PV Output forecasting based on weather classification, SVM and ANN

Abstract

The expansion in solar power is expected to be dramatic soon. A number of solar parks with high capacities are being setup to harness the potential of this renewable resource. However, the variability of solar power remains an important issue for grid integration of solar PV power plants. Changing weather conditions have affected the PV output. Thus, developing methods for accurately forecasting solar PV output is essential for enabling large-scale PV deployment. This paper has proposed a model for forecasting PV output based on weather classification, using a solar PV plant in Maharashtra, India, as the sample system. The input data is first classified using RBF-SVM (Radial Basis Function Support Vector Machines) into three types based on weather conditions, namely, sunny, rainy and cloudy. Then, the neural network model corresponding to that weather type has been applied to forecast the solar PV output. The obtained results for the overall model is studied for its effectiveness and are compared with existing research.