Water Quality Prediction And Analysis Of Urban Sewage Treatment Plant Based On Intelligent Optimized Extreme Learning Machine

Water is an indispensable element for human life and production,which is the most important material resources for human survival and development.The accumulation of pollutants in water will have a serious influence on the survival of human beings and animals.Therefore,the water resources protection measures of China are also increasingly strengthened.The main task requirement of the sewage treatment is not only to meet the relevant national standards,but also to reduce the operation cost of the process.However,the process of wastewater treatment is often affected by some dynamic time-varying factors,thus showing some characteristics such as uncertainty,time-delay,complexity and non-linearity.In addition,some key water quality factors cannot be monitored online,which makes the control technology in the sewage treatment process unable to fully play its regulatory role in the first time.And it has become the shortcoming of the development of sewage treatment technology.Thus,this paper proposes the approaches to predict the inlet and the outlet water quality of the sewage treatment plant.In order to provide a reference for the sewage treatment process control system and reduce the operation cost as much as possible with ensuring the purification efficiency.In recent years,machine learning algorithm has been widely used in various fields,and also attracted the attention of a large number of researchers in the field of water quality prediction and analysis.It depends on its good generalization performance,low computational difficulty,fast learning speed and strong adaptive learning ability.In this paper,the main research task is to establish the inlet and outlet water quality prediction models of sewage treatment plant based on the extreme learning machine(ELM).Combined with previous research results,the main works is in the following aspects.First,analysis the database.The accuracy of sampling data from municipal wastewater treatment plants was analyzed,and preprocesses the data with corresponding data preprocessing technology.In addition,this paper researched water quality evaluation index correlation,to simplify the establishment of the models.Second,establishment the ELM model.The ELM is an easy-to-use and effective machine learning algorithm based on single hidden layer feedforward neural network(SLFNs).This method has the advantages of fast computation speed and strong generalization performance.In this paper,the ELM model was established to predict the water quality of sewage treatment plants.Thirdly,establishment of optimized ELM model.There is no explicit selection mechanism for the determination of random initial weights and thresholds of ELM networks,which makes it difficult to determine initial network parameters with good global performance.Therefore,genetic algorithm(GA)and particle swarm optimization algorithm(PSO)are proposed to optimize the ELM(GA-ELM and PSO-ELM),respectively.It enhances the stability of the algorithm and improves the prediction accuracy and generalization by choosing the optimal weight and threshold of the ELM.Fourthly,analysis of prediction results.The prediction model is established based on the water quality factors from the daily monitoring records of the urban sewage treatment plant in Chengdu,Sichuan province,and chemical oxygen demand(COD)are taken as examples.Root mean square error(RMSE),standard mean absolute error(MAE),mean absolute percentage error(MAPE)and determination coefficient(R~2)were used as evaluation criteria to analyze the results.The ELM model for predicting inlet COD values with RMSE,MAE,MAPE and R~2 were 3.318,2.437mg/L,1.621%and 0.950,respectively.The predict GA-ELM model and PSO-ELM model for predicting outlet COD values with MAE,MAPE,RMSE and R~2 were 0.700,0.474mg/L,4.073%,0.916,and 0.421,0.340mg/L,3.072%,0.975,respectively.The experimental results of the ELM model show that the high prediction precision and strong generalization ability,which could meet the inlet water quality prediction of the urban sewage treatment plant.And optimized the ELM model can overcome the shortcoming of traditional algorithm,which can effectively predict the outlet water quality of sewage treatment plant.It provides reliable reference for the sewage treatment control system and makes its operation more stable.