摘要
采用改进的BP网络对灌溉用水量进行了预测,针对BP网络的不足,采用遗传算法对网络初始权重进行了优化,并采用LM(Levenberg-Marquardt)算法进行了误差逆传播校正。通过引入遗传算法和LM算法,网络比传统的BP网络无论从精度和训练时间上都有了较大的改进。最后对湖北省宜昌市东风渠灌区进行实例分析,证明了该方法的有效性。
Forecast of irrigation water use based on neural network was studied. Genetic algorithm was used to optimize the initial weight and Levenberg-Marquardt (LM) algorithm was used to reduce the error. Case study was conducted for Dongfengqu Irrigation District in Hubei Province and the availability of the forecast method has been approved.
出处
《灌溉排水学报》
CSCD
北大核心
2004年第2期59-61,共3页
Journal of Irrigation and Drainage
关键词
神经网络
灌溉
用水量
预测
LM算法
遗传算法
BP neural network
LM algorithm
genetic algorithm
irrigation water use
forecast.
