Servicios Personalizados
Revista
Articulo
Indicadores
Citado por SciELO
Accesos
Links relacionados
Similares en SciELO
Compartir
Revista Técnica energía
versión On-line ISSN 2602-8492versión impresa ISSN 1390-5074
Resumen
GALLO, Angel; PEREZ, Fabián y SALINAS, Diego. Data Mining and Short-Term Projection of Power Demand in the Ecuadorian Electric System. Revista Técnica energía [online]. 2021, vol.18, n.1, pp.72-85. ISSN 2602-8492. https://doi.org/10.37116/revistaenergia.v18.n1.2021.461.
This article presents a computational tool developed in the Python programming language for data mining and short-term projection of the electrical power demand of the National Interconnected System (SNI), using the predictive approach of the Random Forest machine learning algorithm.
The implementation of the Hyperopt function to define the main hyperparameters of the Random Forest algorithm together with the application of feature engineering allows to fit a suitable machine learning model for the data series. This algorithm is implemented in tasks to mitigate missing values and outliers to structure complete databases free of deviations.
The procedure for data mining and demand projection shows the reliability and versatility of using the computational tool, obtaining relevant results, such as the reduction of anomalies in the data series to improve the precision in the projected electrical demand curves.
Palabras clave : Machine learning; Data mining; Electrical Power; Short-term load forecasting; National Interconnected System..