关键词: 配电网, 线损, RBF神经网络, 差分进化, 自适应二次变异

Abstract: In view of the problem that the structure of medium voltage distribution network is complex, operation data is incomplete, conventional power loss calculation methods are difficult to implement, a practical method of calculating power loss in distribution system is presented. By establishing the corresponding RBF Neural Network model, the method takes advantage of the strong regression ability of RBF Neural Network to map complex non-linear relation between power loss and feature parameters of distribution net, and memorizes the rule of power loss varying with distribution circuit structure and operation parameters. Adopting improved Adaptive Second Mutation Differential Evolution（ASMDE） algorithm to optimize integrally the structure parameters of RBF Neural Network, the method overcomes the shortcomings that conventional differential evolution algorithm is easy to fall into local optimum and the hidden layer and output layer structure parameters are determined separately. The simulation results prove the validity and practicability of the proposed method.

Key words: distribution systems, power loss, RBF Neural Network（RBFNN）, differential evolution；adaptive second mutation