基于CA-GRU的污水处理厂出水总氮浓度预测研究  

Research on Prediction of Total Nitrogen Concentration in Wastewater Treatment Plant Effluent Based on CA-GRU

作　　者：吴婧 廖明潮 WU Jing;LIAO Mingchao(College of Mathematics and Computer Science,Wuhan Polytechnic University,Wuhan 430000,China)

机构地区：[1]武汉轻工大学数学与计算机学院,湖北武汉430000

出　　处：《自动化仪表》2024年第4期97-100,105,共5页Process Automation Instrumentation

摘　　要：为了精确预测污水处理厂出水总氮浓度,以呼玛县某污水处理厂公开监测的污水出水水质数据为样本进行了研究。提出了一种基于卷积注意-门控循环单元(CA-GRU)网络的混合模型。首先,使用时间滑动窗口,将数据转换成连续的特征图以作为输入,并从中提取抽象特征。然后,将这些特征映射到网络模型中。最后,通过门控循环单元(GRU)网络模型获得预测值。试验结果显示,CA-GRU模型的均方根误差(RMSE)为0.172,平均绝对百分比误差(MAPE)为0.010。该结果比GRU网络模型低0.108、0.016,比卷积神经网络(CNN)-GRU模型低0.027、0.005,比Attention-GRU模型低0.065、0.007。该结果表明,CA-GRU模型预测效果良好,利用CNN等模型有利于减少冗余信息的干扰。CA-GRU模型能够充分提取污水水质数据在时间和空间上的特征、更准确地预测出水水质总氮含量,具有较高的应用价值。To accurately predict the total nitrogen concentration of effluent from a wastewater treatment plant,the publicly monitored effluent wastewater quality data from a wastewater treatment plant in Huma county is used as a sample for the study.A convolutional attention-gated recurrent unit(CA-GRU) network hybrid model is proposed.Firstly,using a time sliding window,the data is converted into a continuous feature map as input,from which abstract features are extracted.Then,these features are mapped into a network model.Finally,the gated recurrent unit(GRU) network model is used to obtain the predicted values.The experimental results show that the root mean square error(RMSE) of the CA-GRU model is 0.172 and the mean absolute percentage error(MAPE) is 0.010.This result is lower than the GRU network model by 0.108,0.016,lower than the convolutional neural network(CNN)-GRU model by 0.027,0.005,and lower than the Attention-GRU model by 0.065,0.007.This result shows that the CA-GRU model prediction effect is good,and the use of models such as CNN is conducive to reducing the interference of redundant information.The CA-GRU model can fully extract the characteristics of wastewater water quality data in time and space,and can more accurately predict the total nitrogen content of effluent water quality,which is of high value for application.

关 键 词：污水 出水总氮浓度预测 混合模型 门控循环单元 卷积神经网络 时间滑动窗口 

分 类 号：TH-3[机械工程]