A Collaborative AGC Control Strategy for Energy Storage and Other Power Sources Based on Deep Reinforcement Learning

Aiming at the issues of insufficient active power regulation capacity in grid-connected large-scale wind-solar bases and high assessment pressure at the interface of line corridors, a multi-energy cooperative automatic power generation control strategy is proposed, which incorporates the control performance standard and restores the state of charge (SOC) of energy storage to the optimal normal setpoint throughout the day. A load frequency control model is developed for the interconnected power grid to distinguish the size of regional control deviation values, and respectively design reward functions including control performance criteria, cross-sectional over-limit control, and multiple energy storage unit SOC recovery to the optimal normal setpoint of the corresponding period. Based on the random optimization dispatching results of the thermal power units in the previous two stages, the normal setpoint of the energy storage SOC that has the best effect in suppressing the fluctuations of wind and solar power and the power fluctuations of the connection line at each time period throughout the day is obtained. In addition, a multi-time scale state perception system is constructed, and an optimal adaptive coordinated frequency control strategy for energy storage is formulated. Through the simulation incorporating the actual wind and solar output fluctuation data from the Inner Mongolia East Power Grid, it is verified that when the regional control deviation is within a small range, the SOC values of multiple energy storage units can be restored to the optimal normal setting values by the proposed method. During the control process, the SOC of each energy storage unit gradually becomes consistent. The simulation results indicate that the proposed method not only satisfies the control performance specifications, but also enables the energy storage systems to operate continuously and efficiently.