Portable Detection System Of Power Equipment Based On Acoustic Vibration Signal

Large power equipment such as high-voltage parallel reactor,power transformer in the power grid system always continuously emits low-frequency vibration due to the influence of electromagnetic force in operation.After a long time of continuous operation,mechanical failures will gradually occur inside the power equipment.Over time,the degree of failure will become heavier,which may lead to local short-circuits,fires,and large-scale power outages.Therefore,it is of great practical significance to develop portable detection system for power equipment so as to discover potential failures as early as possible.This paper designs and implements a portable detection system for acoustic vibration signals of power equipment.Based on the JavaFX graphical framework,the software for the data acquisition is developed,and a pilot test of acoustic vibration signal collection is conducted to test the function and performance of the system.This paper analyzes the acoustic vibration signal of the high-voltage parallel reactor,confirming that the signal is composed of the fundamental frequency of 100 Hz and its higher harmonics,and most signals have the strongest frequency components at 100 Hz,while the difference between various signals lies in the frequency amplitude of each frequency component.The acoustic vibration signals of the normal reactor and the faulty reactor are compared,and it was found that the energy of the normalized average acoustic vibration signals of the faulty equipment relatively concentrates at the fundamental frequency of 100 Hz,while the intensity of the frequency component of the normal equipment is more scattered at each harmonic.Based on the analysis above,some classification detection experiments using neural network approach are conducted with the acoustic vibration signals of various simulation equipment and different working conditions,collected by the portable detection system.The experiments indicate that,the classification detection using convolutional neural network in time domain is most effective,whose accuracy,average precision,average recall,and average F measure reaches 95.10%,95.73%,95.87%,and 95.79%,respectively,demonstrating effectiveness of the collection method of the portable detection system.In addition,the data set of mechanical fault simulation platform of the Qian Peng diagnosis company is used to carry out classification and detection experiments.Among them,the classification detection using convolutional neural network in time domain is most effective,whose accuracy,average precision,average recall,and average F measure reaches 96.86%,97.44%,96.70%,and 96.78%,respectively,demonstrating the feasibility of using neural network to diagnose power equipment faults.Based on the acquisition software of acoustic vibration signals and the diagnosis algorithm of neural network,the diagnosis software of power equipment is designed and implemented.The software of power equipment diagnosis based on acoustic vibration signal combining the hardware of portable detection system constituted the portable detection system together for acoustic vibration signal of power equipment,which is able to help power equipment inspection staff carry out power equipment fault detection.