Detection Method Research And Application On The Series Arc Fault

With the rapid development of modern economy and social civilization, the grid scale is expanding while industrial production and daily life consumption is also increasing. The electrical fire would not be ignored because it is one of the biggest potential risks to the personal safety and property loss. Arc fault is one of the major causes of electrical fires. The reasons of arc fault could be electric switch contact loosening, line insulation aging and poor contact. The normal current and arc fault current is similar due to the limit of load line, thus, traditional fault protection device cannot protect the arc fault. Therefore, the research of arc fault detection and protection device is of great importance.In this thesis, the low voltage series arc fault is the research object. Based on experimental platform arc fault experiments with common loads used in house were done. Current signal and the load terminal voltage signals of the normal and arc fault conditions was obtained, while analysis arc fault signal and draw the characteristic based on arc current and load of terminal voltage.Zero rest width method is used to detect arc fault based on arc current. Zero rest width mathematical models are presented in this paper. The results of zero rest width calculation show that the zero width method can identify the same load arc fault and cannot discriminate effectively current signal singularity that comes from different load types, different rated power and working mode conversion. The load terminal voltage signal analysis shows that the load terminal voltage is not affected by the distribution lines, while the normal operation of the current signal singularity. Therefore, we suggest that the load terminal voltage arc fault detection method.In order to study the characteristics of series arc fault, the arc fault diagnosis based on the load terminal voltage signal is realized. Firstly, using the wavelet method reduce the noise of the load terminal voltage. Secondly, based on wavelet transform multi-resolution analysis for load voltage signal is done. Thirdly, the details signal which from db5 wavelet five layer decomposition is extracted and details signal modulus maxima is calculated, which is regard as the load terminal voltage signal eigenvector. Then the support vector machine(SVM) classification model parameters is calculated using cross validation and grid search method. Support vector machine training samples are load terminal voltage signal eigenvector while the support vector machine classification model diagnose arc fault.Based on the controller dsPIC33FJ64GS606, zero rest width combine with the peak width detection method is taken to achieve arc fault detection; the arc fault detection device is completed.