A Signal Processing and Machine Learning Framework for Enhanced TDS Simulation

Total dissolved solids (TDS) is one of the most significant indicators of regulated wastewater effluents, and it is critical in assessing and improving the performance of any wastewater treatment facility. The literature review reveals that the pattern of TDS and other water quality parameters generally shows noisy and complicated features; hence, capturing their pattern is tough. This paper proposes a new approach to noise reduction by Ensemble Empirical Mode Decomposition (EEMD), incorporating machine learning algorithms like Artificial Neural Network (ANN) and Support Vector Machine (SVM) to improve the accuracy of TDS prediction. The data on temperature, pH, turbidity, and TDS were collected from the Bangabandhu Water Treatment Plant (BWTP) in Khulna, Bangladesh. In the current work, traditional and EEMD-based approaches were developed, and the performances of these were evaluated based on MSE, RMSE, and MAE. It is presented that the EEMD-based approach reduced the error considerably over the traditional models, while a maximum reduction of 89.46% and 91.04% in MSE are achieved in ANN and SVM, respectively. It further suggested that the proposed approaches could be applied to other wastewater treatment plants. Practical deployment of this model will help the operators make real-time decisions in optimized treatment processes.