Framework for Optimizing Neural Network Hyper Parameters for Accurate Wind Production Forecasting

To address the escalating issue of burnt fossil fuels, there has been a persistent rise in the integration of Renewable Energy Sources (RES) within the energy production sector. However, this transition has led to a destabilization of the electrical grid due to an inherent mismatch between the production of renewable energy and the corresponding demand. Consequently, the accurate day-ahead forecasting of RES production has emerged as a pivotal component in energy planning and dispatch. Recognizing the efficacy of neural networks in production forecasting, this paper introduces a comprehensive framework aimed at optimizing the hyperparameters of these networks. The primary objective of this framework is to minimize the time required for model selection while ensuring an automated and generic approach to the modeling process. By fine-tuning the hyperparameters of neural networks, researchers and practitioners can enhance the precision of RES production forecasts. The optimization process allows for the identification of the optimal configuration, resulting in more accurate predictions. This, in turn, assists energy planners and operators in effectively managing the challenges associated with the integration of renewable energy into the electrical grid. The proposed framework offers a systematic and efficient method for determining the hyperparameters of neural networks. By reducing the time needed for model selection, it enables swift decision-making processes and facilitates the adoption of an automatic and generic approach to RES production forecasting. This research endeavors to contribute to the ongoing efforts to address the destabilization of electrical grids caused by the growing penetration of renewable energy sources.