基于小波分析的BP神经网络潮水位预测方法

Back Propagation Neural Networks Based on Wavelet Multi-Resolution Analysis for Tide Level Forecast

分析了潮水的特性,提出了基于小波多分辨分析的BP神经网络潮水位预测方法。通过小波分解与重构技术,将潮水位序列分解成不同层次,得到趋势项、周期项和随机项,利用BP神经网络对每一项进行预测,最后再重构得到原潮水位序列的预测值。实例验证,基于小波分析的BP神经网络潮水位预测方法比单独用神经网络对潮水位进行预测更有效。

This paper introduced a method of BP neural networks based on MRAfor tide level forecast.Bywavelet decomposing,tide level series are decomposed into manyseries according to scales.Trend term,cycle term and stochastic term are gotin this way.Then the artifical neural network is used in multi-scale forecasting of these coefficients.Finally,by composing these results,we get the forecasted tide leveltime series.The method is proved more effective than BP neural network by two tested examples.