On the Viability of Stochastic Economic Dispatch for Real-Time Energy Market Clearing

Haoruo Zhao,Mathieu Tanneau,Pascal Van Hentenryck

2024-10-24

Abstract:Over the past decade, the rapid adoption of intermittent renewable energy sources (RES), especially wind and solar generation, has posed challenges in managing real-time uncertainty and variability. In the U.S., Independent System Operators (ISOs) solve a security-constrained economic dispatch (SCED) every five minutes to clear real-time electricity markets, co-optimizing energy dispatch and reserve to minimize costs while meeting physical and reliability constraints. All SCED formulations in the U.S. are deterministic and mostly consider a single time period, limiting their effectiveness in managing real-time operational uncertainty from RES intermittency. This limitation is highlighted by the recent introduction of multiple short-term ramping products in U.S. markets, aiming to bridge the gap between deterministic and stochastic SCED formulations. While stochastic formulations address uncertainty in a unified and endogenous manner, their adoption has been hindered by high computational costs and, to a lesser extent, the availability of probabilistic forecasts. This paper revisits these concerns and demonstrates that stochastic economic dispatch is now a viable technology for real-time market clearing. It introduces the stochastic look-ahead dispatch (SLAD) formulation for real-time market clearing and presents an accelerated Benders' decomposition to solve it efficiently. Extensive experiments on a real, industry-sized transmission grid demonstrate the computational scalability of the proposed approach, with SLAD instances being solved in under 5 minutes. Furthermore, results show that SLAD provides more than 50% additional savings compared to flexiramp products and is more robust to the forecasting methodology. Therefore, SLAD is a promising approach for uncertainty management in real-time electricity markets.

Optimization and Control