This research presents the novel methodology for atmospheric dispersion modeling of radioactive effluents in the case of nuclear accidents at nuclear power plants utilizing Artificial Neural Networks (ANNs). The primary objective is to enhance the accuracy and efficiency of emergency response strategies in the event of nuclear incidents. The study integrates conventional dispersion models’ dataset for nuclear accidents, meteorological parameters, and radiation measurements to train the ANN models. The methodology involves development of datasets of atmospheric dispersion modeling using coupled WRF-CALPUFF system, pre-processing the data, and training ANNs to predict the spread of radioactive contamination, evacuation times, and radiation doses.

Key findings indicate that the ANN models significantly outperform traditional models in terms of prediction accuracy and speed. The trained networks demonstrate lower mean absolute error (MAE) and higher correlation coefficients (R) compared to conventional models’ outputs when tested on unseen datasets. The research underscores the potential of ANNs in real-time decision support systems, capable of providing rapid and reliable predictions during nuclear emergencies.

Furthermore, the study explores advanced machine learning techniques, integration of geospatial data, and human behavior modeling to further enhance the predictive capabilities of the ANN models. The results highlight the practical applicability of ANNs in resource-constrained environments and emphasize the importance of ethical considerations in deploying predictive models in high-stakes scenarios.

This comprehensive approach provides a robust framework for developing predictive models tailored to specific types of nuclear incidents, ultimately contributing to improved emergency preparedness and response capabilities.
 

AI,Safety,Nuclear power plant,Artificial neural network,machine learning,Emergency response,atmospheric dispersion,emergency decision support system