Research On Fault Diagnosis Of Low Voltage DC Series Arc

With the application of renewable energy and the increase of system capacity,the power supply network load topology in power generation,transmission,distribution and utilization system is becoming more and more diverse,and the network topology structure is constantly changing,so the power system protection needs to be improved urgently.DC series arc fault,as a potential huge hidden danger of electric fire in DC system,is not easy to extinguish due to the lack of zero-crossing and periodicity of DC circuit,which leads to electric fire and endangers the safe,reliable and stable operation of power grid.Research on fault diagnosis and key technology of DC series arc is an effective way to avoid electric fire in DC system,and is of great significance to safe operation and maintenance of distribution and utilization power grid.According to the basic theory of DC arc,the interaction between switching arc and its system is analyzed.By introducing classical arc black box models,such as Cassie model,Mayr model and KEMA model,the low-voltage DC series arc fault system is modeled and simulated based on Simulink,and compared with the experimental data,the Cassie model is determined to be the DC series arc model closest to this system.Cassie model is applied to low voltage DC simulation system.Modeling and simulation analysis of single load and multi-load DC power system under different working conditions are carried out to capture the changes of series arc voltage and current in time domain.Based on UL1699-2008 AFCI standard,a DC series arc fault detection platform is built independently for low-voltage residential DC power supply system.The DC arc fault experiment under single load and multi-load of different fault points is carried out,and the DC arc fault database and DC series arc characteristic database under typical load are established.Based on the experimental study of low voltage DC arc fault current,the chaotic characteristics of DC series arc in multi-temporal scales are quantitatively analyzed by introduc:ing the method of chaos theory.The research shows that the DC series arc presents a weak chaotic state,that is,the arc has a certain randomness,and the influence parameters have a diversified trend.Based on this analysis,it provides a priori basis for the research of DC arc fault characteristics.In order to study the fault characteristics of DC series arc and realize the diagnosis of DC series arc faults,the fault characteristics under typical loads are analyzed in time domain,frequency domain and time frequency domain respectively,and the effective eigenvalues are extracted by comparing the normal operating conditions and fault arc characteristics of the system.The BP neural network algorithm is introduced to recognize the pattern of DC series arc fault and realize system diagnosis and fault detection.The diagnostic accuracy of multi-load series arc fault in low-voltage power system under different working conditions is 90.38%.