Abstract: As the proportion of renewable energy continues to increase, problems such as voltage overruns and network losses have an increasingly serious impact on the safe and efficient operation of active distribution networks.Therefore, the reactive power of the inverter must be used reasonably.However, in the active distribution network, the active and reactive power are coupled with each other, and the large scale of operating variables leads to insufficient computing power, which makes it difficult for the traditional distribution network optimization model to operate and mantain.In order to improve the problems of voltage over-limit and network loss in active distribution network, this paper adopts an active and reactive power coordination control strategy based on priority reinforcement learning and applies it to the active distribution network model.Reinforcement learning control strategies are used in interactions between grids and utilize prioritized scene playback to improve efficiency.Finally, this paper takes the active distribution network of a city in Hebei South Power Network as an example to test and analyze the strategy through simulation. The simulation results show that the strategy has obvious advantages in eliminating voltage over-limit, reducing network loss, and maximizing system economy, effectiveness, and superiority.