Enhancing Water Quality Prediction using Adaptive Neuro-Fuzzy Inference System with Neural Network Integration for Operational Safety

Accurately predicting water quality parameters plays a crucial role in managing wastewater treatment systems, helping to safeguard public health and ensure a safer working environment for plant operators. In this study, we introduce a hybrid model that merges the strengths of Adaptive Neuro-Fuzzy Inference Systems (ANFIS) with Artificial Neural Networks (ANN) to enhance the prediction of Biochemical Oxygen Demand (BOD)—a key indicator of water quality. By combining ANFIS’s ability to handle fuzzy, imprecise data with the learning power of ANN, the model is well-equipped to deal with the inherent uncertainties and nonlinear behavior of wastewater data. The model was developed and tested using daily data from an industrial sewage treatment plant (STP), capturing the dynamic and often unpredictable nature of real-world operations. Its performance was assessed using a range of evaluation metrics including MSE, RMSE, MAE, SSE, and the R² score. Results were impressive, with R² values reaching 0.9854 for training, 0.9850 for validation, and 0.9718 for testing. These outcomes underscore the model’s strong predictive capability and highlight its potential to support more efficient, reliable, and cost-effective wastewater management practices.