Akademik Veri Yönetim Sistemi
Decision Making Using AI in Energy and Sustainability, Gülgün Kayakutlu,M. Özgür Kayalıca, Editör, Springer, London/Berlin , İstanbul, ss.105-117, 2023
Concerning the impacts of climate change, renewable energy systems are becoming more prominent by providing more environmentally friendly energy. However, these systems are very dependent on meteorological parameters. In this study, it is aimed to forecast hourly downward surface solar radiation (DSSR), which is the main determinant of power output from PV systems, using Particle Swarm Optimization (PSO) and Long Short-Term Memory Recurrent Neural Networks (LSTM). The hourly data of 12 meteorological features in Karaman, Turkey, for the period between January 2020 and December 2021 were obtained from the European Centre for Medium-Range Weather Forecasts’ (ECMWF) ERA5 reanalysis data sets. PSO is utilized to select the most significant features from this data. These selected variables are put into the LSTM model as inputs. The established hybrid forecasting model is examined for 1 hour, 2 hours, 4 hours, and 6 hours ahead. Furthermore, to measure the performance of the forecasting model, the Mean Squared Error (MSE) and Root Mean Squared Error (RMSE) are calculated. As a result, the lowest error rate is acquired for 1 hour ahead of time horizon with 8496 (W/m2) MSE and 92 (W/m2) RMSE. This study also exhibits that the proposed hybrid model may provide very accurate results to forecast hourly DSSR with selected variables in case it is not possible to measure the DSSR values directly.