Comprehensive evaluation approach of power quality based on neural network ensemble and subordinate degree

Based on the standards of power quality in China, a neural network ensemble (NNE) model was constructed for comprehensive evaluation of power quality. Back Propagation (BP) neural network with same topology structure were applied to all subnets of this ensemble model. The number of ensemble subnets are 20 respectively, determined with the incremental method. A large number of samples based on the random-distribution theory were produced to train these networks, and the NNE output results were analyzed according to the subordinate degree rule. The simulation test result shows that the generalization ability of NNE is superior to simplex BP neural network. Meanwhile, the model was applied to analyzing the power quality for a 0.38kV distribution network. The results show that the proposed method can evaluate the PQ grade correctly and help find out the key factor which impacts the power quality.