Robust Risk Management in Hydro-Electric Pumped Storage Plants

The hydro-electric pumped storage plant management can be formulated as a multi-stage stochastic optimization problem. Recent research has identified efficient policies that minimize the risk of the plant value in a finite horizon framework by aggregating the spot price information into epochs. The most widely accepted models describe the spot price as being affected by a number of factors, each one following a mean-reverting type stochastic process. Yet, the conditions that affect those factors are rather volatile, making the adoption of a single probability model dangerous. Indeed, historical data show that the estimation of the stochastic models parameters is heavily dependent on the set of samples used. To tackle this problem, we follow a robust risk management approach and propose a method to compute optimal dispatch policies that minimize the Robust CVaR of the final plant value. We compare the optimal-Robust CVaR to the optimal-CVaR policies in probability models constructed using spot price data from EPEX for the period between 2005 and 2010. The computational overhead of computing optimal-Robust CVaR policies is linear in the number of scenarios and the computational complexity can be adjusted to achieve the desired level of accuracy in planning.