Influent Quality and Quantity Predictionin Wastewater Treatment Plant: ModelConstruction and Evaluation
Rui Wang, Zhicheng Pan, Yangwu Chen, Zhouliang Tan, Jianqiang Zhang
Abstract
Influent quality and quantity were important factors that caused the abnormal operation of WWTP. In this study, the prediction models of influent quality and quantity were established based on four machine learning methods of Linear Regression, Ridge Regression, ElasticNet Regression and Lasso Regression. The meteorological conditions (precipitation and air temperature) and influent indicators (influent quantity, COD, and NH 3 -N) were used as training data. The influent quantity prediction of the models were evaluated using the historical data obtained from a WWTP located in western China, and the results showed that the normal rates of influent quantity were ranged from 98.9%-100%. The highest accuracy was obtained with Ridge method which was 86.19% .