基于PIWT-IPSO-BP的污水厂出水COD含量的预测模型  

Predictive Model of COD Content in Wastewater Plant Effluent Based on PIWT-IPSO-BP

作　　者：张净[1] 窦慧芸 蒋武 刘晓梅 ZHANG Jing;DOU Hui-yun;JIANG Wu;LIU Xiao-mei(School of Electrical and Information Engineering,Jiangsu University,Zhenjiang 212013,Jiangsu Province,China;Zhenjiang Water Industry Limited Liability Company,Zhenjiang 212000,Jiangsu Province,China;Jiangsu Kemo Information Technology Co.,Zhenjiang 212000,Jiangsu Province,China)

机构地区：[1]江苏大学电气信息工程学院,江苏镇江212013 [2]镇江市水业有限责任公司,江苏镇江212000 [3]江苏科茂信息技术有限公司,江苏镇江212000

出　　处：《中国农村水利水电》2024年第9期15-20,28,共7页China Rural Water and Hydropower

基　　金：国家重点研发计划项目(2019YFC1606600)

摘　　要：在农业灌溉的领域中,化学需氧量(Chemical Oxygen Demand,COD)的测定是衡量水体中有机物污染程度的一个重要指标。当COD浓度超过60mg/L时,其对土壤质量和农作物的生长产生的负面影响成为不容忽视的问题。这一现象可能会严重影响农作物的产量和质量,进而对农作物生产的可持续性构成挑战。因此,有必要精确预测污水处理厂出水COD浓度的变化趋势,从而促进其在农业灌溉中的有效应用。研究结合了改进的小波变换、改进的粒子群优化(Improved Particle Swarm Optimization,IPSO)算法和反向传播BP(Back Propagation,BP)神经网络作为预测模型。鉴于COD受到众多因素的影响,这些因素之间存在复杂的耦合关系,采用PCA进行特征提取。考虑到数据采集的过程中不可避免的噪声干扰,应用小波降噪对原始数据进行处理,以确保数据质量,提高模型准确性。在此基础上,基于BP神经网络算法构建污水处理厂出水COD的预测模型。为了解决BP神经网络参数选择可能遇到的盲目性问题,引入改进的粒子群算法对模型进行参数优化,以提高预测精度。实验结果表明,提出的PIWT-IPSO-BP模型预测效果良好,其平均绝对误差、均方根误差和决定系数分别为0.222、0.386和0.984。该模型在一定程度上改善了数据噪声、多因子制约等问题,为污水循环利用技术应用于农业灌溉方面提供了参考依据。In the realm of agricultural irrigation,the assessment of chemical oxygen demand(COD)stands as a crucial parameter,serving as a significant indicator of the level of organic matter pollution within water bodies.When the concentration of COD surpasses the threshold of 60 milligrams per liter,the adverse effects that this elevation incurs upon the quality of soil and the subsequent growth of agricultural crops manifest as a significant concern that demands attention.This phenomenon harbors the potential to exert a substantial and profound impact on both the quantity and the quality of agricultural crop yields,thereby posing a formidable and daunting challenge to the enduring sustainabili-ty and viability of crop production practices.Consequently,it is imperative to precisely anticipate the trajectory of COD concentration trends in the effluent discharged from wastewater treatment facilities,thereby facilitating its effective utilization in agricultural irrigation practices.This research combines the augmented wavelet transform technique with an improved particle swarm optimization(IPSO)algorithm as well as a sophisticated back propagation neural network to construct a combinatorial prediction model.Considering the multitude of factors influ-encing COD and the intricate interrelationship among these factors,principal component analysis(PCA)is employed for comprehensive fea-ture extraction.Recognizing the inevitable noise interference during data acquisition,wavelet noise reduction techniques are implemented to preprocess the raw data.This ensures data quality and enhances model accuracy.On this basis,the prediction model of COD in effluent of sewage treatment plant is constructed based on BP neural network algorithm.To overcome the potential issue of blindness inherent in parame-ter selection for the BP neural network,an improved particle swarm algorithm is introduced to optimize the parameters of the model to im-prove the prediction accuracy.The empirical findings confirm the exemplary predictive effectiveness of t

关 键 词：化学需氧量 预测模型 小波变换 粒子群优化算法 BP神经网络 

分 类 号：TV11[水利工程—水文学及水资源] X832[环境科学与工程—环境工程]