Stochastic Programming for Valuing Energy Storage Providing Primary Frequency Control

This paper introduces a novel modelling of the economics of energy storages providing frequency containment reserves. The focus lies both on the practical operation of day-ahead markets and the stochastic and autocorrelation characteristics of the grid frequency as they are practical concerns for any energy storage operator. Based on yearly series of grid-frequency measurements we put in light seasonal trends in the grid frequency. After having removed these trends we build an ARMA-GARCH model with fat-tail feature to explain the remaining autocorrelation. The overall model can be used to generate scenarios of 15-minute time-step frequency deviations and we explain how these scenarios can be used along with stochastic programming to dispatch and compute the revenues of a storage providing frequency containment reserves. By considering two different cases, battery storage and variable-speed pumped-storage, the paper emphasizes the importance of the technical characteristics of energy storages providing frequency containment reserves.