基于机器学习的海水混凝建模与应用  

Machine Learning Based Modeling of Seawater Coagulation and Its Application

作　　者：钟骅 朱力[2] 方昱杰 金伟剑[2] 许力[1] ZHONG Hua;ZHU Li;FANG Yujie;JIN Weijian;XU Li(College of Electrical Engineering,Zhejiang University,310013;Hangzhou Water Treatment Technology Development Center,310012,Hangzhou,China)

机构地区：[1]浙江大学电气工程学院,310013 [2]杭州水处理技术研究开发中心有限公司,浙江杭州310012

出　　处：《水处理技术》2021年第4期101-105,共5页Technology of Water Treatment

基　　金：国家重点研发计划项目(2017YFC0403701)。

摘　　要：海水混凝过程的建模对海水淡化的水质控制和加药控制有重要意义。针对海水混凝大滞后的特点,提出一种基于机器学习序列模型的建模方法。采用门控循环单元(Gate Recurrent Unit,GRU)编码器和线性网络解码器等结构化单元,建立出水浊度与入水浊度、流量、絮凝剂及助凝剂用量等因素的关系模型。基于实际运行数据的数值实验验证了所建模型的有效性,也为絮凝剂及助凝剂的减量控制提供了可靠的依据。The model of seawater coagulation process is of great significance for water quality control and dosing control of seawater desalination.A modeling method based on machine learning and sequence modeling is proposed because of the characteristics of large time delay and time-varying of seawater coagulation.Using structured units such as GRU and linear network decoder,a model between turbidity of inflow,flow rate,flocculant and dosage of coagulant aid was established.The numerical experiments based on the actual operation data verify the validity of the model,and provide a reliable basis for the reduction of flocculant and coagulant aids.

关 键 词：海水淡化 预处理 混凝沉淀 序列模型 机器学习 

分 类 号：P747[天文地球—海洋科学]