Clone Immune Network Classification Algorithm for Fault Diagnosis of Power Transformer

Abstract

In this paper, a new optimization algorithm called Clone Immune Network Classification Algorithm (CINC), is proposed for fault diagnosis of power transformers. The algorithm has merged the merits of population-based immune algorithm and network-based immune algorithm. The characteristics of training fault samples are studied and extracted by memory antibody set. Consequently, CINC can be used to find a limited number of antibodies which can represent all training fault samples distributed structures and features, which helps to realize dynamic classification. Then the testing fault samples are classified with the k-nearest neighbor method (KNN). Compared with previous immune network model and immune algorithm, this one can prevent prematurity, keep variety and avoid local optimal. Many fault samples have been tested by CINC algorithm, and its results are compared with those obtained by IEC three-ratio method (TRM) and BP neural network (BPNN) respectively. Comparison results show that the proposed algorithm is feasible and practical. The algorithm is of fast convergence rate and high diagnosis correctness.