摘要
针对传统河流水质预测模型预测精度较低,泛化能力弱的问题.本文在基于传统的时间序列模型进行水质预测的基础上,引入了LSTM神经网络,建立了ARIMA和LSTM组合模型以及SARIMA和LSTM组合模型用于河流水质预测的研究.结果表明,ARIMA和LSTM组合模型的预测精度比单一的ARIMA模型提高了约7%,SARIM和LSTM组合模型比单一的SARIM A模型的预测精度提高了约6%,比ARIM A和LSTM组合模型的预测精度提高了约2%.本文建立的组合模型算法,使得河流水质预测精度得到明显的提高,并且能够较好的应对复杂河流水环境的变化.
In view of the low prediction accuracy and weak generalization ability of the traditional river water quality prediction model.Based on the traditional time series model for water quality prediction,this paper introduces LSTM neural network,establishes ARIMA and LSTM combined model and SARIMA and LSTM combined model for river water quality prediction.The results show that the prediction accuracy of ARIMA and LSTM combination model is about 7%higher than that of single ARIMA model,the prediction accuracy of SARIMA and LSTM combination model is 6%higher than that of single SARIMA model,and 2%higher than that of ARIMA and LSTM combination model.The combined model algorithm established in this paper can improve the accuracy of river water quality prediction obviously,and can better cope with the changes of complex river water environment.
作者
胡衍坤
王宁
刘枢
姜秋俚
张楠
HU Yan-kun;WANG Ning;LIU Shu;JIANG Qiu-li;ZHANG Nan(University of Chinese Academy of Sciences,Beijing 100049,China;Environmental Resources Business Department,Shenyang Institute of Computing Technology,Shenyang 110168,China;Liaoning Ecological Environment Monitoring Center,Shenyang 110161,China;Fuxin Ecological Environment Protection Center,Fuxin 123000,China)
出处
《小型微型计算机系统》
CSCD
北大核心
2021年第8期1569-1573,共5页
Journal of Chinese Computer Systems
基金
国家水体污染控制与治理科技重大专项(2018ZX07601001)资助。
作者简介
胡衍坤,男,1994年生,硕士研究生,研究方向为神经网络和数据挖掘,E-mail:h19801235207@163.com;王宁,男,1974年生,研究员,CCF会员,研究方向为计算机应用;刘枢,男,1980年生,高级工程师,研究方向为环境监测;姜秋俚,男,1981年生,博士,高级工程师,研究方向为环境监测;张楠,女,1982年生,硕士,工程师,研究方向为环境监测。
