Prediction of solar radiation and power using machine learning techniques

In recent years, as the impacts of climate change and particularly global warming have become more evident, the significance of energy efficiency and renewable energy has grown substantially. In the field of renewable energy, there are numerous applications of energy systems, including solar energy and wind energy, as well as geothermal, biomass, wave, and hydropower energy. Examining the economic and environmental sustainability before the installation of renewable energy systems is crucial to achieve optimal efficiency from these systems. In this context, reliability and energy forecasting studies in renewable energy systems have emerged as effective areas of research within the context of sustainability. Machine learning methods are employed for reliable energy forecasting and assessment in energy systems. In the conducted study, various machine learning models such as Artificial Neural Networks, LightGBM (Light Gradient Boosting Machines), and LSTM (Long Short-Term Memory) have been employed for the energy prediction of a solar energy system planned to be installed. The machine learning models used were compared by evaluating their prediction performance using various error metrics, and the LightGBM model, which yielded the most successful results, was selected. The selected model achieved a Mean Absolute Percentage Error (MAPE) value of 21.29% on the test data. Using the predictions made with the chosen LightGBM model for solar irradiance, hourly power output and daily energy production estimates for a solar panel system under specific parameters were conducted for the following day.