Abstract: A 10kV distribution network line loss evaluation method based on improved BP neural network is proposed.Through learning the training sample sets by improved BPNN algorithm,a line loss evaluation model is built and then the line loss of test sample sets can be calculated to assess the line loss of 10kV distribution network,Adaptive genetic algorithm is substituted for Gradient descent method in the process of training BP neural network, weights and biases of BP neural network are global searched so that the defects that running into local minimum value easily and slow convergence rate can be overcome,and the line loss calculation accuracy will be improved.Finally,effectiveness of AGA-BPNN algorithm line loss evaluation methord is proved through specific examples.Compared with traditional BPNN,it has better accuracy , astringency and generalization.