RT info:eu-repo/semantics/bachelorThesis
T1 Técnicas de inteligencia artificial para la predicción de energía renovable
A1 Ramírez Cabrera, Aitor Lorenzo
AB A greater social concern about climate change and the future of our planet togetherwith efficiency improvements in producing renewable energies have made this a risingsector in the last decade. Thus, to estimate the daily produced energy is important for itsintegration into the electric system, especially for isolated systems as the Canary Islands.The main problem in the forecasting of the energy that will be generated in the next dayslies in the dependence of renewable energy on atmospheric conditions, which makes thesehave a highly variable nature.In this work, we propose to combine time series of the produced renewable energy,synoptic meteorological conditions and machine learning techniques, specifically, artificialneural networks (ANN) to forecast the daily energy generated in each island.Some statistical metrics (RMSE, R2 and BIAS) are used to compare the results ofthree different types of ANN trained with different data inputs:DA: where the inputs are just the time series of generated energy.SP: where the inputs are just the atmospheric sea level pressure in an area aroundthe Canary Islands.DAP: which is a combination of the other two, and the inputs are the atmosphericsea level pressure and the time series of generated energy.Those results are also compared with those obtained for the simplest persistence method(ZOH) and with a traditional method for the prediction of time series (ARMA).Finally, very varied results where obtained, with some cases in which the ANNs significantly improve the ZOH results, and other cases in which there is no significant improvement. In general, for solar energy, DA and DAP are the most reliable, being in all casesbetter than the ZOH. For wind energy, DA and DAP are also the most reliable, exceptingfor the cases of La Palma and Gran Canaria in which the best model is SP. Unlike forsolar energy, these models do not provide a significant improvement with respect to theZOH.
YR 2021
FD 2021
LK http://riull.ull.es/xmlui/handle/915/25009
UL http://riull.ull.es/xmlui/handle/915/25009
LA es
DS Repositorio institucional de la Universidad de La Laguna
RD 04-may-2024