A Soft Measurement Model of Wastewater Treatment Process Based on BP Neural Network with Improved Particle Swarm Optimization Algorithms

Considering that the BP neural network-based soft sensor model of wastewater treatment process is affected by the nonlinear characteristics of the system and has such problems as low convergence speed and local mini-mum, a hybrid soft measurement PSO-BP model for predicting the effluent chemical oxygen demand (COD_eff) and the effluent solid suspended matter (SS_eff) in the wastewater treatment process is developed based on BP neural network with improved particle swarm optimization algorithms. This model is compared with models based on genetic algorithm-BP neural network (GA-BP model) and BP neural network. The research results show that when using the PSO-BP model to predict COD_eff, the root mean square error (RMSE) and the coefficient of determination (R²) are 3.995 5 and 0.640 1, respectively. When it is used to predict SS_eff, RMSE and R² are 0.650 3 and 0.681 1, respectively. Compared with the BP model and the GA-BP model, the prediction performance of the PSO-BP model on COD_eff is improved by 4.49% and 0.44% respectively, the prediction performance of SS_eff is improved by 40.11% and 24.77% respectively.