Competition and Cooperation Among Wind Farms Baosen Zhang Stanford CMU Electricity Conference Ramesh Johari Ram Rajagopal Wind Power • • • • Significant increase in wind power 30% renewable by 2030 in CA, 50% by 2050 Taken as negative load Must participate in the market if subsides end Uncertainty limits wind participation 1/8 Overcoming Randomness • Spatial diversity: coalitions 100 550 90 500 80 450 70 400 60 350 50 300 40 250 30 200 20 150 10 100 0 0 10 20 30 40 50 60 70 80 50 0 10 20 30 40 50 60 70 80 5 Wind Farm 1 Wind Farm 4 1100 4.5 x 10 4 1000 3.5 900 3 800 2.5 700 2 1.5 600 1 500 400 0 0.5 10 20 30 40 50 60 10 Wind Farm 70 80 0 0 10 20 30 40 50 60 70 80 50 Wind Farm 2/8 Overcoming Randomness • Spatial diversity: coalitions • What about market power? CA Energy Crisis 2/8 Overcoming Randomness • Spatial diversity: coalitions • What about market power? • Gil & Lin,PJM, Trans. Power Systems, 2013 – 10% of wind can reduce the day-ahead price by half 2/8 Overcoming Randomness • Spatial diversity: coalitions • What about market power? Contribution: • Tradeoff between uncertainty and market power • Optimal size of coalitions 2/8 Market Setup • Two-stage market Day-Ahead Real-Time • N wind farms with output 𝑊1 , … , 𝑊𝑁 • Day-ahead profit for farm i Revenue Penalty 𝜋𝑖 𝑊𝑖 = 𝑝 𝑊1 , … , 𝑊𝑁 𝑊𝑖 − 𝑞𝐸[ 𝑊𝑖 − 𝑊𝑖 Bid Day-ahead Price + ] Real-time Shortfall 3/8 Market Setup • Two-stage market Day-Ahead Real-Time • N wind farms with output 𝑊1 , … , 𝑊𝑁 • Day-ahead profit for farm i 𝜋𝑖 𝑊𝑖 = 𝑝 𝑊1 , … , 𝑊𝑁 𝑊𝑖 − 𝑞𝐸[ 𝑊𝑖 − 𝑊𝑖 Bid Day-ahead Price + ] Real-time Shortfall 3/8 Aggregate Price Curve • Two kinds of generators: Traditional and Wind • Wind farms bid quantity • Aggregate price curve for traditional generators Fraction of Total Demand 20% Wind PJM 2008 Data Price ($/MW) 4/8 Linear Approximation • Price as function of wind 𝑁 𝑝 𝑊1 , … , 𝑊𝑁 = 1 − 𝛼 𝑊𝑖 𝑖=1 Fraction of Total Demand 𝛼 = 3.2 0 0.17 0.34 0.51 0.68 0.83 1 Normalized Price ($/MW) 5/8 Independent Wind Farms • Wind farms are i.i.d. • Large N regime, 𝛼 = 3.2 • Grand coalition max 1 − 𝛼 𝑊 𝑊 − 𝑞𝐸[(∑ 𝑊𝑖 − 𝑊)+ ] • Optimal solution: 𝑊∗ ≤ 1 2𝛼 = 15% • Withholding because of market power 6/8 Independent Wind Farms • Wind farms are i.i.d. • Large N regime, 𝛼 = 3.2 • Grand coalition≤ 1 2𝛼 = 15% • Individual competition max 1 − 𝛼∑𝑊𝑖 𝑊𝑖 − 𝑞𝐸[( 𝑊𝑖 − 𝑊𝑖 )+ ] • Optimal solution: total bid ≤ 15% • Withholding because of uncertainty 6/8 Independent Wind Farms • Wind farms are i.i.d. • Large N regime, 𝛼 = 3.2 • Grand coalition ≤ 1 2𝛼 = 15% • Individual competition ≤ 1 2𝛼 = 15% • Intermediate size Group size of 𝑁 achieves a penetration level of 1 𝛼 = 30% • Socially optimal 6/8 Correlated Wind Farms • Identical but not independent • Can still achieve 1 𝛼 • Solve an optimization to find the optimal group size NREL Data Optimal Size of 30 7/8 Conclusion • Tradeoff between market power and uncertainty • Optimal size of coalitions • Future directions: more detailed models 8/8