Comparison of Machine Learning Methods for Short-Term Solar Irradiation Forecast

The transitory character of solar power has led researchers to forecast solar irradiance to exploit this green energy befitting the electrical demand. This can help the photovoltaic (PV) generators to participate in energy markets generating revenue without stressing the power system as a whole. Accurate forecasting based on weather inputs can reduce large-scale backup requirements. Here, three types of regression: random forest regression, gradient boosting regression, and adaptive boosting regression method are studied on hourly weather data of Singapore for one year to forecast solar irradiance in the short-term. The hyperparameters are set after a thorough trial. The effectiveness of the methods which are being proposed is evaluated based on Mean Square Error (MSE), Root Mean Square Error (RMSE), and regression score indices that indicate random forest regression as a better method than the other two.