Prediction of Residual Chlorine Variation in Secondary Water Supply Systems Based on Elman Neural Network Model
Chlorine is the most widely used disinfectant in urban drinking water systems,capable of effectively controlling bacterial growth in water.The residual chlorine in secondary water supply systems is a crucial index for ensuring the water quality safety of the"last mile".Establishing a predictive model for the variation of residual chlorine in secondary water supply systems to enable early warning is of significant importance for ensuring the safety of end-point drinking water.This study used correlation analysis to identify turbidity and temperature as indices strongly related to residual chlorine.Based on the 2022 data of residual chlorine,turbidity,and temperature from secondary water supply monitoring points in Shanghai,machine learning methods were applied to predict future residual chlorine levels.Three models were established using multiple indicators of current residual chlorine-turbidity,residual chlorine-temperature,and a single indicator of residual chlorine,and evaluated using certainty coefficient(R2),mean absolute error(MAE),mean bias error(MBE),and root mean square error(RMSE)metrics.The results showed that the relative errors of the three models were generally controlled below 10％,and all prediction models could meet the actual needs of residual chlorine prediction at secondary water supply monitoring points.The models were ranked in descending order of prediction accuracy as follows:residual chlorine-turbidity prediction model,residual chlorine self-prediction model,and residual chlorine-temperature prediction model.
machine learningprediction modeltime seriessecondary water supplyresidual chlorine
万方数据
维普
