The importance of renewable energy sources like solar energy in reducing carbon emissions and other greenhouse gases has contributed to an increase in grid integration. However, the intermittent nature of solar power causes reliability issues and a loss of energy balance in the system, which are barriers to solar energy penetration. This study proposes a unique three-step approach that identifies weather parameters with moderate to strong correlation to solar radiation and uses them to predict solar energy generation. The combination of an on-site weather station and a reliable local weather station produces relevant data that increases the accuracy of the forecasting model irrespective of the machine learning algorithm used. This data source combination is tested, along with two other scenarios, using the exponential Gaussian Process Regression machine learning algorithm in MATLAB. It was found to be the most effective algorithm with a Normalized Root Mean Square Error of 1.1922, and an R2 value of 0.66.
Keywords Renewable energy sources, variable renewable energy, machine learning, solar energy forecasting, Gaussian Process Regression