小波变换集遗传算法神经网络的径流预测建模

为获得更精确的径流预报结果,利用dmey小波变换对径流时间序列分解为高频信号和低频信号,再使用遗传算法优化的BP神经网络分别对其进行预测,最后利用dmey小波逆变进行重构,以此建立径流总量预测模型。通过对柳江径流总量进行实例分析,并与遗传算法优化的神经网络模型、BP神经网络模型及传统的时间序列分析方法对比,该方法获得更准确的预测结果。研究结果表明该模型能充分反映径流时间序列趋势,预报稳定性好,预报准确率高,为径流时间序列预测提供一个有效建模方法。

Wavelet transformation combined neural network based on genetic algorithm for runoff prediction

In order to obtain more accurate forecasting results of runoff,using the dmey wavelet transform,the runoff time sequence is decomposed into high frequency part and low frequency part,then using the back propagation neural network based on the genetic algorithm(GANN),both of the two subparts are predicted respectively,finally,using the inverse dmey wavelet transform,both of the two predicted results are reconstructed to form the future behavior of the runoff series.The method has been compared with three individual forecasting models,such as GANN,back propagation neural network(BPNN) and simple moving average(SMA).It is demonstrated that the presented method is superior to the other models presented in this research in terms of the same evaluation measurements.Therefore the nonlinear ensemble model proposed in this paper can reflect the runoff time series trend well,and have good forecasting stability and accuracy,and can be used as an alternative forecasting tool for rain-runoff.