A day-ahead Multi-Approach Machine L earning Technique for Photovoltaic Power Forecasting

The need to curb high dependence on fossil fuel caused much attention on the use of renewable energy especially solar energy as power source. However, the use of this source comes with its disadvantage of random uncertainties. Forecasting therefore is a safe way of insuring reliable power supply and satisfying the economic performance. It also gives an edge to the grid managers to plan ahead to reduce uncertainties. This work was built on Support Vector Regression (SVR) and Artificial Neural Network (ANN) machine learning techniques. Three methods SVR- linear Kernel, SVR-Gaussian Kernel and ANN were simulated using solar radiation, temperature, pressure and power production capacity of the site in Scikit-learn python library as input parameters.