In this paper, a predictive control system based on an adaptive network-based fuzzy inference system (ANFIS) was employed to develop models for predicting and controlling the performance of a paper-making wastewater treatment process. The system includes an ANFIS predictive model and an ANFIS controller. In order to improve the network performance, fuzzy subtractive clustering, euclidean distance clustering, and principal component analysis (PCA) were used to identify model architecture and extract and optimize the fuzzy rule of the model. For the developed predictive model, when predicting, mean absolute percentage error (MAPE) lay 6.06% adopting ANFIS, root mean square normalized error (RMSE) was 24.4485 and R was 0.9731. The control model, taking into account the difference between the predicted value of chemical oxygen demand (COD) and the set point, can effectively change the additive dosages. In order to verify the developed predictive control model, a paper-making wastewater treatment process was picked up to support the operational performance. When the influent COD value or inflow flow rate was changed, the dosage could be accurately adjusted to make the effluent COD remain at the set point, and its MAPE was only 5.19%. The results indicated that reasonable forecasting and controlling performances had been achieved through the developed system.