Application of GA-PSO-LSTM Model in Membrane Fouling Prediction in Wastewater Treatment Process

Membrane bioreactor (MBR), as a new wastewater treatment and water recycling technology, is widely used in the field of water treatment. However, membrane contamination is an unavoidable problem in MBR operation, which seriously affects its utilization. Membrane permeability is an important indicator of membrane contamination, and the researcher proposed a membrane permeability prediction model based on long short-term memory neural network (LSTM). Firstly, the researcher used principal component analysis (PCA) to determine the process variables related to membrane permeability based on the actual operation data. Then, the membrane permeability prediction model was established based on the LSTM neural network, which has many parameters that need to be set, and the researcher used genetic algorithm (GA) to optimize these parameters, and the GA-LSTM model was trained and tested. The results showed that the model was able to predict the membrane water permeability with an RMSE of 0.78958. However, the convergence speed of the model was slow. However, the convergence speed of the model is slow. For this reason, the researcher introduced the particle swarm algorithm (PSO), and designed a hybrid algorithm of GA and PSO to optimize the initial parameters of LSTM, and established the GA-PSO-LSTM model for predicting the membrane water permeability. The results show that GA-PSO-LSTM has higher prediction accuracy than GA-LSTM, the RMSE of the model is only 0.21007, and the convergence speed is faster. GA-PSO-LSTM can accurately predict the membrane contamination, so that the maintenance personnel can carry out the maintenance and cleaning in time to prolong the life of the membrane, and to improve the effect of the application of membrane technology.