基于VMD-AVOA-LSSVM模型的月降水量预测研究

Study on Prediction of Monthly Precipitation Using VMD-AVOA-LSSVM Model

为寻求新型耦合模型来提高月降水量预测的效率,以最小二乘支持向量机(LSSVM)为基础,先基于变分模态分解(VMD)进行月降水序列的分解降噪得到分量,再使用新型非洲秃鹫优化算法(AVOA)优化模型参数,以各个分量为输入构建了VMD-AVOA-LSSVM月降水量预测模型。该模型在华北平原5个典型站点的试验结果表明,与其他对照模型相比,VMD-AVOA-LSSVM模型的综合性能最好;训练误差为R_(RMSE)=4.250 mm/月、M_(MRE)=68.028%、R^(2)=0.995;检验误差为R_(RMSE)=5.593 mm/月、M_(MRE)=64.320%、R^(2)=0.993,说明所提出的VMD-AVOA-LSSVM模型可为相关水资源管理提供参考。

In order to find a new coupling model to improve the efficiency of monthly precipitation prediction,a monthly precipitation prediction model named VMD-AVOA-LSSVM was constructed using each component from the variational modal decomposition(VMD)as input.Firstly,the monthly precipitation sequence was decomposed and denoised based on VMD.And then the parameters of least squares support vector machine(LSSVM)were optimized by using the new African vultures optimization algorithm(AVOA).The experimental results of the model in five typical stations in North China Plain show that the comprehensive performance of VMD-AVOA-LSSVM ranks first compared with other control models;The evaluation indicators in training period are R_(RMSE)=4.250 mm/month,M_(MRE)=68.028% and R^(2)=0.995 while the ones in test period refer to R_(RMSE)=5.593 mm/month,M_(MRE)=64.320% and R^(2)=0.993.The proposed VMD-AVOA-LSSVM model can provide reference for relevant water resources management.