Managing Wind-based Electricity Generation with Storage and Transmission Capacity

Abstract

The electricity industry is currently considering co-locating wind generation and electricity storage facilities. This configuration would allow wind-farms to: (i) store wind generation that exceeds transmission capacity; (ii) time-shift sales of generated power to periods of more favorable prices (i.e., from night to day); and (iii) buy power from the market for future resale. We model this problem --- managing a wind-storage system with limited transmission capacity and the option of buying from the market --- as a finite-horizon Markov decision process that assumes uncertain wind and stochastic prices that can be negative, and apply it to wind and price models calibrated to data. Past literature on wind-storage systems either assumes deterministic wind and price dynamics, based on a single joint historical path, or neglects some relevant market or operational features, such as negative prices (a unique feature of the electricity markets) or the buying option. We find that storage can significantly increase the value of wind generators (typically by more than 30%), but using a deterministic sample path approach could misvalue storage by ±50 %. Likewise, ignoring negative prices could be detrimental when negative prices occur at least 5% of the time (with a loss of about 5%), and omitting the buying option could result in a much larger loss of value (of about 17% when negative prices occur 5% of the time), an observation for which we provide theoretical support. We also find that known static threshold heuristic policies can be more than 30% suboptimal, while a state-dependent multi-threshold heuristic policy that we develop is near optimal (being less than 0.05% suboptimal for many scenarios), even when negative prices are prevalent.