Monitoring effluent quality of wastewater treatment plant by clustering based artificial neural network method

The artificial neural network (ANN), as a data-driven approach, is a powerful tool for forecasting effluent quality of wastewater treatment. However, selecting appropriate input variables is a major challenge in developing ANN models. Recent studies in various fields have highlighted the usefulness of different clustering methods in identifying appropriate input variables, which, however, has largely been unexplored in classifying wastewater quality parameters. This study was carried out to fill this knowledge gap. Three ANN models were developed with different clustering methods, to forecast effluent quality of Tabriz city's wastewater treatment plant. Model A used principal component analysis (PCA) for input selection, model B used those variables identified by non-linear mutual information (MI) measure. In model C, the self-organizing map (SOM) method was used as an artificial intelligence (AI)-based method to cluster data and impose the representative parameters of each cluster as inputs of ANN. Model C presented a more favorable and optimal ANN structure in comparison with models A and B and showed up to 8 % and 23% increment in determination coefficient (DC) efficiency criterion respectively. While the number of parameters involved in the wastewater treatment process are quite many, the proposed model by employing an AI-based clustering method could successfully predict the effluent quality using the minimum number of essential input parameters. Thus, this study highlights the superiority of the SOM technique in selecting dominant input variables for ANN modeling of WWTP efficiency performance, not only because of the enhanced performance of the model with respect to various indicators but also because such a superior result was achieved by an optimal ANN architecture.