Assessment and modeling of benzene micropollutant in surface waters proximal to coal-fired thermal power plants

Micropollutants of priority have gained significant attention due to their rising concentrations and persistence in aquatic environments. Among these significant micro-pollutants is benzene, released into the environment because of mining activities, thermal power plants, and certain industrial operations, thereby contributing to environmental contamination. This study monitored benzene concentrations in the Orhaneli River over a year, employing an artificial neural network (ANN) approach to formulate a predictive model for benzene micropollutant concentrations. To provide input to the ANN, Principal Component Analysis (PCA) was employed, utilizing parameters such as precipitation, temperature, humidity, pH, flow rate, dissolved oxygen, conductivity, total organic carbon, and total nitrogen. Eleven backpropagation algorithms were compared, with the Levenberg-Marquardt algorithm identified as the optimal training algorithm. For the selected training algorithm, an optimized number of 6 neurons in the hidden layer provided the best predictions for benzene concentrations. The three-layered optimized ANN model achieved an impressive correlation coefficient (R2) of 0.9889.