基于CNN-LSTM的工业出水水质预测模型
Prediction Model of Industrial Effluent Quality Based on CNN-LSTM
杨潞霞 1王梦冉 1林兴亮 1付一政 2王智瑜1
作者信息
- 1. 太原师范学院计算机科学与技术学院,山西晋中 030619
- 2. 中北大学材料科学与工程学院,山西太原 030051;山西省煤矿矿井水处理技术创新中心,山西太原 030006
- 折叠
摘要
工业废水含有多种污染物,提前预测工业废水水质从而快速对其进行相应处理具有重要意义.为此,研究提出了一种新的卷积神经网络(convolutional neural network,CNN)和长短期记忆网络(long short-term memory,LSTM)融合的工业废水水质污染物指标预测模型(CNN-LSTM).为了更好地捕捉工业废水数据的时序性和动态性,模型设置了多个滑动窗口.使用CNN算法将时间序列数据进行高维特征提取,利用LSTM模型学习时间序列数据的时序特征,建立CNN-LSTM工业废水预测模型,并对废水水质中的化学需氧量(CODCr)、氨氮、总磷(TP)3 项指标进行预测分析.结果表明,与CNN和LSTM两个基准模型相比,CNN-LSTM预测模型的平均绝对值误差率(MAE)和均方误差率(MSE)均较小,预测效果较优.该模型能较好地实现对工业废水出水水质的准确预测,可为工业废水水质的在线监测和精准控制提供有效的、可行的技术支持和决策依据.
Abstract
Industrial wastewater contains a variety of pollutants,so it is of great significance to predict the quality of industrial wastewater in advance so as to treat it quickly.For this reason,a new predictive model(CNN-LSTM)for industrial wastewater quality pollutants based on the fusion of convolutional neural network(CNN)and long short-term memory(LSTM)was proposed in this paper.In order to better capture the time sequence and dynamics of industrial wastewater data,multiple sliding windows were set up in the model.CNN algorithm was used to extract high-dimensional features of time series data,and LSTM model was used to learn the time series features of time series data.CNN-LSTM industrial wastewater prediction model was established,and three indexes of biological oxygen(CODCr)content,ammonia nitrogen content and total phosphorus(TP)content in wastewater quality were predicted and analyzed.The results showed that the mean error rate(MER)and mean square error rate(MSE)of CNN-LSTM model were smaller than those of CNN and LSTM model.The model can accurately predict the effluent quality of industrial wastewater,and can provide effective and feasible technical support and decision-making basis for on-line monitoring and precise control of industrial wastewater quality.
关键词
卷积神经网络(CNN)/长短期记忆网络(LSTM)/工业出水水质预测/滑动窗口方法/预处理/归一化Key words
convolution neural network(CNN)/long short-term memory(LSTM)/industrial effluent quality prediction/sliding window method/pretreatment/normalization引用本文复制引用
基金项目
山西省重点研发计划(202102010101008)
山西省研究生教育改革研究项目(2023JG163)
出版年
2024
国家科技期刊平台
NSTL
万方数据