摘要
本文在总结大量洪水预报实践经验的基础上 ,提出了一种峰值识别理论及相应的改进BP算法 (ErrorBackPropagationwithPeakRecognizer ,简称BPPR) .该理论及算法在修改网络权重时 ,偏重大值误差 ,即大值误差对权重的修改起主要作用 .这种BPPR算法使人工神经网络洪水预报模型对洪峰的预报精度显著提高 。
A modified BP algorithm with peak recognition theory (BPPR) is proposed for improving the calculation of artificial neural network. By this algorithm the error correction of flood peak value is mainly depending on the modification of weight according to the error of the big values. It is useful to improve peak value recognition the accuracy of flood forecasting.
出处
《水利学报》
EI
CSCD
北大核心
2002年第6期15-20,共6页
Journal of Hydraulic Engineering
基金
自然科学基金资助项目 (5 980 90 0 7)
关键词
人工神经网络
峰值识别理论
洪水预报
artificial neural network
peak value recognition theory
flood forecasting
accuracy of forecasting
