Desarrollo de un modelo estadístico basado en la correlación de Pearson para anticipar condiciones precursoras de fallos eléctricos en un sistema de distribución de prueba representativo

Abstract

The continuous growth in electricity demand and the increase in load capacity have increased the likelihood of failures in electrical distribution systems. In this study, a statistical model based on Pearson's correlation was developed to predict electrical failures in the IEEE 33-bus test system. The system was modeled within OpenDSS, and various operational and failure scenarios were simulated to generate an electrical database. Subsequently, the recorded electrical variables were analyzed to apply Pearson's correlation, selecting those with the strongest statistical relationship to the occurrence of failures. The highest positive correlation was found in the maximum current (0.1646) and active losses (0.1645). A logistic regression model was implemented to evaluate the model's ability to anticipate failures. The results showed a 100% recall for the failure class, an ROC curve with an AUC of 0.40, and overlapping detection probabilities. This demonstrates that the model has a limited ability to differentiate between certain operating conditions but shows that this methodology is feasible as a complementary tool for preventive maintenance planning.