ELNET 2015 - Katedra informatiky FEI VŠB-TUO · The conference ELNET 2015 was held on 24th November 2015 at V SB-Technical University of Ostrava, ... Michal Kolcun (Technical University

ELNET 2015Proceedings of the 12th Workshop

Faculty of Electrical Engineering and Computer ScienceVSB – Technical University of Ostrava

ISBN 978–80–248–3858–8

ELNET 2015http://www.cs.vsb.cz/elnet/

12th WorkshopOstrava, 24th November 2015Proceedings of papers

Organized by

VSB – Technical University of OstravaFaculty of Electrical Engineering and Computer Science

ELNET 2015c© Radomır Gono, editor

ISBN 978–80–248–3858–8

This work is subject to copyright. All rights reserved. Reproduction or publication ofthis material, even partial, is allowed only with the editors’ permission.

Technical editors:

Peter Chovanec [email protected]

Michal Kratky [email protected]

Faculty of Electrical Engineering and Computer Science,VSB – Technical University of Ostrava

Page count: 90Impression: 100Edition: 1st

First published: 2015

This proceedings was typeset by PDFLATEX.

Published by Faculty of Electrical Engineering and Computer Science, VSB – Technical

University of Ostrava

Preface

The conference ELNET 2015 was held on 24th November 2015 at VSB-TechnicalUniversity of Ostrava, Czech Republic. This is the twelfth conference.

The conception of ELNET conferences was a response to increasing interestin Energy and Power Systems and related aspects in the Czech Republic andSlovakia, in the last few years. An important point is the interdisciplinary natureof key topics of the conference:

– Energy and Power Systems– Distributed Power Generation– Fault Diagnosis– Power Breakdown Analysis– Survivable Network System Analysis– Energy Data Storing and Analysis– Visualisation– Structure and Grow of Networks

ELNET is a workshop intended for meeting of promoters of Energy andPower Systems and related aspects. It is focused on theoretical and technicalfoundations of information technologies, time-proven methods and developmenttrends. It also serves as a place for discussion about new ideas.

Conference provided an excellent opportunity for faculty, scholars, and prac-titioners to meet renowned researchers and to discuss innovative ideas, resultsof research, and best practices on various conference topics.

I would like to cordially thank the authors and PC members for their effort,materialised in this volume. Special thanks go to the Organising Committeemembers for their arduous editing work.

In conclusion, I would like to thank all contributing authors for their excellentresearch papers.

November 2015 Zdenek HradılekProgram Committee Chair

ELNET 2015

Organization

Evaluation Committee

Chair:Zdenek Hradılek (VSB – Technical University of Ostrava, Czech Republic)

Members:Vaclav Snasel (VSB – Technical University of Ostrava, Czech Republic)Stanislav Rusek (VSB – Technical University of Ostrava, Czech Republic)Ales Horak (Masaryk University in Brno, Czech Republic)

Program Committee

Anna Gawlak (Technical University Czestochowa, Poland)Radomır Gono (VSB – Technical University of Ostrava, Czech Republic)Przemyslaw Janik (Technical University Wroclaw, Poland)Michal Kolcun (Technical University Kosice, Slovak Republic)Zbigniew Leonowicz (Technical University Wroclaw, Poland)Zbynek Martnek (University of West Bohemia, Czech Republic)Harald Schwarz (BTU Cottbus, Germany)Jerzy Szkutnik (Technical University Czestochowa, Poland)Petr Toman (VUT Brno, Czech Republic)Ladislav Varga (Technical University Kosice, Slovak Republic)Jirı Tuma (CVUT Praha, Czech Republic)

Organizing Committee

Peter Chovanec (VSB – Technical University of Ostrava, Czech Republic)Michal Kratky (VSB – Technical University of Ostrava, Czech Republic)Yveta Geleticova (VSB – Technical University of Ostrava, Czech Republic)

VII

Workshop Location:

Campus of VSB – Technical University of Ostrava17. listopadu 15, 708 33 Ostrava–Poruba, Czech Republic24th November 2015

http://www.cs.vsb.cz/elnet/

VIII

Sponsor

Workshop ELNET 2015 is supported by Skupina CEZ.

http://www.cez.cz/

Table of Contents

Determining the Reliability Indicators of Industrial Local DistributionSystem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1Jirı Drholec, Radomır Gono

Energy Balance Model of Biogas Station . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7Jirı Jansa, Zdenek Hradılek

Ultracapacitors in Hydrogen Energy Storage . . . . . . . . . . . . . . . . . . . . . . . . . . 15Michal Ney, Zdenek Hradılek

Modeling of a technological centre in Ostrava using the softwareEMTP-ATP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22Tomas Mozdren, Stanislav Rusek, Radomır Gono, Veleslav Mach

Analysis of Data Measured in Tosanovice Biogas Station . . . . . . . . . . . . . . . 28Ladislav Novosad, Zdenek Hradılek, Petr Moldrık

Regulation of Cogeneration Unit in Industry . . . . . . . . . . . . . . . . . . . . . . . . . 37Michal Spacek, Zdenek Hradılek

Dynamic Model of Asynchronous Machine . . . . . . . . . . . . . . . . . . . . . . . . . . . 45Martin Kral, Radomır Gono

Actual Results of the Reliability Computation in 2015 . . . . . . . . . . . . . . . . . 53Radomır Gono, Stanislav Rusek, Pavel Bednar, Peter Chovanec,Michal Kratky

Utilization of Signature Methods for Range Query Processing overOutage Database . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62Peter Chovanec, Michal Kratky

Scalable GPU Range Query Processing over Outage Database . . . . . . . . . . 69Pavel Bednar, Michal Kratky

Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

X

Determining the Reliability Indicators ofIndustrial Local Distribution System

Jirı Drholec, Radomır Gono

Department of Electrical Power Engineering,FEECS, VSB – Technical University of Ostrava,

17. Listopadu 15/2172, 708 33 Ostrava-Poruba, Czech [email protected], [email protected]

Determining the Reliability Indicators of Industrial Local Distribution System

Jiří Drholec, Radomír Goňo

Department of Electrical Power Engineering VŠB - Technical University of Ostrava, 17. listopadu 15, 708 33 Ostrava – Poruba

[email protected], [email protected]

Abstract. This paper deals with a collection and processing of information from the dispatcher logs related to the operation of the local distribution system in the area of an industrial plant. The first part focuses on development of equipment passportization and processing of documentation for system opera-tion. Additionally, the work deals with calculation of reliability indicators of overhead lines and components. The final part is the comparison of selected equipment reliability indicators of the local distribution system and the power distribution system of the Czech Republic.

Keywords: Reliability, local distribution system, industrial plant, control cen-tre, distribution plant, failure rate, mean time to repair, faultless operation prob-ability.

1 Introduction

This paper deals with collection and processing of data from control centre logs in the company concerned with industrial power engineering [1], which is a complex and extensive enterprise focused specially on the needs of metallurgical and steel industry. Its scope of activities includes provision of services within the fields of water econo-my, combined heat and power production, gas engineering, power engineering and production of technical gases. Additionally, the work deals with calculation of relia-bility indicators of the local distribution system (LDS) in the premises of the industri-al plant which focuses on pig iron and metallurgy production and processing, as well as on engineering production [7]. The final part lays emphasis on comparison of reli-ability indicators between two distribution systems.

2 Collection of data on outages and failures of power supply

The most data on reliability of power supply within LDS can be obtained by monitor-ing and evaluation of its operation. Efficient records of downtime, outages and fail-

c© Radomır Gono (Ed.): ELNET 2015, pp. 1–6, ISBN 978–80–248–3858–8.VSB – Technical University of Ostrava, FEECS, 2015.

2 Jirı Drholec, Radomır Gono

ures with subsequent analyses of their consequences is based on procurement of the set of data for evaluation of various reliability indicators. However, application of the reliability theory requires gathering and processing of large amount of monitoring data. That is why collection of primary data is vital to ensure their further sorting and processing.

As far as the heating plant and operator of LDS is concerned, the primary data is collected by the head supervisor of control centre maintaining permanent presence at this site. The records contain mainly this type of data: - information about the location and cause of failure, - circumstances preceding the failure occurrence, - course of the failure, - cause of failure and its further impact operation of LDS, - type of equipment/component affected, - time of failure occurrence / extinction.

All events associated with operation of LDS are recorded by the Head Operations Officer at the control centre manually, using the Operation Log that is available to the Head Supervisor of Power Supply Control (VSME) as the material for sign off during the shift in the Lotus Notes interface. The application is based on a text document only and it is not designed for filtering of any items or events. The report on a failure occurring during particular shift will be generated by the attending supervisor, this report is necessary. Once completed, the report will be forwarded to the person au-thorised and responsible for evaluation of reliable LDS operation for further pro-cessing. That will help with prevention of accidental omissions of less serious fail-ures. For the list of data and the set of numerical codes to be included in such failure report, see the distribution system operation principles [8].

3 Implementation of methodology for operation data

The entire LDS is required to provide highly reliable service with a very low failure rate to avoid any failures in supply of electric power to end users, supported by the method of operation and backup employed [4].

3.1 Scope of power distribution system

Precise enumeration of potential reliability indicators shall be based on accurate in-formation about the number of elements and their voltage levels. That is why these elements need to undergo the passportization process. The local distribution system in question contains approximately 1400 cells and MV and HV power cabinets in 52 distribution plants within the LDS operated by the heat production facility of this company.

Determining the Reliability Indicators of Industrial Local Distribution ... 3

a) Cable runs, cable ducts and bridges Fig. 1 shows an overview of cable lines 6 kV, 22 kV and 110 kV. The lengths of cable ducts are in kilometres and these include all the cable outlets, cable connections and cables linked with the transformer for power transmission. These do not include the cable lines to technological and distribution transformers, motors and technological cable connections operated by other plants of the industrial enterprise.

Fig. 1. Overview of the scope and installation of cable network rated for 6, 22 and 110 kV

b) Equipment on MV and HV distribution plants The summary of equipment on MV and HV distribution plants makes use of the sin-gle-pole diagrams for these plants with subsequent passportization of all switches, disconnectors and other fittings in particular cells. For the passportization chart, refer to Table 1. The number of elements and line lengths are further complemented by other essential data for evaluation of failure rate of the system concerned, i.e. which is the element age.

Table 1. Passportization of equipment on MV and HV distribution plants

Device Number of ele-ment

Device Number of ele-ments

Bus disconnector 6 kV 1438 Outlet disconnector 22 kV 72

Bus disconnector 22 kV 132 Outlet disconnector 110 kV 1

Bus disconnector 110 kV 12 Transformer 110 / 22 kV 9

Disconnector with ZN 110 kV 6 Transformer 22 kV / LV 3

Power switch 6 kV 826 Transformer 22 / 6 kV 20

Power switch 22 kV 72 Bus bar 6 kV 826

Power switch 110 kV 5 Bus bar 22 kV 78

Outlet disconnector 6 kV 430 Bus bar 110 kV 11


3.2 Processing of LDS operation documents and calculation of element

reliability indicators

Gathering and subsequent analysis of a substantial amount of materials referring to operation of the LDS for the period of 2010 – 2014 allowed compiling the fiver-year database to document the overall behaviour of the system with respect to its opera-tion, maintenance and failures, including time stamps for each occurrence [3].

The number of events linked to operation of the industrial LDS directly to be pro-cessed amounts to approximately 2 000 each year. That is up to 10 000 over the five-year period. This is evidently not just a routine transcription of records on shift changeover by the VSME into electronic form, yet rather a specialised procedure conducted by the expert familiar with the system wiring. He must be also able to make a decision on specific categorisation of each event to filter anomalies that re-main concealed to anyone not skilled in the field of electric power engineering.

Modest mathematical tool was employed for enumeration of the essential reliabil-ity indicators for elements over the five-year period; these are listed in Table 2 [2] [3].

Table 2. Essential reliability indicators within the LDS observed

Device / element

Failure frequency

Mean time to repair

Mainte-nance

frequency

Mean time for mainte-

nance

λp [year-1]* τp [hr] λu [year-1]* τu [hr]

Bus disconnector 6 kV 0.004 9.077 0.231 15.011 Bus disconnector 22 kV 0.008 7.450 0.241 48.790 Bus disconnector 110 kV 0.000 0.000 0.000 0.000

Disconnector with ZN 110 kV 0.000 0.000 0.600 48.466 Power switch 6 kV 0.025 5.538 0.238 13.845 Power switch 22 kV 0.039 5.702 0.289 62.522 Power switch 110 kV 0.000 0.000 0.000 0.000

Outlet disconnector 6 kV 0.002 6.358 0.193 15.810 Outlet disconnector 22 kV 0.000 0.000 0.244 63.282 Outlet disconnector 110 kV 0.000 0.000 0.000 0.000 Transformer 110 / 22 kV 0.333 4.403 1.111 74.679 Transformer 22 kV / LV 0.200 0.361 1.733 23.132 Transformer 22 / 6 kV 0.130 0.481 0.970 90.282

Cable 6 kV 0.032 8.104 0.284 108.640 Cable 22 kV 0.022 11.417 0.209 53.474 Cable 110 kV 0.000 0.000 3.380 51.922

Power line 110 kV 0.037 0.700 0.186 47.443 Bus bar 6 kV 0.000 0.000 0.014 15.421 Bus bar 22 kV 0.003 3.900 0.054 14.011 Bus bar 110 kV 0.000 0.000 0.018 5.250

* cable lines are described by failure frequency units [year-1 / km]

Determining the Reliability Indicators of Industrial Local Distribution ... 5

4 Reliability results comparison

The importance of failure database with the option to determine reliability indicators is also based on the fact that there are actually certain databases showing failures of distribution networks within the Czech Republic, yet there is absolutely no failure database using data about industrial power distribution network to enable calculation of the essential reliability indicators. Unlike the reliability indicators for distribution networks within the Czech Republic, those are obviously different and cannot be subject to mutual comparisons.

Industrial grids differ by pattern, the method for connection into the grid or even geographical conditions, etc. Furthermore, power grids in heavy industry are also exposed to many adverse effects, e.g. presence of water, jolts, shaking, vibrations, heavier dust formation, high ambient temperatures or even occurrence of corrosive, chemical or other contaminating materials. That is why the element reliability results are very valuable for the existing voltage level of 6, 22 and 110 kV not subject to analysis yet.

Table 3 provides comparison of selected reliability indicators that have been enu-merated by means of the operating design principles of 22/80 ČEZ for the period of 1975 – 1990 [5], updated results from the period of 2000 – 2014 [6] and the five-year period of 2010 – 2014 with respect to the local distribution system.

Table 3. Comparison of results for essential reliability indicators

Damaged equipment ČEZ 22/80

1975 – 1990 Period

2000 – 2014 LDS

2010 – 2014

Cable 6 kV λ [year-1] not included in

the database not included in the database

0.032* τ [h] 8.104

Cable 22 kV λ [year-1] 14.5 4.661 0.022*

τ [h] 215 5.710 11.417

Transformer 110 kV / MV λ [year-1] 0.04 0.058 0.333

τ [h] 1300 0.231 4.403

Transformer MV / MV λ [year-1] not included in


0.130

τ [h] 0.481

Transformer MV / LV λ [year-1] 0.030 0.006 0.200

τ [h] 2500 5.303 0.361

Power switch 6 kV λ [year-1] not included in


0.025

τ [h] 5.538

Power switch 22 kV λ [year-1] 0.015 0.012 0.039

τ [h] 30 23.580 5.702 * With respect to the scope of LDS cable system, the resultant values of failure fre-

quencies on cables rated 6 and 22 kV refer to each 1 km rather than 100 km, which is the case of the ČEZ cable system.


5 Conclusion

Gathering and subsequent analysis of a large amount of materials referring to opera-tion of the industrial LDS in the period of 2010 – 2014 allowed compiling the data-base to document the overall behaviour of the system with respect to its operation, maintenance and failures, including time stamps for each occurrence. That will enable enumeration of the reliability indicators for significant elements within the system and their comparison against reliability indicators for other distribution systems.

Calculation of these reliability indicators has been performed using an approach different to the regular standard. Those are values of the mean time to repair and the mean time to maintenance. Both values have been calculated using the actual comple-tion of maintenance operations on equipment, as initially defined by the record on shift changeover for VSME rather than the values contained in the charts of internal regulations – the Revision Code to determine periodical performance of check-up and revision operations.

ACKNOWLEDGEMENTS

This research was partially supported by the SGS grant from VSB - Technical Uni-versity of Ostrava (No. SP2015/192) and by the project TUCENET (No. LO1404).

References

1. Hradílek, Z.: Elektroenergetika distribučních a průmyslových sítí. Ostrava. VŠB -TU Ostra-va (2008)

2. Brown, R. E.: Electric power distribution reliability. New York, USA. Marcel Dekker (2002)

3. Todinov, M. T.: Reliability and Risk Models: Setting Reliability Requirements. Chichester. John Wiley & Sons Ltd., England (2005)

4. Barlow, R. E. & Proschan, F.: Statistical theory of reliability and life testing: probability models. New York, USA. Holt, Rinehart and Winston (1975)

5. Tůma, J., Rusek, S., Martínek, Z., Chemišinec, I., Goňo, R.: Spolehlivost v elektroenergeti-ce. Praha. ČVUT Praha (2006)

6. Goňo, R., Rusek, S., Krátký, M., Slivka, M.: Component Reliability Parameters of Distribu-tion Network. In Proceedings of the 8th International Scientific Symposium on Electrical Power Engineering (Elektroenergetika 2015), Stará Lesná, Slovakia. Košice. Technical Uni-versity of Košice (2015) 376-379

7. Drholec, J., Goňo, R.: Operation of Industrial Local Distribution System. In Proceedings of the 16th International Scientific Conference Electric Power Engineering (EPE), Kouty nad Desnou, Czech Republic. Ostrava. VŠB-TU Ostrava (2015) 478-483

8. ČEZ Distribuce, a. s. Pravidla provozování distribuční soustavy 2014 [on line], 2014 [cito-váno 2015-11-15] Dostupné z: http://www.cezdistribuce.cz/cs/energeticka-legislativa/pravidla-provozovani-ds/ppds-2014.html.

Energy Balance Model of Biogas Station

Jirı Jansa, Zdenek Hradılek



Energy Balance Model of Biogas Station

Jiří Janša1, Zdeněk Hradílek

2

1 VŠB – TU Ostrava, Department of Electrical Power Engineering, 17. listopadu 15,

708 33 Ostrava, http://fei1.vsb.cz/kat410/

tel: +420 597 329 321, email: [email protected], 2 VŠB – TU Ostrava, Department of Electrical Power Engineering, 17. listopadu 15,

708 33 Ostrava, http://fei1.vsb.cz/kat410/

tel: +420 597 325 919, email: [email protected],

Abstract. Model of biogas station operates by entering annual amount of each input raw

materials. To raw materials are assigned biogas yield per tonne of input material. The model

then calculates from the specified amount the total annual energy input of the fuel in MWh.

From this annual energy is then determined by the approximate input of a biogas plant. The

next step is to select a particular cogeneration units from a list containing the cogeneration

units burning biogas from leading suppliers ie. TEDOM, Schnell, Jenbacher and others. After

selecting cogeneration units (CHPU) according electric power will be displayed the other pa-

rameters. It´s electrical performance, thermal performance, electrical, thermal and total efficien-

cy, and above all power needed in the fuel. User can select from 1 to 10 CHPU. Energy Bal-

ance of biogas station is in the next part. Data obtained from measurements on specific biogas

plant was used for verification of the model. On this biogas plant was measured electrical and

thermal parameters.

Keywords: model, energy balance, biogas station, cogeneration unit,

Model of Biogas station

Model of biogas station operates by entering annual amount of each input raw materi-

als. To raw materials are assigned biogas yield per tonne of input material. The model

then calculates from the specified amount the total annual energy input of the fuel in

MWh. From this annual energy is then determined by the approximate input of a bio-

gas plant. The next step is to select a particular cogeneration units from a list contain-

ing the cogeneration units burning biogas from leading suppliers ie. TEDOM, Schnell,

Jenbacher and others. After selecting cogeneration units (CHPU) according electric

power will be displayed the other parameters. It´s electrical performance, thermal

performance, electrical, thermal and total efficiency, and above all power needed in

the fuel. User can select from 1 to 10 CHPU. Below the table with the parameters we

will see if I still have some power in the fuel remain, or vice versa missing. Energy

Balance of biogas station is in the next part. We see a number of days in the individual

month. Furthermore, it shows the number of days of shutdowns cogeneration units,

which we can choose. In the event that it does not do is set automatically shutdowns

for 6 hours after 500 hours and one day shutdown for larger service intervention every

6 months of operation. Another parameter is operating hours for each month and the

monthly fuel consumption and its energy content. According to the parameters of each


8 Jirı Jansa, Zdenek Hradılek

CHPU is calculated produced electricity, produced thermal energy, self-consumption

of electricity, self-consumption of heat, amount of usable electricity (supply into the

network) and the amount of usable heat. From these values are then plotted monthly

charts for the electrical parameters - generated electricity, own consumption, usable

electrical energy; and for heat parameters - generated heat, own consumption of heat

and the amount of usable heat.

Model verification

This article focuses mainly to verification of this model, due to assess the extent to

which manufacturers the indicated value differs from the actual values measured in

real operation. For verification of the model has been used data obtained from meas-

urements at a specific biogas plant. On this biogas plant was measured electric and

thermal parameters. These measurements are described in more detail in the next

section.

Biogas station description

The biogas plant (BGS) in which the measurement was carried out, is located on the

territory of the Moravian-Silesian Region. BGS is in the area of the farm, which is

engaged in pig holding and processed mainly corn silage and pig manure. The main

reason for this location was a source of pig manure, which serves as a source of fluids

for needs wet fermentation technology. One other reason this location was the possi-

bility of using waste heat for heating the adjacent stables, office buildings, and in the

summer in a newly built after-harvest line used to dry corn. The biogas plant consists

of two fermenters, each with a usable volume of 1630 m3, and after-fermenter with

usable volume of 2090 m3. Installed electrical output has a value of 1,090 kW and

thermal output has 1,080 kW. For converting biogas into electrical energy are respon-

sible four cogeneration unit. These are the three identical compression ignition unit

with an output of 250 kW and one spark ignition unit with an output of 340 kWe.

Energy Balance Model of Biogas Station 9

Fig. 1 Cogeneration unit Agrogen 222

Measurement of electrical parameters

Measurements were realized with the help of an automatic digital measuring device

ENA network analyzer 500 from the company ELCOMwhich at one minute intervals

measured and saved in the memory of the rms value of the phase voltage, current and

power factor of each phase. The remaining variables were calculated using the device,

ie. active, reactive and apparent power.


Fig. 2 The involvement of the measuring device into switchboard

Voltage was measured directly on the busbar in the switchboard RH1 and currents

using clamp current transducers MT-UNI using the already installed current instru-

ment transformers 1500 / 5A.[1]

Own consumption of heat and heat measurement

We obtained mass flow rates of heat transfer media from the measurements by ultra-

sonic flowmeter. We measured the temperature of the outlet and return using thermo-

couples, we have identified specific heat capacity and searched by the type of flowing

medium and temperature measured by the software Engineering Equation Solver.

These data were used to calculate the heat output, which transported the individual

pipes.


Fig. 3 Measurement of flow and temperature

Because the flows of biogas into CHPUs were on different terms than those given

standard, were recalculated based on the temperature and pressure of gas in so called

normal conditions, a temperature of 0 ° C and a pressure of 101 325 Pa. Energy inputs

in biogas supplied to CHPUs units were calculated based on recalculated flow of

biogas to normal conditions and the calorific value of biogas determined by analyzing

the chemical composition of biogas. Biogas plant operator had to prepare an analysis

of the chemical composition of biogas in the past

On the basis of power of biogas entering the CHPU and power generation at the gen-

erator was determined CHPUs electric efficiency.[2]

Comparison of the model with the measured data

Parameters of cogeneration units, which are located in our measured BGS were insert-

ed into the model. These are the values provided by the manufacturer of the cogenera-

tion units.[3],[4] Results of comparing between the model and actual measured data

are presented in the following tables.


Table 1 Fuel consumption

Fuel consumption Schnell ZV250-V5 Agrogen BGA222

Model 549 kW 886 kW

Measurement 560 kW 864 kW

Difference 11 kW -22 kW

Difference 2.003643 % -2.48307 %

The measurement shows that fuel consumption is about 11kW more than the value

stated by the manufacturer, for weaker CHPU with an electrical output 250 kW. This

is a deviation roughly 2%. In contrast, the stronger CHPU with an electrical output

340 kW, the measured value is lower by 22 kW, is around 2.5% of the total power

consumption of the fuel. This is due a reduction of the performance of CHPU to 340

kW of maximum 350 kW. This reduction in electric power has been carried out by the

supplier due to exceeding permitted installed capacity, and since the beginning of

commissioning the BGS is operated as follows. However, in the model were retained

the original nominal values. The manufacturer states that the CHPU can be operated

with a reduced output, but does not mention the individual parameters CHPU with

reduced output.

Table 2 Electric output

Electric output Schnell ZV250-V5 Agrogen BGA222

Model 250 kW 350 kW


Difference 1 kW 10 kW

Difference 0.4 % 2.857143 %

Electrical power was maintained during the measurement to the set value when the

deviation is less than 0.5% and it is 1kW. In the second CHP the measured value dif-

fers from the nominal value of 10 kW, it is less than 3%, which is, as mentioned earli-

er, due to the operation of CHP at lower power.

Table 3 Electric efficiency

Electric efficiency Schnell ZV250-V5 Agrogen BGA222

Model 45.5 % 39.5 %

Measurement 44.5 % 39.4 %

Difference -1 % -0.1 %

Difference -2.35643 % -0.3836 %

Measured electrical efficiency in both cases was lower than indicated by the manufac-

turer. This may be caused by other than normal ambient conditions (temperature,

pressure, humidity) in real measurements.


Table 4 Heat output

Heat output Schnell ZV250-V5 Agrogen BGA222

Model 220 kW 350 kW


Difference -3 kW 9 kW

Difference -1.36364 % 2.571429 %

Heat output was for the first CHPU slightly smaller than the value specified by the

manufacturer, which could be due to heat loss. In real terms was not measured directly

at the output of the CHPU because of the lack of space for measuring instrumentsIn

the second cogeneration unit is on the contrary achieved a higher heat output than the

nominal, which may be due to the non-contemporary measurement of two cooling

circuits. Also, it is to some extent due to the operation of CHPU at lower electrical

power than nominal.

Table 5 Heat efficiency

Heat efficiency Schnell ZV250-V5 Agrogen BGA222

Model 40.07286 % 39.50339 %

Measurement 38.75 % 41.55093 %

Difference -1.32286 % 2.04754 %

Difference -3.30114 % 5.183201 %

Heat efficiency values correspond to the measured heat output, i.e. the first CHPU has

lower efficiency and conversely the second has a higher efficiency by 5% (related to

the nominal value of the thermal efficiency).

Table 6 Total efficiency

Total efficiency Schnell ZV250-V5 Agrogen BGA222

Model 85.6102 % 79.0068 %

Measurement 83.2143 % 80.9028 %

Difference -2.3959 % 1.896 %

Difference -2.79863 % 2.399802 %

The overall efficiency indicated by the manufacturer for the first cogeneration unit

Schnell ZV250-V5 is 85%. We obtained the value of 83% from the measurements,

which represents a deviation of less than 3%. In the second CHP unit Agrogen

BGA222 manufacturer indicates the total efficiency of 79%. Resulted from measuring

the total efficiency of 81%, the deviation is almost 2.5%, but in this case the meas-

urement showed greater efficacy than the nominal. This deviation may be caused by

measurement inaccuracies and potentially lower energy content of the fuel, than which

it was calculated. For a more accurate comparison would have to be measured by the


composition of the biogas at the time of measuring other parameters, as opposed sup-

ports for calculation when we calculate with the regular analysis of the composition of

the biogas.

Conclusion

In this article, we examined by the verification of the model BPS using measurements

on real BGS. In the model we have chosen the same CHPU as in the specific BPS,

which was measuring. It is a cogeneration unit Schnell ZV250-V5 and Agrogen

BGA222. After comparing the measured parameters of electric, heat output and power

consumption of the fuel, we also compared the calculated parameters based on these

measured values and it is electrical, thermal and total efficiency. As expected, the

measured values differ slightly from the values of the model based on the

manufacturer rated values. However, the deviation is at best 0.4%, and in the worst

case 5.2%. The average deviation is then about 2.4%. We have worked mainly on

verification of the model parameters, which serve as the basis for the energy balance

of biogas plants. This model is and will continue to evolve toward more detailed

energy balance of BPS, ie. Operating hours, the amount of produced electricity, the

amount of generated heat, own consumption of electricity, own consumption of heat....

The model should devote further appreciation of the possibilities of using untapped

heat from cogeneration units, which is discharged into the environment, and thus to

increase the overall efficiency of the biogas plant.

Acknowledgement

This work was supported by the Ministry of Education, Youth and Sports of the Czech

Republic (No. SP2015/192).

References

1. Janša J., Moldřík P. “Electrical Measurement on Cogeneration units from BGS Lodě-

nice“, 2014.

2. Janša J., Janša J., “Heat Flow Measurement Report from BGS Loděnice,” Ostrava,

2015.

3. “Produktübersicht 2012/2013,” SCHNELL motor, June 2012 4. “Technical data Gensets BGA,” AGROGEN gasmotoren, last print 2010-01-20

Ultracapacitors in Hydrogen Energy Storage

Michal Ney, Zdenek Hradılek



Ultracapacitors in Hydrogen Energy Storage

Michal Ney, Zdenek Hradilek

VSB – Technical University of Ostrava, Faculty of Electrical Engineering and Computer Science, Department of Electric Power Engineering, 17. listopadu 15,

708 33 Ostrava - Poruba, Czech Republic [email protected]

http://fei1.vsb.cz/kat410/index1.htm

Abstract Nowadays, the count of renewable energy sources such as wind and photovoltaic plants is increasing. But there`s one significant problem, the power supplied from these sources is unstable according to actual weather conditions and without any regulation. So if there`s no enough load in grid, when weather conditions are good, produced energy isn`t used properly. On the other hand, in time of peak load, weather conditions may be bad and there`s deficiency of en-ergy, that must be compensated by the traditional energy sources such as ther-mal and nuclear plants. To prevent this energy unbalance is necessary to up-grade renewable sources with some kind of energy storage. There are various energy storage technologies using different physical principles. On mechanical principle works pumped-hydroelectric plants, which are the most widespread, compressed air storage used with gas turbines and also flywheels. On chemical principle works broad range of accumulators such as classic lead-acid batteries, modern lithium-based batteries, high temperature batteries (sodium-sulfur) or flow batteries. Fuel cells can convert chemical energy of liquid or gaseous fuel directly to electric energy without any other energy conversion like thermal and mechanical in traditional plants. As a fuel for fuel cells can be used petrol, methanol, ethanol, natural gas, biogas and hydrogen. Exactly hydrogen fuel cells are useful for energy storage, because hydrogen can be produced by water electrolysis. On clearly electrical principle works ultracapacitors and supercon-ducting coils, which store energy in electric or magnetic field. In this paper are described possibilities of cooperation of hydrogen technology and ultracapaci-tors in energy storage systems.

Keywords Electrolysis, Energy Storage, Fuel Cell, Hydrogen, Ultracapacitor.

1 Introduction

Hydrogen energy storage is based on hydrogen production by water electrolysis, hy-drogen storage (in gaseous or solid state) and electricity production by fuel cells. See simplified block diagram below.


16 Michal Ney, Zdenek Hradılek

Fig. 1. Hydrogen energy storage block diagram

2 Water electrolysis hydrogen production

Water electrolysis is based on direct current conduction in aqueous solution, which leads to water splitting and separation of hydrogen and oxygen. Practically used elec-trolyzers are mostly PEM (Proton Exchange Membrane) type. Demineralized water is lead to anode, where is decomposed to oxygen, hydrogen anions and electrons. Oxy-gen is lead out as a waste product (or may be stored) and hydrogen anions go through solid electrolyte (PEM) to cathode. Electrons go through power supply circuit to cath-ode, where together with hydrogen anions form hydrogen, which is lead out as a prod-uct. Real electrolyzers use several cells in serial connection, which form so-called stack. Stacks can be used separately (Acta AES) or can be integrated to compact unit together with power supply, control system, water tank etc. (Hogen). [1]

Fig. 2. PEM electrolytic cell principle

Ultracapacitors in Hydrogen Energy Storage 17

Fig. 3. Real electrolyzers – Hogen GC600 (left) and Acta AES500 (right)

3 Hydrogen storage

Produced hydrogen can be stored in gaseous state in pressure vessels or in solid state in metal-hydride containers. Metal-hydride containers are based on chemical binding of gaseous hydrogen into solid material (nickel, magnesium, lanthanum, iron and titanium alloys). Charging reaction is exothermic, produced heat must be lead out by the cooling water circuit. Discharging reaction is endothermic, there must be heat supplied to the container by the water circuit (in case of high-temperature metal-hydrides). Low-temperature metal-hydrides can loosen hydrogen by the normal ambient temperature. [2]

Fig. 4. Metal-hydride principle and real container Hbond-1500L


4 Fuel cell electricity production

For conversion of chemical energy of stored hydrogen back to electric energy is used PEM type fuel cell. Its principle is basically reverse electrolysis. Fuel (hydrogen) is lead to anode and oxidant (oxygen from ambient air) to cathode. From anode to cathode go hydrogen anions through solid electrolyte (PEM) and electrons through external load (power output). On cathode electrons together with hydrogen anions and oxygen form water (waste product). Fuel cell consists of bipolar plate and membrane electrode assembly. Bipolar plate is made of graphite or graphite polymers and contains gas channels for fuel and oxidant input. Membrane electrode assembly consists of polymer proton exchange membrane with catalytic and diffusion layer. This membrane works as a solid electrolyte. In practical applications is not only one cell, but there are several cells serial connected in stack (module) similarly to electrolyzer. Stacks can be also used separately (Nedstack) or can be integrated to compact unit together with auxiliary devices like cooling, control system etc. (Ballard). [3]

Fig. 5. PEM fuel cell principle


Fig. 6. Real fuel cell modules – Ballard Nexa 1.2 kW (left) and Nedstack FCS 8 kW (right)

5 Ultracapacitors

Ultracapacitors handle capacitance up to 10,000 F, energy density orderly 10 Wh/kg, long lifetime, high efficiency up to 95 %, low internal resistance orderly 1 mΩ, extremely short charge and discharge times, high charge and discharge cycles count and short time current capacity up to 4,500 A. Ultracapacitors contain anode and cathode made from aluminum foil with carbon powder layer for significant increase of electrode surface. Electrolyte may be in liquid or gel state. While ultracapacitor is discharged, ions are evenly distributed in electrolyte. While voltage source is connected to electrodes, cations are drawn to anode and anions to cathode. On electrodes arise double layer with mirrored charge distribution, the so-called Helmholtz double layer. Operation voltage of ultracapacitor cell is between 1.5 and 3.8 V, modules of serial connected cells with total voltage up to 190 V are available.

1 – DC source 2 – Electrode leads 3 – Electrodes 4 – Helmholtz double layer 5 – Electrolyte 6 – Separator

Fig. 7. Ultracapacitor principle


6 Ultracapacitor and hydrogen technology cooperation

There are many possibilities how to use ultracapacitors, in hydrogen technology is suitable parallel cooperation with fuel cell because of its soft load characteristics. Fuel cell voltage at nominal current may be almost a half of its no load voltage. This fact may cause unreliable function of DC/AC inverters during current peaks (load switching on). By using ultracapacitor (with appropriate DC/DC converter) this current peaks can be supplied from its capacity and final output voltage is harder. Similar situation is in parallel cooperation of two (or more) fuel cells, when no loaded fuel cell (G1) is parallel connected with other fuel cell under load (G2). In the moment of connection, each fuel cell has different voltage that cause current peak from no loaded cell (G1) with higher voltage to cell under load (G2) with lower voltage. During this transient process, final voltage is stabilized at value between initial values and energy of current peak is absorbed by fuel cell G2. Current absorption by fuel cell is possible by reverse saturation of its anode. If we use parallel ultracapacitor with cell G2 and then connect cell G1, energy of current peak during potential balancing is absorbed by the ultracapacitor and output voltage is no significant influenced by this transient process. [4]

Fig. 8. Fuel cell load characteristic (Nedstack FCS 2 kW)


GG1

GG2 Load (grid)C

Fig. 9. Ideological diagram of parallel cooperation of fuel cells with ultracapacitor

7 Conclusion

Ultracapacitors may have useful technical contribution to hydrogen fuel cell technolo-gy, but practical applications are limited by still high price of ultracapacitors. Overall hydrogen energy storage is expensive technology due to high prices of used devices. Further research in this field of energy storage technologies is necessary to its utiliza-tion in renewable sources, backup sources, smaller island grids etc.

Acknowledgment

This work was supported by the SGS grant from VSB – Technical University of Os-trava No. SP2015/192.

References

1. http://www.siei.org/mainpage.html 2. http://www.labtech-hydrogen.com/common_files/brochure.pdf 3. http://www.ballard.com/ 4. Minařík D., Mlčák T., Sokanský K., Využití superkapacitoru pro stabilizaci chodu palivo-

vého článku, EPE 2009

Modeling of a technological centre in Ostravausing the software EMTP-ATP

Tomas Mozdren, Stanislav Rusek, Radomır Gono, Veleslav Mach


17. Listopadu 15/2172, 708 33 Ostrava-Poruba, Czech [email protected], [email protected], [email protected],

[email protected]

Modeling of a technological centre in Ostrava using the software EMTP-ATP

Tomáš Mozdřeň1, Stanislav Rusek1, Radomír Goňo1 Veleslav Mach1

1 Department of Electrical Power Engineering VŠB – Technical University of Ostrava, 17. listopadu 15, 708 33, Ostrava - Poruba [email protected], [email protected], rado-


Abstract This paper deals with the analysis of the energy production units, which are introduced in the laboratories of the technological centre in Os-trava-Vitkovice and according to which it will be created a dynamic mo-del of the technological centre. The following dynamic model is created in the program of the EMTP ATP. The current status of the dynamic mo-del based on the schema of the block in which they are arranged all of the non-traditional energy sources.

Keywords: Dynamic model, technological centre, software, AC bus, DC bus

1 Introduction

The current demand for sources of electric power keeps rising and the existing de-posits of fossil fuels are unfortunately exhaustible. This is the reason for accelerated efforts towards search and application of new renewable energy sources for humankind. The reasons mentioned above then supported evolution of the ENET project dealing with research on utilisation and application options covering non-traditional sources of energy in the laboratory of the technological centre in Ostrava-Vitkovice ("TCO") since the year 2010.

2 Connections and technology TCO

Simplified electrical diagram of the technological centre is shown on Fig. 1. Individ-ual energy sources are split for better clarity and drawn into the technological blocks, which is bringing together according to their basic functional principles, or the project-ed follow-up to other technologies in the TCO. The primary power and sheer perfor-mance, TCO also Center is provided through the connection on the local distribution


Modeling of a technological centre in Ostrava using ... 23

system. Backbone source of electrical performance, TCO is a cogeneration unit with an asynchronous generator with an output of 100 kW cooperating with gas piston internal combustion engine. This unit is supplied with gas from the pyrolytic process unit con-tinuously optimises typically 250-450 kg of fuel material per hour (depending on the composition). The topology of the distribution system of electricity TCO is made up of the main AC bus with the possibility of connection to the AC backup bus. The whole distribution system is designed and implemented as a fully self-contained with the even-tual options to go into operation in the island of 3x230/400V and, if necessary, with the ability to power the adjacent industrial buildings. An important part of the TCO is the equipment different accumulation systems. This accumulation is ensured by different types of batteries, which are always connected with the main DC bus DC 400V. This allows the electric power into individual parts and equipment of a divorce by means of inverters. The link between the DC bus and bus with standard network parameters is ensured by a power semiconductor inverter, which along with the detention of trans-formers and chokes TR1+TL1 a TR2+TL2 allow bidirectional energy transfer between bus. If the center is disconnected from the supply from the network and are exhausted its own storage capacity, is in the backup diesel generator, which is designed to ensure a steady supply of energy and the safe shutdown of all technologies, in particular the pyrolysis unit.

Fig. 1 Connection of technology centre

24 Tomas Mozdren, Stanislav Rusek, Radomır Gono, Veleslav Mach

2 Software for modelling of transients

For the assembly, considered a dynamic model is used, the software that is available and is used by Department of Electrical Power Engineering VŠB-TUO. This software program is the EMTP-ATP (Electro Magnetic Transients – Alternative Transients Pro-gram) that is designed mainly for modelling of transients and creation of alternate dia-grams of electric machinery, distribution lines and components. The specialty behind development of this program is the opportunity to simulate transients of electromagnetic and electromechanical nature. It is adapted for design of circuits using algebraic, partial and differential equations. This program also helps with design of extensive and com-plicated electric grids as well as control systems. The calculation process itself makes use of a simplified modification of the Newton-Raphson method.

3 The current status of the dynamic model of TCO

This section is created by the partial dynamic model in the environment of the EMTP ATP, which is shown on Fig. 2. The following model is based on the block diagram, on which are displayed the individual energy sources in TCO.

Fig. 2 Dynamic model of TCO


The topology model of electric energy is made up of TCO main power AC bus 3x230/400V and backup AC bus 3x230/400V, the voltage is shown on Fig. 3. On main bus are connected as the main source of the cogeneration unit with an output of 100kW, Stirling engine power 10kWe. If the centre is disconnected from the supply from the network and are exhausted its own storage capacity, are in the backup diesel generator and UPS to ensure a steady supply of energy and the safe shutdown of all technologies, in particular the pyrolysis unit.

The connection between the DC bus voltage 400V on the surface and with the stan-dard bus network parameters is ensured by using the reversible power semiconductor AC/DC converters and inverters DC/AC in this case, the pairs of U10, U20 and U11, U21 which along with the detention of transformers and chokes TR1+TL1 and TR2 + TL2 allow bidirectional energy transfer between DC and AC bus.

Power inverter AC / DC is modeled as 6 pulse controlled rectifier with surge pro-tectors for thyristors, where the inverter output where the output of the inverter is the value of the DC voltage 400V showing on Fig. 4. In contrast to the three-phase inverter DC / AC modeled using transistors with surge protections. Transistors are switched PWM (Pulse Width Modulation) for the proper functioning of the inverter are output phase voltage and current values shown in Fig. 5 and Fig. 6.

The main DC bus 400 V are connected, two main storage units, which consist of five series-connected accumulator blocks and provides a voltage reference level of the DC voltage 400V. On the main DC bus 400V are attached two main storage units, which consist of five into a series of blocks and supplying the united battery voltage to the reference level of voltage DC 400V.

Fig. 3 The voltage of the main AC bus

Fig. 4 The voltage of the main DC bus

26 Tomas Mozdren, Stanislav Rusek, Radomır Gono, Veleslav Mach

Fig. 5 Phase voltage waveforms of the inverter output

Fig. 6 Phase current waveforms of the inverter output


4 Conclusion

The connection is described using a block scheme of TCO in which are displayed all the energy resources in the laboratories of the technological centre. In the second part of the posts is described software, which is modeled after the Technology Center. The following section displays the current dynamic model, which is created in the environ-ment of the EMTP ATP output values from a dynamic model for a sample placed in the chart. The dynamic model are need to be modeled DC/DC converters, photovoltaic cells and hydrogen technologies.

Acknowledgements

This research was partially supported by the SGS grant from VSB - Technical Uni-versity of Ostrava (No. SP2015/192) and by the project TUCENET (No. LO1404).

References

[1] D. Minarik, S. Rusek and R. Gono, “Implementation of accumulation technologies of electric energy to hybrid energy system”, – project, In Electric Power Engineering 2012, ISBN 978 - 80 - 214 - 4514 – 7

[2] T. Mozdren, S. Rusek, R. Gono, D. Minarik, P. Moldrik: The Input Analysis of Parameters and Expected Operation Modes of Energetic Technologies in Technological Centre ENET – project, In Electric Power Engineering 2015, ISBN 978-1-4673-6787-5

[3] M. Uher, “Optimization of network operation with renewable resources with the use of a dynamic model”, Dissertation thesis, Ostrava, 2014.

[4] EMTP-ATP, “About ATP”, 2015, [cit. 2015-20-2]. Avalaible at WWW: <http://www.emtp.org/>

[5] TCO, “Information and technical documention”

Analysis of Data Measured in Tosanovice BiogasStation

Ladislav Novosad, Zdenek Hradılek, Petr Moldrık


17. Listopadu 15/2172, 708 33 Ostrava-Poruba, Czech [email protected], [email protected], [email protected]

adfa, p. 1, 2011.

© Springer-Verlag Berlin Heidelberg 2011

Analysis of Data Measured in Tošanovice Biogas Station

Ladislav Novosád, Zdeněk Hradílek, Petr Moldřík

Department of Electrical Power Engineering

VŠB- Technical University of Ostrava

17. listopadu 15/2172, 708 33 Ostrava-Poruba, Czech Republic

[email protected],[email protected],

[email protected]

ABSTRACT. This article analyses the data measured on a biogas station

(BGS). Its aim is to describe the technology used in a given biogas station for

the generation of electric power. It also describes the actual measurement,

which took place at the TOZOS Tošanovice BGS, which located near the vil-

lage of Horní Tošanovice. The measurement took 51 days and to record the da-

ta, was used an automatic digital measuring device. Finally, this article discus-

ses the analysis of measured courses, particularly voltage, active power and re-

active power. The courses show that power supply was stable throughout the

measured period; however, some blackouts were observed. These blackouts

were caused by a shutdown of one or several cogeneration units.

Keywords: Biogas station, BGS, co-generation unit, CGU, measurement analy-

sis

1 INTRODUCTION

Measurements were carried out in TOZOS Tošanovice biogas station (BGS) or, to be

more precise, on co-generation units installed in this BPS. The actual biogas station is

located near the village of Horní Tošanovice. The station was put into operation on

1 December 2008. Biogas produced by the controlled fermentation of renewable re-

sources (raw materials – mainly corn silage and manure slurry) is used to produce

electricity. The electricity produced is supplied to the distribution network, thermal

energy is used to heat barns and the adjacent agricultural and administrative buildings

in the cooler months. Unused thermal energy is discharged to the surroundings

through forced ventilation using electric fans.

Electricity is produced in three Schnell co-generation units (CGU) equipped with

Scania combustion engines. The total installed electric power is 750kW (2x265 +

250kW). The power of individual CGUs is led through the respective generator


Analysis of Data Measured in Tosanovice Biogas Station 29

distributor to the main distributor NN RH1 located in the utility room next to the en-

gine room. Furthermore, the electricity is led to a transformer station 0.4/22kV 1,000

kVA and then to the distribution grid. (DG).

Fig. 1. CGU installed in Tošanovice BGS

2 DESCRIPTION OF MEASUREMENT

Period of measurement: since 3 September 2015 to 23 October 2015;

a total of 51 days

Place of measurement: distributor: RH1 – TOZOS Tošanovice BGS

The actual measurement was carried out using a digital automatic measuring device –

BK ELCOM ENA 330 Network Analyser. Using this device, phase voltages and cur-

rents were recorded at one-minute intervals. Voltage was measured directly on the

RH1 distributor busbars and currents were measured via clamp current transducers.

The measurement of electrical quantities included the assessment of power quality

according to ČSN EN 50160. Voltage, total harmonic distortion, flicker, voltage

decreases and unbalance were primarily evaluated. BK ENA 330 Analyser meets the

requirements for measuring instruments and measurement procedure given by the

ČSN EN 50160, ČSN EN 61000-4-7, ČSN EN 61000-4-15 and ČSN EN 61000-4-30

standards.

30 Ladislav Novosad, Zdenek Hradılek, Petr Moldrık

3 MEASURED COURSES

The measured characteristics of individual quantities are stated below.

3.1 Voltage Courses

Fig. 2. Voltage course at various phases during the measured period

Fig. 3. Histogram of voltage at various phases during the measured period

The above diagrams show voltage courses at various phases during the entire measu-

red period. In terms of the voltage, we can say that the voltage showed large fluctuati-

ons throughout the measurement. In terms of supply voltage variation, none of the

phases exceeds the values specified in a relevant standard (± 10% Un) which corre-

sponds to the values from the range 207V–253V. The diagrams show that there is no

significant difference between voltages in the individual phases, i.e. only a small

asymmetry between the phases can be seen.


Fig. 4. Harmonic higher order voltage spectrum with indicated permissible limits according to

ČSN EN 50160

Figure 4 shows the harmonic higher order voltage spectrum with indicated permissi-

ble limits according to ČSN EN 50160. None of the limits was exceeded; the closest

to them was the fifth and twenty-third exceeding a half of the allowed limits.

3.2 Electric Current Courses

Fig. 5. Current course at various phases during the measured period


Fig. 6. Histogram of currents at various phases during the measured period

The above diagrams of Figure 4 and 5 show the courses of currents at different phases

during the entire period. In terms of their courses, we can say that the current ranged

from approximately 950 to 1,000 kA during most of the measurement period. The

diagrams also show that there is a small current asymmetry between the phases, while

the smallest average values are shown by the first phase and contrarily, the largest

contrary by the third phase. Short decreases in current can be also observed in the

diagrams, associated with decreases in output and caused probably by switching off

some of the CGUs.

3.3 Courses of Active Power

Fig. 7. Course of the total active power during the measured period


Fig. 8. Course of the total active power during the measured period

Figure 7 shows the course of the total active power supplied to the grid by all CGUs

during the measured period. The total course of power is created by the addition of

individual active powers from all phases. The diagram shows that one overall power

outage as well as a larger active power supply decrease occurred during the measured

period. The decreases were characterized by short duration and the fact that power

supplied during these was always stabilized around the value of ca. 200kW. Such a

course would correspond to the switching off of two of the three installed CGUs.

The causes of individual outages has not yet been identified (an operation book has

not yet been provided).

Figure 8 shows the histogram of active power in the individual phases. Again, the

histogram shows that there is a small asymmetry between the phases. Two sets of

values of the supplied power can be seen as well. The first group includes powers

supplied during normal operation (ca. 250kW) and the second group includes powers

of the values of ca. 140–150kW. Their size confirms the occurrence of regular decrea-

ses in the overall power of about 550kW during the outage.


Fig. 9. Course of the total active power and power in different phases during a short decrease

Fig. 10. Course of voltage in various phases during a short decrease

Figures 8 and 9 show the courses of quantities during the aforementioned frequent

decreases in power. Given that the decreases had always the same or similar course,

one decrease was randomly chosen for demonstration purpose. The course of power

clearly shows that there was a sudden decrease in power supplied to the grid by ca.

500kW on a given day at 8:05am. This decrease was followed by a period of 5 min

when a reduced power was maintained and the supply of active power to the grid was

gradually fully restored. The entire process lasted from the decrease to the full resto-

ration approximately 15 minutes.

Regarding the course of voltage, it is clear that a short-term increase in voltage occur-

red during the outage. However, these peaks are not different from the voltage course

in the grid during the monitored period in terms of their course. Because of the

method of measurement (1-min intervals), it cannot be accurately determined whether

the cause of the CGU switching off was e.g. an overvoltage in DS, which the CGU

protections responding to.


3.4 Courses of Reactive Power

Fig. 11. Course of reactive power in various phases during the measured period

Fig. 12. Histograms of reactive power in various phases during the measured period

The above diagrams and histograms clearly show the asymmetry between the phases.

It is evident that the reactive power of phase 1 was around the values of approximate-

ly 10 kVAr. Phase 2 reactive power ranged between approximately 5 kVAr values

and phase 3 reactive power around the value of 15 kVAr. Moreover, it can be obser-

ved that phase 2 has a very often ringing into a capacitive character. This phenome-

non is also seen in phase 1, although in a much lesser extent.


CONCLUSION

This article is primarily focused on the overall analysis of the courses of active power,

reactive power and voltage supply. The courses show that the power supply was even,

although some irregular outages of CGUs occurred. The unbalance of individual

quantities between the phases was also found, the most apparent on the course of

reactive powers. This phenomenon cause may be either the actual construction of the

generators, or the wiring of internal consumption or technology in the biogas station.

Variations in the voltage occurred throughout the measurement, but they do not exce-

ed the standard allowable limits.

Currently, the measurement is continued to clarify the causes of short CGU blackouts

and to compare them with an operation book.

This paper shall be also used as a resource presenting information about stability of

performance of co-generation units for further research purposes. The latter will be

dealing with cooperation among other renewable power resources in order to provide

options for stabilization of voltage fed into 22kV networks.

Work was partially supported by SGS grant VŠB-TU Ostrava No. SP2015/192.

References

1. Data measured in TOZOS Tošanovice BGS

2. TOZOS Tošanovice BGS operation book and operation rules

3. L. Novosad, Z. Hradilek, “Power analysis of co-generation units at biogas station”, Pro-

ceedings of the 16th International Scientific Conference on Electric Power Engineering

(EPE) 2015 Dlouhé Stráně, 2015, ISBN 978-1-4673-6788-2

4. L. Novosád, “Využití biomasy jako akumulační zdroj energie pro tepelné a elektrické

sítě.” 2014. Diplomová práce. Vysoká škola báňská - Technická univerzita Ostrava. Fa-

kulta elektrotechniky a inf.. Vedoucí práce prof. Ing. Zdeněk Hradílek, DrSc.

Regulation of Cogeneration Unit in Industry

Michal Spacek, Zdenek Hradılek



Regulation of Cogeneration Unit in Industry

Michal Špaček1, Zdeněk Hradílek2, 1VŠB – TU Ostrava, Katedra elektroenergetiky, 17. listopadu 15, 708 33 Ostrava,

http://fei1.vsb.cz/kat410/ tel: +420 597 329 321, email: [email protected],

2VŠB – TU Ostrava, Katedra elektroenergetiky, 17. listopadu 15, 708 33 Ostrava, http://fei1.vsb.cz/kat410/

tel: +420 597 321 235, email: [email protected],

Abstract. Recently, we are trying to deal with energy as efficiently as possible. When it comes with nature-friendly technologies and energy saving. One possi-ble solution is replacement of cogeneration units (KGJ) instead of gas boilers. When natural gas or other methane-rich gases, such as biogas are burned in KGJ. The gas is burned in a combustion engine which is connected to the shaft by an electrical generator that produces electricity. Engine cooling and flue gas-es thermal energy is obtained. Another possible application of KGJ is masking the peak demand in the electricity supply when we try to reduce consumption peaks during the day.

Keywords: cogeneration unit, electric output, voltage, Viessmann Vitobloc 200,

1 Introduction

One of these energy savings is applied in the company NC Line in Suchdol nad Od-ra. When it will be gas boiler and gas burners replaced by cogeneration unit with an accumulation tank. Thermal and electrical energy is produced using cogeneration units burning natural gas. To the KGJ is connected pipe for hot water outlet into the tech-nology and accumulation tank (AKU). Part of the heat (1.3) is thus supplied directly into the technology of heat and a part of (2/3) is supplied to the accumulation tank. Scheme of KGJ technology and accumulation tank is displayed in Figure 1. Warm water of main heating circuit of cogeneration unit operates with a thermal gradient of 85/65 ° C. For measuring the amount of heat energy is connected calorimeter. For accumulation of thermal energy was used accumulation tank with a capacity of 60 cubic meters which is thermally insulated.

To achieve greater engine power of cogeneration unit is installed turbo compressor. Compressed fuel (natural gas) is cooled using by closed water circuit with dry cooler. In cooling circuit is necessary to maintain the temperature gradient of 38/35 ° C.


38 Michal Spacek, Zdenek Hradılek

While designing was calculated that total power will be consumed by the company. And there will be no overflows into the network. Over the last six months it has started to variable electricity consumption in the company. When occurred overflows of elec-tricity from the cogeneration unit to the network. Firm has required a reserved perfor-mance of 0 kW. Against these overflows the company wants to implement regulation of cogeneration unit.

Fig. 1. Visualization of technology of cogeneration units with accumulation tank and other

auxiliary devices

2 Cogeneration unit

Cogeneration unit was used Viessmann Vitobloc 200 EM-199 / 263.

CHP parameters: max. electrical output: max. heat output: engine output:

199 kWe 263 kWt 210 kW

fuel consuption: 53 Nm3/h min. overall efficiency: 89,6 % temp. gradient of heating circuit: 65/85°C temp. gradient of compressor circuit: 10/50°C temp. gradient of fuel mixture 35/38°C Dimensions LxWxH: 3,58x1,60x2,00 m

Regulation of Cogeneration Unit in Industry 39

Fig. 2. Cogeneration unit Viessmann Vitobloc 200 EM-199/263

Module of block of thermal power station (BTE module) is a unit ready for connec-

tion to a gas source, to electrical network and to heating circuit. It includes air-cooled synchronous generator for production of three-phase current 400 V, 50 Hz. The heat-ing circuit is operated with temperature gradient 85/65 ° C. BTE module may be oper-ated electrically or thermally, in the operating range of 50 - 100% of rated electric power (corresponding to 60 - 100% of the thermal output). [3]

In the company cogeneration unit is operated in the heat mode at 100% of rated power. The unit is about electrical output of 199 kW and therefore falls into the cate-gory of production facilities with an output of over 100 kW connected to distribution network (DN). In case of danger and reliable operation of the electricity system is necessary for supervisory control to temporarily restrict or shut down active power supply of the electricity produced. The source is able adequately (quickly and accu-rately) to respond to a command from dispatching of transmission distribution net-work (TDN) to limit stepped active power in mode 0, 50, 75 and 100% of the installed capacity. This is with the help of regulating switchboard (RTU).

3 Measurement on cogeneration unit

In block thermal power station is installed equipment Vitobloc Gateway which pro-vides communication with a control computer (PLC) over RS 485. In control comput-er are recorded values. These basic parameters of electrical quantities are recorded at minute intervals.

The obtained data are dated July 2, 2014, when a command to switch of KGJ was at 4:00 am and then was shutdown at 12:00 pm .


In the course of electric power Fig. 3 can be observed in the first half of the opera-tion time considerably frequent changes. In the second half of the running time are already not recorded such power changes that occurs during each start KGJ and lasts so long when machine is not stabilized. Before the end of KGJ shutdown from 11:40 is possible to observe a significant instability of performance. That is caused by water temperature increasing in the main heating circuit. Return water increased because of charge of accumulation tank. When sudden change of inlet temperature must been set internal parameters of KGJ, resulting in the instability of the produced electric power.

Fig. 3. Course of electric power during operation from KGJ

Fig. 4 shows the current waveform from the generator at each stage. At which is possible to observe current unbalance and it especially in L2 phase.

Fig. 4. Courses of currents from KGJ during operation


From the measurement of voltage on KGJ may be observed voltage fluctuations that are caused by character of network load. To the network are connected metal lasers, which are characterized by instability collection.

Fig. 5. Course of electric voltage from KGJ during operation

4 Connecting of cogeneration units to the network

Wiring diagram of cogeneration unit is in Figure 6. Company is powered from dis-tribution network of 22kV, via two current transformers T1 and T2. Each of them has the performance of 500kVA. These transformers can operate parallel or separately. Currently they are working separately, and for their parallel cooperation is used longi-tudinal fuse switch F20. Transformers operate separately, mainly because of the la-sers, which have a very variable consumption from the network, which cause fluctua-tions in line voltage. In the company have two types of network - "dirty" and "clean" for this reason. Clean network is supplied from the transformer T1, to which is con-nected the lighting, offices and other electronics depend on network. On dirty network (transformer T2) are connected lasers, welding centers and the cogeneration unit.


Fig. 6. Monopolar diagram of a cogeneration unit connected into the network

5 Control and connection of cogeneration unit

Methods of regulation may be a large amount. One of them is continuous regulation that is used most often. Other method is jump regulation, in which we perform the regulation in individual steps (for example, 50%, 60%, 70%, 80%, 90% of rated pow-er 100). At cogeneration units is necessary to keep power regulation between 50% and 100% of rated power. Below 50% of rated power unit has very poor efficiency. The best method of operation of the unit is at a maximum (100%) power, in which has the largest ratio electricity to heat energy. When designing units from the viewpoint of regulating electric power is better to suggest more units. These would be triggered progressively according to the desired performance and should be regulated by the rated power. In our case, the unit will be operated at nominal power. [1] [2]

This cogeneration unit is controlled by the thermal regime. When is operated only at 100% of rated power. This type of unit can be operated from 50% to 100% of the rated electric power.

In the first phase on supply will be mounted current transformer about parameters of 800 / 5A, with Accuracy Class 0.5. This output current will be connected to the converter together with the potentials from individual phases. The converter will con-


vert the calculated electric performance on 4-20mA. When 4 mA will be correspond to -200 kW and 20 mA will be 525kW. The output current loop will be connected to the control machine, which will give impetus about operation of cogeneration unit. A simplified diagram is shown in Figure 7. Further to the control system will be con-nected various information about the operation of cogeneration unit. Between which there will include: a power failure, gas leaks, underlying disorder of cogeneration unit etc. This information will be sent via GSM communication via SMS to prepro-grammed numbers.

Fig. 7. Simplified circuit diagram of regulation

7 Conclusion

The aim of this article is verifying the regulation of cogeneration unit not only from a theoretical point of view, but also from a practical point of view. One of the most important factor devise an algorithm that will be turn on and subsequently turn off operation of the unit. Another factor is also the operation time at which is not recom-mended to turn off the unit under running time of two hours. An important aim of the solution is the analysis of the control options relative to variable loads in industrial


network. An important factor is the need to devise an algorithm to regulate the opera-tion of the cogeneration source.

Acknowledgment

The presented research was supervised by prof. Ing. Zdeněk Hradílek, DrSc. from Department of Electrical power engineering, VSB - TU Ostrava, Czech republic.

References

1. Dvorský, E.: “The combined production of electricity and thermal energy,” BEN technical literature, Praha 2005, ISBN 80-7300-118-7

2. Krbek, J., Ochrana, L., Polesný, B.: “Industrial Power,” PC DIR Real Brno, 1996 3. Dokumentation BHKW Vitobloc 200, 2012-02 / V.1, Viessmann Group, ESS Energie

Systems and Service GmbH 4. Krbek, J., Ochrana, L. , Polesný, B.: “Heating and cogeneration,” PC-DIR Real, s.r.o.,

Brno, 1999, ISBN 80-214-1347-6. 5. Z. Hradílek, “Elektroenergetika průmyslových a distribučních zařízení,” 1.vyd. VŠB – TU

Ostrava. 2008. ISBN 978-80-7225-291-6.,p.315, 2008. 6. Pravidla provozování distribučních soustav Příloha 4 - Pravidla pro paralelní provoz

zdrojů se sítí provozovatele distribuční soustavy [online]. [cit. 2015-04-29]. Available from: http://www.eon-distribuce.cz/file/cs/electricity/regulations/PPDS_2009_4.pdf

Dynamic Model of Asynchronous Machine

Martin Kral, Radomır Gono



Dynamic Model of Asynchronous Machine

Martin Král1, Radomír Goňo1

1Department of Electrical Power Engineering, FEECS, VŠB - Technical University of Ostrava, 17. listopadu 15, 708 33 Ostrava – Poruba


Abstract: This paper deals the mathematical description of an asynchronous machine. Differential equations are then used to create a dynamic model of asynchronous machine. For the possibility of evaluation the quality of the created model are created at the conclusion the basic characteristics of representing the asynchronous machine.

Key words: Three phase Induction Machine, Dynamic Model of Asynchronous Machine.

1 Introduction

Asynchronous motors, especially with a squirrel-cage motor, are for many years the most famous electric motors on our planet. It happened because of their structural simplicity, low cost, high reliability and efficiency.

Mathematical model of the asynchronous motor is difficult and some researchers have developed the model by transferring into d and q axis. However, more accurate model can be derived by using α, β, γ axis. Therefore, this paper is aimed to obtain dynamic model of the motor which has been transferred into three axes.

2 Mathematical model of asynchronous machine

Due to the design we consider asynchronous machine (ASM) for the nonlinear system with a number of parameters. We tried to find a system of differential equations that would sufficiently and accurately describe the properties of the machine. When we design a mathematical model, we use a series of assumptions that simplify the Assembly model. In particular, the following assumptions:

• Stator and rotor winding is three-phase, coils of phases are spread out

symmetrically along the air gaps. • Three-phase stator and rotor winding is connected into the star. • Magnetic induction along the air gap is ideal sine. • Magnetic circuit has a linear characteristic. • Resistances are constant. • Losses in magnetic circuits are zero [3].


46 Martin Kral, Radomır Gono

Fig. 1. Three-phase symmetric induction motor [1]

The flux linkages are found as functions of the corresponding currents in the stator and rotor windings using the self- and mutual-inductances. According to the Fig. 1 we can write:

∙ ∙ ∙ ∙ ∙ ∙ ∙ ∙ ∙ ∙ ∙ ∙ ∙ ∙ ∙ ∙ ∙ ∙ ∙ ∙ ∙ ∙ ∙ ∙ ∙ ∙ ∙ ∙ ∙ ∙ ∙ ∙ ∙ ∙ ∙ ∙

(1)

Where , , , , , are stator and rotor sefl-inductances,

these inductances are and other inductances are mutual inductances between stator-stator, stator-rotor and rotor-rotor windings [1].

Dynamic Model of Asynchronous Machine 47

Mutual inductance between the stator windings (or rotor) are negative, as the axis of the winding makes an angle α = 2π/3 (cos2π/3 =-1/2) [2].

Fig. 2. The mutual inductances stator windings

The mutual inductances between stator windings are:

∙ cos , (2)

– turns of stator The angel between the rotor winding is the same as between the stator winding,

therefore:

∙ cos , ∙ . (3)

The matrix of self- and mutual inductances and :

,

′ ∙′

′′

, ′ ∙ .

(4)

The mutual inductances between the stator and rotor windings are periodic

functions of the electrical angular displacement , and the period is 2 . Assume that the mutual inductances are sinusoidal functions such that

∙ , ∙ 23 ,

∙ , ∙ 2

3 , ∙ , ∙ , ∙ 2

3 , ∙ 23 ,

∙ , ∙ , ∙ , [1].

(5)


The matrix of stator-rotor mutual inductances: ′ = ∙

=+ 23 − 23− 23 + 23+ 23 − 23

(6)

The matrix of rotor-stator (transposition) mutual inductances: ′ = ∙

=− 23 + 23+ 23 − 23− 23 + 23

(7)

The matrix of flux linkages:

′ = ′′ ′ ∙ ′ (8)

The equations of stator and rotor voltage:

Stator voltage - = ∙ + Rotor voltage - ′ = ′ ∙ ′ +

(9)

From the equation listed above we are able to calculate stator and rotor current. The equation describes the electromagnetic torque: = − 2 ∙ ∙

∙+ 23 − 23− 23 + 23+ 23 − 23

′′′ (10)

The equation of torsional-mechanical: = ∙ ∙ − ∙ − ∙ ∙ , = [1]. (11)


3 Simulation of three phase asynchronous machine in MATLAB

Simulink

To create a dynamic model, we took advantage of the differential equation, which is described above. As has been written, these differential equations describe the three phase asynchronous machine, if the rotor voltages are ( ′ ) = 0, this is a machine with a squirrel-cage.

Fig. 3. Dynamic model of three phase asynchronous machine made in MATLAB Simulink


When using the dynamic model of asynchronous machine, I modelled in two modes, the motor and the generator. Simulation mode can by changed by using value of TL (torque of load), if is TL > 0 asynchronous machine is in the motor mode, if is TL

< 0 asynchronous machine is in the generator mode. The parameters of the machines in the simulation: Stator voltages: 230 , rotor voltages ′ 0 (squirrel-cage), 2

(poles), 0.3 (resistance of stator circuit), 0.2 (resistance of rotor circuit), 0.035 (mutual inductance), 0.003 (stator inductance), ′ 0.003 (rotor inductance), 0.003 (magnetic induction),

0.02 ∙ (moment of inertia), 50 (frequency), 40; 40 ∙ (torque of load).

The simulation results:

Fig. 4. Electromagnetic torque

Fig. 5 Electromagnetic torque

Fig. 6. Rotor current

Fig. 7. Stator current


Fig. 8. Rotor voltage

Fig. 9. Stator voltage

Fig. 10. Slip

Fig. 11. Torque-speed

4 Conclusion

At the beginning, the basic differential equations describe the asynchronous machine. From these equations, there are compiled differential equations, which are useful for the description of dynamic model of asynchronous machine.

In the practical part, there was created a dynamic model of asynchronous machine in MATLAB Simulink, which uses the previously described differential equations, which simulate the operation of the asynchronous machine. For simulating there were chosen the parameters of the real machine with a squirrel-cage, these parameters are also listed in this work.

Of the simulations, it is obvious that we did in the dynamic model two changes of the torque load that is imposed on the shaft of the asynchronous machine. At time t = 0.5 s we put on the shaft of the machine load torque TL = 40Nm (motor region). It is clear from the charts that act as a real model, the electromagnetic torque and slip has increased, the angular velocity is decreased. At time t = 1s we put on the shaft of the machine load torque TL = - 40Nm (generator region). From the graph it is clear that


the model behaves as the real, the electromagnetic torque and slip is decreased, the angular velocity is increased. Acknowledgement: This research was partially supported by the SGS grant from VSB - Technical University of Ostrava (No. SP2015/192) and by the project TUCENET (No. LO1404).

References

1. Victor Giurgiutiu, Sergey Edward Lyshevski Micromechatronics: Modeling, Analysis, and Design with MATLAB, Second Edition CRC Press, 2009. 920 s. ISBN 9781420065626.

2. Sergey Edward Lyshevski Electromechanical Systems and Devices CRC Press, 2008. ISBN 9781420069723.

3. Josef Běloušek Trakční pohony s asynchronním motorem Doctoral thesis (VUT Brno), 2013. 136 s.

Actual Results of the Reliability Computation in2015

Radomır Gono1, Stanislav Rusek1, Pavel Bednar2, Peter Chovanec2, MichalKratky2

1 Department of Electrical Power Engineering2 Department of Computer Science

FEECS, VSB – Technical University of Ostrava,17. Listopadu 15/2172, 708 33 Ostrava-Poruba, Czech Republic

radomir.gono, stanislav.rusek, pavel.bednar, peter.chovanec,

[email protected]

Actual Results of the Reliability Computation in 2015

Radomír Gono1, Stanislav Rusek1, Pavel Bednář2, Peter Chovanec2, Michal Kratky2

1Department of Electrical Power Engineering 2Department of Computer Science

VŠB - Technical University of Ostrava, 17. listopadu 15, 708 33 Ostrava – Poruba radomir.gono, stanislav.rusek, pavel.bednar, peter.chovanec,

[email protected]

Abstract. The paper deals with the computation of distribution network com-ponents reliability parameters. Actual value of the component reliability param-eters in distribution network is used for the reliability computation and also for reliability-centered maintenance system. Reliability indices are possible to re-trieve only from accurate databases of distribution companies. Such a database includes records of outages and interruptions in power networks. The main problem for an analysis of these databases is the heterogeneity feature: data-bases of various distributors differ from one another. It is impossible to retrieve reliability parameters from this data in a direct way. In this paper there is ap-plied a framework for the retrieving of parameters from various outage data-bases in the Czech and Slovak republics. There are also actual results.

Key Words: Component reliability, failure rate, mean time to repair, distribu-tion network, and outage database

1 Introduction

This work deals with the component reliability. It is necessary to observe outages and interruptions in the transmission and distribution of electrical energy for retrieving the component reliability [1]. Furthermore, electrical energy unsupplied to consumers is possible to compute. A statistical significance of an outage database depends on the number of records in the database. A larger database would describe the real condition of network equipment more accurately. Therefore, it is necessary to merge databases of various distributors and distribution areas. The main problem of the merging is the heterogeneity feature: databases of various distributors differ from one another, be-cause they have different database systems and also different approaches for evalua-tion of outages and interruptions in their networks.

In [2] there is introduced a framework that makes it possible to retrieve parameters from these various databases. This idea is developed and new results are shown here.


54 Radomır Gono et al.

2 History of Outage Monitoring

Component failure rates tend to vary with a component work life. A bathtub curve is commonly applied to represent the time-dependent failure rate changes of a compo-nent. Many parameters in the field of reliability vary for a specific component and the condition in which a component works. These random variables are represented by probability distribution functions [3], [4], [5], [6].

Failure rates of overhead distribution equipment are, in general, very system specif-ic due to their dependence on geography, weather, animals and other factors [7]. Typ-ical reliability values for pieces of distribution equipment have been introduced in [8], [9], [10], [11], [12], [13].

Outage monitoring in the former Czechoslovakia started in 1975 according to regu-lations 2/74 [14]. These regulations unified interruptions, outages and damaged equipment monitoring options for all distribution companies in Czechoslovakia.

Unfortunately, database building has ceased since 1990 because of political and so-cial changes. The expert group, CIRED Czech, has introduced a discussion on relia-bility issues. The first calls for integration of particular outage databases were already claimed at the first meeting of this group in 1997. In 1999, distributors opted for uni-fied monitoring of global reliability indices and the reliability of selected pieces of equipment [15]. Data for the reliability computation is centrally processed and ana-lyzed at the Technical University of Ostrava. This data has been handled and pro-cessed since the year 2000.

3 Reliability Analyses

A majority of reliability computations is performed in the following way. The reliabil-ity computation of the whole system is executed on the basis of components reliability that is included in the system. That is the reason why the reliability is computed in two phases. The first phase represents the retrieving of component reliability parameters and the second phase is the reliability computation itself. Other phases may include the evaluation of computed results and an improvement of the supply quality.

In virtue of experience, it is necessary to state that in most cases, the retrieving reli-ability parameter is far more complicated than the reliability computation itself.

3.2 Input Data for Computations

There are various methods for input data retrieval which are based on the type of an examined object, available data of an examined object, etc. Reliability is divided into two basic groups in compliance with the method of input data retrieval:

• Empirical reliability – input data for the reliability computation is retrieved from data on equipment, or similar equipment operating under similar conditions

• Predetermined reliability – the probability of outage-free operation is expressed on the basis of knowledge about component status.

Actual Results of the Reliability Computation in 2015 55

Obviously, incorrect input data leads to poor results, even when a correct computa-tion method is applied. Moreover, in many cases of reliability computations in electri-cal power engineering, we face the problem of insufficient data size for a component, e.g. an insufficient number of historical records.

3.3 Reliability Computation

In the case of empirical reliability, we need data on operations and outages of compo-nents occurring in the reliability diagram, or data on components of the same type operating under similar operating conditions. The more extensive the database, the more reliable the results are. In the case of power system components, data must be available for outages of breakers, disconnectors, transformers, lines, etc. for a set type and voltage level. Moreover, there is another type of data necessary for the reliability computation. We need to have knowledge of the power network itself. For example, we must know the number of pieces of equipment for a set type, the total length of a line type, voltage level and so on.

Consequently, retrieval of the failure rate for a power system is the basis of the em-pirical reliability computation. This method is mostly employed in retrieving reliabil-ity parameters for the reliability computation because the application of predetermined methods requires different approaches to each power system component.

On the other hand, empirical methods require accurate records of outages. Conse-quently, for statistically significant results of reliability computations, data on outages dating back to many years in the past is required. It is possible to compute basic relia-bility parameters of particular components from this database - annual failure rate and time to repair.

The number of outages per period is retrieved from the database. The period is usually defined depending on requirements concerning the reliability computation. An additional value necessary for the failure rate computation is the number of compo-nents for a set type and area. This value is possible to retrieve from the equipment owner (usually system operator). As the numbers of components change in the real power network during a period, we update it annually. Other important information is possible to retrieve in more detailed databases, e.g. the most frequent cause of outag-es, areas of the greatest amounts of undelivered energy, etc.

Regulations 2/74 include reliability parameters for basic equipment. These parame-ters were set in 1980 and are very outdated. It is necessary to update these parameters using an analysis of outage databases.

3.4 Heterogeneous Outage Data

In the case of electrical power networks, each distributor produces incompatible out-age data. Although a data model of this data may be the same (e.g. relational data model), such data is not necessarily compatible. For example, sets of relations for two distributors belong to different relation schemes. Moreover, each scheme includes different attributes expressing the same feature of an entity type.


A common way of addressing the problem is to develop a common relation scheme and different data transform into the relation. It enables querying and analysis. We have selected 31 attributes [16]. For the component reliability only few attributes are necessary:

• Distribution Company - anonymous code of distributor • Outage Identification - unique code of event • Outage Type - accidental, planned or forced • Equipment Voltage - 0.4 kV, 22 kV... • Outage Cause - foreign influences, causes before starting operation... • Equipment Type - overhead line, underground line... • Failed Equipment - specific device - conductor, switch, pole, fuse... • Failed Equipment Type - further specification - wooden pole, steely pole... • Amount of Failed Equipments • Producer - Siemens, ABB... • Production Year - age of the component • Beginning of outage • End of outage - time of restoration of supply to all consumers • End of equipment failure - time of repair of the device • Failure Type - with or without equipment damage Some other attributes are included for continuity of supply analyses and some for

future expansion purposes.

6. Results

The basic reliability data of particular elements may be computed from the data-base of outages and interruptions stored at the VSB – Technical University of Ostrava. The results include the rates and mean durations of equipment outages.

6.1 Database Range

The actual data collection includes outage data from distributors from the Czech Re-public and one from the Slovak Republic. We have retrieved data from eight distribu-tion areas (Table 1).

Distributors have delivered their data in xls files twice a year. Today database con-tains more than 400 thousand records (from 2000 to 2015) on voltage levels 110 kV, MV and partially LV.


Table 1. Database range (months)

Region1 Region2 Region3 Region4 Region5 Region6 Region7 Region82000 - - - 1 - 12 - - 1 - 12 1 - 122001 1 - 12 - 1 - 12 1 - 12 - 1 - 12 1 - 12 1 - 122002 1 - 12 - 1 - 12 1 - 12 - 1 - 12 1 - 12 1 - 122003 1 - 12 - 1 - 12 1 - 12 - 1 - 12 1 - 12 1 - 122004 1 - 12 - 1 - 12 1 - 12 1 - 12 1 - 12 1 - 12 1 - 122005 - - 1 - 12 1 - 12 1 - 12 1 - 12 1 - 12 1 - 122006 - 1 - 12 1 - 12 1 - 12 1 - 12 1 - 12 1 - 12 1 - 122007 - 1 - 12 1 - 12 1 - 12 1 - 12 1 - 12 1 - 12 1 - 122008 - 1 - 12 1 - 12 1 - 12 1 - 12 1 - 12 1 - 12 1 - 122009 - 1 - 12 1 - 12 1 - 12 1 - 12 1 - 12 1 - 12 1 - 122010 - 1 - 12 1 - 12 1 - 12 1 - 12 1 - 12 1 - 12 1 - 122011 - 1 - 12 1 - 12 1 - 12 1 - 12 1 - 12 1 - 12 1 - 122012 - 1 - 12 1 - 12 1 - 12 1 - 12 1 - 12 1 - 12 1 - 122013 - 1 - 12 1 - 12 1 - 12 1 - 12 1 - 12 1 - 12 1 - 122014 - 1 - 12 1 - 12 1 - 12 1 - 12 1 - 12 1 - 12 1 - 122015 - 1 - 6 1 - 6 1 - 6 1 - 6 1 - 6 1 - 6 1 - 6

6.2 Framework Results

The graphic representation of all distribution regions reliability indices from the above-mentioned data for the 22 kV cable is given in Fig. 1. From the significant differences in particular years it is possible to observe the contribution of our anal-yses. The divergence of reliability indices is eliminated during long-term observation.

Fig. 1. The value tendency of reliability indices of the 22 kV cable


These parameters could update reliability indices from old Regulations 2/74 [14]. There is a comparison of both databases, 1975 - 1990 and 2000 - 2015, in Table 2.

Table 2. Comparison of results

In Table 2, we can observe that the current reliability indices are rather more supe-

rior. One of the results of analyses is structuring failures according to their causes (Fig.

2). The most common cause of outages is “Operation and maintenance causes”.


Causes before starting operation Operation and maintenance causes

Foreign influences Forced outage

Cause not explained Other causes

Fig. 2. Structuring outages according to their causes

It is possible to provide also comparison of distribution regions - REAS (Fig. 3). The Energy Regulatory Office could find these results useful for justifying of renewal costs among distribution system operators.

Fig. 3. Comparison of distribution regions


0

20000

40000

60000

80000

100000

120000

140000

160000

180000

200000

0 - 1 min 2 - 3 min 4 - 10 min 11 - 60 min 61 min - 1 month

Failure duration

Nu

mb

er o

f fa

ilu

res

Fig. 4. Number of outages distributed according to their duration

Fig. 4 shows distribution of outages according to their duration. The most of outag-es are longer than 1 hour and shorter than 1 month.

We can also obtain other information important for operators, such as the faulty equipment series from a specific producer, areas of the greatest amounts of unsupplied energy, etc.

7. Conclusion

A statistical significance of an outage database depends on the number of records in the database. A larger database would describe the real condition of the network equipment more accurately. Therefore, it is necessary to merge databases of various distributors. The main problem of the fusion is the heterogeneity feature: databases of various distributors differ from one another.

The framework result may include the rates and mean durations of equipment out-ages. We can also obtain other significant information for operators. The result proves the framework is appropriate for analyzing such data. We compared the new results to the original results in this paper.

Acknowledgements: This research was partially supported by the SGS grant from

VSB - Technical University of Ostrava (No. SP2015/192) and by the project TUCENET (No. LO1404).


References

1. R.E. Barlow, & F. Proschan, Statistical theory of reliability and life testing: probability models (New York, USA: Holt, Rinehart and Winston, 1975)

2. M. Kratky, R. Gono, S. Rusek, & J. Dvorsky, A framework for an analysis of failures data in electrical power networks. Proc. PEA Conf. on Power, Energy, and Applications, Gaborone, BW, 2006, 45-46

3. W.H. Beyer, Crc standard mathematical tables (Boca Raton, USA: CRC Press, 1984) 4. R. Ramakumar, Engineering reliability: fundamentals and applications (Upper Saddle River,

USA: Prentice-Hall, 1996) 5. T. Gonen, Electric power distribution system engineering (New York, USA: Mcgraw-Hill

College, 1985) 6. S. Asgarpoor & M.J. Mathine, Reliability evaluation of distribution systems with non-

exponential down times, IEEE Transactions on Power Systems, 12(2), 1997, 579-584 7. R.E. Brown & J.R. Ochoa, Distribution system reliability: default data and model validation,

IEEE Transaction on Power Systems, 13(2), 1998, 704-709 8. W.F. Horton, S. Goldberg, & R.A. Hartwell, A cost/benefit analysis in feeder reliability

studies, IEEE Transaction on Power Delivery, 4(1), 1989, 446-452 9. R. Brown, S. Gupta, S. Venkata, R. Christie, & R. Fletcher, Distribution system reliability

assessment using hierarchical Markov modeling, IEEE Transaction on Power Delivery, 11(4), 1996, 929-1934

10. R. Brown, S. Gupta, S. Venkata, R. Christie, & R. Fletcher, Distribution system reliability assessment: momentary interruptions and storms, IEEE Transaction on Power Delivery, 12(4), 1997, 1569-1575

11. H.L. Willis, Power Distribution Planning Reference Book (Boca Raton, USA: CRC Press, 1997)

12. P. Save, Substation reliability - practical application and system approach, IEEE Transac-tion on Power Systems, 10(1), 1995, 380-386

13. D. Karlsson, H.E. Olovsson, L. Walliin, & C.E. Slver, Reliability and life cycle cost esti-mates of 400 kV substation layouts, IEEE Transaction on Power Delivery, 12(4), 1997, 1486-1492

14. J. Piskac, & J. Marko, Regulations for electric power system no. 2 – failure statistics at electricity distribution (Prague, CZ: CEZ, 1974)

15. Distribution companies of the Czech Republic, Distribution network grid code, appendix no. 2 - methodology of reliability determination of electric power supply and distribution network equipments (Prague, CZ: ERU, 2005)

16. R. Goňo, M. Krátký, & S. Rusek, Analysis of Distribution Network Failure Databases. Przegląd elektrotechniczny (Electrical Review), 86(8), 2010, 168-171

17. R. Cimbala, J. Kurimský, I. Kolcunová: Determination of thermal ageing influence on rotating machine insulation system using dielectric spectroscopy, Przegląd Elektrotech-niczny, Vol. 87, no. 8 (2011), p. 176-179

Utilization of Signature Methods for RangeQuery Processing over Outage Database*

Peter Chovanec, Michal Kratky

Department of Computer Science,FEECS, VSB – Technical University of Ostrava,

17. Listopadu 15/2172, 708 33 Ostrava-Poruba, Czech Republicpeter.chovanec, [email protected]

Utilization of Signature Methods for RangeQuery Processing over Outage Database?

Peter Chovanec, Michal Kratky

Department of Computer ScienceVSB – Technical University of Ostrava, Czech Republic

peter.chovanec, [email protected]

Abstract. The reliability computation is applied for the maintenance ofequipments in power networks. The reliability computation is calculatedfrom a database of outages in electrical power networks. In the case ofthe outage database described in this paper, data are stored in a relationwith 35 attributes. The reliability computation requires to process up tohundreds of range queries over this data. Therefore, the efficient process-ing of these queries is necessary. Since multidimensional range queriesare used, a multidimensional data structure, the R-tree, has been ap-plied in our previous work. The efficiency of the R-tree decreases withthe increasing dimension of an indexed space. We depict an utilizationof the signature methods for more efficient range query processing. Theapproach, called the Signature R-tree, is utilized and compared with theR-tree.

Key words: power networks, reliability computation, outage data, R-tree, Signature R-tree

1 Introduction

Institutional changes taking place all over the world drastically effect the ap-proach to power supply quality. It is developing towards a purely commercialmatter between suppliers and their customers. The supply that does not complywith agreed qualitative parameters will lead to trade disputes and financial set-tlements. Undelivered energy, including its valuation, has arrived on the scene.The two following aspects of supply quality may be considered:

1. Supply reliability – relating the availability of electricity in the given location.2. Voltage quality – relating to the purity of characteristics of the voltage wave-

form, including the absolute level of voltage and frequency.

This document deals with the first aspect in more detail. Worldwide centersof reliability computation1 provide databases of information about the availabil-ity of electronic and non-electronic components and distribution functions for

? This work was supported by the Ministry of Education, Youth and Sports of theCzech Republic (SGS, No. SP2015/192).

1 For example Alion System Reliability Center, http://src.alionscience.com/


Utilization of Signature Methods for Range Query Processing ... 63

various failure types. They include the result failure rate and we can retrieveinformation about the producer, operation conditions, etc. These databases areapplicable to the availability prediction of complicated systems. However, thesedatabases do not include data about power equipments.

The IEEE standards define a host of reliability indices applied to distributionreliability. IEEE P1366 [17] explains the reliability indices applied to measure-ment of distribution system reliability, and a way of calculating reliability indices.Although authors introduce a discussion about some factors influencing these in-dices, reliability parameters for power system equipment are not depicted. TheCanadian Electrical Association2 introduces a collection of reliability parame-ters for power system equipment. This is useful for North America; however, it isalmost impossible to compare conditions and equipment in North America andCentral Europe.

In many cases, it is necessary to compute electrical energy not supplied toconsumers; probability computation of not supplied energy is only possible onthe basis of the reliability computation results. Consequently, we need to observefailures and outages in the transmission and distribution of electrical energy3 forretrieving the component reliability [1].

In [13], we introduced a framework for retrieving reliability parameters in dis-tribution networks. Consequently, we improved the approach by a new embeddedDBMS, called RadegastDB, presented in [11]. In [6], we depicted preliminary re-sults of multiple range queries over the outage database. Since we store outagedata in a relation with 35 attributes and we use multidimensional range queriesto query the relation, a multidimensional data structure, the R-tree, is utilized asa storage of the data. In this paper, we show results of the Signature R-tree [14,7] and a comparison with the R-tree over the outage data.

This paper is organized as follows. In Section 2, we briefly describe Rade-gastDB and the R-tree. In Section 2.3, we outline principles of the SignatureR-tree. In Section 3, we introduce the outage database and describe the reli-ability computation. In Section 4, we put forward preliminary results of theapproach. In the last section, the paper content is resumed and the possibilityof a future work is outlined.

2 Database System for Handling Outage Data

2.1 Introduction

In [13], we have introduced a framework for storage and querying outage data [9,8]. Databases of various distributors are transformed into the common relationscheme with 35 attributes. Since then, several works have been presented [12,2, 4]. In [11], we introduced a new data storage based on multidimensional data

2 http://www.canelect.ca/3 We have used the term ’outage database’ instead of the preferred phrase ’database

of failures and outages in the transmission and distribution of electrical energy’ inthis paper.

64 Peter Chovanec, Michal Kratky

structures [15]. A variant of the R-tree [10], called the R∗-tree [3], has beenapplied for the implementation. In [6], we depicted preliminary results of mul-tiple range queries over the outage database. In [5], we introduced completealgorithms, cost model, and results of multiple range queries.

2.2 R-tree and its Variants

Since 1984 when Guttman proposed his method, R-trees [10] have become themost cited and most used as reference data structure in this area. The R-treeis a height-balanced tree based on the B+-tree with at least 50% utilizationguaranteed. This data structure supports point and range queries and someforms of spatial joins as well. A general structure of the R-tree is shown inFigure 1.

R1 R2

R3 R4 R5 R6

p2 p4 p10 p6 p9 p1 p7 p3p8 p5 p11

R1

R2

R3

R4

R5

R6

p2p4

p8

p10

p6

p9

p1

p5

p7

p3

p11

Fig. 1. A planar representation and general structure of the R-tree

It is a hierarchical data structure representing spatial data by the set ofnested n-dimensional minimum bounding rectangles (MBR). If N is an innernode, it contains pairs (Ri, Pi), where Pi is a pointer to a child of the node N .If R is the inner node MBR, then the boxes Ri corresponding to the childrenNi of N are contained in R. Boxes at the same tree level may overlap. If N is aleaf node, it contains pairs (Ri, Oi), so called index records, where Ri contains aspatial object Oi.

Many variants of the R-tree have been proposed during the last decades.Although original algorithms of the R-tree tried to minimize the area covered byMBRs, R∗-tree [3] takes other objectives into account, e.g. the overlap amongMBRs. R+-tree [16] was introduced as a variant that avoids overlapping MBRsin intermediate nodes of the tree and an object can be stored in more than oneleaf node.

2.3 Signature R-tree

The range query algorithm traverses the tree from the root node and it followsrelevant items in each node. The item is relevant if its MBR is intersected bythe query rectangle. The algorithm recursively traverses all subtrees and it is


finished after all relevant subtrees are processed. The R-tree has some featureswhich can significantly slow down the range query processing. MBR of nodesoften overlap and cover the dead space (the space without any items). It canlead to the retrieval of nodes without any relevant items.

In [7, 14], the Signature R-tree was introduced for efficient processing of rangequeries. The Signature R-tree is a variant of the R-tree including multidimen-sional signatures for a more efficient filtration of irrelevant tree nodes. A generalstructure of the Signature R-tree is presented in Figure 2. Each leaf node ofthe tree is described by a multidimensional signature, that is built from signa-tures of particular node items for each dimension. Each inner node item includesthe MBR as well as the multidimensional signature. Consequently, such a treecontains two hierarchies, the hierarchy of MBRs and hierarchy of signatures.

T

Rl:R

h S

indexed tuples

index – hierarchy of MBRs and n-dimensional

signatures

Rl:R

h S...

Rl:R

h S R

l:R

h S... R

l:R

h S R

l:R

h S...

T... T T......

...

T T... T T.........

... ...

n-dimensional signature of tuples in

the region

super-region n-dimensional signature of tuples in the super-region

region (MBR)

tuples in the region

...

Fig. 2. Structure of the Signature R-Tree

3 Reliability Computations

The majority of reliability computations is performed in the following way. Thereliability computation of the whole system is executed on the basis of compo-nents reliability that are included in the system [8]. That is the reason why thereliability is computed in two phases. The first phase represents the retrievingof component reliability parameters and the second phase is the reliability com-putation itself. Other phases can include an evaluation of computed results andan improvement of the supply quality.

In virtue of experience, it is necessary to state that in most cases, retrievinga reliability parameter is far more complicated than the reliability computa-tion itself. It consists from a set of non-trivial queries over the data collection,e.g. Figure 3 shows a form for the reliability computation generating 120 rangequeries.


Fig. 3. A form of the reliability computation

4 Preliminary Results

In our experiments4, we compare the R-tree and the Signature R-tree over theoutage database. As a storage, we used embedded RadegastDB presented in [11].The embedded DBMS has been implemented in C++. The outage database in-cludes approximately 362,259 records with 35 attributes. In our test, we measureprocessing times of the passportization computation for all distributors for dam-aged equipments and equipment types. It typically includes 670 range queriesper one computation. The efficiency of the reliability computation has been mea-sured by the number of logical accesses and the throughput of queries in the datastructure.

We compare performance of the passportization computation in the years2013, 2014, and 2015. All tests have been executed 10×, the average results areshown in Tables 1 and 2.

As we can see, the number the logical accesses in the case of the SignatureR-tree is 3–6× lower, and consequently, the query processing is approximately1.4× to 2.8× more efficient for the Signature R-tree compared to the R-tree.

5 Conclusion

The outage database is a collection of outages in power networks in the Czechand Slovak Republics. Its existence is necessary for the reliability computation

4 The experiments were executed on an Intel Xeon E5430 2.66Ghz, 12.0 MB L2 cache;8GB of DDR333; Windows 2003 Server R2.


Table 1. The computation of passportization for damaged equipments and variousyears

Year Logical Accesses Throughput [q/s]

R-tree Signature R-tree R-tree Signature R-tree

2013 47.65 7.15 10,736 28,267

2014 33.37 7.30 15,279 30,630

2015 29.49 6.96 18,020 36,683

Table 2. The computation of passportization for equipment types and various years

Year Logical Accesses Throughput [q/s]

R-tree Signature R-tree R-tree Signature R-tree

2013 21.07 5.62 23,263 41,618

2014 15.08 5.74 34,359 51,904

2015 12.02 4.66 41,020 57,391

of a wholesale-consumer connection; therefore, the demand for this computa-tion increases. A significant number of complex queries is necessary to processduring the computation; a sophisticated storage of the data and efficient queryprocessing are necessary. In [11], we introduced a new embedded DBMS, calledRadegastDB, for handling the outage database. The R-tree data structure hasbeen used as a storage of the data. In this paper, we compared the SignatureR-tree with the R-tree: the query processing is approximately 1.4× to 2.8× moreefficient for the Signature R-tree compared to the R-tree.

References

1. R. E. Barlow and F. Proschan. Statistical Theory of Reliability and Life Testing:Probability Models. Holt, Rinehart and Winston, Inc., 1975.

2. R. Baca, M. Kratky, and V. Snasel. Bulk-loading of Compressed R-tree withFailure Data. In Proceedings of the 4th Workshop ELNET 2007. FEECS, VSB –Technical University of Ostrava, 2007.

3. N. Beckmann, H.-P. Kriegel, R. Schneider, and B. Seeger. The R∗-tree: An efficientand robust access method for points and rectangles. In Proceedings SIGMOD 1990,pages 322–331. ACM Press, 1990.

4. P. Chovanec and M. Kratky. Benchmarking of Lossless R-tree Compression forData of Failures in Electrical Power Networks. In Proceedings of the 7th Workshopof ELNET, Czech Republic, 2010.

5. P. Chovanec and M. Kratky. On the Efficiency of Multiple Range Query Processingin Multidimensional Data Structures. In Proceedings of the 17th International


Database Engineering & Applications Symposium, IDEAS ’13, pages 14–27, NewYork, NY, USA, 2013. ACM.

6. P. Chovanec, M. Kratky, and P. Bednar. Querying Outage Data using MultiQueries - Preliminary Results. In Proceedings of the 9th Workshop of ELNET,Czech Republic, 2012.

7. P. Chovanec and M. Kratky. Efficiency Improvement of Narrow Range QueryProcessing in R-tree. In Proceedings of the Dateso 2009 Annual InternationalWorkshop on DAtabases, TExts, Specifications and Objects, volume 471. CEURWorkshop Proceedings, 2009.

8. R. Gono and S. Rusek. Analysis of Power Outages in the Distribution Networks.In Proceedings of the 8th International Conference on Electrical Power Quality andUtilisation (EPQU2003), Cracow, Poland, 2003.

9. R. Gono, S. Rusek, and M. Kratky. Reliability analysis of distribution networks.In Proceedings of the 9th International Conference on Electrical Power Quality andUtilisation, EPQU 2007. Barcelona, Spain. IEEE Press, 2007.

10. A. Guttman. R-Trees: A Dynamic Index Structure for Spatial Searching. In Pro-ceedings of the International Conference on Management of Data, ACM SIGMOD1984, Boston, USA, pages 47–57. ACM Press, 1984.

11. M. Kratky, R. Baca, and P. Chovanec. Efficiency of the Embedded DatabaseSystem for Handling Outage Data. In Proceedings of the 8th Workshop of ELNET,Czech Republic, 2011.

12. M. Kratky, R. Gono, and S. Rusek. A Framework for Querying and IndexingElectrical Failure Data. In Proceedings of ELNET 2006. Ostrava, Czech Republic,2006.

13. M. Kratky, R. Gono, S. Rusek, and J. Dvorsky. A Framework for an Analysis ofFailures Data in Electrical Power Networks. In Proceedings of the InternationalConference on Power, Energy, and Applications Conference, ELNET/PEA 2006.IACTA Press/IASTED, 2006.

14. M. Kratky, V. Snasel, J. Pokorny, and P. Zezula. Efficient Processing of NarrowRange Queries in the R-Tree. In Proceedings of the tenth International DatabaseEngineering & Applications Symposium, IDEAS 2006. IEEE Computer SocietyPress, 2006.

15. H. Samet. Foundations of Multidimensional and Metric Data Structures. MorganKaufmann, 2006.

16. T. K. Sellis, N. Roussopoulos, and C. Faloutsos. The R+-Tree: A Dynamic IndexFor Multi-Dimensional Objects. In Proceedings of VLDB 1997, pages 507–518.Morgan Kaufmann, 1997.

17. The Institute of Electrical and Electronics Engineers. Guide for elec-tric distribution reliability indices, http://ieeexplore.ieee.org/xpl/

articleDetails.jsp?arnumber=1300984, 2003.

Scalable GPU Range Query Processing overOutage Database*

Pavel Bednar, Michal Kratky

Department of Computer Science,FEECS, VSB – Technical University of Ostrava,

17. Listopadu 15/2172, 708 33 Ostrava-Poruba, Czech Republicpavel.bednar, [email protected]

Scalable GPU Range Query Processing overOutage Database?

Pavel Bednar and Michal Kratky

Department of Computer ScienceVSB – Technical University of Ostrava, Czech Republic

pavel.bednar, [email protected]

Abstract. The reliability computation is applied for the maintenance ofequipments in power networks. The reliability computation is calculatedover a database of outages in electrical power networks. In the case ofthe outage database considered in this paper, data are stored in a tableincluding 35 attributes. The reliability computation requires to processup to hundreds of range queries over the database. Therefore, the ef-ficient processing of these queries is necessary. Since multidimensionalrange queries are used, a multidimensional data structure, the R∗-tree,has been applied in our previous works. In this paper, we introduce animproved method for querying the outage database using a GPU deviceand compare it with its CPU counterpart.

Key words: power networks, reliability computation, outage data, R∗-tree, GPU, CUDA

1 Introduction

Institutional changes taking place all over the world drastically effect the ap-proach to power supply quality. It is developing towards a purely commercialmatter between suppliers and their customers. The supply that does not complywith agreed qualitative parameters leads to trade disputes and financial settle-ments. Undelivered energy, including its valuation, has arrived on the scene. Thetwo following aspects of supply quality can be considered:

1. Supply reliability – relating the availability of electricity in a location.2. Voltage quality – relating to the purity of characteristics of the voltage wave-

form, including the absolute level of voltage and frequency.

This document deals with the first aspect in more detail. Worldwide centersof reliability computation1 provide databases of information about the availabil-ity of electronic and non-electronic components and distribution functions forvarious failure types. They include the result failure rate and we can retrieve

? This work was supported by the Ministry of Education, Youth and Sports of theCzech Republic (SGS, No. SP2015/192).

1 For example Alion System Reliability Center, http://src.alionscience.com/


70 Pavel Bednar, Michal Kratky

information about the producer, operation conditions, etc. These databases areapplicable to the availability prediction of complicated systems. However, thesedatabases do not include data about power equipments.

The IEEE standards define a host of reliability indices applied to distributionreliability. IEEE P1366 [18] explains the reliability indices applied to a measure-ment of distribution system reliability, and a way of calculating reliability indices.Although authors introduce a discussion about some factors influencing these in-dices, reliability parameters for power system equipments are not depicted. TheCanadian Electrical Association2 introduces a collection of reliability parame-ters for power system equipments; this is useful for North America; however, itis almost impossible to compare conditions and equipments in North Americaand Central Europe.

In many cases, it is necessary to compute electrical energy not supplied toconsumers; probability computation of not supplied energy is only possible onthe basis of the reliability computation results. Consequently, we need to observefailures and outages in the transmission and distribution of electrical energy3 forretrieving the component reliability [1].

In [15], we introduced a framework for retrieving reliability parameters indistribution networks. Consequently, we improved the approach by a new em-bedded DBMS, called RadegastDB, presented in [12]. In [7], we depicted prelim-inary results of multiple range queries over the outage database. Since we storeoutage data in a relation with 35 attributes and we use multidimensional rangequeries to query the relation, a multidimensional data structure, the R∗-tree, isutilized as a storage of the data. We presented the improvements over the outagedatabase in [7, 6, 8, 13].

The architecture of GPU’s (Graphics Processing Unit) is suitable for vectorand matrix algebra operations. That leads to the wide usage of GPUs in thearea of information retrieval, data mining, image processing, data compressionand so on. In this paper, we show some preliminary results of an improvedalgorithm [13] processing a range query using a GPU device and compare themwith the results of the R∗-tree over the outage data. Results presented in [13]showed the GPU performance is up to 3× higher than its CPU counter part.We develop an optimization of the GPU algorithm resulting even in the higherGPU performance over CPU.

This paper is organized as follows. In Section 2, we briefly describe our stor-age framework called RadegastDB and the R∗-tree. In Section 2.3, we outlineprinciples of the GPU range query algorithm. In Section 3, we introduce theoutage database and describe the reliability computation. In Section 4, we putforward results of the approach. In the last section, the paper content is resumedand the possibility of a future work is outlined.

2 http://www.canelect.ca/3 We have used the term ’outage database’ instead of the preferred phrase ’database

of failures and outages in the transmission and distribution of electrical energy’ inthis paper.

Scalable GPU Range Query Processing over Outage Database 71

2 Database System for Handling Outage Data

2.1 Introduction

In [15], we have introduced a framework for storage and querying outage data [10,9]. Databases of various distributors are transformed into a common relationscheme with 35 attributes; other works have been presented [14, 2, 5]. In [12], weintroduced a new data storage based on multidimensional data structures [16].A variant of the R-tree [11], called the R∗-tree [3], has been applied for theimplementation. In [7], we depicted preliminary results of multiple range queriesover the outage database. In [6], we describe complete algorithms, the cost model,and results of multiple range queries.

2.2 R-tree and its Variants

Since 1984 when Guttman proposed his method, R-trees [11] have become themost cited and most used as reference data structure in this area. The R-treeis a height-balanced tree based on the B+-tree with at least 50% utilizationguaranteed. This data structure supports point and range queries and someforms of spatial joins as well. A general structure of the R-tree is shown inFigure 1.

R1 R2

R3 R4 R5 R6

p2 p4 p10 p6 p9 p1 p7 p3p8 p5 p11

R1

R2

R3

R4

R5

R6

p2p4

p8

p10

p6

p9

p1

p5

p7

p3

p11

Fig. 1. A planar representation and general structure of the R-tree

It is a hierarchical data structure representing spatial data by the set ofnested n-dimensional minimum bounding rectangles (MBR). If N is an innernode, it contains pairs (Ri, Pi), where Pi is a pointer to a child of the node N .If R is the inner node MBR, then the boxes Ri corresponding to the childrenNi of N are contained in R. Boxes at the same tree level may overlap. If N isa leaf node, it contains pairs (Ri, Oi), so called index records, where Ri containsa spatial object Oi. Each node of the R-tree contains between m and M entriesunless it is the root and corresponds to a disk page. Other properties of theR-tree include the following:

– Whenever the number of node’s children drops below m, the node is deletedand its descendants are distributed among the sibling nodes. The upperbound M depends on the size of the disk page.


– The root node has at least two entries, unless it is a leaf.

– The R-tree is height-balanced; that is, all leaf nodes are at the same level.The height of an R-tree is at most blogm Nc−1 for N index records (N > 1).

Many variants of the R-tree have been proposed during the last decades.Although original algorithms of the R-tree tried to minimize the area covered byMBRs, R∗-tree [3] takes other objectives into account, e.g. the overlap amongMBRs. R+-tree [17] was introduced as a variant that avoids overlapping MBRsin intermediate nodes of the tree and an object can be stored in more than oneleaf node.

2.3 Range Query using GPU

In the case of the multidimensional range query, we have two primitive oper-ations: IsInRectangle, returning true if a tuple is in the query rectangle, andIsIntersected, returning true if a rectangle (or an MBR – minimal boundingrectangle) intersects the query rectangle. In our previous work [4], we show theGPU range query processing can be up to 10× faster than conventional CPUalgorithms. We state to achieve the maximal performance of the GPU when theworkload has to be effectively distributed across GPU’s processing units and weneed to minimize data transfers between a host (CPU) and a device (GPU).

In the case of the outage database, we can minimize data transfers if wepreload a database to the GPU’s main memory (we suppose all data can fit intothe GPU’s main memory). Moreover in the case of the reliability computation,we process many range queries in a single computation; more range queries areprocessed in a batch.

The GPU performance highly depends on its utilization. The architectureof GPU is best suitable for parallel operations over the same data. We benefitthis situation in the case of batched query executing where many range queriesare compared over the same data. The GPU’s hardware parameters varies a lotnot only among different architectures but also among various GPU’s with samearchitecture. To handle such variety we do not use only one universal GPU kernelbut we choose the most beneficial based on GPU’s and input data parameters.We have implemented some cache strategies for storing as most as possible ofrange quries in memories with lower latencies. We also made minor performanceoptimalizations to improve the kernel performance.

3 Reliability Computations

The majority of reliability computations is performed in the following way. Thereliability computation of the whole system is executed on the basis of compo-nents reliability that are included in the system [9]. That is the reason why thereliability is computed in two phases. The first phase represents the retrieving


of component reliability parameters and the second phase is the reliability com-putation itself. Other phases can include an evaluation of computed results andan improvement of the supply quality.

In virtue of experience, it is necessary to state that in most cases, retrievinga reliability parameter is far more complicated than the reliability computa-tion itself. It consists from a set of non-trivial queries over the data collection,e.g. Figure 2 shows a form for the reliability computation generating 120 rangequeries.

Fig. 2. A form of the reliability computation

4 Experimental Results

In our experiments4, we compare the range query processing over the outagedatabase using CPU and GPU. As a storage, we used our storage frameworkcalled RadegastDB [12] to be implemented using C++ and CUDA.

The outage database includes approximately 370,000 records with 35 at-tributes. In our test, we measure processing times of the passportization compu-tation for individual distributors (with the REAS xx abbreviation). It typicallyincludes 1,000 range queries per one computation. The efficiency of the reliabilitycomputation has been measured by the database time.

4 The experiments were executed on an Intel i5-2450P 3.2Ghz, 6.0 MB L2 cache; 8GBof DDR3; nVIDIA GeForce GTX690; Windows 10 Proffesional.


0

10000

20000

30000

40000

50000

60000

70000

80000

REAS1 REAS2 REAS3 REAS4 REAS5 REAS6 REAS7 REAS8 REAS9 REAS10 REAS11 REAS12

QU

ERIE

S /

SECO

ND

Passportization 2014(poškozené zařízení)

CPU GPU (v. 2014) GPU (v. 2015)

Fig. 3. Performance comparison for damaged equipment

As we stated before, the reliability computation consists of hundred or thou-sands range queries per each computation; we take an advantage of processing allrange queries in a single batch. The results of the batch processing are depictedin Figures 3 and 4 where we show the throughput of particular algorithms forpassportization. The higher throughput means the lower query processing timeand better performance. We compare CPU, our previous GPU implementation(GPU v.2014), and improved GPU algorithm (GPU v.2015).

In both cases, the GPU algorithm overcomes CPU. We have achieved a min-imum of 2× higher throuput for GPU in the worst case. In the best case, theGPU’s throughput was up to 6.5× higher over CPU.

In Figure 3, there was 670 queries in a batch. The improved GPU algorithmoutperforms the previous GPU implementation up to 50%. In Figure 4, therewas only 360 queries in a batch resulting in the minor GPU improvement ofapproximately 5% over the previous implementation.

Evidently the results confirmed the GPU is more suitable for batched queryexecution as CPU in the case of outage database. The overall performance is upto 6.5× higher as the CPU implementation. Utilizing the performance optimiza-tion and caching strategies lead to the improvement of the new GPU algorithmfrom 5% to 50%.

5 Conclusion

The outage database is a collection of outages in power networks in the Czechand Slovak Republics. Its existence is necessary for the reliability computation


0

10000

20000

30000

40000

50000

60000

70000

80000

REAS1 REAS2 REAS3 REAS4 REAS5 REAS6 REAS7 REAS8 REAS9 REAS10 REAS11 REAS12

QU

ERIE

S /

SECO

ND

Passportization 2014(druh zařízení)

CPU GPU (v. 2014) GPU (v. 2015)

Fig. 4. Performance comparison for device type

of a wholesale-consumer connection; therefore, a demand for this computationincreases. A significant number of complex queries is necessary to process duringthe computation; a sophisticated storage of the collection and query processingare necessary. In [12], we described a utilization of our embedded storage frame-work, called RadegastDB, for handling the outage database. We put forward theGPU performance over the outage database in [13]

In this paper, we compared CPU and GPU variants of range query algo-rithms over the outage database. We introduced an improved GPU range queryalgorithm and compared with our previous implementation and the CPU imple-mentation. We summary the results: (1) The GPU performance for passporti-zation outperforms CPU up to 6.5× in the best case. (2) The improved GPUalgorithm brings speed-up of 5% to 50% compared the previous implementationdepending on the batch size.

References

1. R. E. Barlow and F. Proschan. Statistical Theory of Reliability and Life Testing:Probability Models. Holt, Rinehart and Winston, Inc., 1975.

2. R. Baca, M. Kratky, and V. Snasel. Bulk-loading of Compressed R-tree withFailure Data. In Proceedings of the 4th Workshop ELNET 2007. FEECS, VSB –Technical University of Ostrava, 2007.

3. N. Beckmann, H.-P. Kriegel, R. Schneider, and B. Seeger. The R∗-tree: An efficientand robust access method for points and rectangles. In Proceedings SIGMOD 1990,pages 322–331. ACM Press, 1990.


4. P. Bednar, P. Gajdos, M. Kratky, and P. Chovanec. Processing of range query usingsimd and gpu. In New Trends in Databases and Information Systems, volume 185of Advances in Intelligent Systems and Computing, pages 13–25. Springer BerlinHeidelberg, 2013.

5. P. Chovanec and M. Kratky. Benchmarking of Lossless R-tree Compression forData of Failures in Electrical Power Networks. In Proceedings of the 7th Workshopof ELNET, Czech Republic, 2010.

6. P. Chovanec and M. Kratky. On the Efficiency of Multiple Range Query Processingin Multidimensional Data Structures. In Proceedings of the 17th InternationalDatabase Engineering & Applications Symposium, IDEAS ’13, pages 14–27, NewYork, NY, USA, 2013. ACM.

7. P. Chovanec, M. Kratky, and P. Bednar. Querying Outage Data using MultiQueries - Preliminary Results. In Proceedings of the 9th Workshop of ELNET,Czech Republic, 2012.

8. P. Chovanec, M. Kratky, and P. Bednar. Improved physical design of outagedatabase. In Proceedings of the 10th Workshop of ELNET, Czech Republic, 2013.

9. R. Gono and S. Rusek. Analysis of Power Outages in the Distribution Networks.In Proceedings of the 8th International Conference on Electrical Power Quality andUtilisation (EPQU2003), Cracow, Poland, 2003.

10. R. Gono, S. Rusek, and M. Kratky. Reliability analysis of distribution networks.In Proceedings of the 9th International Conference on Electrical Power Quality andUtilisation, EPQU 2007. Barcelona, Spain. IEEE Press, 2007.

11. A. Guttman. R-Trees: A Dynamic Index Structure for Spatial Searching. In Pro-ceedings of the International Conference on Management of Data, ACM SIGMOD1984, Boston, USA, pages 47–57. ACM Press, 1984.

12. M. Kratky, R. Baca, and P. Chovanec. Efficiency of the Embedded DatabaseSystem for Handling Outage Data. In Proceedings of the 8th Workshop of ELNET,Czech Republic, 2011.

13. M. Kratky and P. Bednar. Range query processing using gpu over outage database.In Proceedings of the 11th Workshop of ELNET, Czech Republic, 2014.

14. M. Kratky, R. Gono, and S. Rusek. A Framework for Querying and IndexingElectrical Failure Data. In Proceedings of ELNET 2006. Ostrava, Czech Republic,2006.

15. M. Kratky, R. Gono, S. Rusek, and J. Dvorsky. A Framework for an Analysis ofFailures Data in Electrical Power Networks. In Proceedings of the InternationalConference on Power, Energy, and Applications Conference, ELNET/PEA 2006.IACTA Press/IASTED, 2006.

16. H. Samet. Foundations of Multidimensional and Metric Data Structures. MorganKaufmann, 2006.

17. T. K. Sellis, N. Roussopoulos, and C. Faloutsos. The R+-Tree: A Dynamic IndexFor Multi-Dimensional Objects. In Proceedings of VLDB 1997, pages 507–518.Morgan Kaufmann, 1997.

18. The Institute of Electrical and Electronics Engineers. Guide for elec-tric distribution reliability indices, http://ieeexplore.ieee.org/xpl/

articleDetails.jsp?arnumber=1300984, 2003.

Author Index

Bednar, Pavel, 53, 69

Drholec, Jirı, 1

Gono, Radomır, 1, 22, 45, 53

Hradılek, Zdenek, 7, 15, 28, 37

Chovanec, Peter, 53, 62

Jansa, Jirı , 7

Kral, Martin, 45

Kratky, Michal, 53, 62, 69

Mach, Veleslav, 22Moldrık, Petr, 28Mozdren, Tomas, 22

Ney, Michal, 15Novosad, Ladislav, 28

Rusek, Stanislav, 22, 53

Spacek, Michal, 37

Editor: Radomır Gono

Title: ELNET 2015

Place, year, edition: Ostrava, 2015, 1st

Page count: 90

Edit: VSB – Technical University of Ostrava,17. listopadu 15, 708 33 Ostrava-Poruba, Czech Republic

Impression: 100

ISBN 978–80–248–3858–8

Documents

ELNET 2015 - Katedra informatiky FEI VŠB-TUO · The conference ELNET 2015 was held on 24th November 2015 at V SB-Technical University of Ostrava, ... Michal Kolcun (Technical University