Ciência de dados para identificação da relação entre variáveis climáticas e interrupções em redes de energia elétrica

Abstract

The growing demand for electrical energy and the need for increased reliability in distribution networks highlight the importance of improving failure detection and prediction methods. This study investigates the relationship between climatic variables and operational faults through exploratory data analysis techniques applied to two heterogeneous datasets: an operational dataset containing records of events in power networks, and a climatic dataset composed of meteorological records from INMET. After data cleaning and temporal–spatial integration, histograms, heatmaps, time series, and scatter plots were employed to identify patterns and correlations. The results indicate that precipitation, maximum temperature, and wind gusts are critical factors, with extreme events showing a stronger association with failures than moderate events. The heterogeneous spatial distribution of these events suggests regions of greater vulnerability, while the moderate correlation among climatic variables reinforces their combined influence. The analyses confirm that integrating operational and meteorological data is both feasible and informative, providing initial support for the future development of Machine Learning–based predictive models and for more effective preventive maintenance and power system management strategies.