1. penetration of wind power
Comparing deterministic and stochastic models for electricity market clearing with high penetration of wind power
penetration of wind power
Ali Daraeepour - Dalia Patiño-Echeverri
Nicholas School of the Environment - Duke University
CEDM Annual Meeting, May 24, 2017

2. Classic market clearing can be improved
Motivation:
Classic market clearing can be improved
High penetration of wind energy resources and market inefficiencies
Day-ahead market
Balancing market
Offer day-ahead expected production
Lower forecast error
4 hour ahead
Deviations from day-ahead wind schedules are settled
Long Startup time
Day-ahead committed generation can be
Insufficient / inefficient
Costly Adjustments in the Real-time
Real-time Wind Curtailment

3. Two possible improvements to mkt clearing
1. Deterministic + Flexibility Reserve
II. Stochastic Market Clearing
Wind forecast error
Deviation offers
Day-ahead
energy offers
Wind
forecast
Flexibility reserve requirements
Day-ahead
energy offers
Wind scenarios
Day-ahead Market
Stochastic Market Clearing
Deviation offers
Energy and
reserve schedules
Energy and flexibility
reserve schedules
Balancing Market
Balancing Market

4. Objective
To assess the performance of stochastic market clearing, relative to deterministic + flexibility reserves
Environmental benefits
Wind energy integration
Reduction of air emissions
Economic outcomes
Reduction in costs of fossil fuels
Electricity prices
Market efficiency
Need for uplift payments
Convergence of day-ahead and real-time prices

5. Day-ahead Market Clearing
Method for comparing both mkt designs
Simulation of hourly operations of both markets
over one year
under two different scenarios of wind penetration
using a Unit Commitment / Economic Dispatch model
Production cost based / or assuming perfect competition
Transmission constraints not binding
Wind power and curtailment offered at no cost
Electricity demand is deterministic and inelastic
Real time commitment looks two hour ahead
Day D
Day D+1
Day-ahead Market Clearing
UC + EDC
Balancing Market and Operation
UC+EDC
Uplift Calculation
Balancing Market and Operation
UC + EDC
Day-ahead Market Clearing
Uplift Calculation

6. Ensuring the same reliability in both designs
1. Deterministic + Flexibility Reserve
II. Stochastic Market Clearing
Informed by the same reliability standard = no load shedding in one year
Wind forecast error
Reserves rule
Day-ahead
energy offers
Wind
forecast
Flexibility reserve requirements
Day-ahead
energy offers
Wind scenarios
Deviation offers
VOLL
Day-ahead Market
Stochastic Market Clearing
Deviation offers
Energy and
reserve schedules
Energy and flexibility
reserve schedules
Balancing Market
Balancing Market

7. Method Specify a Reliability Standard
Making the reserves rule and VOLL consistent
Specify a Reliability Standard
Estimate the minimum VOLL that ensures this reliability standard
Identify the dynamic flexibility reserve requirement rule
Step 1: Reliability Standard  Maximum annual allowable load-shedding equals zero
Step 2: Minimum VOLL  Run system operation with stochastic market clearing and different VOLL
Find the minimum VOLL that ensures reliability across all scenarios
Step 3: Identify the dynamic flexibility reserve requirement rule

8. Flexibility Reserve Requirement (d,t) = α × WPSTD (d,t)
Method
Determining a Dynamic flexibility reserve requirement rule
Flexibility Reserve Requirement (d,t) = α × WPSTD (d,t)
Wind Production Standard Deviation
Proportion of Uncertainty covered by Flexibility reserves
Identify the minimum α that ensures the reliability standard
By trying different values until the minimum requirement for the reliability standard is specified

9. Method
Inform both market clearing designs with the same uncertainty characterization
SynTiSe
4 years historical data on day-ahead wind power forecast error
MCMC model
30 scenarios for day-ahead forecast errors
50 scenarios for day-ahead forecast errors
Add to day-ahead forecasts
50 scenarios for day-ahead hourly wind
Stochastic :
Use scenarios set directly
Deterministic:
Use expected value of wind production scenarios as a day-ahead forecast of wind
Use standard deviation of wind power production scenarios to calculate flexibility reserve requirement

10. Capacity (MW) & share (%) following reserve capability
Test Grid & Data
12% scaled version of PJM’s fossil-fired generation mix
heat rate and capacity data from EPA-NEEDS
Installed capacity of thermal resources = MW
Expected Peak = MW Reserve margin =15.5%
Fuel prices from Energy Information Administration (EIA)
Technology
# of units
Capacity (MW) & share (%)
following reserve capability
Quick start
Nuclear ST(i) 4
4616 (23%) No
Coal ST 19
8727 (44%)
Yes
NGCC (ii) 14
2996 (15%)
Oil CT (iii) 8
631 (3%)
NGCT 22
3030 (15%)
BPA’ synchronous demand and wind data
Three case studies with different wind penetration levels
Case 1: 6%
Case 2: 12%
Case 3: 21%

11. Results
Fundamental difference between two models is in their DA scheduling of wind
Deterministic always schedules the expected value (i.e., the forecast)
Stochastic schedules different quantities
Sometimes is less than the expected value
Sometimes is a value between the expected value and the maximum
Sometimes is the maximum value
Depending on the ratio of expected wind to load
At 21% wind penetration, expected wind is more than 50% of load,
 Schedules of less than expected value are more common
Wind penetraion
21%
12%
When expected wind is not much compared to load,
 schedule it all

12. Results
Wind integration

13. Results
Reductions in fossil-fired generation
 12% wind penetration

14. Results
Cost savings achieved by stochastic clearing from less use of fossil-fuels

15. Results
Day-ahead energy prices

16. Results
Real-time energy prices

17. Results
Generator’s revenue

18. Conclusions
1) Stochastic market clearing increases wind integration, lowers emission, and fossil fuel costs
Under case 2, annual costs are reduced by 1.36% (i.e. 500 Million USD)
2) Benefits are mostly due to better day-ahead wind energy schedules
3) Higher wind integration in the stochastic case lowers the day-ahead prices and fossil-fired generation revenues
5) Less flexible resources incur significant losses from implementing the stochastic market clearing
6) Revenues to generators are lower under stochastic market clearing
 If not able to recover fixed costs, higher payments from capacity market will be needed