Short-term photovoltaic power forecasting with weighted support vector machine

Abstract

The output power of the solar photovoltaic (PV) arrays has the property of uncertainty, and usually fluctuates with the changes of solar radiation and the ambient temperature. It is important to forecast the output power of the PV power station so as to coordinate the relationship between the conventionality power supply and the grid-connected PV power station. In this paper, a weighted Supported Vector Machine (WSVM) is adopted to forecast the short-term PV power, in which the 5 days with the most similarity to the day to be forecasted were selected as the training samples, and the weights of the samples for the WSVM are designed based on the similarities together with the time interval. The proposed algorithm is experimentally validated and the results empirically show that the output power forecasted by use of the WSVM is more efficient than that of the artificial neutral network (ANN) and more practicable.