Short-term Wind Speed Forecasting Based On Data Mining Technology

In recent years, wind power technology becomes a potential renewable energy power generation technology. This paper reviews the present situation and characteristics of wind speed both at home and abroad. Using time series analysis and support vector regression analyzes short-term wind speed forecasting of the wind farm to find a more accurate method .Short-term wind speed forecasting requires a lot of historical data as a sample. The presence of singularities will increase the noise and reduce wind speed forecasting accuracy. So, firstly, this paper finds and fixes the singularity to ensure the rationality and validity of data. Then through canonical correlation analysis, it analyzes various variables to get the variables which greater correlate with wind speed for more accurate forecasting of wind speed.Basing on lots of statistical analysis and disposals with data, this paper forecasts short-term wind speed by the methods of time series analysis and support vector regression machine. When using time series analysis to forecast the wind speed, firstly, it makes the sequence steady, then, it identifies the model and verifies parameter. After that, it establishes a model to forecast the short-term wind speed. When using support vector regression machine to forecast the wind speed, this paper establishes ? ?SVRand v ? SVRto forecast the wind speed by the software Libsvm-2.88. It shows that ? ?SVRhas a better result.At last, this paper compares and analyzes the forecasting result which bases on the method of time series analysis and support vector regression machine. Because the former only considers the pre-historical data to forecast the wind speed and the model is simple, but the latter uses pre-historical and relevant variables to forecast wind speed. So the support vector regression machine gets the better results.