Calculation Method of Partial Discharge Severity Assessment Index Weight Using Factor Analysis Based on Mutual Information

In Partial Discharge (PD) severity evaluation assessment, our previous research constructed a relatively perfect feature set F
_s
= [u
_max
⁺
, u
_max
^-
, N
⁺
, N
^-
, ΔT
⁺
, ΔT
^-
, ΔT
_max
, Q
_acc
, E
_n
] based on the theoretical study of a flurry of UHF signals in different PD severity under four artificial defects in gas-insulated switchgear. To dig the connection between the PD features and the severity of PD, the mutual information theory is introduced to measure the correlation between the features in Fs, and the correlation between features and the four defined PD severity levels, and then we can get the order of the features' importance. If we assume that the minimum of redundant information is the objective function, we can get the more objective weighted value under the PD data analysis. The fuzzy C means clustering algorithm (FCM) is employed to get the clustering center and clustering radius of different PD severity level for all the features to build the membership function of each assessment features avoid division errors due to strict fuzzy partitioning of sample types. Using the fuzzy comprehensive evaluation method, the proposed method possesses good expansibility in the test and it can be very useful in actual applications.