基于VMD-CQPSO-GRU模型的气象干旱预测方法  

Prediction Method of Meteorological Drought Based on VMD-CQPSO-GRU Method

作　　者：刘雪梅 宋文辉 钱峰[2] 王立虎 冯岭 谢文君[2] LIU Xuemei;SONG Wenhui;QIAN Feng;WANG Lihu;FENG Ling;XIE Wenjun(School of Information Engineering,North China University of Water Resources and Electric Power,Zhengzhou 450046,China;Information Center,Ministry of Water Resources,Beijing 100053,China)

机构地区：[1]华北水利水电大学信息工程学院,河南郑州450046 [2]水利部信息中心,北京100053

出　　处：《华北水利水电大学学报（自然科学版）》2021年第4期31-40,共10页Journal of North China University of Water Resources and Electric Power：Natural Science Edition

摘　　要：针对当前气象干旱预测方法中存在的预测准确度低、可信度差等问题,提出了一种基于机器学习的气象干旱预测方法。利用变分模态分解(Variational Mode Decomposition,VMD)将干旱时序信号分解为若干本征模态分量;利用改进的混沌量子粒子群算法(Chaotic Quantum Particle Swarm Optimiztaion,CQPSO)优化门控循环单元(Gated Recurrent Unit,GRU)神经网络超参数;采用“分解-合成”策略构建VMD-CQPSO-GRU智能预测模型。以安阳、郑州、信阳3座城市1951—2018年的月降雨量、月平均气温两种特征为学习样本,预测2019—2020年期间24个月的特征值,预测的干旱级别准确率为86.11%;相比于单一循环神经网络模型,VMD-CQPSO-GRU模型的预测误差降低了71.55%,可信度提高了132.06%。In view of the low prediction accuracy and poor reliability of current meteorological drought prediction methods,a meteorological drought prediction method based on machine learning is proposed.The method of variational mode decomposition(VMD)is used to decompose the drought time series signal into several intrinsic mode function.The improved algorithm of chaotic quantum particle swarm(CQPSO)is used to optimize hyperparameters of the gated recurrent unit(GRU)neural network.The“decomposition-synthesis”strategy is adopted to construct the VMD-CQPSO-GRU intelligent prediction model.Taking the two characteristics of monthly rainfall and monthly average temperature of Anyang,Zhengzhou and Xinyang from 1951 to 2018 as learning samples,the eigenvalues of 24 months from 2019 to 2020 are predicted,and the accuracy of drought level is 86.11%.Compared with the single cyclic neural network model,the prediction error of the VMD-CQPSO-GRU model is reduced by 71.55%,and the reliability is increased by 132.06%.

关 键 词：干旱预测 变分模态分解 深度学习 神经网络 时间序列 

分 类 号：TV124[水利工程—水文学及水资源] P338[天文地球—水文科学]