CAJ | 학술논문

电气设备油中局部放电超声阵列定位是将阵列信号处理技术应用于局放超声检测中的一种新方法，其实质是通过相应的算法得到信号的方位信息，继而对局放源进行定位。研究表明，当超声传感器接收到直达信号时，可对局放源进行准确定位；当接收到混叠(多路径)信号时，会导致定位成功率和定位精度下降，甚至产生虚假定位。因此，有效识别超声直达波与混叠波信号是局放定位成功的关键。该文选取峰值因数、分形盒维数和李雅普指数作为局放超声直达波信号的特征参量；分别基于各特征量进行超声直达波识别的仿真研究，结果表明超声信号的混叠程度会对识别结果造成很大影响，利用BP人工神经网络，提出一种基于多特征量的电气设备油中局放超声直达波识别方法，仿真结果显示其识别成功率接近100%；最后在设定环境下进行油中局放超声直达波识别的实验研究，结果表明，采用单个特征量识别直达波和混叠波的成功率约为90%，基于多特征量的识别成功率为95%，验证了所提方法的有效性。
전기설비유중국부방전초성진렬정위시장진렬신호처리기술응용우국방초성검측중적일충신방법，기실질시통과상응적산법득도신호적방위신식，계이대국방원진행정위。연구표명，당초성전감기접수도직체신호시，가대국방원진행준학정위；당접수도혼첩(다로경)신호시，회도치정위성공솔화정위정도하강，심지산생허가정위。인차，유효식별초성직체파여혼첩파신호시국방정위성공적관건。해문선취봉치인수、분형합유수화리아보지수작위국방초성직체파신호적특정삼량；분별기우각특정량진행초성직체파식별적방진연구，결과표명초성신호적혼첩정도회대식별결과조성흔대영향，이용BP인공신경망락，제출일충기우다특정량적전기설비유중국방초성직체파식별방법，방진결과현시기식별성공솔접근100%；최후재설정배경하진행유중국방초성직체파식별적실험연구，결과표명，채용단개특정량식별직체파화혼첩파적성공솔약위90%，기우다특정량적식별성공솔위95%，험증료소제방법적유효성。
The ultrasonic array position of the partial discharge (PD) in oil is a new method to PD detection. Research suggested that, when the sensor received a direct signal, the PD could be accurately positioned; when the sensor received an aliasing signal, the success rates were reduced. Therefore, the effective identification of the ultrasonic direct and aliasing waves is key to PD positioning. First, peak factor, fractal box dimension and Lyapunov exponent are extracted as the characteristic parameters of the partial discharge (PD) ultrasonic direct wave in this paper; Second, simulation was done to identify the ultrasonic direct wave and superimposed wave using these characteristic parameters respectively, and the results show that the aliasing degree of ultrasonic signals has a significant impact on identification result; so a method based on BP artificial neural network is used to identify the direct wave which takes all characteristic quantities into consideration, the simulation results show the success rates are closed to 100%; Finally, the direct wave identification experiments are made in set environment, it indicates that the success rate of identification using single characteristic parameter is about 90% while the success rate is 95% using multi-characteristic parameters, which demonstrates the validity of this method.