Research on Distribution Network Reconfiguration Based on Deep Q-learning Network

The reconfiguration problem of distribution network is an optimization problem aims to minimize the transmission power loss. Traditional mathematical methods are difficult to solve this kind of nonlinear mixed integer problem. This paper proposes an algorithm based on deep Q-learning network (DQN) which combines deep Q-learning and artificial neural network. A new method for the distribution network reconfiguration is presented based on DQN in this paper, with the objective of minimizing transmission power loss and the consideration of the corresponding constraints. The proposed method has been proved valid by the results from the dynamic reconfiguration of the IEEE 33-bus test system.