Thyristor State Evaluation Method Based on Kernel Principal Component Analysis

Abstract

The reliability of the thyristor is directly related to the safe operation of the DC transmission system. A method for evaluating the state of thyristors based on kernel principal component analysis (KPCA) is proposed, which firstly considers the thyristor test data, operation records, maintenance history, appearance inspection information, states of other components and operating environment. A basic index system for evaluating the aging state of thyristor with 42 parameters is established. Next, a mathematical model was developed by Fisher Discriminant Analysis (FDA). The kernel function of the kernel principal components is then optimized by an improved particle swarm optimization (IPSO) algorithm. The improved KPCA is applied to extract key parameters from the base index system to obtain the reduced dimensional evaluation indicators. The obtained principal component factors are used to determine the weights of the fuzzy composite factors, which are applied for fuzzy evaluation of the thyristor. Finally, 20 thyristors are selected for experimental and theoretical calculations. The results show that the cumulative contribution of the first three principal component variables after dimensionality reduction reaches 93.76%, which is consistent with the state of the thyristor. Compared to the four existing evaluation methods, the results of the method proposed in this paper are more reasonable, which removes the influence of redundant indicators, reduces the amount of data, and provides a reference for the related research on thyristor state evaluation.

Research on evaluation index method of cloud-network convergence capability

“…Recently, there are two main categories of methods for determining weights, qualitative analysis including analytic hierarchy process (AHP) [16][17][18][19] and quantitative analysis including fuzzy cluster analysis 20 , entropy weight method (EWM) [21][22][23] , and index weight determination method based on principal component analysis 24 , etc. Fuzzy clustering is suitable for clustering large-scale data by calculating the uncertainty degree of samples belonging to various categories and performing clustering analysis.…”