基于VMD-PSO-LSTM水质预测模型的应用研究  

Research on application of water quality prediction model based on VMD-PSO-LSTM

作　　者：顾乾晖 曾斌 涂振宇[1] GU Qianhui;ZENG Bin;TU Zhenyu(School of Information Engineering,Nanchang Institute of Technology,Nanchang 330099,China;Flood Control Information Center of Jiangxi Province,Nanchang 330096,China)

机构地区：[1]南昌工程学院信息工程学院,江西南昌330099 [2]江西省防汛信息中心,江西南昌330096

出　　处：《南昌工程学院学报》2021年第4期23-29,共7页Journal of Nanchang Institute of Technology

基　　金：江西省水利厅科技项目(KT201639);江西省科技厅重点研发项目(20151BBE50077)。

摘　　要：河道水质预测精度的准确性有利于解决水污染防治和水质监管等问题。基于变分模态分解(VMD)在处理非平稳序列上的优势、长短时记忆神经网络(LSTM)在处理长时间序列的非线性问题上的优势以及粒子群算法(PSO)能自适应寻优的优势,提出了一种基于VMD-PSO-LSTM模型的河流水质预测方法。以河南南阳当地2020年5月—2020年10月的水质中高锰酸盐数据为验证数据,对水质评价指标中的高锰酸盐含量进行预测。通过与PSO-LSTM和SVR等模型进行对比,实验结果表明,VMD-PSO-LSTM水质预测的结果精确度最高,且具有更快的收敛速度。The accuracy of river water quality prediction is conducive to solving the problems of water pollution prevention and water quality supervision.Based on the advantages of Variational Modal Decomposition(VMD)in processing non-stationary sequences and the advantages of Long and Short-term Memory Neural Networks(LSTM)in processing nonlinear problems in long-term sequences,as well as the ability of Particle Swarm Optimization(PSO)to adaptively search excellent advantages,this paper proposes a method of river water quality prediction based on VMD-PSO-LSTM model.The permanganate content in the water quality from May 2020 to October 2020 in Nanyang,Henan Province was used as the verification data to predict the permanganate content in the water quality evaluation index.Through comparison with models such as PSO-LSTM and SVR,the experiment shows that the results of VMD-PSO-LSTM water quality prediction have the highest accuracy and fastest convergence rate.

关 键 词：水质预测 LSTM神经网路 神经网络 PSO VMD 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]