Simple model for short-term photovoltaic power forecasting using statistical learning approach

In the midst of increasing demand for electricity and rising fuel prices, renewable energy is imposed as one of the promising solutions for energy and environmental problems. In recent years, Photovoltaic power has witnessed a significant increase in the value of investments, according to the state of the art photovoltaic energy is known to be variable and uncertain because it is strongly influenced by meteorological variations, as consequence this type of energy present challenges to the operation of the power system because it disrupt the conventional methods for planning the daily operation of the electric grid. Short-term photovoltaic power forecasting is very important for power system operators, especially for load control in order to balance energy supply and demand. This article presents an offline photovoltaic power forecasting model using locally measured data, this model can give short-term forecasts without the need for weather forecasts, this is interesting for the managers of the electricity grid as well as for individuals who do not have access to weather data and forecasts. We investigate also the performance of two well-known statistical learning approach in photovoltaic forecasting, the Least Square Support Vector Regression (LS-SVR) and Feed Forward Neural Network (FFNN). At the end, and to evaluate the performance of our models, we will use the persistence model as a benchmark. The results of the simulation show that LS-SVR based model gives the best results and outperform the FFNN and persistence models.