基于小波变换预处理的MLP溶解氧预测模型  

MLP dissolved oxygen prediction model based on wavelet transform preprocessing

作　　者：傅炎强 王兆华 帅冬生[3] FU Yanqiang;WANG Zhaohua;SHUAI Dongsheng(School of Mathematics and Computer Sciences,Nanchang University,Nanchang 330031,China;Jiangxi Vocational and Technical College of Finance and Economics,Jiujiang Jiangxi,332000,China;Jiangxi Zhonggan Investment Survey and Design Limited Company,Nanchang 330031,China)

机构地区：[1]南昌大学数学与计算机学院,江西南昌330031 [2]江西财经职业学院,江西九江332000 [3]江西省中赣投勘察设计有限公司,江西南昌330031

出　　处：《南昌大学学报(理科版)》2025年第1期82-87,共6页Journal of Nanchang University(Natural Science)

基　　金：校企合作项目《智慧水环境监测技术研究(HX202109040001)》。

摘　　要：溶解氧的精确预测对于河流污染预防具有关键作用,但现有模型没有充分考虑数据自身具有的时序特性和基于对数据进行有效预处理的角度来提升溶解氧预测的精度。针对上述问题,提出了一种基于小波变换(wavelet)和多层感知器(MLP)的深度学习模型(wavelet_MLP),该模型首先对数据进行三层小波分解的预处理,得到三个低频子序列和一个高频子序列,并对这四个子序列进行数据重构;其次利用多层感知器(MLP)对每个子序列分别进行预测;最后将所有预测得到的子序列整合再预测输出最终结果。模型在江西丰城小港口站点和江西肖江江口站点数据集上与其他八种模型相比在四个指标上总体最优,这验证了所提出模型是有效的。The accurate prediction of dissolved oxygen plays a crucial role in preventing river pollution,but existing models have not fully considered the temporal characteristics of the data itself and improved the accuracy of dissolved oxygen prediction from the perspective of effective data preprocessing.A deep learning model based on wavelet transform(wavelet)and multilayer perceptron(MLP)is proposed to address the above issues.The model first preprocesses the data with three-layer wavelet decomposition to obtain three low-frequency subsequences and one high-frequency subsequence,and then reconstructs the data for these four subsequences;Secondly,using multi-layer perceptron(MLP)to predict each subsequence separately;Finally,integrate all predicted subsequences and predict the final result.The model was generally optimal in four indicators compared to the other eight models on the datasets of Fengcheng Small Port Station and Xiaojiang Jiangkou Station in Jiangxi,which verifies the effectiveness of the proposed model.

关 键 词：时间序列预测 溶解氧预测 小波变换 多层感知器 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]