Design of threshold-based energy storage control policy based on rule-constrained two-stage stochastic program

Assuming that a residential electricity consumer is equipped with solar photovoltaic panels integrated with energy storage while participating in a demand response program with time-varying price, this study focuses on developing a proper control policy for energy storage operations to minimize consumer electricity cost. In particular, this study intends to develop a threshold-based control policy that is designed to adjust the energy storage levels by charging and discharging energy storage to ensure that the energy storage levels are bounded from below by the thresholds across discrete time periods. In this case, the thresholds will be derived so that consumers are able to avoid the peak electricity rate and utilize more solar power generation while meeting the electricity demand. Specifically, the set of rule constraints is developed to enforce logical conditions to energy storage operations under the proposed control policy and integrated with the two-stage stochastic program to find proper thresholds using a real historical data set. Once the thresholds are obtained by solving the proposed rule-constrained two-stage stochastic program, the proposed control policy can be implemented to control energy storage operations. Compared to the existing approaches, the proposed control policy has merits in terms of practical application. Numerical experiments conducted with various residential house data show that the proposed threshold-based control policies result in 1%–4% gap compared to off-line optimal operations in terms of total energy cost for various residential house data. Also, compared to the reinforcement learning-based approaches, the proposed control policies show better performance with less computational time required to train the models.