Estimation of solar radiation using support vector regression

The solar radiation estimation is important parameter of interest in the Photovoltaic system installation, maintenance and performance evaluation. In the present work support vector regression (SVR) is used for the estimation of global horizontal solar radiation (GHSR). The model evaluates the performance of SVR model with different combinations of metrological parameters such as ambient temperature, humidity, atmospheric pressure etc. The data of three years from 2009-2011 is acquired from the National Institute of Solar Energy (NISE) Gurugram, India. The performance metric for estimation is the root means square error (RMSE). The results of SVR model were compared and found superior with other state of the art models like Hidden Markov Model and Artificial Neural Networks. After the analysis of results, it was found that temperature is the most important parameter along with atmospheric pressure, relative humidity, day number and wind speed.