Short-term wind speed prediction based on support vector machine of fuzzy information granulation

Wind power prediction is an effective way to decrease the effects for the large-scale grid-connected wind power generation. Improving accuracy of short-term wind prediction is the key to wind power prediction. There is often a lot of redundant information in the observed values of wind speed to result in large computation and affect the predictive validity. In this paper, a support vector machine (SVM) method was proposed based on information granulating. Firstly, original data was dealt with fuzzy information granulation to form information granulation. Secondly, adopting SVM is widely used in regression prediction to predict the changes of short-term average wind speed in trends and in space. At last, the actual data was compared with the predicted data to verify results. The verification experiments show that granulated data can not only reflect the characteristics of wind but also reduce redundant information. The predicted result was changing in the space of average wind speed. So, this method can predict the short-term wind speed space.