A Novel Load Transfer Method for Distribution Systems Based on Reinforcement Learning

When a fault occurs in a radial distribution system, the lost load should be transferred to the non-fault branches for service restoration. However, current load transfer methods are difficult to develop the optimal load transfer scheme in a short time. This paper proposes a novel load transfer method for distribution systems based on the reinforcement learning algorithm. Firstly, the process of load transfer is analyzed and formulated as the Markov decision model. Secondly, the deep reinforcement learning is applied into the problem of load transfer. The interaction between the intelligent agent and the distribution system is modeled with the objective of maximizing the restored load. Key constraints such as voltage, current and radial topology are considered. Finally, the effectiveness of the proposed method is verified based on the distribution system of an urban power grid.