Estimation of treated wastewater quality is a non-linear problem and cannot be solved statistically. In this respect, it is of great importance to create a new method for predicting treated wastewater quality and usability rate. The main objective of the study is to evaluate the suitability of machine learning modelling as a valid input-output tool to predict some outlet parameters in Kırklareli Advanced Biological Wastewater Treatment Plant (K-WWTP), Thrace Region of Turkey, where water scarcity is an important issue. Accordingly, several machine-learning techniques were tested to assess the usability of wastewater quality for irrigation in the case study area.
[...]
This target combined three artificial intelligence models (Artificial Neural Network - ANN, Adaptive Neuro-Fuzzy Inference System – ANFIS, and Fuzzy Logic-Mamdani - FLM) using various scenarios based on inlet and outlet water quality parameters of K-WWTP measured daily during the last three years. At the same time, an agricultural projection was realized to assess reuse potential of municipal treated wastewater for each selected crop. The best performances were observed with the ANNs model with R2 of 0.83, 0.96, 0.94, 0.80, 0.80, 0.74 and 0.85, and mean squared error (MSE) of 0.001020, 0.000591 μS/cm, 0.000526%, 0.004606 mg/l, 0.007718 mg/l, 0.009034 mg/l, 0.006684 mg/l for pH, Conductivity, Salinity, COD, Total N, Total P, TSS, respectively. The usability of wastewater quality varied between 69% and 72% during the irrigation season, depending on the artificial intelligence models. It was found that approximately 35% of the 20-thousand hectares of agricultural area can be irrigated with treated wastewater. Therefore, the presented method may provide a reliable and effective reference to assess the usability of wastewater use in agriculture.