Juan Hernández Arrieta, Idi Isaac Millán, Javier Sierra Carrillo
Non-conventional renewable energies, such as solar, have great potential for generation, however, there is no accurate and timely information for making decisions that lead to the installation of microgrids in interest zones. For this, an ensemble model has been developed to forecast the main operating variables associated with the distributed resources of potential microgrids. To develop the model, a comprehensive technological surveillance of the strategies, models and techniques used for the valuation and forecast of photovoltaic generation and energy storage resources was carried out, then the model K Nearest Neighbors (KNN) was referenced, selected and programmed for the treatment (purification and imputation) of the data; The energy models for these systems were referenced and selected, using the HOMER PRO software model, where the main operating variables that are the object of forecasting were determined for each of the selected distributed resources. The graphical interface of the ensemble model was programmed for the short-term forecast of these variables, using the selected artificial intelligence forecasting techniques (decision trees, k Nearest Neighbor and Random Forests), and finally the results obtained were validated with the model, versus the commercial software HOMER PRO and another artificial intelligence model (Artificial Neural Network - Multilayer Perceptron).
Generation forecasting, forecasting techniques, artificial intelligence, microgrid, statistical models, assembly model
Cite this paper
Juan Hernández Arrieta, Idi Isaac Millán, Javier Sierra Carrillo. (2022) An Ensemble Model Based in AI Using Past Output Data for Forecast of Operating Variables of Distributed Resources. International Journal of Power Systems, 7, 101-117