Abstract: Total nitrogen in effluent is one of the critical indicators for evaluating the performance of wastewater treatment plants. A BP neural network model was developed to simulate the present nitrogen removal system for wastewater treatment, and an autoregressive integrated moving average (ARIMA) model was creatively applied to realize the short-term prediction of future effluent. The results showed that the simulation average relative error of BP model on training set was 15.9%, and that on test set was 16.5%,which revealed that the stability of model prediction was poor. The average error of the ARIMA model for predicting the total nitrogen value in the coming week was around 4.41%, which showed high prediction accuracy. The combination of the two models could help fast and efficient on-line detection of wastewater treatment plants.