TY - JOUR
TI - Machine learning methods in forecasting solar photovoltaic energy production
AU - Milićević Marina M
AU - Marinović Budimirka R
JN - Thermal Science
PY - 2024
VL - 28
IS - 1
SP - 479
EP - 488
PT - Article
AB - Energy has an effective role in economic growth and development of societies. This paper is studying the impact of climate factors on performance of solar power plant using machine learning techniques for underlying relationship among factors that impact solar energy production and for forecasting monthly energy production. In this context this work provides two machine learning methods: ANN for forecasting energy production and decision tree useful in understanding the relationships in energy production data. Both structures have horizontal irradiation, sunlight duration, average monthly air temperature, average maximal air temperature, average minimal air temperature and average monthly wind speed as inputs parameters and the energy production as output. Results have shown that used machine learning models perform effectively, ANN predicted the energy production of the PV power plant with a correla-tion coefficient higher than 0.97. The results can help stakeholders in determining energy policy planning in order to overcome uncertainties associated with renewable energy resources.