基于EEMD的区域地下水埋深PSO-ELM预测模型

PSO-ELM prediction model of regional groundwater depth based on EEMD

针对区域地下水埋深时间序列预测问题,本文将集合经验模态分解(EEMD)、粒子群算法(PSO)和极限学习机(ELM)组合,构建了EEMD-PSO-ELM地下水埋深预测模型。选用黑龙江省三江平原友谊农场地下水埋深的时间序列数据,首先利用EEMD将地下水埋深时序数据分解成若干个IMF分量,根据各分量均值将IMF分量分组叠加为高频部分、低频部分和余项;然后针对3个新序列分别构建不同的PSO-ELM模型,进而得到3组预测值,最后将预测值叠加就得到原始地下水埋深序列的最终预测值。通过精度检验发现,该组合预测模型预测效果很好;与径向基网络(radial basis function neural network,RBF)、PSO-ELM模型和ELM模型对比分析,实验结果表明,EEMD和PSO均能有效改善ELM神经网络的预测精度,ELM神经网络在区域地下水埋深预测方面有很大的应用前景。

In this paper,an EEMD-PSO-ELM model is established by combining ensemble empirical mode decomposition(EEMD),particle swarm optimization(PSO)and extreme learning machine(ELM)aiming to solve the problem of time series prediction of regional groundwater depth.The groundwater depth time series data of Youyi farm in Sanjiang Plain of Heilongjiang Province is selected.First,the groundwater depth time series data is decomposed into several IMF components by EEMD,and the IMF components are superimposed into high frequency part,low frequency part and remainder term according to the mean value of each component.Then different PSO-ELM models are constructed by the 3 new sequences to obtain the corresponding predicted values.Finally,the predicted value of the original groundwater depth sequence is obtained by the above predicted values.The research results show that the model achieves good prediction effect.The results indicate that EEMD and PSO effectively improve the prediction accuracy of ELM neural network compared with the radial basis function neural network(RBF),PSO-ELM model and ELM model,and the ELM neural network has a great application prospect in the prediction of regional groundwater depth.