Photovoltaic Energy Production Forecasting with Machine Learning
3rd International Conference on Frontiers in Academic Research , Konya, Türkiye, 15 - 16 Haziran 2024, ss.966-977, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Basıldığı Şehir: Konya
- Basıldığı Ülke: Türkiye
- Sayfa Sayıları: ss.966-977
- Ankara Yıldırım Beyazıt Üniversitesi Adresli: Evet
Özet
The release of carbon dioxide (CO2) and the greenhouse gases, which is one of the main reasons
for climate change that we feel more and more every day. Therefore, there is an increasing interest in
renewable energy sources such as photovoltaic (PV) solar systems with the aim of combating the climate
change and achieving cleaner air. The dependence of PV systems on weather conditions causes instability
in PV power output, voltage fluctuations, and interruptions. This disadvantage of PV systems makes them
difficult to be integrated into grids. Predicting output power of PV systems in advance will allow for the
reduction of fossil fuel use, will provide efficient use of renewable energy sources, will lead to efficient
operation of the electricity grid, as well as will ensure optimal management of production and storage, and
consequently will help in reduction of electricity production costs. In this study, firstly, five different
machine learning methods, Decision Tree (DT), Artificial Neural Networks (ANN), Gradient Boosting
(GB), Random Forest (RF), and K Nearest Neighbors (KNN), which learn faster and perform with high
accuracy, overcoming the disadvantages of overlearning and short term power prediction that are common
in machine learning models in the literature, were subjected to training and testing with our own dataset.
Subsequently, the method that best performs the estimation of the power generation (P) was selected. As a
result, it was observed and concluded that the RF model performs better than the other four machine
learning models.