In this work, genetic algorithm coupled with neural networks was used to predict the concentration of pollutants in the Yassa area. The model takes into account the meteorological parameters of the study area and the source over 5 years, period from January 2017 to December 2021. To evaluate the model, two indices are used to indicate the prediction model’s performance; the Squared Correlation Coefficient R2 whose value in the test case is 0.9996 and the Mean Square Error (MSE) whose best value is 0.0044756. We also evaluated the optimization methods and found that compared to the swarm by particle optimization, the genetic algorithm gives a better “fitness function curve” as a function of the number of iterations. The results showed that the GA-ANN coupling presented here is able to provide a reliable and accurate prediction of the distance of the area where the air quality standards and criteria are met. The safe distance where thermal power plant activity respects the air quality criteria is beyond 2200m. The results help not only in the choice of the location of the thermal power plant but also in the monitoring of the air quality in residential areas.