Empirical Data on Empirical Data on Settlement of Weather Settlement of Weather Sensitive Loads Sensitive Loads Josh Bode, M.P.P. Josh Bode, M.P.P. ERCOT Demand Side Working Group ERCOT Demand Side Working Group Austin, TX Austin, TX September 20, 2012 September 20, 2012
33
Embed
Empirical Data on Settlement of Weather Sensitive Loads Josh Bode, M.P.P. ERCOT Demand Side Working Group Austin, TX September 20, 2012.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Empirical Data on Settlement of Empirical Data on Settlement of Weather Sensitive LoadsWeather Sensitive Loads
Josh Bode, M.P.P.Josh Bode, M.P.P.
ERCOT Demand Side Working GroupERCOT Demand Side Working GroupAustin, TXAustin, TX
September 20, 2012September 20, 2012
Presentation Overview
Why is settlement of weather sensitive loads an issue?
Testing accuracy of settlement methods
Empirical results
Using smart meter data and control groups for evaluation
Page 2
Baselines are a tool to estimate demand reductions
Measuring demand reductions is an entirely different task than measuring power production Power production is metered and thus is measured directly.
Demand reductions cannot be metered. They must be estimated by indirect approaches.
In principle, the reduction is simply the difference between electricity use with and without the load curtailment However, it is not possible to directly observe or meter what
electricity use would have been in the absence of the curtailment – the counterfactual.
Instead, the counterfactual must be estimated.
Page 3
The accuracy of baseline estimates for large C&I customers has been studied multiple times
Page 4
KEMA Baseline Analysis for CEC
WG2 Baseline accuracy analysis
(Quantum)
WG2 Report Baseline Accuracy Analysis
(Quantum)
LBNL Study (proxy events)
Ontario Power Authority Study
(FSC)
California Aggregator and
DBP Evaluations(CAEC)
Highly volatile load customer study
(CAEC)
ISO-NE Baseline Study
(KEMA)
PJM Baseline Study
(KEMA)
2003 2004 2005 2006 2007 2008 2009 2010 2011 2012
California Aggregator Programs
(FSC)
Settlement of reductions from weather sensitive loads like AC has been studied less
Weather sensitive loads have demonstrated the ability to support multiple grid functions
4.8 million residential AC units and more than 500,000 water heaters in the U.S. have load control devices
Recent technological innovations enable aggregation and real time visibility of small scale loads
Many load control devices now include over and under frequency relays, providing an automated fail-safe mechanism that is synchronized with the grid
Page 5
Visibility of loads has been tested
Because of the sheer number of AC units, it is not practical to monitor all data points
Real time monitoring of AC units is expensive, even for a sample
Page 6
Feeder Load
AC end use sample
Estimated loads for population
Load control programs have shown the ability to provide contingency reserves
Page 7
0 15 30 45 60 75 90 105 120 135 150
Computer to modem Modem to devices
0%
20%
40%
60%
80%
100%
120%
140%
-4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Nor
mal
ized
Avg
. Cus
tom
er k
W
Time Since Event Start
Average
Fast start up timeFast ramp up to full resource capability
Highly granular dispatch is possible – it is possible to dispatch all or some of the resources in a specific area
No one has tested the built-in under frequency relays which provide a failsafe mechanism, do not rely on central dispatch and should respond even faster
Customers that were curtailed 68-71 over the summer report same comfort and frequency of events as customers that were curtailed once
Water heaters have demonstrated the ability to follow regulation signals
Page 8
Graph is from PJM pilot (Joe Callis) The initial test was for a unit and has since expanded to wider scale testing
However, these loads are highly variable
Average Hourly Residential AC Loads by Temperature
Page 9
0.00
0.20
0.40
0.60
0.80
1.00
1.20
1.40
1.60
1.80
0:00 3:00 6:00 9:00 12:00 15:00 18:00 21:00 0:00
Avg
Dem
and
Per A
C U
nit
(kW
)
Hour Ending
0 CDD
1-2 CDD
3-4 CDD
5-6 CDD
7-8 CDD
9-10 CDD
11-12 CDD
13-14 CDD
15-16 CDD
17-18 CDD
19-20 CDD
21-22 CDD
The fact that some loads are very weather sensitive does not mean they are unpredictable
Testing the accuracy of settlement methods
Page 10
We tested 11 different settlement alternatives for baselines for short curtailments
Type of
EstimatorMethod No. Calculation
Data SourceIndividual
ACAggregate
ACFeeder
HouseData
Within- subject
estimators
Day-matching baseline
1 10-in-10 with a 20% in-day adjustment cap X X X X2 10-in-10 without an in-day adjustment cap X X X X3 Top 3 in 10 without an in-day adjustment cap X X X X
Weather- matching baseline
4 Profile selected based on daily maximum temperature without an in-day adjustment cap X X X X
Regression
5 Treatment variables and no day or hourly lags or leads X X X X6 Treatment variables with a day lag X X X X7 Treatment variables with hourly lags and leads X X X X8 No treatment variables but use of hourly lags and leads X X X X
Between- subject
estimators
Random assignment of load control operations
9 Comparison of Means X X
10 Difference-in-differences X X
Pre-calculated load reduction estimate tables 11
Multiply the number of AC units in each geographic location by the corresponding estimate of demand reductions per AC unit for the corresponding area, hour of day, and temperature bin.
X X
Page 11
To test accuracy, one needs to know the correct values
Page 12
Because the demand reductions values are artificially introduced and
known, we can determine the accuracy of each baseline alternative
For the tables and approaches that relied on control groups, we used a split sample approach
1. Randomly split data into two groups
2. Simulate the reduction in one group
3. Use the second group to produce the counterfactual or baseline
4. Calculate impacts and store
5. Repeat 100 times
Page 13
REPEAT 100 TIMES
GROUP AUSAGE WITH AC CONTROL
▪ Unperturbed electricty use patterns are known
▪ Substract simulated impacts for each event are from actual load, producing and estimate of load with DR
GROUP BUSED TO PROVIDE COUNTERFACTUAL
▪ Unperturbed electricty use patterns are known
▪ Unperturbed electricty use patterns are used to estimate what Group A's electricity use would have been absent AC control - the counterfactual.
Randomly select sample from relevant data
source
Randomly assign to Groups A and B
CALCULATE IMPACT ESTIMATE
▪ Simple comparison of means: difference between average unperturbed electricity use ofGroup B minus average of load with DR for Group A.
▪ Regression: Explain difference in electricity use pattern to produce a more precise estimate of the impact
STORE RESULTS(Actual and estimated impacts)
There are two key issues in assessing accuracy – bias and goodness of fit
Type of Metric Metric Description
Bias Mean Percentage Error (MPE)
The mean percentage error (MPE) indicates the percentage by which the measurement, on average, tends to over or underestimate the true demand reduction.
Goodness-of-Fit
Mean Absolute Percentage Error (MAPE)
The mean absolute percentage error (MAPE) is a measure of the relative magnitude of errors across event days, regardless of positive or negative direction. It is normalized allowing comparison of results across different data sources.
CV(RMSE) This metric is similar to MAPE except that it penalizes large errors more than small ones
Page 14
0.00
0.20
0.40
0.60
0.80
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15Redu
ction
per
AC
unit
(kW
)
Simulated Reduction Baseline estimate
Avg. actual reduction Avg. Baseline Estimate
Limitations of analysis
The analysis of settlement approaches focuses on ancillary services Short term reductions to stabilize the grid
Usually triggered by generation or transmission outages and sometimes unexpected changes in wind or loads
Not always in the hottest hours
The errors in the demand reduction estimates depend on the magnitude of the reduction signal The estimates were based on 50% standard cycling
Air conditioner use is lower in California than in Texas
The direction of the findings likely hold up but the magnitude of the errors will be different
Page 15
How does the magnitude of the demand reduction signal affects accuracy? Example
Feeder characteristics 2,672 accounts on feeder 266 AC load control accounts (10% of
feeder) 292 AC controllable AC units Likely includes commercial Penetration higher than 90% of feeders
Event day characteristics August 24, 2010, max temp 103 F Simulated event period 12:00-14:00
• AC load per unit 0.63 kW• Load Impact 35%• Controllable AC load 183.3 kW (0.63
kW per unit x 292 AC units)• Feeder Impact 64 kW • Actual load without DR 7772 kW• Simulated load with DR 7708 kW
Percent impact on feeder 0.83%!!!
Page 23
0
2,000
4,000
6,000
8,000
10,000
12,000
14,000
0 3 6 9 12 15 18 21 24
Feed
er L
oad
(kW
)
Hour Ending
Baseline (+1% bias)
Actual Load without DR (SCADA)Simulated Load with DR
Smart AC participant house loadControllable AC Load
Control group methods with AC end use data(500 control group, 500 treatment)
Page 24
23% 34% 55% 20% 30% 43% 39% 41% 44%0%
20%
40%
60%
80%
100%
5% Median 95% 5% Median 95% 5% Median 95%
Comparison of Means Diff-in-diff Impact Estimate Tables
Between Subjects Impact Estimate Tables
Typi
cal %
Err
or (C
VRM
SE)
-32% -2%
39%
-16%
0% 14%
-6%
0% 7%
-30%
-20%
-10%
0%
10%
20%
30%
5% Median 95% 5% Median 95% 5% Median 95%
Comparison of Means Diff-in-diff Impact Estimate Tables
BIAS
(MPE
)
Control group methods with whole house data (2,000 control group, 2,000 treatment)
Page 25
17% 24% 52% 15% 20% 27% 39% 41% 44%0%
20%
40%
60%
80%
100%
5% Median 95% 5% Median 95% 5% Median 95%
Comparison of Means Diff-in-diff Impact Estimate Tables
Typi
cal %
Err
or (C
VRM
SE)
-41% -2%
32%
-9% -1%
10%
-6%
0% 7%
-30%
-20%
-10%
0%
10%
20%
30%
5% Median 95% 5% Median 95% 5% Median 95%
Comparison of Means Diff-in-diff Impact Estimate Tables
Between Subjects Impact Estimate Tables
BIAS
(MPE
)
Implications of study Don’t rely on feeder data for settlement
Day matching baselines are the least accurate approach with weather sensitive loads
Day-matching baselines are not well suited for measuring demand reductions from highly weather sensitive loads
More granular meters do not necessarily increase the accuracy of demand reduction measurement because measuring demand reduction is fundamentally different
Complex methods provide limited improvement
Pre-calculated load reduction tables can produce results that on average are correct, but may err for individual days, especially if they are cooler
Methods with control groups and large sample sizes perform best
Page 26
Using Smart Meter Data and Control Groups for
Evaluation
Page 27
Impact estimate tables are not developed in a vacuum
They should be based on a history of results
Ideally this includes systematic testing of load control devices under different condition and with different operation and control strategies
The estimates from the operations underlying the tables need to be unbiased and, ideally, precise
The more data points, the better the results
One may need to account for changes in the customer mix, if relevant
Page 28
With large samples and random assignment, estimation error is virtually eliminated
Actual example for 2011 PG&E SmartAC evaluation
Wide availability of smart meter data and individually addressable devices are a pre-requisite
Page 29
Randomly assign population into 10 groups
For each test event, one group was activated and the other 9 were held as a control groups
For a few events, we tested different operation strategies side-by-side
It enables side by side testing of different operation strategies
Page 30
It also enables side by side testing of different control strategies
Page 31
By using control groups and short events, once can get a substantial history of results
In the PG&E study, each customer only experienced one event, but we obtained results from 7 days, including 3 with side by side testing
It is reasonable to call up to 10 events per customers, especially if the curtailments are short (e.g. 1-2 hrs)
This can yield results under 100 different curtailments to inform the impact estimate tables
Page 32
Page 33
For any questions, feel free to contact
Josh Bode, M.P.P.
Freeman, Sullivan & Co.101 Montgomery Street 15th Floor, San Francisco, CA 94104