Forecasting error-aware optimal dispatch of wind-storage integrated power systems: A soft-actor-critic deep reinforcement learning approach

A significant amount of renewable energy sources, including rapidly advancing wind power technologies, are incorporated into the new power system. Simultaneously, wind farm forecasting error has emerged as the main obstacle to large-scale wind power grid integration due to wind power variability. This presents a significant issue in determining the best way to dispatch and operate wind farm energy storage systems. This research presents a behavioral decision model for energy storage based on soft-actor-critic (SAC) algorithm for energy storage system (ESS) dispatch in order to address this uncertainty issue. By analyzing the output characteristics of wind farms, the wind power prediction model's output serves as the wind farm scheduling plan. The timely charging and discharging of the energy storage system balances the forecasting error between the actual output and the schedule plan, thereby increasing the net-connectedness of wind power. Lastly, the economically optimal capacity configuration of the wind farm energy storage system under the rated power is obtained by combining the energy storage decision-making model with pertinent technical and economic indicators of the energy storage system and the penalty cost of the wind storage system dispatch.