Rainfall-Runoff Simulation Using Simulated Annealing Wavelet BP Neural Networks
Wavelet neural network is a powerful tool for rainfall-runoff (RR) prediction. In this essay, a neural network based on wavelet function was proposed. But due to the probability of reaching local minimum of WNN, an improved simulated annealing neural network SAWNN was used in comparison of the WNN, the SAWNN has the ability of reaching the global minimum by employing the disturbing function and is able to mapping non-linear relations. Results show that the SAWNN has ideal performance in RR simulation and has small training error. It also indicates that the training samples should contain as much samples in different condition as possible.
WNN SAWNN Training error Rainfall-runoff BP Simulated annealing
Yuhui Wang Yunzhong Jiang Xiaohui Lei Wang Hao
School of Environmental Science and Engineering, Donghua University, Shanghai, China, 201620 China I China Institute of Water Resources and Hydropower Research, Beijing, China, 100038
国际会议
长沙
英文
2133-2137
2010-05-11（万方平台首次上网日期，不代表论文的发表时间）