基于串级LSTM深度学习模型的二次供水余氯预测方法

Prediction Method of Residual Chlorine in Secondary Water Supply Based on Cascaded LSTM Deep Learning Model

下载PDF

导出

摘要伴随着城市高层住宅的增多,住宅区内二次供水泵房数量迅速增加。由于二次供水水箱位于城市供水系统末端,水质安全引起社会广泛关注。为提升水箱水质,一些泵房引入自动补氯装置,然而传统自动控制方法在应对二次供水系统中长时间延迟和非线性特性的补氯系统时存在局限性,仅能在线监测水箱余氯水平,过多的余氯可能对人体健康有害,因此,确保自动补氯系统安全运行成为亟待解决的问题。研究提出基于串级LSTM深度学习的神经网络模型,用于分析水箱余氯数据、准确预测水箱出水余氯浓度,并制定相应监测和控制策略。试验验证和实际应用结果表明,该深度学习模型能有效智能预测水箱余氯,为自动补氯系统提供重要的智能控制手段,具有实用意义。 With the increase of high-rise residential buildings in urban areas,the number of secondary water supply pump rooms in residential areas is rapidly increasing.As the secondary water supply tank is located at the end of the urban water supply system,water quality safety has attracted widespread attention from society.To improve the water quality in tanks,some pump rooms have introduced automatic chlorination devices.However,traditional automatic control methods have limitations in dealing with the long time delay and non-linear characteristics of chlorination systems in secondary water supply systems,as they can only monitor the residual chlorine level in tanks.Excessive residual chlorine may be harmful to human health,making it imperative to ensure the safe operation of automatic chlorination systems.This study proposed a neural network model based on cascaded LSTM deep learning to analyze residual chlorine data in tanks,accurately predict the residual chlorine concentration in tank water,and formulate corresponding monitoring and control strategies.Experimental validation and practical application results demonstrated that this deep learning model could effectively intelligently predict residual chlorine levels in tanks,providing important intelligent control means for automatic chlorination systems and holding practical significance.

作者肖磊李中伟刘书明陈春芳吴雪伍丽燕 XIAO Lei;LI Zhongwei;LIU Shuming;CHEN Chunfang;WU Xue;WU Liyan(School of Environment,Tsinghua University,Beijing 100084,China;Changzhou CGE Water Co.,Ltd.,Changzhou 213003,China)

机构地区清华大学环境学院常州通用自来水有限公司

出处《净水技术》 CAS 2024年第8期160-166,共7页 Water Purification Technology

基金国家水体污染控制与治理科技重大专项(2017ZX07201002)。

关键词二次供水水箱补氯 LSTM 深度学习余氯预测时间序列串级网络模型 secondary water supply water tank chlorination LSTM deep learning residual chlorine prediction time series cascade network model

分类号 TU991 [建筑科学—市政工程]