SOLAR RADIATION PREDICTION USING MACHINE LEARNING APPROACH BASED ON VARIOUS METEOROLOGICAL VARIABLES

5. International Anatolian Scientific Research Congress, Hakkari, Türkiye, 21 - 23 Temmuz 2023, ss.1460-1467

Due to the increasing global climate crisis and environmental issues, electricity production using

renewable energy sources has been steadily rising. Solar energy systems, considered a clean energy

source, have rapidly developed in recent years and become a significant renewable energy option.

Solar radiation, an essential parameter in almost all scientific disciplines, varies depending on different

climatic conditions and is negatively affected by adverse weather conditions resulting from various

meteorological factors. The efficiency of solar energy plants depends on the accuracy of solar

radiation prediction. Accurate radiation prediction enhances the efficiency of photovoltaic (PV) plants

and enables their proper, stable, and effective integration with the power grid. In this study,

simultaneous solar radiation values were predicted using a machine learning approach, based on

fundamental meteorological parameters such as wind speed, ambient temperature, atmospheric

pressure, and relative humidity. The data were obtained from a meteorological measurement station in

Hakkari province between 2019 and 2020. The relationships among the input parameters were

evaluated using Extreme Learning Machine (ELM), a machine learning model commonly used in

prediction studies. The solar radiation values were estimated with an accuracy rate of approximately

92% using the Mean Squared Error (MSE) statistical validation method. The ELM algorithm exhibited

high accuracy with minimal errors. Therefore, using the data obtained from the study area and the

ELM algorithm, solar radiation values, which are the most crucial parameter for the design and

planning of PV plants, can be accurately determined without requiring direct measurement.