A Regression Model for Forecasting the Impact on Continuity Indicators and Its Influence on the Number of Critical Days

Objective: To evaluate the effectiveness of simple and multiple linear regression models in predicting the number of monthly critical days in an electric utility company in the state of Rio Grande do Sul, considering climatic variables as predictive factors. Method: The research used data on critical days of electrical interruption and climatic variables (precipitation, atmospheric pressure, temperature, winds, and humidity) collected between 2015 and 2020. Following a descriptive and correlation analysis, multiple linear regression models were applied, and the least significant variables were iteratively removed. Statistical tests assessed the normality, homoscedasticity, and independence of the residuals to ensure the model's adequacy. Results and Discussion: The adjusted models explained a low variability in critical days (adjusted R² up to 32%), with high error rates in predictions. The model identified a significant relationship between critical days and climatic variables (especially precipitation and winds) but with insufficient results for reliable forecasting. This suggests the need for more robust models and new variables to improve accuracy. Research Implications: Practical implications indicate the limitation of linear regression models for accurate predictions of critical days due to high data variability. Practically, it highlights the importance of seeking additional variables and alternative methodologies to support proactive planning and management of electric networks, while theoretical implications add a new perspective on the use of climatic variables as predictors of power interruptions, suggesting that while linear regression is common, it may have limitations. This encourages future investigations with nonlinear or hybrid models to improve predictions in this field. Originality/Value: The originality of the research lies in the attempt to directly correlate specific climatic indicators (precipitation, atmospheric pressure, and wind speed) with critical days, a topic still underexplored in energy reliability forecasts. This approach is innovative by investigating the impact of climatic variables on monthly interruptions, even recognizing that these linear regression models have limitations in capturing all variability of critical days. The relevance and value of the research are in highlighting how climatic factors can directly influence the reliability of the power grid, especially for energy distributors facing frequent interruptions due to adverse climatic events.