Abstract:Wind farms in deep sea scenarios are affected by extreme climates such as tropical cyclones，which results in large-scale wind power ramp events，seriously affecting the safe and stable operation of the electrical power network. An optimal configuration method of energy storage capacity combining deep reinforcement learning with distributionally robust optimization to smooth wind power ramp events is proposed. Firstly，based on the improved swinging door algorithm，wind power ramp events is identified，and the proximal policy optimal algorithm is used to smooth wind power ramp events. Secondly， the model trained based on deep reinforcement learning uses distributionally robust optimization to optimize the capacity allocation of energy storage. Finally，the configuration results of energy storage capacity in different scenarios are compared and analyzed. The simulation results verify the effectiveness of the proposed optimization configuration method.