Research On DC Arc Fault Detection Method For Electric Bicycle

The DC arc fault in the DC power supply system should be detected in time to avoid the adverse consequences caused by it.So it is of significance to protect the safe and reliable operation of DC power supply system.Aiming at the DC arc fault detection technology of electric bicycle's DC power supply system,this thesis presents a method for DC arc fault detection based on SVM.The main work of this paper include:1.An experimental platform for electric bicycle's DC arc fault data acquisition is designed,in which a DC arc generating device is made according to UL1699 B standard,and the hardware and software of the data acquisition equipment is designed and implemented.With this experimental platform a large number of experimental data in different working conditions of electric bicycles when normal operation,charging and DC arc fault's occurring can be collected.2.A design scheme of DC arc fault detection based on SVM is proposed,in which supervised learning was used to train the SVM classifier which has the function of detecting DC arc fault.The trained SVM classifier can judge whether the DC arc fault occurs or not according to the input feature data of DC arc fault.3.Several features which can be used for arc fault detection are proposed through time-domain and frequency-domain analyzing to the DC arc fault experimental data.According these features the training sample set and test set of the SVM classifier for DC arc fault detection are constructed.4.The SVM classifier for DC arc fault detection is designed and trained with the data form the above training sample set.In order to select the appropriate penalty parameter c and kernel parameter g for the SVM,the particle swarm optimization algorithm combined with the K-fold-validation is used to find the optimal combination of these two parameters.In the thesis,the data from test set are used to simulate the actual feature data of DC arc fault which is sampled and extracted in real time during the actual work of the electric bicycle to test the performance and effect of the on-line detection of the SVM classifier for DC arc fault detection.The test results show that the SVM classifier of DC arc fault detection has a good effect on the correct detection rate and mis-operation.