基于HP-ENN-MC模型降雨量预测

Rainfall Prediction Based on HP-ENN-MC Model

综合HP滤波、Elman神经网络、马尔科夫链的优点建立HP-ENN-MC模型对某地区10年内降雨量进行预测.以某地区1990-2015年植物生育期(6-10月)的降雨量数据作为实验训练样本,以2010-2015年(6-10月)的降雨量数据为实验的测试样本,证明HP-ENN-MC模型的实用性.由最后实验结果得到,预测平均相对误差为3.83%.所建模型能够对降雨量准确预测,与Elman、ENN-MC模型相比,HP-ENN-MC模型对降雨量预测更有效.

Objective This paper HP filter, Elman neural network, Markov chain their nature establish HP-ENN-MC model to predict rainfall. Methods According to a regional 1950- 2015 crop growth period （JuneOctober） rainfall data 1990-2009 year （JuneOctober） rainfall as the training sample, 2010-2015 （JuneOctober） rainfall effectiveness as a test sample to verify the model, Results the results showed that the mean prediction relative error of 3.93%. Conclusion The model to predict the effect is well, compared with Elman, HP-ENN-MC model more suitable for rainfall prediction.