Research On Data-driven Sewage Treatment Control Application Based On Machine Learnin

At present,the social demand for water resources is increasing,the sewage discharges and the types of pollutants are increasing,and the water environment is seriously polluted,which causes a serious negative impact on the life safety of the residents and the efficiency of industrial production.As part of environmental protection,the importance of sewage treatment is obvious.In addition,promoting the development of informatization and intellectualization of society is the key concern of builders and researchers,and is also a necessary condition for promoting the development of Industry4.0.With the sustainable development of Smart City and the continuous progress of information technology,many smart products are emerging,as a vital part of Smart Water,intelligent management and control of sewage treatment are highly recommended as a great way to effectively control sewage treatment.Aiming at the safety,stability,energy saving,consumption reduction,and intelligence of water plants,and taking the actual scenarios as the research objects,this paper applies the new-generation information technology and control technology including Machine Learning(ML),Internet of Things(Io T)sensor technology and Model Predictive Control(MPC),makes full use of the unique advantages of the data itself,and devotes to realizing the data-driven intelligent management and control of sewage treatment.The main work of this paper is as follows:(1)In the sewage treatment,Chemical Oxygen Demand(COD)is an important reference indicator to reflect its treatment effect and water quality.At present,the intelligent research of the small-scale sewage treatment plants has not paid enough attention and most water quality prediction models have high feature requirements,this study takes a small-scale fine chemical plant in Qingdao as the research site,effectively combines many sensors,and develops a sewage management platform based on Io T sensor technology to realize the information management of the treatment process and acquire multi-dimensional data in different treatment stages.Since the changing trend of outflow COD is highly related to operation status,this study uses different types of temperature and water inflow data as model inputs without referring to any water quality data and applies three ML models to predict the future trend of outflow COD,which are Long Short-Term Memory(LSTM)neural network,Support Vector Regression(SVR),and Gated Recurrent Unit(GRU)neural network.The experimental results show that the COD trend prediction model built in this study is effective,and the GRU model performs better than the LSTM and SVR and can be better used as algorithmic support for the plant.This study effectively reduces the feature requirement for water quality prediction models,facilitates the modeling of water plants,and provides a new research idea for the trend prediction of outflow COD.(2)In order to effectively solve a series of key problems existing in the field of sewage treatment control,for example,the factors considered by intelligent control are relatively simple,the scenarios targeted by intelligent control are mostly existing simulation environments and it less combined with actual scenarios,expert knowledge,and water quality prediction,this study effectively combines the LSTM-GRU combination model with the Self-Attention mechanism and applies it to the Behavioral Cloning(BC)model based on a lot of existing expert knowledge.And a new sewage treatment predictive control model is proposed,which is called a Multi-output Behavioral Cloning model based on LSTM-GRU and Self-Attention(MOBC/LG-SA)in this paper.Considering the more urgent control demands of large-scale sewage treatment plants,this study takes a large-scale sewage treatment plant in Rizhao as the research site and conducts experiments with urban domestic sewage as the research object.The experimental results show that in the actual sewage treatment scenario,compared with the traditional single model,the MOBC/LG-SA model has the better performance,which can better achieve the water quality prediction and operation control of this plant,so as to achieve the predictive control of sewage treatment,ensure its water quality meets the discharge requirement.The model can effectively reduce safety risks and costs,improve the operation level,so as to improve the sewage treatment efficiency and ensure the stability of sewage treatment.