Automatic monitoring of treated water released from wastewater treatment plants using model-based clustering with density estimation

One of the most promising efforts to fight against the water scarcity threat is to reuse the treated water released from WasteWater Treatment Plants (WWTP). The objective of this paper is to propose an integrated approach for continuously evaluating the performance of wastewater treatment plants (WWTPs), with a focus on treated wastewater quality assessment and reuse of treated water for beneficial purposes like irrigation, aquarium, groundwater recharge, and in river water discharge based on pollution level in treated water. This paper implemented a model-based clustering with density estimation to generate the non-overlapped clusters to categorize the clusters. Cluster analysis using the Euclidean distance resulted in three clusters labeled under a specified category of water polluted: non-polluted, lightly polluted, highly polluted or slightly polluted. Unlike standard clustering algorithms like K-means, hierarchical that produce optimized clusters in statistical terms that deviate from naturally categorized clusters, model-based clustering with density estimation operates on the assumption that each data object originates from the mixture of underlying probability distributions. Water quality parameters like suspended solids (SS) have been considered for the analysis. Our experimental results conclusively show the polluted levels of wastewater from WWTP using a model-based clustering approach. The Dataset used in this work has been derived from the wastewater treatment plant located in Manresa, a town of 100,000 inhabitants near Barcelona (Catalonia). The plant treats a flow of 35,000 m3/day, mainly domestic wastewater, although wastewater from industries located inside the town is received in the plant too. In this research, the plant’s behavior over 527 days are under consideration. Model-based density clustering algorithm discovers 3 clusters, with half lying in size range of 14–89 and a maximum size of 352. With the help of natural clusters generated, our results show that out of 445 days, in 352 days, the treated water is almost non-polluted. By this, we can assess the performance of the wastewater treatment plant.