pnnl methodology to evaluate balancing reserve
TRANSCRIPT
PowerPoint PresentationNader Samaan Pacific Northwest National
Lab
Presentation to PDWG - PCM (Production Cost Model) Data Work Group 12/15/2020
PNNL Team: Tony Nguyen, Quan Nguyen, Allison Campbell, Malini Ghosal, Marcelo Elizondo Hitachi ABB Power Grids Team: Jinxiang Zhu, Hongyan Li and Long Cheng
Generation flexibility assessment as input to Production Cost Model
December 15, 2020 2
existing + planned
• Minute by minute load, wind and solar data
• Forecasting practices for wind, solar, and loads
• Available flexible generation
Operational practices – dispatch,
scheduling, forecasts, and
PNNL’s Grid Reserve And Flexibility Planning Tool (GRAF-Plan)
Hitachi ABB GridVIEW
• Production cost model
• Generation and transmission planning
GRAF-Plan methodology has been applied in several studies in U.S., including , CAISO and WECC, and internationally
Joint on-going TCF project (PNNL and Hitachi ABB Power Grids ) integrating GRAF-Plan within GridView
0 500 1000 1500
Agenda
This presentation includes: PNNL Balancing Reserve Calculation Approach Use Case: WECC ADS PCM 2030 Modeling of Load Following and Regulation Constraints in GridView
December 15, 2020 3
The methodology mimics the real balancing process including scheduling and real-time dispatch It incorporates both the variability and uncertainty factors All sources of uncertainty (load, wind, solar) are included
It reflects forecast errors and their impact on balancing requirements
The methodology evaluates not only the capacity requirements, but also the ramping requirements The approach is flexible to reflect differences between balancing processes in different systems The methodology has been benchmarked against the actually observed balancing requirements It has been used in multiple studies
December 15, 2020 4
Net Load and Generation Requirement
BA Hourly Net Load = Load – Wind – Solar + Interchange Net Load = Generation Requirement Scheduled Generation Requirement = Forecasted Net Load Actual Generation Actual Net Load ± Δ Generation Requirement:
Energy schedules Real time dispatch (Load following) Regulation
December 15, 2020 5
December 15, 2020 6
Average Actual Load
December 15, 2020 7
Real Time Load Schedule
Average Actual Load
(5) RT Forecast Error (Persistent)
1 min
(1) Actual Gen Requirement (Net Load)
(7 ) L
oa d
F ol
lo w
in g
(4) 5 min average
Load Following and Regulation
Load Following is the difference between the hourly energy schedule including 20-minute ramps (red line) and the short-term 5-minute forecast/schedule and applied “limited ramping capability” function (blue line).
Regulation is the difference between the actual generation requirement and the short-term 5- minute dispatch (the red area between the blue and green lines).
December 15, 2020 10
Example of Load Following and Regulation Requirements
Generate hour-ahead and 5-minute ahead forecast errors for load, wind and solar Use actual and forecast data to derive minute by minute load following and regulation requirements for each BA
December 15, 2020 11
Models for Balancing Requirements Uncertainty: Multidimensional Uncertainty Analysis
In existing approaches, the analysis is frequently limited to just one dimension of the uncertainty problem –capacity. But the capacity is not a single sufficient descriptor of the problem. Operational performance of a power system can be demonstrated through four basic metrics, forming the “first performance envelope”.
Capacity (π) indicates the required minute-to-minute amount of generation or change in generation output to meet variations in balancing requirements. Ramp rate (ρ) is the slope of the ramp. Ramp duration (δ) is the duration of a curve’s ramp along the time axis. Energy () is the integration of capacity over time and can be calculated as the area between the analyzed curve and the time axis.
December 15, 2020 12
Assessment of Ramping Requirements
A swinging door algorithm is used to derive the required capability (π), ramp rate (ρ) and ramp duration (δ).
Populate the triads (πi, ρi, δi) into a three-dimensional space.
Determine the boundary values for (π, ρ, δ) to make the probability of being outside the box:
December 15, 2020 13
Concurrent consideration of the capacity, ramp rate and duration requirements.
inout
Simulate HA Forecast Errors –Load and Wind
Load and wind forecast errors are simulated using a random number generator based on the statistical characteristics of the actual forecast errors The distribution of forecast errors is an unbiased Truncated Normal Distribution (TND) Wind HA forecast errors are calculated at the BA level (not at the wind plant level)
December 15, 2020 14
Simulated Solar Forecast Errors
Behavior of solar is different than wind: In the absence of clouds and fog, solar output is very
predictable. Cloud and fog impacts are less predictable and act
quickly. Many days have little appreciable variation caused by
cloud cover. For days with cloud cover, some hours are cloudless.
The clearness index (CI) is obtained by dividing the observed horizontal global solar radiation Rg by the horizontal extraterrestrial solar radiation R:
k = Rg/R Solar forecast errors vs. clearness index.
December 15, 2020 15
Solar forecast errors vs. clearness index.
Simulated Solar Forecast Errors (cont.)
The time and weather conditions during a day can result in the following different solar forecast errors patterns: Night Time — the forecast error is zero, ε = 0; Sunny Day — the forecast error is small or close to 0,
ε→0, when the CI→1; Very Cloudy Day — the forecast error is limited or close to
0, ε→0, when the CI→0; Partly Cloudy Day — the forecast error varies in a wide
range for the intermediate values of CI. The standard distribution of solar forecast errors can be
described as a function of CI. Solar generation profiles including actual solar generation and
ideal solar generation are used to calculate the clearness index:
Solar HA forecast errors are calculated at the solar plant level (not at the BA level) 16
)21=( )(
)( =)( n,..,,t,
Hypothetical distribution of the standard deviation of solar forecast errors depending on the
clearness index.
CI σHA
0 < CI ≤ 0.2 5% 0.2 < CI ≤ 0.5 10% 0.5 < CI ≤ 0.8 7.5% 0.8 < CI ≤ 1.0 5%
Table 1. Standard Deviation of HA Solar Forecast Errors Based on Clearness Index
Level.
std(ε)
CI
stdmax
10.5
WECC 2030 ADS PCM Planning
By 2030, WECC expects a 67% increase in renewables (increases of 23% for wind, 73% for solar, and 178% in behind-the- meter solar) (*) WECC develops Anchor Data Set for 2030 model, including Production Cost Model case in Hitachi ABB GridView software tool PNNL supports this planning activity by providing calculation of flexibility reserves for the year 2030 by applying PNNL developed methodology
December 15, 2020 17
Wind and Solar Summary in WECC 2030
December 15, 2020 18
Summary at the BA level:
Data Resource Wind Capacity (MW) Solar Capacity (MW) Current capacity (WECC State of Interconnection Report) 29,000 23,000
GridView database for WECC 2030 (v. 1.4.5) 35,723 39,908
Data Needs for PNNL Flex Reserve Calculation Tool (GRAF-Plan)
December 15, 2020 19
1-min net load (native load – BTM solar), wind and utility scale solar Hourly HA-Forecast for net load, wind and utility scale solar 5-min real-time forecast for net load, wind and utility scale
solar Native load, BTM, wind and utility scale solar must be all time
synchronized. Based on 2009 weather model and shifted to 2030 time stamp Gridview, took Tuesday and Thursday from 2009 second week
and add them as first two days in 2030 to match the weekday
Forecast error statistics collection
December 15, 2020 20
Analyze data from EIA (load and day-ahead load forecast) Analyze data from CAISO OASIS for: Hour-ahead, real-time and actual load Hour-ahead and actual utility scale solar Hour-ahead and actual wind data
Analyze data from BPA for: Hour-ahead and actual wind data
Forecast error statistics and other assumptions
December 15, 2020 21
Swing Door Tolerance (regulation) 0.3 L10
Swing Door Tolerance (load following) 0.6 L10
PV CSP0 CSP6
Solar - HA Mean StdDev autocorr Mean StdDev autocorr Mean StdDev autocorr
0<CI<=0.2 0 2.5% 0.8 0 2.5% 0.8 0 2.5% 0.8
0.2<CI<=0.5 0 10% 0.8 0 10% 0.8 0 10% 0.8
0.5<CI<=0.8 0 7.5% 0.8 0 7.5% 0.8 0 7.5% 0.8
0.8<CI<=1.0 0 2.5% 0.8 0 2.5% 0.8 0 2.5% 0.8
Wind - HA 0 8% 0.9
Net Load - HA 0 2% 0.9
Wind and Solar - RT Persistence model Forecast at time T = actual value at T-7.5 min, or average from -12.5 to -7.5
Load - RT 0 0.15% 0.6
Confidence Level BA will Hold Capacity to Balance 95% of load following and regulation needs
Forecast error statistics and other assumptions (2)
December 15, 2020 22
Wind Truncation +/- 4.5 std std aggregated at BA level
Solar 0 and max value from clear sky curve at this hour for the plant
Frequency of load following Every 10-min
Number of Monte Carlo runs 30
Monte Carlo random seed value, , for current run, , out of total runs,
• = − × × + × − + • = seed value for this Monte Carlo iteration • = month number (1 to 12) • = total number of Monte Carlo runs (30) • = current Monte Carlo run (1 to 30)
CAISO Net load (Native Load BTM_Solar Utility Scale Solar Wind) (Jan 2030 days)
December 15, 2020 23
Weekend Weekday
Impact of Confidence Level 95% vs 99% CAISO Load Following UP (LFU) Jan 2030
December 15, 2020 24
0
500
1000
1500
2000
2500
3000
3500
4000
4500
5000
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Lo ad
F ol
lo w
in g
U p
in M
CAISO Jan 2030 Load Following Up 95% 99%
CAISO January 2030 Load Following decomposed (95% case) (LF capacity needs by each individual resource)
December 15, 2020 25 Note: the total requirement (dark blue in the graph) is not a simple linear addition of individual components
Hour of the Day
0
500
1000
1500
2000
2500
3000
3500
4000
4500
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Lo ad
F ol
lo w
in g
U p
in M
W
CAISO Load Following January Native Load - BTM Solar Wind Solar Total
-4000
-3500
-3000
-2500
-2000
-1500
-1000
-500
0
M
W
CAISO January 2030 Load Following decomposed (Percent of each resource of total LF capacity)
December 15, 2020 26 Hour of the Day
-5000
-4000
-3000
-2000
-1000
0
1000
2000
3000
4000
5000
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Lo ad
F ol
lo w
in g
in M
CAISO Load Following January Native Load - BTM Solar Wind Solar
CAISO July 2030 Load Following decomposed (LF capacity needs by each individual resource)
December 15, 2020 27
0
500
1000
1500
2000
2500
3000
3500
4000
4500
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Lo ad
F ol
lo w
in g
U p
in M
-4000
-3500
-3000
-2500
-2000
-1500
-1000
-500
0
M
W
Note: the total requirement (dark blue in the graph) is not a simple linear addition of individual components Hour of the Day
CAISO July 2030 Load Following decomposed (Percent of each resource of total LF capacity)
December 15, 2020 28
-4000
-3000
-2000
-1000
0
1000
2000
3000
4000
5000
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Lo ad
F ol
lo w
in g
in M
CAISO Load Following July Native Load - BTM Solar Wind Solar
Hour of the Day
CAISO January 2030 Regulation decomposed (Regulation capacity needs by each individual resource)
December 15, 2020 29
0
100
200
300
400
500
600
700
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Re gu
la tio
n U
p in
M W
-600
-500
-400
-300
-200
-100
0
Hour of the Day
Note: the total requirement (dark blue in the graph) is not a simple linear addition of individual components
CAISO January 2030 Regulation decomposed (Percent of each resource of total Reg capacity)
December 15, 2020 30
-600
-400
-200
0
200
400
600
800
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Re gu
la tio
n in
M W
Hour of the Day
CAISO July 2030 Regulation decomposed (Reg capacity needs by each individual resource)
December 15, 2020 31
0
100
200
300
400
500
600
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Re gu
la tio
n U
p in
M W
-700
-600
-500
-400
-300
-200
-100
0
Hour of the Day
Note: the total requirement (dark blue in the graph) is not a simple linear addition of individual components
CAISO July 2030 Regulation decomposed (Percent of each resource of total Reg capacity)
December 15, 2020 32
-800
-600
-400
-200
0
200
400
600
800
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Re gu
la tio
n in
M W
Hour of the Day
December 15, 2020 33
Hourly load following (up/down) and regulation requirements (up/down) at the BA level
Which gen units can provide load following and/or regulation, it can be set at the following levels: Global level: Unit sub-type at system level or units sub-type at BA level, or Local level : Individual units (will overwrite the global level if provided unless the global
level is smaller) Units excluded in the ADS 2030 model from providing flex reserve: Wind and Utility Scale Solar
could provide 100% of their current simulation hour Pgn as load following down, could provide load following up and spinning reserve if curtailed, not currently used in the model
Nuclear (needs to be set at unit level) Run of the river hydro (reserve contribution factor in GridView is set to zero) Most coal units (exceptions for few BAs like PACE and PACW)
How to Set the Limits of How Much Reserve can be Provided by Hydro Gen Units in GridView
December 15, 2020 34
GridView enforces 3 constraints for each unit to provide reserve: (1) Reserve Contribution Factor (for Hydro units only) How much reserve the unit can provide Up and Down as percentage of the unit
Pmax, in the WECC ADS 2030 case, 0 for run of river hydro units 0.1 for small hydro units Min(0.5,1/# of units) for dispatchable large hydro units As an example 0.25 Reserve Contribution Factor for a 100 MW unit means this unit can
provide: 25 MW up (LF_up + reg_up + Spinning) and 25 MW down (LF_down + reg_down)
(2) Unit Pmax If this 100 MW unit is operating at 90 MW, it can only provide 10 MW up and 25 MW
down
How to Set the Limits of How Much Reserve can be Provided by Hydro Gen Units in GridView
December 15, 2020 35
(3) Limits set by unit ramp rate LF_up and LF_down is bounded by ramp rate per min * 20min Reg_up and Reg_down is bounded by ramp rate per min *10min If this same 100 MW unit can only ramp 60 MW per hour, it can only
provide: 1MW/min *20min = up to 20 MW LF up or down and 1MW/min *10min = up to 10 MW Reg up or down or spinning If it provides 10 MW for Reg up reserve, it cannot provide spinning but
can provide up to 10 MW for LF up
How Much Reserve can be Provided by Thermal Gen Units in GridView
December 15, 2020 36
Thermal units can provide reserve based on minimum of Bid Amount (MW) Ramping Up/Down Rate * Min(AS) Upward: Cap – Dispatch Downward: Dispatch – Pmin
WECC 2030 ADS can provide up to 100% capacity for reserves Bid amount is not defined in the database (e.g. Nuclear or some coal
units can be adjusted by unit bids if don’t want to provide reserves) Ramping rate and ramping minutes are defined in the database
How Much Reserve can be Provided by Wind and Solar Gen Units in GridView
December 15, 2020 37
BTM Solar cannot provide reserve in WECC 2030 ADS Wind and Utility Solar can provide downward reserve based on Hourly generation Amount (MW)
Wind and Utility Solar can provide upward reserve based on Hourly curtailment Amount (MW)
Wind and Utility Solar can provide up to 100% of hourly generation for downward reserves Wind and Utility Solar cannot provide upward reserves
Summary and conclusions
High increase of solar (utility and BTM) and wind generation is expected in WECC by 2030 WECC prepares datasets for Production Cost Model in GridView software, ADS 2030 case to support generation, transmission planning PNNL supports with generating generation flexibility requirements to be used as input in GirdView High resolution time series is key for estimating flexibility requirements
Native load, BTM, wind and utility scale solar must be all time synchronized - 2009 data has been adapted to 2030 scenario 1-minute data was generated for load, solar (BTM and plants), and wind, based on available profiles – interpolation and noise was added to create realistic 1-min data Forecast errors were assumed based on EIA data and operational data from California ISO and BPA
Load following and regulation results have been incorporated in WECC ADS 2030 case at the BA level in Hitachi-ABB GridView
December 15, 2020 38
Contact Information
Nader Samaan, Ph.D., P.E. Team Lead (Grid Analytics) Electricity Infrastructure Group Pacific Northwest National Laboratory P.O. Box 999, MSIN J4-90 Richland, WA 99352
Phone: (509) 375-2954 (W)
Generation flexibility assessment as input to Production Cost Model
Agenda
Net Load and Generation Requirement
Modeling of Variability and Uncertainty (HA-Forecast)
Block Hour-Ahead Net Load Schedules
Real-Time Scheduling
Assessment of Ramping Requirements
Simulated Solar Forecast Errors
Wind and Solar Summary in WECC 2030
Data Needs for PNNL Flex Reserve Calculation Tool (GRAF-Plan)
Forecast error statistics collection
Forecast error statistics and other assumptions (2)
CAISO Net load (Native Load BTM_Solar Utility Scale Solar Wind) (Jan 2030 days)
Impact of Confidence Level 95% vs 99%CAISO Load Following UP (LFU) Jan 2030
CAISO January 2030 Load Following decomposed (95% case) (LF capacity needs by each individual resource)
CAISO January 2030 Load Following decomposed (Percent of each resource of total LF capacity)
CAISO July 2030 Load Following decomposed(LF capacity needs by each individual resource)
CAISO July 2030 Load Following decomposed (Percent of each resource of total LF capacity)
CAISO January 2030 Regulation decomposed(Regulation capacity needs by each individual resource)
CAISO January 2030 Regulation decomposed(Percent of each resource of total Reg capacity)
CAISO July 2030 Regulation decomposed(Reg capacity needs by each individual resource)
CAISO July 2030 Regulation decomposed(Percent of each resource of total Reg capacity)
Modeling of Load Following and Regulation Constraints in GridView
How to Set the Limits of How Much Reserve can be Provided by Hydro Gen Units in GridView
How to Set the Limits of How Much Reserve can be Provided by Hydro Gen Units in GridView
How Much Reserve can be Provided by Thermal Gen Units in GridView
How Much Reserve can be Provided by Wind and Solar Gen Units in GridView
Summary and conclusions
Slide Number 39
Presentation to PDWG - PCM (Production Cost Model) Data Work Group 12/15/2020
PNNL Team: Tony Nguyen, Quan Nguyen, Allison Campbell, Malini Ghosal, Marcelo Elizondo Hitachi ABB Power Grids Team: Jinxiang Zhu, Hongyan Li and Long Cheng
Generation flexibility assessment as input to Production Cost Model
December 15, 2020 2
existing + planned
• Minute by minute load, wind and solar data
• Forecasting practices for wind, solar, and loads
• Available flexible generation
Operational practices – dispatch,
scheduling, forecasts, and
PNNL’s Grid Reserve And Flexibility Planning Tool (GRAF-Plan)
Hitachi ABB GridVIEW
• Production cost model
• Generation and transmission planning
GRAF-Plan methodology has been applied in several studies in U.S., including , CAISO and WECC, and internationally
Joint on-going TCF project (PNNL and Hitachi ABB Power Grids ) integrating GRAF-Plan within GridView
0 500 1000 1500
Agenda
This presentation includes: PNNL Balancing Reserve Calculation Approach Use Case: WECC ADS PCM 2030 Modeling of Load Following and Regulation Constraints in GridView
December 15, 2020 3
The methodology mimics the real balancing process including scheduling and real-time dispatch It incorporates both the variability and uncertainty factors All sources of uncertainty (load, wind, solar) are included
It reflects forecast errors and their impact on balancing requirements
The methodology evaluates not only the capacity requirements, but also the ramping requirements The approach is flexible to reflect differences between balancing processes in different systems The methodology has been benchmarked against the actually observed balancing requirements It has been used in multiple studies
December 15, 2020 4
Net Load and Generation Requirement
BA Hourly Net Load = Load – Wind – Solar + Interchange Net Load = Generation Requirement Scheduled Generation Requirement = Forecasted Net Load Actual Generation Actual Net Load ± Δ Generation Requirement:
Energy schedules Real time dispatch (Load following) Regulation
December 15, 2020 5
December 15, 2020 6
Average Actual Load
December 15, 2020 7
Real Time Load Schedule
Average Actual Load
(5) RT Forecast Error (Persistent)
1 min
(1) Actual Gen Requirement (Net Load)
(7 ) L
oa d
F ol
lo w
in g
(4) 5 min average
Load Following and Regulation
Load Following is the difference between the hourly energy schedule including 20-minute ramps (red line) and the short-term 5-minute forecast/schedule and applied “limited ramping capability” function (blue line).
Regulation is the difference between the actual generation requirement and the short-term 5- minute dispatch (the red area between the blue and green lines).
December 15, 2020 10
Example of Load Following and Regulation Requirements
Generate hour-ahead and 5-minute ahead forecast errors for load, wind and solar Use actual and forecast data to derive minute by minute load following and regulation requirements for each BA
December 15, 2020 11
Models for Balancing Requirements Uncertainty: Multidimensional Uncertainty Analysis
In existing approaches, the analysis is frequently limited to just one dimension of the uncertainty problem –capacity. But the capacity is not a single sufficient descriptor of the problem. Operational performance of a power system can be demonstrated through four basic metrics, forming the “first performance envelope”.
Capacity (π) indicates the required minute-to-minute amount of generation or change in generation output to meet variations in balancing requirements. Ramp rate (ρ) is the slope of the ramp. Ramp duration (δ) is the duration of a curve’s ramp along the time axis. Energy () is the integration of capacity over time and can be calculated as the area between the analyzed curve and the time axis.
December 15, 2020 12
Assessment of Ramping Requirements
A swinging door algorithm is used to derive the required capability (π), ramp rate (ρ) and ramp duration (δ).
Populate the triads (πi, ρi, δi) into a three-dimensional space.
Determine the boundary values for (π, ρ, δ) to make the probability of being outside the box:
December 15, 2020 13
Concurrent consideration of the capacity, ramp rate and duration requirements.
inout
Simulate HA Forecast Errors –Load and Wind
Load and wind forecast errors are simulated using a random number generator based on the statistical characteristics of the actual forecast errors The distribution of forecast errors is an unbiased Truncated Normal Distribution (TND) Wind HA forecast errors are calculated at the BA level (not at the wind plant level)
December 15, 2020 14
Simulated Solar Forecast Errors
Behavior of solar is different than wind: In the absence of clouds and fog, solar output is very
predictable. Cloud and fog impacts are less predictable and act
quickly. Many days have little appreciable variation caused by
cloud cover. For days with cloud cover, some hours are cloudless.
The clearness index (CI) is obtained by dividing the observed horizontal global solar radiation Rg by the horizontal extraterrestrial solar radiation R:
k = Rg/R Solar forecast errors vs. clearness index.
December 15, 2020 15
Solar forecast errors vs. clearness index.
Simulated Solar Forecast Errors (cont.)
The time and weather conditions during a day can result in the following different solar forecast errors patterns: Night Time — the forecast error is zero, ε = 0; Sunny Day — the forecast error is small or close to 0,
ε→0, when the CI→1; Very Cloudy Day — the forecast error is limited or close to
0, ε→0, when the CI→0; Partly Cloudy Day — the forecast error varies in a wide
range for the intermediate values of CI. The standard distribution of solar forecast errors can be
described as a function of CI. Solar generation profiles including actual solar generation and
ideal solar generation are used to calculate the clearness index:
Solar HA forecast errors are calculated at the solar plant level (not at the BA level) 16
)21=( )(
)( =)( n,..,,t,
Hypothetical distribution of the standard deviation of solar forecast errors depending on the
clearness index.
CI σHA
0 < CI ≤ 0.2 5% 0.2 < CI ≤ 0.5 10% 0.5 < CI ≤ 0.8 7.5% 0.8 < CI ≤ 1.0 5%
Table 1. Standard Deviation of HA Solar Forecast Errors Based on Clearness Index
Level.
std(ε)
CI
stdmax
10.5
WECC 2030 ADS PCM Planning
By 2030, WECC expects a 67% increase in renewables (increases of 23% for wind, 73% for solar, and 178% in behind-the- meter solar) (*) WECC develops Anchor Data Set for 2030 model, including Production Cost Model case in Hitachi ABB GridView software tool PNNL supports this planning activity by providing calculation of flexibility reserves for the year 2030 by applying PNNL developed methodology
December 15, 2020 17
Wind and Solar Summary in WECC 2030
December 15, 2020 18
Summary at the BA level:
Data Resource Wind Capacity (MW) Solar Capacity (MW) Current capacity (WECC State of Interconnection Report) 29,000 23,000
GridView database for WECC 2030 (v. 1.4.5) 35,723 39,908
Data Needs for PNNL Flex Reserve Calculation Tool (GRAF-Plan)
December 15, 2020 19
1-min net load (native load – BTM solar), wind and utility scale solar Hourly HA-Forecast for net load, wind and utility scale solar 5-min real-time forecast for net load, wind and utility scale
solar Native load, BTM, wind and utility scale solar must be all time
synchronized. Based on 2009 weather model and shifted to 2030 time stamp Gridview, took Tuesday and Thursday from 2009 second week
and add them as first two days in 2030 to match the weekday
Forecast error statistics collection
December 15, 2020 20
Analyze data from EIA (load and day-ahead load forecast) Analyze data from CAISO OASIS for: Hour-ahead, real-time and actual load Hour-ahead and actual utility scale solar Hour-ahead and actual wind data
Analyze data from BPA for: Hour-ahead and actual wind data
Forecast error statistics and other assumptions
December 15, 2020 21
Swing Door Tolerance (regulation) 0.3 L10
Swing Door Tolerance (load following) 0.6 L10
PV CSP0 CSP6
Solar - HA Mean StdDev autocorr Mean StdDev autocorr Mean StdDev autocorr
0<CI<=0.2 0 2.5% 0.8 0 2.5% 0.8 0 2.5% 0.8
0.2<CI<=0.5 0 10% 0.8 0 10% 0.8 0 10% 0.8
0.5<CI<=0.8 0 7.5% 0.8 0 7.5% 0.8 0 7.5% 0.8
0.8<CI<=1.0 0 2.5% 0.8 0 2.5% 0.8 0 2.5% 0.8
Wind - HA 0 8% 0.9
Net Load - HA 0 2% 0.9
Wind and Solar - RT Persistence model Forecast at time T = actual value at T-7.5 min, or average from -12.5 to -7.5
Load - RT 0 0.15% 0.6
Confidence Level BA will Hold Capacity to Balance 95% of load following and regulation needs
Forecast error statistics and other assumptions (2)
December 15, 2020 22
Wind Truncation +/- 4.5 std std aggregated at BA level
Solar 0 and max value from clear sky curve at this hour for the plant
Frequency of load following Every 10-min
Number of Monte Carlo runs 30
Monte Carlo random seed value, , for current run, , out of total runs,
• = − × × + × − + • = seed value for this Monte Carlo iteration • = month number (1 to 12) • = total number of Monte Carlo runs (30) • = current Monte Carlo run (1 to 30)
CAISO Net load (Native Load BTM_Solar Utility Scale Solar Wind) (Jan 2030 days)
December 15, 2020 23
Weekend Weekday
Impact of Confidence Level 95% vs 99% CAISO Load Following UP (LFU) Jan 2030
December 15, 2020 24
0
500
1000
1500
2000
2500
3000
3500
4000
4500
5000
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Lo ad
F ol
lo w
in g
U p
in M
CAISO Jan 2030 Load Following Up 95% 99%
CAISO January 2030 Load Following decomposed (95% case) (LF capacity needs by each individual resource)
December 15, 2020 25 Note: the total requirement (dark blue in the graph) is not a simple linear addition of individual components
Hour of the Day
0
500
1000
1500
2000
2500
3000
3500
4000
4500
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Lo ad
F ol
lo w
in g
U p
in M
W
CAISO Load Following January Native Load - BTM Solar Wind Solar Total
-4000
-3500
-3000
-2500
-2000
-1500
-1000
-500
0
M
W
CAISO January 2030 Load Following decomposed (Percent of each resource of total LF capacity)
December 15, 2020 26 Hour of the Day
-5000
-4000
-3000
-2000
-1000
0
1000
2000
3000
4000
5000
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Lo ad
F ol
lo w
in g
in M
CAISO Load Following January Native Load - BTM Solar Wind Solar
CAISO July 2030 Load Following decomposed (LF capacity needs by each individual resource)
December 15, 2020 27
0
500
1000
1500
2000
2500
3000
3500
4000
4500
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Lo ad
F ol
lo w
in g
U p
in M
-4000
-3500
-3000
-2500
-2000
-1500
-1000
-500
0
M
W
Note: the total requirement (dark blue in the graph) is not a simple linear addition of individual components Hour of the Day
CAISO July 2030 Load Following decomposed (Percent of each resource of total LF capacity)
December 15, 2020 28
-4000
-3000
-2000
-1000
0
1000
2000
3000
4000
5000
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Lo ad
F ol
lo w
in g
in M
CAISO Load Following July Native Load - BTM Solar Wind Solar
Hour of the Day
CAISO January 2030 Regulation decomposed (Regulation capacity needs by each individual resource)
December 15, 2020 29
0
100
200
300
400
500
600
700
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Re gu
la tio
n U
p in
M W
-600
-500
-400
-300
-200
-100
0
Hour of the Day
Note: the total requirement (dark blue in the graph) is not a simple linear addition of individual components
CAISO January 2030 Regulation decomposed (Percent of each resource of total Reg capacity)
December 15, 2020 30
-600
-400
-200
0
200
400
600
800
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Re gu
la tio
n in
M W
Hour of the Day
CAISO July 2030 Regulation decomposed (Reg capacity needs by each individual resource)
December 15, 2020 31
0
100
200
300
400
500
600
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Re gu
la tio
n U
p in
M W
-700
-600
-500
-400
-300
-200
-100
0
Hour of the Day
Note: the total requirement (dark blue in the graph) is not a simple linear addition of individual components
CAISO July 2030 Regulation decomposed (Percent of each resource of total Reg capacity)
December 15, 2020 32
-800
-600
-400
-200
0
200
400
600
800
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Re gu
la tio
n in
M W
Hour of the Day
December 15, 2020 33
Hourly load following (up/down) and regulation requirements (up/down) at the BA level
Which gen units can provide load following and/or regulation, it can be set at the following levels: Global level: Unit sub-type at system level or units sub-type at BA level, or Local level : Individual units (will overwrite the global level if provided unless the global
level is smaller) Units excluded in the ADS 2030 model from providing flex reserve: Wind and Utility Scale Solar
could provide 100% of their current simulation hour Pgn as load following down, could provide load following up and spinning reserve if curtailed, not currently used in the model
Nuclear (needs to be set at unit level) Run of the river hydro (reserve contribution factor in GridView is set to zero) Most coal units (exceptions for few BAs like PACE and PACW)
How to Set the Limits of How Much Reserve can be Provided by Hydro Gen Units in GridView
December 15, 2020 34
GridView enforces 3 constraints for each unit to provide reserve: (1) Reserve Contribution Factor (for Hydro units only) How much reserve the unit can provide Up and Down as percentage of the unit
Pmax, in the WECC ADS 2030 case, 0 for run of river hydro units 0.1 for small hydro units Min(0.5,1/# of units) for dispatchable large hydro units As an example 0.25 Reserve Contribution Factor for a 100 MW unit means this unit can
provide: 25 MW up (LF_up + reg_up + Spinning) and 25 MW down (LF_down + reg_down)
(2) Unit Pmax If this 100 MW unit is operating at 90 MW, it can only provide 10 MW up and 25 MW
down
How to Set the Limits of How Much Reserve can be Provided by Hydro Gen Units in GridView
December 15, 2020 35
(3) Limits set by unit ramp rate LF_up and LF_down is bounded by ramp rate per min * 20min Reg_up and Reg_down is bounded by ramp rate per min *10min If this same 100 MW unit can only ramp 60 MW per hour, it can only
provide: 1MW/min *20min = up to 20 MW LF up or down and 1MW/min *10min = up to 10 MW Reg up or down or spinning If it provides 10 MW for Reg up reserve, it cannot provide spinning but
can provide up to 10 MW for LF up
How Much Reserve can be Provided by Thermal Gen Units in GridView
December 15, 2020 36
Thermal units can provide reserve based on minimum of Bid Amount (MW) Ramping Up/Down Rate * Min(AS) Upward: Cap – Dispatch Downward: Dispatch – Pmin
WECC 2030 ADS can provide up to 100% capacity for reserves Bid amount is not defined in the database (e.g. Nuclear or some coal
units can be adjusted by unit bids if don’t want to provide reserves) Ramping rate and ramping minutes are defined in the database
How Much Reserve can be Provided by Wind and Solar Gen Units in GridView
December 15, 2020 37
BTM Solar cannot provide reserve in WECC 2030 ADS Wind and Utility Solar can provide downward reserve based on Hourly generation Amount (MW)
Wind and Utility Solar can provide upward reserve based on Hourly curtailment Amount (MW)
Wind and Utility Solar can provide up to 100% of hourly generation for downward reserves Wind and Utility Solar cannot provide upward reserves
Summary and conclusions
High increase of solar (utility and BTM) and wind generation is expected in WECC by 2030 WECC prepares datasets for Production Cost Model in GridView software, ADS 2030 case to support generation, transmission planning PNNL supports with generating generation flexibility requirements to be used as input in GirdView High resolution time series is key for estimating flexibility requirements
Native load, BTM, wind and utility scale solar must be all time synchronized - 2009 data has been adapted to 2030 scenario 1-minute data was generated for load, solar (BTM and plants), and wind, based on available profiles – interpolation and noise was added to create realistic 1-min data Forecast errors were assumed based on EIA data and operational data from California ISO and BPA
Load following and regulation results have been incorporated in WECC ADS 2030 case at the BA level in Hitachi-ABB GridView
December 15, 2020 38
Contact Information
Nader Samaan, Ph.D., P.E. Team Lead (Grid Analytics) Electricity Infrastructure Group Pacific Northwest National Laboratory P.O. Box 999, MSIN J4-90 Richland, WA 99352
Phone: (509) 375-2954 (W)
Generation flexibility assessment as input to Production Cost Model
Agenda
Net Load and Generation Requirement
Modeling of Variability and Uncertainty (HA-Forecast)
Block Hour-Ahead Net Load Schedules
Real-Time Scheduling
Assessment of Ramping Requirements
Simulated Solar Forecast Errors
Wind and Solar Summary in WECC 2030
Data Needs for PNNL Flex Reserve Calculation Tool (GRAF-Plan)
Forecast error statistics collection
Forecast error statistics and other assumptions (2)
CAISO Net load (Native Load BTM_Solar Utility Scale Solar Wind) (Jan 2030 days)
Impact of Confidence Level 95% vs 99%CAISO Load Following UP (LFU) Jan 2030
CAISO January 2030 Load Following decomposed (95% case) (LF capacity needs by each individual resource)
CAISO January 2030 Load Following decomposed (Percent of each resource of total LF capacity)
CAISO July 2030 Load Following decomposed(LF capacity needs by each individual resource)
CAISO July 2030 Load Following decomposed (Percent of each resource of total LF capacity)
CAISO January 2030 Regulation decomposed(Regulation capacity needs by each individual resource)
CAISO January 2030 Regulation decomposed(Percent of each resource of total Reg capacity)
CAISO July 2030 Regulation decomposed(Reg capacity needs by each individual resource)
CAISO July 2030 Regulation decomposed(Percent of each resource of total Reg capacity)
Modeling of Load Following and Regulation Constraints in GridView
How to Set the Limits of How Much Reserve can be Provided by Hydro Gen Units in GridView
How to Set the Limits of How Much Reserve can be Provided by Hydro Gen Units in GridView
How Much Reserve can be Provided by Thermal Gen Units in GridView
How Much Reserve can be Provided by Wind and Solar Gen Units in GridView
Summary and conclusions
Slide Number 39