THE APPLICATION OF GENETIC ALGORITHM TO OPTIMIZE XGBOOST ALGORITHM IN MBR SIMULATION

Abstract

Membrane bio-reactor (MBR) is one of the mainstream technologies of modern wasted-water treatment, but membrane fouling is a key factor which restricts the development of MBR. The formation of membrane fouling directly leads to a decrease in membrane flux. As one of the important parameters to measure membrane fouling, membrane flux is the focus and difficulties of membrane fouling research. In this paper, XGBoost algorithm is used to simulate and predict MBR membrane flux. In view of the shortcomings of XGBoost with many parameters, combined with the characteristics of genetic algorithm (GA), such as strong global search ability and fast convergence speed, the parameters of XGBoost are optimized and adjusted. By analyzing the prediction results of GA-XGBoost algorithm, and comparing with the experimental data, the results show that GA-XGBoost neural network prediction model is superior to XGBoost algorithm prediction model in the prediction of MBR membrane flux, with higher accuracy.