The Application of Genetic Algorithm Backpropagation Neural Network Model on the Prediction and Optimization of Wastewater Treatment System

Abstract

Prediction and optimization on water quality parameters (WQPs) have become more and more important to the wastewater treatment system (WWTs). In this study, the genetic algorithm backpropagation neural network model (GA-BPNN) had been used to predict and optimize WQPs of a low-strengthen complex wastewater treatment system (LSCWWTs). Results showed that the correlation coefficients between the predicted values and measured values were R2 =0.946 for COD, R2=0.962 for BOD, R2=0.933 for TN, R2=0.985 for NH3-N, R2=0.969 for TP, and R2=0.968 for SS, indicating the predictive values by the GA-BPNN model well fitted the mesured values of effluent WQPs. The optimal effluent WQPs were COD=27.6mg/L, BOD=7.1mg/L, TN=5.4mg/L, NH3-N=0.9mg/L, TP=0.11mg/L and SS=9.25mg/L, respectively. And the corresponding operating parameters were MLSS=3045.4mg/L, MLVSS=2405.9mg/L, T=23.2 °C, R=1.4, SRT=12.5d, HRT=17.3h, CODin =643.3mg/L, BODin=342.2mg/L, TNin=54.2mg/L, NH3-Nin=45.3mg/L, TPin=4.9mg/L, SSin=452.6mg/L, which could be beneficial to the operation optimization of LSCWWTs.