Detection of pollutant source in groundwater using hybrid optimization model

摘要

Abstract Detection of groundwater pollution source is an inverse problem. To solve this inverse problem, it has been posed into an optimization problem. In this study, a hybrid optimization model has been developed for detection of groundwater pollution sources in terms of its source characteristics, namely, source location, source strength and duration of activity of pollution source. In this hybrid optimization model, the Genetic Algorithm model has been linked with the Gradient Descent optimization model. The global convergence property of Genetic Algorithm has been utilized to find the optimal solution near global minima. This solution is then used as an initial guess to the Gradient Descent optimization model to find the global minima. The performance of developed hybrid model has been evaluated for two-dimensional case for error free and erroneous concentration data. Performance results show a significant improvement in the prediction error of groundwater pollution source parameters. The prediction error using GA model was found to be equal to 10.806%, 13.930% and 25.211% in source location, duration of activity and source strength, respectively, while using hybrid optimization model the prediction error were improved to 0.461%, 7.178% and 9.999%. The novelty of the present work is that it requires the observed concentration data of one observation well only for complete identification of pollution source. Also, a prior knowledge of probable locations of potential pollution sources is not required in this model.