| Abstrakt: |
In view of the characteristics of strong randomness and low direct prediction accuracy of daily power of runoff hydropower stations, this paper uses the Extreme-point Symmetric Mode Decomposition (ESMD) to smooth the power sequence. Combined with the Least-Square Support Vector Machines (LSSVM), a combined prediction model based on ESMD-LSSVM is established. The daily power time series of a runoff hydropower station in Northwest China in 2020 is selected for example analysis, and compared with the prediction results of single model SVM, LSSVM, BP and combined model ESMD-SVM, ESMD-BP. The results show that: ➀ PACF analysis shows that the feature vectors of each subsequence after ESMD decomposition are different, which reflects the complexity and variability of the daily power of runoff hydropower stations. ➁ Compared with the single model, the combined model has stronger generalization ability and more accurate prediction of the power mutation point in the time series. ➂ The daily power of ESMD-LSSVM combined model has good prediction effect, which provides a new method reference for the daily power time series prediction of runoff hydropower station. [ABSTRACT FROM AUTHOR] |
