Series arc fault detection of DC bus of photovoltaic system based on 1DCNN and D-S multi-information fusion
DC bus is the main route of photovoltaic system output energy.Due to long-term exposure,weathering and other effects,cables,connectors and other components deteriorate,so the possibility of arcing in the DC bus of photovoltaic system rises sharply,and is easy to cause fire,electric shock and other accidents.In photovoltaic sys-tems,series arc faults will cause the loop current to drop,which can not be recognized by conventional overcurrent protection.Therefore,this paper proposes a method based on deep learning and Dempster-Shafer(D-S)to identify series arc faults,which uses a one-dimensional convolutional neural network(1DCNN)to identify the arc of the detection data based on the current and voltage signals of shunt capacitors.On this basis,the recognition result based on a single sensing data is used as the evidence,and the reliability distribution is calculated by using the D-S multi-information synthesis rule,and finally the decision rule is used to determine whether a series arc fault oc-curs.The results show that the accuracy of series arc recognition based on current and voltage signals of shunt ca-pacitors is 97.19%and 94.98%,respectively,while the recognition accuracy of DC series arc fault detection of photovoltaic system based on 1DCNN and D-S multi-information fusion can be increased to more than 99%.
photovoltaic system1DCNNseries arc faultD-S multivariate information fusionfault detection
NETL
NSTL
万方数据
维普
