Análise de métodos de inteligência artificial na previsão de dados pluviométricos

Abstract

This study investigates the application of artificial intelligence methods in forecasting rainfall data, focusing on time series analysis and its implications for the energy sector. Three predictive models were evaluated: support vector machines (SVM), random forest and ARIMA, comparing their effectiveness in modeling climate patterns and energy consumption. The research used historical data on precipitation and electrical consumption, applying processing, normalization and statistical analysis techniques to validate the models. The results demonstrated that the SVM presented the best performance, with lower predictive error and greater ability to adapt to seasonal fluctuations and extreme weather events, such as the El Niño phenomenon. Random forest achieved an intermediate performance, while ARIMA showed difficulties in capturing non-linear patterns and brupt variations in the data. The comparative analysis reinforces the importance of advanced techniques for predicting energy consumption, showing that models based on machine learning are more effective than raditional statistical approaches. It is concluded that the application of artificial intelligence can improve energy management and assist in the formulation of strategies to mitigate the impacts of climate variables on electrical demand.