Fault Diagnosis of Power Circuits Based on SVM Ensemble with Quantum Particles Swarm Optimization Fault Diagnosis of Power Circuits Based on SVM Ensemble with Quantum Particles Swarm Optimization

Based on least squares wavelet support vector machines (LS-WSVM) ensemble with quantum particle swarm optimization algorithm (QPSO), a systematic method for fault diagnosis of power circuits is presented. Firstly, wavelet coefficients of output voltage signals of power circuits under faulty conditions are obtained with wavelet lifting decomposition, and then faulty feature vectors are extracted from the disposed wavelet coefficients. Secondly, a boosting strategy is adopted to select faulty feature vectors automatically for LS-WSVM-based multi-class classifiers, QPSO is applied to select the optimal values of the regularization and kernel parameters of multi-class LS-WSVM. So the multi-class LS-WSVM ensemble model with boosting for the power circuits fault diagnosis system is built. The simulation result of push-pull circuits shows that the fault diagnosis method of the power circuits using LS-WSVM ensemble with QPSO is effective.