Energy Reports · 2024 · 22 citations · 34 references
ComputingData ScienceForecastingMachine LearningTime Series AnalysisEnsemble ForecastingForecasting For Renewable GenerationMultivariate ForecastingRenewable EnergyTime Series ForecastingWind ForecastingEnergyLong Short-term Memory NetworksMulti-input Multi-output ModelsMulti-input Multi-output SystemsMulti-step ForecastingMulti-step-ahead ForecastingSupport Vector RegressionWind Power Forecasting
Wind energy has gained more presence in Mexico, specifically in the Isthmus region of Oaxaca. Due to the intermittency of environmental conditions, predicting power generation across various wind farms in the area is essential for making informed decisions. However, there is currently a lack of strategies that provide energy predictions for wind farms in this region over a specific period, particularly using a multi-step forecasting approach. This paper proposes a methodology and implementation for forecasting energy generation in wind farms within the Isthmus region. The methodology includes stages for data analysis and exploration, preprocessing, configuring regression models, evaluation and simulation, and multi-step forecasting (24-hour period). Five regression algorithms were analyzed: Linear Regression (LR), Support Vector Regression (SVR), Multiple-SVR (M-SVR), General Regression Neural Network (GRNN), and Long Short-Term Memory (LSTM). Additionally, multi-step forecasting strategies such as recursive and Multi-Input Multi-Output (MIMO) were examined. Among these models, the LR and M-SVR models using the MIMO strategy yielded the best results in this study, achieving a Root Mean Square Error (RMSE) of 0.10 and a Mean Absolute Error (MAE) of 0.08. We also analyze daily forecasts to demonstrate the monthly model performance fluctuations during a whole year. Furthermore, the proposed model is based on actual wind conditions in the area, enhancing its effectiveness and feasibility. • Methodology for wind energy forecasting in the Isthmus region. • Machine and deep learning algorithms were configured using a heuristic strategy. • The MIMO multi-step ahead forecasting strategy yielded the best results in this study. • Daily forecasts demonstrated the monthly model performance fluctuations during a year.
34
Scikit-learn: Machine Learning in Python
Fabián Pedregosa, Gaël Varoquaux, Alexandre Gramfort et al. · arXiv (Cornell University) · 2012 · 63.7K citations · Full text
International Conference on Artificial Neural Networks
Habiboulaye Amadou Boubacar · Neural Networks · 1990 · 2.1K citations