Fault pattern recognition method for the high voltage circuit breaker based on the incremental learning algorithms for SVM

In order to recognize faults of the high voltage circuit breaker (HVCB) in the whole fault state space precisely and minimize the impact of the lack of fault data on the accuracy of fault recognition, a method of fault recognition was proposed based on the incremental learning algorithm for SVM. Firstly, the incremental learning algorithm for SVM was analyzed theoretically, and the state monitoring variables were determined by the current signal and voltage signal of control unit and the vibration signal of the switching for HVCB. Based on the fault mechanism, the feature extraction method for the monitoring variables was proposed. Secondly, four common faults, including the spring loosening, the core jamming, the coil aging and the abnormal electrical power supply, were simulated. Then the fault feature was extracted, and the fault data samples as well as the incremental learning data samples were established. After training the feature variables based on the incremental learning algorithms for SVM, the fault recognition model was acquired and its accuracy was validated through exerting the new fault features into the model. Finally, it is shown that the incremental learning algorithms for SVM can be used to recognize the faults of HVCB effectively, and its recognition accuracy can be improved by continuous learning of the new samples.