Using Neural Networks to Develop a Database of Failures and Emergencies at Hydroelectric Power Stations

The article provides an example of employing a neural network and a natural language model to develop the database of failures and emergencies at hydroelectric power stations around the world that is available at JSC Vedeneev VNIIG. Using particular examples in conjunction with the t-SNE machine learning algorithm for visualization and the DBSCAN data clustering algorithm, the study shows an approach for enhancing the database. This technique enables a remarkable improvement in the selection of analog objects when justifying accident scenarios.