Research on Reactive Voltage Optimization of Multi-DC Sending Power Grid Based on Reinforcement Learning

The traditional DC reactive voltage regulation is mainly based on the single DC reactive voltage control, and it lacks the consideration of the interaction effect on the multi-DC reactive voltage. This paper presents a method of multi-DC reactive voltage optimization based on reinforcement learning. In the interaction between the action strategy and the multi-dc state, the Q-value function corresponding to each state-action is obtained, and the optimal reactive voltage control strategy is formed under various operating states of multi-DC. At the same time, because the Q-value function based on this method contains the global response information of the power grid, it can realize the unified coordination of reactive voltage between DC and DC, converter station filter and DC near region conventional units, and give the global optimal control strategy within the power grid. The effect of multi-DC reactive voltage optimization control is improved. Based on the actual data of power grid, the simulation results show the effectiveness and rationality of this method.