Efficiency Impact of Convergence Bidding in the California Electricity Market Ruoyang Li*, Alva Svoboda**, Shmuel Oren* *University of California at Berkeley **Pacific Gas and Electric Co. Presented at Conference on the Economics of Energy and Climate Change Toulouse, France September 8 ‐ 9, 2015
Scope
Pricing Mechanisms The prevailing mechanism for pricing electric energy in US electricity markets operated by Independent System Operators (ISO) is Locational Marginal Prices (LMP), defined as the incremental (least) cost of supplying a marginal MW of power to the specific location while meeting all security constraints
4
Over ‐ generation, congestion and no storage capability can lead to negative prices 6
Two ‐ settlements Electricity Markets
California ISO Market Timeline Day Ahead Market (DAM) T - 7 days 10:00 13:00 Bids Publish DAM Process Clear the Submitted Results Begins Market SIBR CMRI Applications: Triggers the • SIBR - Scheduling and Infrastructure Business Rules Real Time • CMRI – California ISO Market Results Interface Market • ADS – Automated Dispatch System • SLIC – Scheduling and Logging for ISO of California – Outages • MRI-S – Market Results Interface-Settlements Real Time Market (RTM ) Beginning at midpoint of T-1 after 13:00 T-75min each 5min period Bids Receive Settlements RTM Process Clear the Submitted Dispatches MRI-S Begins Market SIBR ADS
Typical Daily Schedule
Example of DA ‐ RT Price Spread 07 ‐ 2015 10
Convergence (Virtual) Bidding (CB) CB CB also also enables enables ma market et partici participan ants ts ex executing ph physic ical al tr trade ades to to op opt fo for RT RT pri prices es in instead of of DA DA pri prices. es. It It Al Also so in incr creases ma market et liqui quidi dity ty by by enabl enabling ng participan participants ts wi with no no asse assets ts to to ta take posi positions tions arbitr arbitraging aging the the DA DA ‐ RT RT spr spread ad 11
Theoretical Benefits of CB
Typical Submitted and Cleared CB Volumes 13
Con Conver ergence nce bidding bidding vo volume mes and and we weight hted price price di differ erences ences Q4 Q4 2014 2014 5 000 $25 Virtual supply Weighted price difference between day ‐ ahead and real ‐ time Virtual demand 4 000 $20 Weighted price difference for virtual supply Average hourly megawatts Weighted price difference for virtual demand 3 000 $15 2 000 $10 ($/MWh) 1 000 $5 0 $0 ‐ 1 000 ‐ $5 ‐ 2 000 ‐ $10 ‐ 3 000 ‐ $15 Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec 2014 07 ‐ 2015 14
The Efficient Market Hypothesis
Our Approach • Use the perspective of a trader attempting to construct an optimal portfolio of virtual hourly positions. • Construct and estimate a statistical time series model of the returns from historical data • Obtain an optimal portfolio (based on appropriate constraints on risk) • Evaluate the arbitrage profits of the optimal portfolio using in sample and out of sample data • Asses market efficiency based on profitable arbitrage opportunity. • Alternative approach proposed by Jha and Wolak [2013] was based on evaluating the “implied trading cost” for which one cannot reject the null hyphotesis of “no arbitrage” (based on average return = uniform portfolio assumption )
Portfolio Optimization
Portfolio Optimization (cont’d)
Portfolio Optimization with VaR Constraints
Portfolio Optimization with VaR Constraints (cont’d)
Portfolio Optimization with CVaR Constraints
Portfolio Optimization with CVaR Constraints (cont’d)
Risk Measures: VaR vs. CVaR
Time Varying Forward Premium
Time Varying Forward Premium (cont’d)
Time Varying Forward Premium (cont’d)
Model Description
Model Description (cont’d)
Data
Optimal Numbers of States and Clusters
Transition probabilities of the Post ‐ CB GMHMM
Post ‐ CB Posterior State Probabilities
Marginal Distribution of Post ‐ CB DA ‐ RT Spread
Regression Analysis
In Sample and Out of Sample Tests
Performance Under a VaR constraint
Performance Under a CVaR Constraint
Conclusions
Recommend
More recommend
