Prediction of Solar Energy Potential with Machine Learning and Deep Learning Models
International Conference on Applied and Industrial Mathematics, ICAIM 2023, Greater Noida, India, 24 - 26 March 2023, vol.482, pp.11-25, (Full Text)
- Publication Type: Conference Paper / Full Text
- Volume: 482
- Doi Number: 10.1007/978-981-96-3460-6_2
- City: Greater Noida
- Country: India
- Page Numbers: pp.11-25
- Keywords: 68T07, ANN, Discrete wavelet analysis, GRU, Hybrid forecasting, LSTM, Solar irradiance forecasting
- Süleyman Demirel University Affiliated: Yes
Abstract
Replacing fossil fuel-based energy with carbon-free sources is essential for achieving climate neutrality in Europe, aligned with the European Green Deal. The southern regions of Turkey, especially Serik, Antalya, are well-suited to support this strategy due to their significant solar energy potential. This research investigates the solar energy potential of Serik, a region with one of Turkey’s highest renewable energy reservoirs. Accurately forecasting solar energy potential requires understanding meteorological, geographical, and technological factors, particularly advancements in machine learning and deep learning models. This study employs Nonlinear Autoregressive Neural Networks (NAR), Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Seasonal Autoregressive Moving Average (SARMA), and Discrete Wavelet Analysis, as well as the NARX model, incorporating Wavelet outputs as inputs. The integration of Discrete Wavelet Transform with neural networks has significantly enhanced forecast accuracy. The SARMA model also performed well over the test period. This paper serves as a guide for selecting optimal approaches to forecast medium- and long-term solar energy potential, contributing to improved strategies for solar energy utilization.