摘要
采用均匀设计的方法安排数值试验,比较不同的遗传算子组合及数据输入模式对神经网络径流预报精度的影响。研究发现,与未归一数据输入模式相比,归一化数据输入模式使网络预报的精度明显提高;不同算子组合优化神经网络初始权重径流预报精度差别较大,未归一网络的优化效果较归一网络好;同时采用数据归一输入模式与遗传算法优化神经网络初始权重未产生优化效果叠加。
GA and mode of data-in can offect the ANN's runoff forecast accuracy by optimizing the ANN's initial weights and biases. In this pa per, the numerical experiment was designed by the Uniform Design to compose the difference in runoff forecast accuracy between various combinations of genetic o perators and mode of data-in. The conclusion indicats as follow: the first, the accuracy of ANN's runoff forecast is improved obviously through making data bet ween 0 and 1,the second, there is great difference in effect of forecast accurac y among various combinations of genetic operators to optimize the ANN's initial weights and biases, and the optimizing effect of ANN which don't make the data b etween 0 and 1 is better than ones which make the data between 0 and 1.The last, there is no effect of superposition between data-in and optimizing initial we ights and biases by GA.
出处
《水电能源科学》
2005年第3期26-28,i003-i004,共5页
Water Resources and Power
关键词
径流预报
人工神经网络
遗传算子
输入模式
runoff forecast
artifical neural network
genetic operator
data-in mode
