Application of Improved Kernel Principal Component Analysis Support Vector Machine Model in Power Transformer Condition Assessment

摘要

Considering the fact that membership functions and factor weights are highly depend on man in fuzzy synthetic evaluation, a model for power transformer condition assessment based on Support Vector Machine (SVM) is presented. In order to eliminate the false feature and retain the true, the information data of the transformer is preprocessing with Kernel Principal Component Analysis (KPCA) first, the results become the inputs of SVM to form the KPCA+SVM model for transformer evaluation. To enhance the effect of evaluation, an improved KPCA+SVM model is proposed. Mixtures of kernels and parallel optimized strategies are used. The example shows the effectiveness and superiority of the improved model.