltage to oscillate strongly too. Due to the stronger connection between Zealand and Radsted the voltage at Zealand is slightly lower than in case 1. This has no consequences for the voltages in the rest of the Nordic system.
The fault does however upset the grid frequency, because SG Zealand exhibits strong rotor speed oscillations, which is visible in strong active power oscillations as shown in Figure 7.9. These power oscillations can only propagate through the line between Zealand and Sweden (called ‘line DK-S’ in Figure 7.9) into the rest of the Nordic power system and be absorbed by other bulk power plants.
0
0.2
0.4
0.6
0.8
1
1.2
0 1 2 3 4 5 6time [s]
volta
ge [
pu]
Future OffshoreNystedRadstedZealand
)LJXUH�����&DVH������9ROWDJH�DW�GLIIHUHQW�ORFDWLRQV�LQ�WKH�V\VWHP�
91
12x267 TR F6212
������ &DVH����� In the case simulated here, the future offshore wind farm employs its grid frequency stabilizer to counteract the frequency oscillations caused by the short circuit [36]. This emulates the situation that in future wind turbines will be involved in the stabilization of the power system. In the first instance these turbines have to tackle their own drive train oscillations before they can contribute to grid frequency stabilization. Therefore the rise in voltage (Figure 7.10) just after the clearance of the fault is similar to that in the previous case. The voltage dip after simulation time 3 seconds is much less severe, which is due to the pitching actions of the grid frequency stabilizer. Comparing Figure 7.9 with Figure 7.11 it becomes visible that the oscillations in the rotor speed and hence the active power of SG Zealand and in line DK-S, are noticeably dampened by the control actions of the grid frequency stabiliser in the future offshore wind farm.
-200
0
200
400
600
800
1000
0 1 2 3 4 5 6
time [s]
P [
MW
]Future OffshoreNystedSG Zealandline DK-S
)LJXUH���� &DVH������$FWLYH�SRZHU�DW�GLIIHUHQW�ORFDWLRQV�LQ�WKH�V\VWHP�DV�VKRZQ�LQ�)LJXUH�����
0
0.2
0.4
0.6
0.8
1
1.2
0 1 2 3 4 5 6time [s]
volta
ge [
pu]
Future OffshoreNystedRadstedZealand
)LJXUH����� &DVH������9ROWDJH�DW�GLIIHUHQW�ORFDWLRQV�LQ�WKH�V\VWHP�
92
12x267 TR F6212
������ &RPSDULVRQ�RI�WKH�*ULG�)UHTXHQF\�5HVSRQVH�RI�&DVH����&DVH�����DQG�&DVH����� As noted above, the voltage variations caused by the faults, simulated in the different cases, has a negligible impact on the Nordic power system. The frequency and hence the active power flow through the system gets affected though. An effective means of comparing the consequence of the different scenarios on the Nordic power system is comparing the speed responses of SG Zealand, i.e. the grid frequency.
Figure 7.12 shows the speed of SG Zealand for the cases 1, 2.0 and 2.1. The comparison of case 1 and case 2.0, which both do not include any active frequency damping by the installed wind turbines, shows that in the case of larger wind power installation the speed of SG Zealand gets more upset. As shown in Figure 7.4 and Figure 7.8 this is not caused by the stronger connection between the fault location and Zealand. The voltage at Zealand dips during the fault almost equally low in both cases. Instead it is the power that the wind farms have to inject into the Nordic system to dampen the drive train oscillations. While in case 1 most of the installed wind power disconnects during the fault, hence does not contribute to the excitation of oscillations, there are two wind farms that stay connected in case 2.0. The turbines of both wind farms exhibit only relatively lightly damped drive train oscillations, which cause corresponding power oscillations that excite rotor speed oscillations in SG Zealand.
-200
0
200
400
600
800
1000
0 1 2 3 4 5 6
time [s]
P [
MW
]Future OffshoreNystedSG Zealandline DK-S
)LJXUH����� &DVH������$FWLYH�SRZHU�DW�GLIIHUHQW�ORFDWLRQV�LQ�WKH�V\VWHP�DV�VKRZQ�LQ�)LJXUH������
0.996
0.997
0.998
0.999
1
1.001
1.002
0 1 2 3 4 5 6 7 8 9 10
time [s]
SG
Zea
land
spe
ed [
pu]
case 1
case 2.0
case 2.1
)LJXUH����� &RPSDULVRQ�RI�WKH�URWRU�VSHHG�RI�6*�=HDODQG�IRU�WKH�FDVHV��������DQG�����
93
12x267 TR F6212
The comparison of case 2.0 and 2.1 in Figure 7.12 shows how the future offshore wind farm can contribute to the damping of such oscillations when it employs its pitch angle grid frequency stabilizer [36]. As described above the rotor speed oscillations in SG Zealand cause power fluctuations that propagate through the entire Nordic power system. Hence the other synchronous generators in the system experience rotor speed oscillations too, as they have to absorb the power fluctuations.
)LJXUH������5RWRU�VSHHG�RVFLOODWLRQV�LQ�WKH�V\QFKURQRXV�JHQHUDWRUV��L��FHQWUDO�6ZHGHQ���LL��ZHVWHUQ�1RUZD\�DQG��LLL��)LQODQG�IRU�WKH�FDVHV��������DQG����� Figure 7.13 shows the speed of three arbitrarily chosen synchronous generators, which are in different, relatively remote locations in the system. Also here the three cases are compared with each other. The same observations that have been made with the speed of SG Zealand can be made here too. The most important observation is that higher levels of wind power penetration lead to increased excitations of the power system. Also here it can be observed that the future offshore wind farm with its grid frequency stabilizer is capable of impacting positively on the damping of system-wide oscillations. When quantifying the damping effect of the future offshore wind farm it has to be kept in mind that its rating is only about a tenth of the rating of SG Zealand, and much less compared to the total active power transmitted through the system (43000 MW).
0.998
0.9985
0.999
0.9995
1
0 1 2 3 4 5 6 7 8 9 10
time [s]
cent
ral S
wed
en S
G s
peed
[p
u]
case 1
case 2.0
case 2.1
0.9975
0.998
0.9985
0.999
0.9995
1
0 1 2 3 4 5 6 7 8 9 10
time [s]
wes
tern
Nor
way
SG
spe
ed
[pu]
case 1
case 2.0
case 2.1
0.9975
0.998
0.9985
0.999
0.9995
1
0 1 2 3 4 5 6 7 8 9 10time [s]
Fin
land
SG
spe
ed [
pu] case 1
case 2.0
case 2.1
94
12x267 TR F6212
�� &21&/86,216� Stability studies have been performed based on the WP5 objectives:
- Identify and quantify potential system stability problems in particular related to large-scale integration of intermittent renewable energy generation into the power system
- Identify and evaluate various solutions to eliminate/reduce the problems In the following the conclusions from each major stability study performed in WP5 will be presented. ���� )UHTXHQF\�VWDELOLW\� The frequency stability studies have been carried out with three different tools/methods in order to verify the achieved results, and find potential differences between dynamic simulations and stepwise stationary simulations. Generally the correspondence between the different methods was good. The frequency stability studies have focused on the ability of the Nordel power system to maintain frequency control within the specified requirements during normal operation. Only control problems related to under-frequency have been considered, i.e. the requirements for the Nordel system states that the frequency during normal operation should be higher than 49.9 Hz. The reason for this is that no major technical problems are expected related to over-frequency. Wind turbines can be stopped or wind farm output can be reduced effectively to deal with generation surplus. It has also been investigated if the power flow between different regions in the Northern European system (Nordel plus West Denmark and North-West Germany) are acceptable. The case studies are chosen based on simulations performed with the WILMAR Joint Market Model (JMM) developed in WP6. The output from the JMM in terms of production, load, wind power, etc. for two consecutive hours of interest are used as input to the frequency stability simulations. The WP5 analyses simulate in more detail the variation in flow, production, load, and system frequency within the hour between the two consecutive instances given from WP6. This way the WP5 simulation may reveal overloads, unacceptable frequency deviations, or lack of balancing power. A 2001 case served as a test case to verify the simulation methods applied to simulate frequency stability. These methods included activation of power on the balancing power market. Together with the simulated 2010 cases the analysis illustrates potential problems that can be revealed with the frequency stability studies, but not with the WILMAR JMM. One possible problem in the analyses was related to the different network representation in JMM and in the frequency stability simulations.
95
12x267 TR F6212
Due to including the network impedances the power flow was not determined by the transmission limits as in the WILMAR JMM simulations, but by the physical properties of the grid. This way the flow deviated from the WILMAR JMM simulations and in some cases transmissions between regions became overloaded. One solution to this is to include network impedances in the JMM in order to reveal potential overloads. This would not guarantee that the overloads are avoided within the hour and still the WP5 type analysis would be useful. However, it will decrease the probability for heavy overload and one can argue that potentially overloads anyway could be treated by moderate trading on the balancing power market. The main conclusions from the case studies are summarized below: - Location of balancing power: In one of the simulated 2010 cases the location of balancing
power enhanced the overload of a certain transmission corridor. This illustrates the importance of not only having enough reserves, but also that the location of the reserves can become increasingly critical.
- Drop of system frequency to unacceptable levels: This is always to some extent possible in the present day system, as it is the transmission system operator that manually activates bids on the balancing power market to avoid unacceptable system frequencies during normal operation. In one of the 2010 cases the Stepwise Power Flow simulations showed that that balancing power could not be activated fast enough to keep the frequency above 49.9 Hz. Nevertheless, the amount of balancing power activated during a simulated hour was always well below 50 % of the 4400 MW requirement for balancing power in the Nordel system (if West Denmark is disregarded).
Solutions to the observed problems and recommendations regarding integration measures are either related to improvements in technical control, new market arrangements or network reinforcements: - Continue to develop market based solutions t5hat enable demand (flexible loads) to take part
in power balancing and reserve markets. - Develop Market based automatic generation control (AGC) as a technical solution to perform
faster balancing control. - Require wind power to take part in the primary control. This increases the cost of wind power
as part of the potential production can not be used. It does, however, increase the system bias and this way the frequency stability.
- Increase requirements for on-line reserves. This will improve frequency stability but is also an expensive solution.
- Enhancing the transmission capacities to reduce local congestions. This means costly investments and is often a long process.
96
12x267 TR F6212
���� 6PDOO�VLJQDO�VWDELOLW\� Through simulations it has been showed how large-scale integration of wind power in Norway may influence the damping of inter-area oscillation modes in the Nordel system. There are two major inter-area oscillations in the Nordel system. One mode (0.3 Hz mode) is characterized by generators in Finland oscillating against the rest of the system, while the other main inter-area mode (0.58 Hz mode) is characterized by generators in Norway oscillating against generators in Sweden and Denmark. The simulations showed that it is only the oscillation mode between Norway and Sweden that is influenced to a significant degree from large-scale wind power integration in Norway. The reason for that the inter-area oscillation mode between Finland and Sweden is not influenced by large scale wind power integration in Norway is simply because the 0.3 Hz mode is not influenced by the generators in Norway. The simulations indicate that the different wind power technologies will influence the damping of the 0.58 Hz mode differently. Basically, wind turbines with the traditional squirrel cage induction generators will improve the damping of this mode, while wind turbines with the doubly fed induction generators or full frequency converters will decrease the damping of the 0.58 Hz mode. The worst impact on damping of the 0.58 Hz mode is from wind turbines with the direct drive synchronous generators and frequency converters. In this case the damping of the 0.58 Hz mode is decreased from 6.6 % in the case without wind power in Norway to 3.8 % in the case with 5000 MW of wind turbines in Norway. It is important to keep in mind that wind turbines can be controlled in different ways, and that the power and speed control strategies of the wind turbines may influence the system damping. However, large-scale wind power integration will influence the system damping, and the influence might be significant unless the power control systems are designed to provide damping. There exist well known solutions to damp inter-area oscillation modes, referred to as power system stabilisers. Power system stabilisers are usually designed as low cost auxiliary control loops, which can be realised as an extra control loop in the excitation system of a generator, a FACTS device like SVC, or a HVDC-link. ���� 7UDQVLHQW�VWDELOLW\� In the transient stability studies it was considered how a future 198 MW offshore wind farm in Eastern-Denmark would influence the post-fault conditions in the Nordel system, in the case of a 100 ms symmetrical three phase short circuit in Eastern-Denmark (not far from the new offshore wind farm). It was also tested if a grid frequency controller in the future 198 MW wind farm could improve the post-fault conditions in the Nordel system. The main finding can be summarized as:
97
12x267 TR F6212
- The simulations shows that the future offshore wind farm at Nysted will cause increased stress on the voltage locally in South-East Denmark.
- It will also upset the grid frequency, as it causes strong rotor speed oscillations in the
synchronous generator modelling East-Denmark (these oscillations are observed throughout the Nordel system).
- The grid frequency controller in the new wind farm at Nysted will give positive damping of
these oscillations, and improve the voltage response at Nysted after the fault has been disconnected.
98
12x267 TR F6212
�� 5()(5(1&(6�
1. Bakken, B.H: “Technincal and economic aspects of operation of thermal and hydro power system”, Doctoral Dissertation at the Norwegian University of Science and Technology, 1997
2. Kundur, P: ”Power System Stability and Control”, McGraw-Hill, Inc. 1994 3. Bakken, B.H, Petterteig, A: “Alternatives to reduce reserve requirements and reserve costs
in the Nordel system”, SINTEF TR A6034, February 2005 4. ”AVTAL om driften av det sammankopplade nordiska kraftsystemet (Systemdriftsavtal)”,
Nordel 2004-04-01 (in Swedish) 5. Lindahl, S: ”Frekvenshållning innom Nordel-systemet”, Nordel Annual Report 1983 (in
Swedish) 6. ”Tertialrapport 3/2003 fra Landssentralen”, Statnett, 2004-01-07 (in Norwegian) 7. ”Automatiska driftsåtgärder och gällande regler rörande driftreserven i det synkrona
nordiska kraftsystemet som består av Danmark öster om Stora Bält, Finland, Norge och Sverige”, Nordel’s Operations Committee 1982, approved by Nordel 1982-09-02 (in Swedish)
8. ”Tidsrespons for frekvensstyrede reserver i Nordel-nettet”, Report to Nordel’s Operations Committee from System Operation Working Group (NOSY), January 1982 (in Norwegian)
9. Nordel Annual Report 2003 10. Bakken, B.H: “Krav til reserver – En sammenlikning av ulike europeiske regelverk”,
SINTEF AN 01.12.106, January 2002 11. http://www.pti-us.com/pti/software/psse/index.cfm 2005-11-01 12. Bakken, B.H, Petterteig, A, Haugan, E, Walther, B: ”Stepwise Power Flow – A new Tool
to Analyse Capacity Shortage and Reserve Requirements”, 15th Power Systems Computation Conference, Liege, Belgium, August 2005
13. Zimmerman, R.D, Gan, D: ”Mathpower – A MATLAB Power System Simulation Package”, User’s Manual, PSERC, December 1997
14. http://www.mathworks.com/ 2005.03.23
15. Nørgaard, P, Holttinen, H: “A Multi-Turbine Power Curve Approach”, Nordic Wind Power Conference, 1-2 March 2004, Chalmers University of Technology
16. Bjørndal M, Jørnsten K. Zonal Pricing in a deregulated electricity market. The Energy Journal, Vol 22, No. 1. p 51-73.
17. Hagstrøm, E, Norheim, I, Uhlen, K: “Large Scale Wind Integration in Norway and Impact on Damping in the Nordic Grid”, Nordic Wind Power Conference, 1-2 March 2004, Chalmers University of Technology
18. Hagstrøm, E, Norheim, I, Uhlen, K: “Large Scale Wind Integration in Norway and Impact on Damping in the Nordic Grid”, Wind Energy Journal, Vol. 8, NO. 3, July-September 2005.
99
12x267 TR F6212
19. Slootweg, J.G : “Wind power, Modeling and Impact on Power System Dynamics”, Ph.D, Power Systems Loaboratory, Delft University of Technology, 2003
20. http://www.eltra.dk/media(14949,1033)/Evaluering_av_kraftmarkedet_vinter_2003.pdf 2005.09.23
21. http://www.gams.com 22. Lindgren, E, Söder, L: “Minimizing Regulation Costs in Multi-Area Markets”
Power Systems Computation Conference PSCC'05, Aug 2005, Liège, Belgium 23. http://www.statnett.no/Resources/Filer/Prosjekter/Nea%20-
%20Järpströmmen/Nea_jarpstrommen_Melding_sept2004.pdf 2005.10.20
24. Vindmøller tilsluttet net med spændinger under 100 kV. Teknisk forskrift for vindmøllers egenskaber og regulering. Elkraft System og Eltra. TF 3.2.6. November 2004.
25. Vindmøller tilsluttet net med spændinger over 100 kV. Teknisk forskrift for vindmølleparkers egenskaber og regulering. Elkraft System og Eltra. TF 3.2.5. December 2004.
26. DIgSILENT GmbH. Manuals DIgSILENT PowerFactory Version 13. 2004, www.digsilent.de (available August 2005)
27. ENERGI E2. Nysted Havmøllepark. 2005, http://uk.nystedhavmoellepark.dk/frames.asp (available July 2005)
28. Energistyrelsen. Stamdataregister for Vindmøller. 2004, http://www.ens.dk/sw1954.asp (available August 2004)
29. SEAS Wind Energy Centre. Nysted Offshore Wind Farm at Rødsand, Annual Status Report for the Environmental Monitoring Program. 2001, http://www.offshore-wind.de/media/article000336/Annual%20Status%20Report%20for%20the%20Environmental%20Monitoring%20Programm,%20Rodsand%202001.pdf (available 2004)
30. Sørensen P., Hansen A.D., and Rosas P.A.C. Wind Models for Simulation of Power Fluctuations From Wind Farms. Journal of Wind Engineering & Industrial Aerodynamics , pp. 1381-1402, 2002,
31. Hansen A.D., Sørensen P., Blaabjerg F., and Bech J. Dynamic Modelling of Wind Farm Grid Interaction. WIND ENGINEERING , VOL: 26, ISSUE: 4, pp. 191-208, 2002,
32. Sørensen P., Hansen A., Christensen P., Meritz M., Bech J., Bak-Jensen B., and Nielsen H. Simulation and Verification of Transient Events in Large Wind Power Installations. VOL: Risø-R-1331(EN), PUBLISHER: Risø National Laboratory, (Roskilde, Denmark), 2003,
33. Jauch C, Hansen A.D., Sørensen P., and Blaabjerg F. Simulation Model of an Active-Stall Fixed-Speed Wind Turbine Controller. WIND ENGINEERING , VOL: 28, ISSUE: 23, pp. 177-198, 2004,
34. Jauch, C., Sørensen, P., and Bak-Jensen, B. Simulation Model of a Transient Fault Controller for an Active-Stall Wind Turbine. WIND ENGINEERING , VOL: 29, ISSUE: 1, pp. 33-48, 2005,
35. Jauch, C., Cronin, T., Sørensen, P., and Bak-Jensen, B. A Fuzzy Logic Pitch Angle Controller for Grid Frequency Stabilisation. Wind Energy , VOL: submitted for publication, 2005,
100
12x267 TR F6212
36. A.J.Wood and B.F.Wollenberg. Power generation operation and control. Second Edition. John Wiley & Sons, Inc. New York 1996.
