Application of BPNN model optimized by GA to distributed rainfall interpolation

The genetic algorithm(GA) is integrated with back propagation algorithm(BPA);a hybrid GABP algorithm is applied to study distribution rainfall interpolation.And there is some inherent limitation that the back propagation neural network(BPNN) is easily to converge locally and initial connection weight and threshold value is set by randomness;so that GA is mainly used to optimize the initial connection weight and threshold value of BPNN.Finally,GA-BPNN model is used to estimate the rainfall in Yichang,Hubei province;the testing result shows that the estimation precision and the robustness of GABP model are improved more greatly than using classical BPNN to rainfall interpolation.The average relative error(ARE) of BPNN is 27.68%,whereas the ARE of GA-BPNN model is 18.93%.