Deep Reinforcement Learning-Based Method for Joint Optimization of Mobile Energy Storage Systems and Power Grid with High Renewable Energy Sources

Abstract

The joint optimization of power systems, mobile energy storage systems (MESSs), and renewable energy involves complex constraints and numerous decision variables, and it is difficult to achieve optimization quickly through the use of commercial solvers, such as Gurobi and Cplex. To address this challenge, we present an effective joint optimization approach for MESSs and power grids that consider various renewable energy sources, including wind power (WP), photovoltaic (PV) power, and hydropower. The integration of MESSs could alleviate congestion, minimize renewable energy waste, fulfill unexpected energy demands, and lower the operational costs for power networks. To model the entire system, a mixed-integer programming (MIP) model was proposed that considered both the MESSs and the power grid, with the goal of minimizing costs. Furthermore, this research proposed a highly efficient deep reinforcement learning (DRL)-based method to optimize route selection and charging/discharging operations. The efficacy of the proposed method was demonstrated through many numerical simulations.