A Hybrid Model for Hydroturbine Generating Unit Trend Analysis

According to the nonlinear and nonstationary characteristics of hydroelectricity systems, an hybrid prediction model based on wavelet transform and support vector machines is proposed in this paper for the trend analysis of hydroturbine generating unit (HGU). Firstly, the nonlinear and nonstationary time series are decomposed through wavelet transform into several subseries with obvious tendency characteristics. Then the tendency of these subseries is forecasted respectively with least squares support vector machines (LS-SVM), an extension of standard support vector machines, in which the kernel functions and model parameters are chosen appropriately. Finally, these prediction outputs are summed as the prediction results of original time series. The hybrid prediction model is applied in the peak-peak value of vibration time series of some HGU. The comparison between method based only on LS-SVM and hybrid model shows that the performance of hybrid method outperforms the former.