Research On Characteristics And Recognition Methods Of Series Fault Arc In Low Voltage Bolted Electrical Connector

Low voltage bolted electrical connector is the most common used device in industrial power supply and distribution system.It is widely used to achieve electrical connection between cable and bus,cable terminal and electrical equipment.Electrical connection looseness fault and the resulting series fault arc are main reasons of electrical fire.And there is still no effective prevention method.Therefore,it is of great significance to research fault features and recognition methods of series fault arc.It can be used to develop new type arc fault circuit interrupter and prevent electrical fire.The series fault arc in low voltage bolted connector was taken as research object.A kind of looseness fault experimental system and a new type series fault arc experimental system were developed successfully.The electrical connection looseness fault experiments of low voltage bolted electrical connector were carried out.The thermo-electric characteristics of the looseness fault were analyzed.And a new looseness fault identification method based on current energy entropy and Probabilistic Neural Network(PNN)was proposed.The series fault arc experiments in typical industrial load such as three phase motor load,three phase motor with frequency converter load,industrial computer load and so on were carried out for the first time.Different operation states such as no-load state,full-load state,starting state,braking state,and load fluctuation state were considered for both three phase motor load and three phase motor with frequency converter load.Difference between steady arc and fault arc was also considered.The fault features of series fault arc were extracted by using wavelet analysis,then a fault feature database was established.Three kinds of fault arc identification model were established based on PNN,support vector machine(SVM)and BP neural network respectively.And they were optimized by using loop optimization method,particle swarm optimization(PSO)algorithm and genetic algorithm(GA)respectively.A new series fault arc identification method based on twentytwo kinds of fault features and the fault arc identification model was proposed,and its effectiveness was verified with lots of experimental tests.Some main conclusions are listed as follows:(1)Wire temperature can be used to reflect overheat status when bolted electrical connector works in tight state.Looseness fault will lead to contact temperature,screw nut temperature and wire temperature rise rapidly.And the temperature difference between contact temperature and wire temperature increases with the increase of looseness level.Screw nut temperature is more accurate than wire temperature to characterize overheat status when the bolted electrical connector works in loosening state.(2)Mechanical vibration will lead to increase of contact temperature.Once a high current bolted electrical connector operates in vibration state,even a very slight electrical connection looseness fault will also bring serious fire hazard or cause fire.Therefore,we should pay more attention to looseness fault of the electrical connector working in vibration state.(3)The contact voltage waveform will distort seriously once the looseness fault occurs.The feature can be used to identify electrical connection looseness fault.(4)Wavelet energy entropy of current signal can be used as looseness fault feature.The optimized PNN can be used to identify looseness fault of low voltage bolted electrical connector.(5)Twenty-two features of fault-phase current signal can be used as fault feature of series fault arc.These features include time-domain features and time-frequency domain features.The time-domain features are average value,standard deviation,variance,average value of absolute value integration,root-mean-square,and correlation coefficient.The time-frequency domain features are average value,module maximum value,standard deviation,wavelet entropy and wavelet energy of three wavelet decomposition coefficients.These coefficients are the second level detailed coefficient and approximate coefficient obtained by using db4 wavelet decomposition.(6)The proposed fault arc identification models can effectively identify series fault arc occurred in three phase motor load,three phase motor with frequency converter load,industrial computer load and resistive-inductive load.