Use of support vector machine for wind speed prediction

The penetration of wind energy has been encouraged significantly throughout the world. The wind power is a clean, inexhaustible, and almost a free source of energy. But the integration of wind parks with the power grid has resulted in many challenges for the unity in terms of commitment and control of power plants. As wind speed and wind direction fluctuate frequently, the accurate long-term and short-term forecasting of wind speed is important for ascertaining the wind power generation availability. To deal with wind speed forecasting, many methods have been developed such as physical method, which use lots of physical considerations to reach the best forecasting precision and other is the statistical method, which specializes in finding the relationship of the measured power data. Wind speed can be predicted by using time series analysis, artificial neural network, Kalman Filter method, linear prediction method, spatial correlation models and wavelet, also by using the support vector machines. In this paper, the SVM is used for day ahead prediction of wind speed using historical data of wind speed at site. It is observed that the Mean Absolute Percentage Error (MAPE) is around 7% and correlation coefficient is close to 1. This justifies the ability of SVM for wind speed prediction task.