Power transformer condition assessment using support vector machine with heuristic optimization

This work investigates the practical application of support vector machine (SVM) to power transformer condition assessment. Partiuclarly, this paper proposes to integrate the SVM algorithm with two heuristic optimization algorithms which are particle swarm optimization algorithm (PSO) and genetic algorithm optimization (GA). These two optimization algorithms are used for efficiently and effectively determine the optimal parameters for SVM. The resulatant two hybrid algorithms, i.e. SVM-PSO and SVM-GA can improve the performances of the original SVM algorithm on classifying the incipient faults in power transformers. Extensive case studies and statistic comparison among the original SVM, SVM-PSO, and SVM-GA over multiple datasets are also provided. Calculation results may demonstrate the effectiveness and applicability of the two hybrid algorithms in improving the classification accuracy of SVM for condition assessment of power transformer.