Abstract:

In order to overcome the nonlinear change of the amplitude of turbo-generator unit in malfunction,a forecasting model combining GM (Gray Model)with ARMA (Auto Regression Moving Average),a forecasting model based on the fractal collage theorem and the fractal interpolation,and a forecasting model based on the leastsquare support vector machine are established,which are then respectively fitted with the daily average vibration peak-peak values of a 200 MW unit．At the same time,the forecasting performances of the three proposed models are analyzed and compared,with a suitable model obtained for the fault forecasting of turbo-generator units.

Key words: turbo-generator, forecasting, time series, gray system , fractal, least squares support vector machine