Research On Intelligent Detection Technology Of Electrical Fire Based On Multi-information Fusion

Safety utilization of electric power has become the eternal theme with the development of the power industry in our country. How to detect the cause of electrical fires, reduce the rate of electrical fires accident and the loss of property are the focus of safety utilization of electric power nowadays. In the system of low voltage power distribution of the building, the existing protection equipment for preventing electrical fires only can detect the overcurrent caused by overload and short-circuit, residual current caused by grounding fault, when the series arcing fault apt to cause electrical fires occurs, due to that the fault current value is lower than the action threshold of traditional protective devices, falling to prevent electrical fires completely and effectively. Therefore, the number of fires caused by electrical fault is higher than other reasons.At first, the research background and significance of this subject and the main causes of electrical fires are introduced. The development status of protection equipments and technical measures against electrical fires at home and abroad are summed up, the standards and norms about preventing electrical fires are consulted, and deeply analyzed the key problems which have to be settled urgently.Secondly, experimental platform for simulating the arcing fault in the system of low voltage power distribution is built; current transformer and data acquisition module are used to acquire the arcing fault waveform of resistive load and resistive-inductive load, the waveform of the switched power supply of computer and the waveform when the load changed. The waveform of arcing fault is validated by power quality analyzer, the harmonic, K-coefficient and total harmonic distortion are analyzed. The initial mathematical analysis of waveform of arcing fault which were done by multi-function oscilloscope.Thirdly, to effectively draw the similarities and dissimilarities between the waveforms of fault arcing and that of normal circuit in real life, the arcing fault waveform of resistive load and resistive-inductive load, the waveform of the switched power supply of computer and the waveform when the load changed are analyzed using MATLAB, with methods of FFT, WT, statistical analysis. The spectrum and the wavelet high-frequency coefficients of the foregoing waveforms are analyzed and extracted; the sum, the maximum, the median, the mean value, the standard deviation and the slope factor of each period of the waveforms are calculated, represented the characteristics of each waveform with tables and charts and advantages/disadvantages of the above methods to recognize arcing fault.Fourthly, to compensate the flaws of the existing techniques against electrical fires, a technique of intelligent detection of electrical fires based on multi-information fusion is present in this paper. Using fuzzy logical ratiocination and hyperball CMAC neural network to establish the system model, the extracted information is fused to detect the arcing fault in the system of low voltage power distribution of the building that is regularly incurs electrical fires, justified the stability of the multi-information model through simulations and consequently demonstrated that the multi-information method could compensate the defects of single criterion and detect the electrical fires more thoroughly and accurately.In conclusion, all works that have done are summed up, and the next problem and research direction of intelligent detection technology of electrical firess are proposed.