Upload
emily-nicholson
View
213
Download
0
Tags:
Embed Size (px)
Citation preview
1
R.12-03-014: Energy Division Workshop – Operating Flexibility Modeling
Nathaniel Skinner Senior Analyst, Generation & Transmission Planning
California Public Utilities Commission
September 19, 2012
Remote Access
2
Call in #: Passcode:866-758-1675 3481442Note: *6 to mute/unmute
Upon entry to the call, please place yourself on mute, and remain on mute unless you are asking a question
WebExMeeting Number: 741 769 312Meeting Password: LTPP
https://van.webex.com/van/j.php?ED=189577152&UID=491292852&PW=NNGQ4MGM0MTBk&RT=MiM0
Agenda
3
Time Item
10:00 - 10:10 Introduction
10:10 – 10:45 Revisiting the “Step 0” and “Step 1” methodologies
10:45 - 11:15 Review of activities between 2010 LTPP settlement and today
11:15 – 12:15 Review probabilistic methodology for evaluating flexibility
12:15 – 1:10 Lunch
1:10 – 2:00 Should and how can operating flexibility criteria be understood within the context of NERC/WECC standards
2:00 – 2:30 Considering & modeling resources to meet operating flexibility needs
2:30 – 2:45 Break
2:45 – 3:00 Scenarios and Assumptions and informing other processes
3:00 – 3:15 Thoughts on the Flexibility Procurement Modeling Challenge
3:15 – 3:45 Q&A Session
3:45 – 4:00 Wrap-up / Next steps
Workshop Purpose
• Review past (2010 LTPP) Operating Flexibility modeling approaches
• Examine proposed modeling approaches for the 2012 LTPP
• Begin framing discussion for meeting any needs identified for end of 2013 decision
4
Roadmap
5
PLANNING STANDARDS (6/21)
Straw Proposal
Workshop
Comments
Proposed Scenarios (8/2)
Workshop
Comments
SCENARIOS
2012 – System 2013 - System
Modeling
Need Determination
Meeting Needs
AUTHORIZATION System Needs
Local Area Needs Determination / Authorization Based on 2011/12 CAISO TPP
SDG&E Application & Track 1 LTPP (LA Basin + BC/Ventura)
Bundled Rules / Plans
CAISO 2013/14 TPP
Op Flex Meeting 1
Op Flex Meeting 2
Op Flex Meeting #
Slide 6
Operating Flexibility Analysis for R.12-03-014
Mark Rothleder, Executive Director, Market Analysis and Development
Shucheng Liu, Principal Market Developer
CPUC, Workshop September 19, 2012
Slide 7
Description of Past Method and Model
Slide 8
Study process quantifies operational requirements and evaluates fleets ability to meet operating requirements.
RenewablePortfolios
Variable Resource Wind / Solarand Load
Profiles
FlexibilityRequirements
(Regulation, Balancing)
DevelopProfiles
Shortages
Infrastructure Needs
Costs, EmissionsImport/Export
Capacity Factor
Statistical Analysis/model
Productionsimulation
Slide 9
Study Methodology
• Variability and uncertainty of renewable resources and load largely determine system regulation and load following requirements
• Currently, load is the dominant uncontrollable variable • 33% RPS introduces two additional uncontrollable variables that
impact load-following and regulation requirements – Wind, and
– Solar
• System regulation and load following requirements depend on three factors: – Forecast quality: Load, Wind and Solar forecast errors
– Interaction between load, wind and solar: net variability
– Market timeline: how fast the market re-commit and re-dispatch the controllable resources
Slide 10
CAISO Scheduling Process
MW
tOperating Hour
Hour Ahead Schedule
Day Ahead Schedule
Hour AheadAdjustment
Load Following
Generation Requirement
Regulation
Hour Ahead ScheduleAnd Load Following
Slide 11
Calculating hourly load-following requirement
• Load Following is defined as the difference between the 5-min forecasted net load and the hour ahead forecasted net load
• Determine the 95th percentile of maximum load-following requirement
swhrhahrha
swrtfrtf
lf GLGLG ,1,1,
,min5,min5,
Slide 12
Calculating hourly regulation requirement
• Regulation is defined as the difference between the net load (load– wind– solar) and the 5 minute forecasted net load (load–wind–solar)
• Determine the 95th percentile maximum regulation requirement
srtf
wrtfrtf
sa
waa
r GGLGGLG min5,min5,min5,
Slide 13
Flow chart for calculating load-following requirements
Slide 14
Flow chart for calculating regulation requirements
Slide 15
Review of activities between 2010 LTPP settlement and today
Slide 16
Additional sensitivity and analysis performed since July 2011
• PRM Analysis Deep Dive analysis of PRM– All-Gas reserve margin overstated PRM– All-Gas case PRM~21% instead of 41%
• Step 1 Sensitivity– Assessed and bounded impacts of forecast errors– Assessed drivers of flexible ramp
• 5 minute simulation– Similar but slightly reduced violations observed
• Regional modeling and coordination– Improved modeling of the GHG improves the regional flows– If not import constrained regional coordination can improve
access to flexible resources
Slide 17
Operational criteria within the context of NERC/WECC standards
Slide 18
Balancing Authority ACE limit (BAAL) – System control opposes frequency deviations
-500
-300
-100
100
300
500
59.80 59.85 59.90 59.95 60.00 60.05 60.10 60.15 60.20
Frequency
AC
E (
MW
)
Balancing Authority ACE Limit RADAR
CurrentLast 5 minsLast 6-15 minsLast 16-30 mins
-485.00 MW / 0.1 Hz
CISO 4/23/12 15:03
000000000014:55 15:0215:0115:0014:5914:5814:5714:56Time
Consecutive Minutes Exceeding Limit(s)
00000000014:44 14:45 14:5214:5114:5014:4914:4814:4714:46
000000000014:5414:35 14:4314:4214:4114:4014:3914:3814:3714:36
014:53
60.00Hz Scheduled Frequency
15:03:4415:03
BIAS
ACE
Frequency
33.08
ACPS1
60.022
191
BAAL shall not be exceeded for more than 30 consecutive clock-minutesNumber of BAAL Exceedances
in last 30 minutes =
BAAL shall not be exceeded for more than 30 consecutive clock-minutes
CTRL-T to start timerCTRL-S to stop timer
0
• BAAL is designed to replace CPS2
• BAAL relaxes area regulation needs
• ACE is allowed to be outside BAAL for up to 30 minutes
Slide 19
Control Performance Standard Scores (CPS1) Scores January 2009 through April 2012
0
10
20
30
40
50
60
70
80
90
100
110
120
130
140
150
160
170
180
190
200
Per
cen
t (%
)
CPS1
CPS 1 Scores – January 2009 through April 2012
Began operating to BAAL
Slide 20
The assessment of a balancing authority control performance is based on three components
• Control Performance Standard (CPS1) - measures the control performance of a BA's by comparing how well its ACE performs in conjunction with the frequency error of the Interconnection
• Balancing authority Ace Limit (BAAL) - is a real-time measure of Area Control Area and system frequency which cannot exceed predefined limits for more than 30-minutes
• Disturbance Control Standard (DCS) - is the responsibility of the BA following a disturbance to recover its ACE to zero if its ACE just prior to the disturbance was greater than zero or to its pre-disturbance level if ACE was less than zero - within 15 minutes
• Control Performance Rating
Pass is when CPS1 ≥ 100%; BAALLimit ≤ 30 minutes & DCS = 100%
Slide 21
A Stochastic Model for Analyzing Ramping Capacity Sufficiency
Slide 22
A stochastic model is needed to assess the probability of upward ramping capacity sufficiency.
• A deterministic production simulation case adopts only one of the many possible combinations of input assumptions
• A stochastic model can evaluate various input combinations based on probability distributions of the stochastic input variables
• Monte Carlo simulation determines insufficient ramping capacity probability
• It complements the deterministic production simulation
Slide 23
Available ramping capacity depends on the balance of supply and demand.
Supply curve is constructed based on variable cost of each generation unit
Slide 24
Uncertainties in supply and demand affect availability of ramping capacity.
Slide 25
Available ramping capacity of each generation unit is determined based on the following factors:
• Maximum and minimum capacity
• Unit availability (due to forced and maintenance outages)
• Dispatch level
• Ramp rate
• Ramp time allowed (10 or 20 minutes)
Slide 26
Ramping capacity shortage may occur due to variations in both availability and requirement.
Slide 27
This stochastic model considers uncertainties in some of the key inputs, including:
• Load forecast
• Inter-hour load ramp
• Requirements for regulation-up and load following-up
• Generation by wind, solar, and hydro resources
• Availability of generation units (due to forced and maintenance outages)
Slide 28
A model is developed for a time period in which all hours have similar conditions.
• No unit commitment• No chronologic constraint (such as min run time and min
down time, etc.)• Independent with identical probability distribution
functions for each hour in the period• Insufficient ramping capacity probability for each hour
determined through Monte Carlo simulations• Insufficient ramping capacity probability for the whole
year calculated based on Binomial distribution
Slide 29
Probability distributions are developed based on data from the Plexos production simulation model.
• Hourly load forecast
• Hourly regulation and load following-up requirement
• Hourly wind, solar, and hydro generation
• Uniform distribution functions based on forced and maintenance outage rates of each generation unit
Slide 30
Inter-hour load ramp is calculated based on hourly load forecast.
Upward direction only
A new stochastic variable
Met by 60-min ramping capability
A part of load
)0max( 1 ttt LoadLoad,RampInter-Hour
Slide 31
These are examples of probability distribution functions of stochastic variables.
5.0% 95.0% 0.0%3.8% 96.2% 0.0%
38,935 69,949
35
,00
04
0,0
00
45
,00
05
0,0
00
55
,00
06
0,0
00
65
,00
07
0,0
00
75
,00
08
0,0
00
0
1
2
3
4
5
6
7
Val
ue
s x
10
^-5
Fit Comparison for Total CA LoadRiskBetaGeneral(1.5122,1.5704,36266,70059)
Input
Minimum36,429.66Maximum69,948.41Mean 52,846.29Std Dev 8,409.00Values 630
BetaGeneral
Minimum36,266.00Maximum70,059.00Mean 52,843.49Std Dev 8,360.86
1-in-2 forecast
5.0% 94.1% 0.9%3.1% 94.2% 2.6%
38,935 69,949
35
,00
04
0,0
00
45
,00
05
0,0
00
55
,00
06
0,0
00
65
,00
07
0,0
00
75
,00
08
0,0
00
0
1
2
3
4
5
6
7
Val
ue
s x
10
^-5
Fit Comparison for CA Load ExtendedRiskBetaGeneral(1.9596,2.7721,35699,77352)
Input
Minimum36,429.66Maximum76,569.35Mean 53,035.74Std Dev 8,593.85Values 636
BetaGeneral
Minimum35,699.00Maximum77,352.00Mean 52,949.29Std Dev 8,569.90
1-in-10 forecast
1-in-2 forecast
Exploring the probability to have load higher than 1-in-2 forecast
Slide 32
Examples of probability distribution functions of stochastic variables. (cont.)
50. % 90.0% 5.0%6.0% 90.0% 4.1%
645 880
40
05
00
60
07
00
80
09
00
10
00
11
00
12
00
0.000
0.001
0.002
0.003
0.004
0.005
0.006
0.007
0.008
0.009
Fit Comparison for Total RegURiskLogLogistic(354.15,383.65,9.9920)
Input
Minimum 506.1000Maximum1172.3400Mean 746.1442Std Dev 77.5920Values 630
LogLogistic
Minimum 354.1500Maximum +∞Mean 744.1946Std Dev 72.2464
5.0% 90.0% 5.0%2.0% 82.9% 15.1%
3169 6993
10
00
20
00
30
00
40
00
50
00
60
00
70
00
80
00
0
1
2
3
4
5
6
7
8
Val
ue
s x
10
^-4
Fit Comparison for Hydro GenRiskBetaGeneral(16.221,1.9444,-7509.4,7572.3)
Input
Minimum1275.9663Maximum7500.7600Mean 5975.4646Std Dev 1214.3980Values 630
BetaGeneral
Minimum-7509.4000Maximum7572.3000Mean 5957.9751Std Dev 1065.0696
Slide 33
Examples of probability distribution functions of stochastic variables. (cont.)
5.0% 90.0% 5.0%6.0% 89.8% 4.2%
647 4774
-10
00
01
00
02
00
03
00
04
00
05
00
06
00
07
00
0
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
Val
ue
s x
10
^-4
Fit Comparison for Wind GenRiskInvGauss(2611.5,10804.9,RiskShift(-489.50))
Input
Minimum 197.6900Maximum6509.0300Mean 2121.9544Std Dev 1245.0564Values 630
InvGauss
Minimum-489.5000Maximum +∞Mean 2122.0000Std Dev 1283.8807
5.0% 90.0% 5.0%3.6% 88.8% 7.6%
4236 9454
10
00
20
00
30
00
40
00
50
00
60
00
70
00
80
00
90
00
10
00
0
0
1
2
3
4
5
6
Val
ue
s x
10
^-4
Fit Comparison for Solar GenRiskBetaGeneral(6.3810,1.1541,-2999.5,9688.9)
Input
Minimum2491.6537Maximum9688.6904Mean 7751.9740Std Dev 1620.0039Values 630
BetaGeneral
Minimum-2999.5000Maximum9688.9000Mean 7745.5041Std Dev 1564.1544
Slide 34
Correlations among the stochastic variables are enforced.
LoadLoad Ramp
Wind Gen
Solar Gen
Hydro Gen
RegU LFU
Load 1 0.2884 -0.0947 -0.1997 0.4302 0.3801 0.0722
Load Ramp
0.2884 1 -0.3782 0.6156 0.0779 0.2064 -0.3193
Wind -0.0947 -0.3782 1 -0.1618 0.2855 -0.0108 0.0609
Solar -0.1997 0.6156 -0.1618 1 0.0254 -0.1101 -0.5064
Hydro 0.4302 0.0779 0.2855 0.0254 1 0.3094 -0.1283
RegU 0.3801 0.2064 -0.0108 -0.1101 0.3094 1 0.1415
LFU 0.0722 -0.3193 0.0609 -0.5064 -0.1283 0.1415 1
This is an example of the correlation matrix
Slide 35
Generation units in the stochastic model have the following characteristics from the Plexos model.
• From input data– Maximum and minimum capacity– Ramp rate– Forced outage and maintenance outage rates
• From simulation results– Average generation cost (to determine an initial dispatch order)
Slide 36
Generation unit availability is stochastically determined in each iteration of the Monte Carlo simulations.
• Forced and maintenance outages are determined independently for each generation unit
• Each of the outages is determined based on the unit’s outage rate and a draw using a uniform distribution function
• A maintenance outage allocation factor is used to represent seasonal pattern of maintenance schedules
• A unit is unavailable when any one of the outages occurs
Slide 37
Contributions of a generation unit to energy and ramping capacity are subject to:
• 10-min upward ramping capacity constraint
• 20-min upward ramping capacity constraint
• 60-min upward ramping capacity constraint
• Maximum capacity constraint
),10min( iiii MinCapMaxCapRampRateAS
),20min( iiiii MinCapMaxCapRampRateLFUAS
iiiii MaxCapLdRampLFUASE
energy dispatch total upward ancillary service contribution
load following up contribution inter-hour load rampi i
i i
E AS
LFU LdRamp
),60min( iiiiii MinCapMaxCapRampRateLdRampLFUAS
Slide 38
Total contributions by all generation units should meet energy and ramping capacity requirements.
• 10-min upward ancillary service requirement
• 20-min upward ramping capacity requirement
• 60-min upward ramping
• Energy balance
i ASi
AS Req
i i AS LFUi
AS LFU Req Req
i ii
E LdRamp Load
upward AS requirement load following-up requirement
inter-hour load ramp total loadAS LFU
Load Ramp
Req Req
Req Load
i i i AS LFU Load Rampi
AS LFU LdRamp Req Req Req
Slide 39
The model seeks a least-cost solution to meet energy and all ramping capacity requirements.
• Generation units are dispatched economically to meet load first
• Remaining qualified ramping capacity is used to meet upward ancillary service, load following, and inter-hour load ramp requirements
• Energy dispatch and ramping capacity contributions are co-optimized when there is insufficient ramping capacity initially
Slide 40
Monte Carlo simulation determines insufficient ramping capacity probability.
• Monte Carlo simulation is conducted using this stochastic model
• Insufficient ramping capacity results are presented in a probability distribution format
• The key results are the probability to have ramping capacity shortage each hour and the probability distribution of the volume of the shortages
Slide 41
This example has a 0.8% probability to have 20-min ramping capacity shortage each hour.
0.0% 0.8% 99.2%
-4,661 -1
-6,0
00
-4,0
00
-2,0
00
02
,00
04
,00
06
,00
08
,00
0
0.00
0.05
0.10
0.15
0.20
0.25
20- min Ramping Capacity Sufficiency
20-min Ramping Capacity Sufficiency
Minimum -4,660.87Maximum 7,267.15Mean 1,400.35Std Dev 1,158.94Values 5000
Results for the Super-Peak period.
Slide 42
The highest 20-min ramping capacity shortage is 4,661 MW in this example.
100.0%105 4,661
05
001
,00
01
,50
02,0
002
,50
03,0
003
,50
04
,00
04,5
005
,00
00.00
0.02
0.04
0.06
0.08
0.10
0.12
0.14
20- min Ramping Capacity Shortage
20-min Ramping Capacity Shortage
Minimum 105.39Maximum 4,660.87Mean 1,331.02Std Dev 1,142.42Values 39 / 5000Filtered 4961
Results for the Super-Peak period.
Slide 43
The probability to have 10-min ramping capacity shortage each hour is 0.1%.
0.0% 0.1% 99.9%
-2,180 -1
-3,0
00-
2,0
00-
1,0
00
01
,00
02
,00
03
,00
04
,00
05
,00
06
,00
0
0.00
0.02
0.04
0.06
0.08
0.10
0.12
10- min Ramping Capacity Sufficiency
10-min Ramping Capacity Sufficiency
Minimum -2,179.62Maximum 5,262.96Mean 1,618.69Std Dev 649.86Values 5000
100.0%219 2,180
20
04
00
60
08
001
,00
01,2
001
,40
01,6
001
,80
02,0
002
,20
0
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
10- min Ramping Capacity Shortage
10-min Ramping Capacity Shortage
Minimum 219.17Maximum 2,179.62Mean 1,334.97Std Dev 762.50Values 6 / 5000Filtered 4994
Results for the Super-Peak period.
Slide 44
Monte Carlo simulation results for all periods are summarized as follows:
10-min 20-min 10-min 20-min# of Hours in the Period 630 630 2298 2298Probability of Shortage 0.12% 0.78% 0.04% 0.16%
Max Shortage (MW) 2,180 4,661 1,420 3,855
Super-Peak Summer Off-Peak
Example Case
Slide 45
Cumulative probabilities of ramping capacity shortage are calculated using Binomial distribution.
i 10-min 20-min
1 81.3% 100.0%2 49.9% 99.8%3 23.6% 99.1%4 8.9% 97.2%5 2.8% 93.0%6 0.7% 85.8%7 0.2% 75.4%8 0.0% 62.7%9 0.0% 49.0%
10 0.0% 35.9%11 0.0% 24.6%12 0.0% 15.9%13 0.0% 9.6%14 0.0% 5.5%15 0.0% 2.9%16 0.0% 1.5%17 0.0% 0.7%18 0.0% 0.3%19 0.0% 0.1%20 0.0% 0.1%21 0.0% 0.0%22 0.0% 0.0%
Example Case
It is the probability to have at least i hours with ramping capacity shortage in year 2020.
Slide 46
Expected number of hours with ramping capacity shortage in 2020 are calculated based on the probabilities.
10-min 20-min
1.68 8.59
Example Case
It is insufficient ramping capacity expectation.
Slide 47
What we learned from this approach:
• It does– Use probability distributions to capture uncertainties in key input
factors– Implement ramping constraints and flexibility requirements– Present insufficient ramping capacity events in probabilistic
format
• It does not– Decide unit commitment– Impose chronological constraints
• It can be improved to– Use multi-year synchronized historical data to capture more
variations of input stochastic variables
Slide 48
A Study to Support Meeting Assembly Bill (AB) 1318
Slide 49
Plexos simulations are conducted about performance of local capacity requirement resources in 2020.
• Base model: High-Load scenario in 2011 LTPP study– Reduced min run time of some demand response resources
from 4 to 1 hour
• LCR resources– Added 3,173 MW local capacity requirement (LCR) resources
• A sensitivity case– Reduced capacity of event-based demand response resources
Slide 50
LCR resources are added to SCE and SDG&E zones.
• 3,173 MW LCR resources based on the ISO OTC study– Los Angeles Basin: 2,370MW– Big Creek Ventura areas: 430MW– San Diego: 373MW*
• LCR resources added as a combination of CCGT and GT units– SCE: 2 x 500 MW CCGT units– SCE: 18 x 100 MW LMS100 GT units– SDG&E: 1 x 373 MW CCGT unit
* It assumes that San Diego proposed generation is included already. So the total need in San Diego should be 373 MW plus
Pio Pico = 300MW Quail Brush = 100MW Escondido Energy Center = 45 MW
Slide 51
LCR resources have generic operating characteristics comparable to newer exiting resources.*
Resource Max/Min Capacity
(MW)
Full-Load Heat Rate (Btu/kWh)
Ramp Rate
(MW/min)
Forced Outage Rate (%)
Maintenance Rate
(%)
Start-up Time (hour)
Start-up Cost ($)
SCE NEW GT 100/40 9,191 12.0 7.24 10.0 1,200
SCE NEW CCGT 500/200 7,000 7.5 4.96 10.0 2 44,520
SDGE NEW CCGT 373/200 7,000 7.5 4.96 10.0 2 44,520
Gateway (CCGT) 530/265 7,000 10.0 10.00 10.0 2 24,411
Sentinel (GT) 106/43 9,191 12.0 10.00 10.0 1,000
* Forced outage rates of the new resources are based on NERC GADS 2006-2010 average EFORd, CCGT for all MW sizes and GT for 50 plus MW
0
2
4
6
8
10
12
14
16
18
Freq
uenc
y
Ramp Rate (MW/min)
Histogram of ISO CCGT Unit Ramp Rates
New CCGT
05
101520253035404550
Freq
uenc
y
Ramp Rate (MW/min)
Histogram of ISO GT Unit Ramp Rates
New GT
Slide 52
LCR resources have high capacity factors and contributions to ancillary services and load following.
Resource 1 2 3 4 5 6 7 8 9 10 11 12 Annual
SCE NEW GT 9.5 11.2 10.0 9.8 12.0 16.5 20.3 17.9 7.9 10.0 8.0 10.2 11.9
SCE NEW CCGT 53.1 60.0 61.4 64.2 59.4 64.1 73.7 83.4 80.9 66.9 61.1 68.3 66.4
SDGE NEW CCGT 49.2 62.1 55.9 20.4 72.6 76.5 69.0 87.4 83.7 50.9 37.8 20.3 57.1
Gateway (CCGT) 52.0 45.6 55.3 48.7 45.5 56.1 62.8 55.2 60.1 56.2 60.3 60.7 54.9
Sentinel (GT) 22.1 20.3 17.2 18.3 21.1 19.6 20.4 19.1 11.6 16.2 16.0 12.1 17.8
GT Average 10.9 10.7 8.0 10.8 10.9 12.0 11.2 9.5 6.6 8.4 9.3 10.4 9.8
CCGT Average 48.5 45.9 40.6 39.8 36.1 40.2 62.0 65.4 55.1 51.0 49.6 51.9 49.4
Resource LF Down LF Up Non-Spin Reg-D Reg-U Spin
SCE NEW GT 23.9 537.3 1.9 32.1 320.0 914.8
SCE NEW CCGT 1,888.0 849.2 0.5 101.8 11.6 577.2
SDGE NEW CCGT 264.9 217.8 0.0 202.7 78.6 56.4
Monthly Capacity Factors
Ancillary Service and Load Following Contribution (GWh)
Slide 53
LCR resources are highly utilized.
• Example - utilization of SCE NEW CCGT resource– Max capacity: 1,000 MW (2 x 500 MW)– Total possible generation in 2020: 8,784 x 1,000 = 8,784 GWh– Annual capacity factor: 66.40% (5,833 GWh generation)– Upward AS and load following contribution: 1,439 GWh– Generation plus AS and LF contribution: 7,271 GWh– Total utilization rate: 82.78%– Forced and maintenance outage rate: 14.96%– Unutilized capacity: 2.26%
Slide 54
Additional flexible capacity is needed in addition to the LCR resources.
• Up to 1,251 MW 20-min ramping capacity shortage identified
• Consistent with the findings in previous study– Need for 4,600 MW LMS100 GT capacity with zero outage
identified in previous study– 3,173 MW LCR capacity added– 1,251 MW residual shortage– Additional demand response usage from reduced minimum run
time
Slide 55
In the sensitivity case event-based demand response capacity is reduced by 40.7%.*
Region Original DR Capacity
(MW) Reduced DR Capacity
(MW) Reduction
(MW)
PG&E 1,687 732 955
SCE 2,827 1,977 850
SDGE 302 146 156
TOTAL 4,816 2,855 1,961
* Energy usage limits for the demand response resources are not reduced so some of the resources may be deployed for more hours.
Slide 56
Demand response resources are used frequently in summer months.
Case Jul Aug Sep Oct Sum
Original DR Capacity 44 22 3 2 71
Reduced DR Capacity 47 23 2 2 74
Number of Hours DR Resources Deployed
Slide 57
Ramping capacity shortage increases with reduction of demand response capacity.
• Utilization of LCR resources shows small changes
• 20-min ramping capacity shortage is increased from 1,251 MW to 3,212 MW– 1,961 MW increase in shortage with 1,961 MW reduction in
demand response capacity
• Demand response capacity is “flexible”– No ramp rate constraint– Not dispatchable (min capacity = max capacity)– Freeing up flexible dispatchable capacity once deployed
Slide 58
Methodologies for Next Phase of Study
Slide 59
Three methodologies were used in the previous phase of study.
• Production cost simulation– Detail, chronological, unit commitment, flexibility constraints– Slow, deterministic
• Loss of Load Probability (LOLP) calculation– Probabilistic, fast, multi-year historical data, probabilistic outputs– No flexibility constraint, non-chronological
• Stochastic simulation modeling– Stochastic, fast, flexibility constraints, probabilistic outputs– Non-chronological, no unit commitment, single-year data
Slide 60
The methodologies for the next phase of study should be able to:
• Use stochastic modeling approach to capture uncertainties in key input variables– Load, renewable and hydro generation, flexibility requirements,
resource availability, etc.
• Enforce operational constraints– Chronological linkages between intervals– Ramping constraints and flexibility requirements
Slide 61
The methodologies for the next phase of study should be able to: (cont.)
• Identify capacity and flexibility shortages– Differentiating capacity shortages from flexibility shortages– interpreting in terms of operational performance standards
(BAAL, CPS1, etc.)
• Propose solutions to meet the performance standards– Quantifying in term of capabilities not technologies
• Have manageable calculation/simulation time
62
Scenarios and Assumptions
Nathaniel Skinner Senior Analyst, Generation & Transmission Planning
California Public Utilities Commission
September 19, 2012
Scenarios Purpose• Inform policy-makers by providing
information on a broad range of plausible future scenarios
• Inform bundled procurement plans and positions
• Inform the transmission planning process and analysis of operating flexibility
• Limit the range of analysis to conform with resource constraints, while meeting policy objectives for the current LTPP63
Problem Statement
Scenarios should be developed to answer the following primary questions:
•What new infrastructure needs to be constructed to ensure adequate reliability?
•What mix of infrastructure minimizes cost to customers over the planning horizon?
64
Planning Area & Period
65
• CAISO controlled transmission grid & distribution systems
• Period 1 – 10 years ahead– Today through 2022
• Period 2 – 20+ years ahead– 2023 through 2034– Simplified analysis to understand impacts of
choices made to meet Period 1 needs
Example Scenario: 1 - Base• Reflect expectations of the future with little change
from existing policies• Key Assumptions:
– Mid load, mid inc. EE, mid small PV, low CHP– Mid DR, high probability additions, commercial RPS– Retirements: Low nuclear, low hydro/wind/solar, mid
other
How to get there: No change to business as usual and programs achieve results consistent with forecast expectations
66
Thoughts on the Flexibility
Procurement Modeling ChallengeCPUC Workshop
9/19/2012
Arne Olson, E3 Partner
Agenda
What Question Are We Trying to Answer?
What Tools Are Available Today?
What Would a Hybrid Model Look Like?
68
What Question Are We Trying to Answer?
Today’s Planning Problem Has Two Related Questions
1. How many MW of dispatchable resources do we need to (a) meet load, and (b) meet ramping requirements on various time scales?
2. What is the optimal mix of new resources, given the characteristics of the existing fleet?
Answer is a Matter of Matching Demand and Supply
Demand is a function of a number of variables:
• Load
• Load forecast error
• Load variability
• Expected renewable production
• Expected renewable production forecast error
• Expected renewable production variability
Supply is the ability of the fleet of dispatchable resources to respond on the appropriate time scale
71
Problem is Stochastic in Nature
Load is stochastic, variableand uncertain
• Often characterized as “1-in-5” or “1-in-10”
• Subject to forecast error
Renewable output is also stochastic, variable and uncertain
Supplies can also be stochastic
• Hydro endowment varies from year to year
• Generator forced outages are random
Need to know size, probability and duration of any shortfalls
72
Stochastic Modeling Must be Robust
Need will occur during “tail” events for both demand and supply
• Need enough iterations to accurately capture low-probability events
Flexibility need shortages will be related to capacity shortages
• Inflexible capacity can be a substitute for flexible capacity under some circumstances
73
Loss of Load Events
Need Must Be Defined on a Number of Time Scales
Peak Load is an annual phenomenon
• Loss of Load can occur in 50-250 hours per year
• Typically measured on an hourly basis
Ramping needs must be defined over much smaller time increments
• 5 hours
• 1 hour
• 20 minutes
• 5 minutes
• 1 minute
74Source: Russ Philbrick, PES General Meeting, Detroit, July 2011
What Tools are Available Today?
Planning Models
Calculate Loss-of-Load Probability (LOLP) and related metrics to determine the probability that resources will be inadequate to meet peak loads
• Conducted to support calculation of a reserve margin or to determine least-cost expansion plan
Advantages
• Stochastic model that considers the full range of load conditions
Limitations
• Ignores operations
• In the past, planning to prevent loss of load has provided enough flexibility
76
Operations Models
Production simulation models minimize operating costs subject to constraints on unit availability, transmission availability, etc.
• Conducted to determine the total cost associated with meeting electric load over 8760 hours
Benefits
• Treats details of generator operating limitations, transmission, and time-sequential behavior
Limitations
• Deterministic
• Significant unnecessary detail
• Lengthy run time
77
Enhanced Planning Models
E3’s LOLP/ELCC Model incorporates reservesinto LOLP framework
Does not treat flexibility characteristics explicitly
• Flexibility need modeled as fixed hourly parameter
• Does not measure ability of fleet to meet flexibility needs
• Requires reference to a portfolio that is assumed to be sufficiently flexible (All-Gas Case)
78
E3 LOLP Model Flow Chart
5
Hourly load
Generator Model
Net Load Model
Ancillary Service Model
Hourly wind
Hourly solar
Thermal fleet
Hourly Reg. Up
Hourly LFU
Outage Probability
Table
Net Load Mean & Variance
Spin, LFU & Reg. Up
Requirement
LOLP Model
MW of Need ELCCTarget PRMLOLP/ LOLE
Spinning Reserve
Hydro NQC
Import limits
What Would a “Hybrid” Model Look Like?
Planning for Reliable Operations
A model is needed that plans for reliable system operations, satisfying:
• Capacity Requirement – according to traditional metrics for capacity planning
• Flexibility Requirement – accounting for the limitations of the fleet in time sequential operations
Measures taken to meet need in one category may satisfy need in the other
• The model should select the least-cost array of portfolio/operational changes to meet flexibility
• Differentiated value of resource types over different time scales should be captured
• Storage provides fast response ramping over short time periods, CCGTs provide capacity and ramp, etc.
80
Flexibility Requirement – Defining new set of metrics
Attempts to quantify unit flexibility (NERC’s IVGTF Task 1.4)
Still unanswered: How could these metrics be used in a procurement process?• How would standards be determined and adopted?
• Is there a way to compare flexibility with adequacy using a common currency?
81
LOLP
FOR
ELCC
Inadequate Ramp Resource Probability (IRRP)
Ramp Availability Rate (RAR)
Effective Ramping Capability (ERC)
Power Reliability Metrics: Ramp Reliability Metrics:*
*Proposed by Lannoye et al.
Best solution to satisfy need
Two options to mitigate flexibility violations:
A useful model will be able to quantify the trade-offs between these options
• What measure or combination of measures satisfies need?82
Flip a switch
Operational changes can mitigate flexibility shortfalls
• Reserve scheduling, “pre-curtailment” of renewables
Grab a shovel
Steel in the ground can help to meet both capacity and flexibility requirements
• Fast, expensive resources vs. cheaper, slower ones
The “Sledgehammer” Approach: Stochastic Production Simulation
Minutely time step resolution
Monte Carlo for forecast errors
Requires large datasets• Detailed load, wind, solar datasets
• Individual unit specifications
• Scheduled and forced outages
• Hydro and import conditions
83
Run time: full stochastic simulation may be impractical
Year-long simulation does not capture long-term uncertainties, important for planning analysis
Flexibility of system depends on chosen reserve requirements – possibility of “false violation”
Difficult to incorporate expansion decision
Challenges
One Path Forward: Reduced-Form Production Simulation Modeling
Three key modifications to production simulation modeling framework:
1. Stochastic operations: Run thousands of draws of a single day per month to accurately characterize long-term uncertainty
• Preserve time-sequential unit commitment and operations over 24 hours
2. Endogenous reserves: Include endogenous, minutely specification of reserve flexibility requirements to avoid false violations and accurately characterize fast-ramping resource
3. Expansion decisions: Incorporate operational and expansion decisions (with fixed costs) to find optimal solutions
Requires elimination of all detail that doesn’t help answer question at hand in order to minimize run time (e.g., transmission)
84
+
Thank You!Energy and Environmental Economics, Inc. (E3)
101 Montgomery Street, Suite 1600
San Francisco, CA 94104
Tel 415-391-5100
Web http://www.ethree.com
Arne Olson, Partner ([email protected])
86
Wrap Up & Next Steps
Nathaniel Skinner & Noushin KetabiSenior Analysts, Generation & Transmission Planning
California Public Utilities Commission
September 19, 2012
Operating Flexibility #3
• Planning for an early November third workshop– Would explore the path forward for operating
flexibility analyses and interpreting these needs into the LTPP
– In person or fully over WebEx?
87
CalendarSeptember
19: Track II Final Scenarios to be sent to service list
24: Track I Briefs Due
October
5: Track II Final Scenarios Comments Due LTPP/Energy Storage Workshop Comments Due
12: Track I Reply Briefs Due
19: Track II Final Scenarios Reply Comments Due LTPP/Energy Storage Workshop Reply Comments Due
88
CalendarNovember
2: Track III Rules Comments Due [PENDING ALJ RULING]
20: Track II Scenarios PD
December
12: Track I PD
20: Track II Scenarios Decision on Agenda
January
TBD Commission Meeting: Track I Decision on Agenda
89
90
Thank you!For Additional Information:
www.cpuc.ca.govwww.GoSolarCalifornia.ca.gov
www.CalPhoneInfo.com