CN103048593A

CN103048593A - Identification method of insulation defect type of gas-insulated switchgear

Info

Publication number: CN103048593A
Application number: CN2012105352760A
Authority: CN
Inventors: 王吉文; 高峻; 刘昌界; 李燕; 肖拥东; 国伟辉
Original assignee: State Grid Corp of China SGCC; Haozhou Power Supply Co of State Grid Anhui Electric Power Co Ltd
Current assignee: State Grid Corp of China SGCC; Bozhou Power Supply Co of State Grid Anhui Electric Power Co Ltd
Priority date: 2012-12-12
Filing date: 2012-12-12
Publication date: 2013-04-17
Anticipated expiration: 2032-12-12
Also published as: CN103048593B

Abstract

The invention discloses a method for identifying insulation defect types of gas-insulated switchgear, which comprises the steps of: 1) collecting partial discharge ultra-high frequency signals of gas-insulated switchgear; 2) filtering and reducing noise of partial discharge ultra-high-frequency signals; 3. ) Obtain the envelope of the partial discharge ultra-high frequency signal; 4) Extract the pulse peak V _top from the envelope, the first valley V _v1 after the peak, the first peak V _p1 after the peak, and the second valley V _v2 after the peak Five eigenvalues of peak V _p2 appear after the sum peak; 5) Use the above five eigenvalues to form a set of feature quantities, and take the three feature quantities with the highest discrimination in the set of feature quantities to construct a three-dimensional feature space; 6) Use The sample data of the partial discharge signal corresponding to the typical partial discharge type is used to train the sample space training classifier; 7) the trained sample space training classifier is used to classify and identify the partial discharge type mapped to the three-dimensional feature space.

Description

A method for identifying insulation defect types of gas insulated switchgear

技术领域technical field

本发明涉及一种信号检测方法，具体涉及一种GIS局部放电信号检测方法。The invention relates to a signal detection method, in particular to a GIS partial discharge signal detection method.

背景技术Background technique

气体绝缘开关设备（Gas Insulated Switchgear，GIS）设备以其占地面积小，可靠性高等优点，已在电力系统中得到了广泛的应用。气体绝缘组合电器在绝缘缺陷发生绝缘击穿前会产生局部放电。局部放电是GIS绝缘缺陷的征兆和表现形式。对GIS局部放电的现象进行检测，能够较早发现其内部的绝缘缺陷，以便采取适当措施，从而防止其进一步发展造成事故。Gas insulated switchgear (Gas Insulated Switchgear, GIS) equipment has been widely used in power systems due to its small footprint and high reliability. Gas-insulated combined electrical appliances will generate partial discharges before insulation breakdown occurs due to insulation defects. Partial discharge is a symptom and manifestation of GIS insulation defects. Detecting the phenomenon of partial discharge in GIS can discover its internal insulation defects earlier, so that appropriate measures can be taken to prevent its further development from causing accidents.

GIS设备中的局部放电是由一系列具有极短上升时间的电流脉冲而造成的，并且其发生时会伴随着声、光、化学产物等多种物理化学现象。因此，通过探测这些物理化学现象便能够监测到GIS设备中的局部放电。Partial discharge in GIS equipment is caused by a series of current pulses with a very short rise time, and it will be accompanied by various physical and chemical phenomena such as sound, light, and chemical products. Therefore, partial discharge in GIS equipment can be monitored by detecting these physical and chemical phenomena.

对GIS的局部放电（Partial discharge，PD，以下简称“局放”）现象进行监测和识别，是发现早期绝缘故障、预防事故的重要手段。在监测GIS局放的诸多方法中，超高频（Ultra-High Frequency，UHF）法具有灵敏度高、抗干扰能力强、可识别故障类型和可对局放进行精确定位等优点，是近20年来国内外研究的热点。Monitoring and identifying the partial discharge (PD, hereinafter referred to as "PD") phenomenon of GIS is an important means to discover early insulation faults and prevent accidents. Among the many methods for monitoring partial discharges in GIS, the Ultra-High Frequency (UHF) method has the advantages of high sensitivity, strong anti-interference ability, identifiable fault types and precise positioning of partial discharges. Research hotspots at home and abroad.

GIS中的绝缘缺陷有多种类型，如悬浮电极、自由金属颗粒、绝缘子表面污损等，不同的缺陷产生的放电特征不同，对GIS的危害程度也不一样，为了准确地评估GIS的绝缘状态，正确识别放电的类型至关重要。There are many types of insulation defects in GIS, such as suspended electrodes, free metal particles, insulator surface contamination, etc. Different defects have different discharge characteristics, and the degree of harm to GIS is also different. In order to accurately evaluate the insulation state of GIS , it is crucial to correctly identify the type of discharge.

目前常用于识别局放类型的方法主要是基于相位分布图谱，这种方法没有局放脉冲波形信息，另外还需要提供工频相位同步信息。At present, the method commonly used to identify the type of PD is mainly based on the phase distribution spectrum. This method has no PD pulse waveform information, and also needs to provide power frequency phase synchronization information.

发明内容Contents of the invention

本发明的目的是提供一种气体绝缘开关设备绝缘缺陷种类的识别方法，在不需要工频相位同步信息的情况下，该识别方法基于局放脉冲的波形，将局部放电放的UHF信号（频率范围为300MHz-3000MHz）作为由高频信号调制的低频脉冲信号解调出低频信号以获得局部放电信号的包络线，从而从局部放电超高频包络线中提取出相应的包络特征，并通过识别模式来实现超高频局部放电类型的识别，其能快速准确地判断GIS局部放电的类型，有效地提高了诊断GIS绝缘缺陷的效率及准确度，对于评估GIS的绝缘状态及制定合理的维修策略至关重要，避免由于GIS绝缘故障而导致的安全事故的发生。The object of the present invention is to provide a method for identifying the type of insulation defect of gas insulated switchgear. In the case of no need for power frequency phase synchronization information, the identification method is based on the waveform of the partial discharge pulse, and the UHF signal (frequency The range is 300MHz-3000MHz) as the low-frequency pulse signal modulated by the high-frequency signal to demodulate the low-frequency signal to obtain the envelope of the partial discharge signal, so as to extract the corresponding envelope features from the partial discharge ultra-high frequency envelope, And realize the identification of UHF partial discharge type through the identification mode, which can quickly and accurately judge the type of GIS partial discharge, effectively improve the efficiency and accuracy of diagnosing GIS insulation defects, and help evaluate the insulation state of GIS and formulate reasonable The maintenance strategy is very important to avoid safety accidents caused by GIS insulation failure.

为了实现上述发明目的，本发明提供了一种气体绝缘开关设备绝缘缺陷种类的识别方法，包括下列步骤：In order to achieve the purpose of the above invention, the present invention provides a method for identifying the type of insulation defect of a gas insulated switchgear, comprising the following steps:

（1）采集气体绝缘开关设备的局部放电超高频信号；(1) Collect partial discharge UHF signals of gas insulated switchgear;

（2）对采集到的局部放电超高频信号进行滤波降噪以获得高信噪比的局部放电超高频信号；(2) Filter and denoise the collected partial discharge UHF signal to obtain a partial discharge UHF signal with a high signal-to-noise ratio;

（3）获得局部放电超高频信号的包络线；(3) Obtain the envelope of the partial discharge UHF signal;

由于采集到的局部放电超高频信号表现为高频振荡信号，为了能够得到局部放电超高频信号在时域中的总体趋势，则需要获得局部放电超高频信号的包络线，获得局部放电超高频信号的包络线可以采用希尔伯特变换法、检波—滤波法、高通绝对值解调法和样条曲线法等；Since the collected partial discharge UHF signal appears as a high-frequency oscillation signal, in order to obtain the overall trend of the partial discharge UHF signal in the time domain, it is necessary to obtain the envelope of the partial discharge UHF signal and obtain the local The envelope curve of the discharge UHF signal can use Hilbert transform method, detection-filter method, high-pass absolute value demodulation method and spline curve method, etc.;

（4）从包络线中提取脉冲峰值V_top，峰后初现波谷V_v1，峰后初现波峰V_p1，峰后次现波谷V_v2和峰后次现波峰V_p2五个特征值；(4) Extract five eigenvalues of the pulse peak value V _top , the first after-peak valley V _v1 , the first after-peak peak V _p1 , the second after-peak valley V _v2 and the second after-peak peak V _p2 from the envelope;

不同类型的局部放电超高频信号的包络线在外形上均相似于振荡衰减的双指数函数，但是振荡频率、衰减时间常数外往往会存在着明显的差异，此时则需要通过提取几个关键点的数据来识别局部放电的类型，本技术方案提取脉冲峰值V_top，峰后初现波谷V_v1，峰后初现波峰V_p1，峰后次现波谷V_v2和峰后次现波峰V_p2这五个特征值；The envelopes of different types of partial discharge UHF signals are similar to the double exponential function of oscillation attenuation in appearance, but there are often obvious differences in oscillation frequency and attenuation time constant. At this time, it is necessary to extract several The key point data is used to identify the type of partial discharge. This technical solution extracts the pulse peak V _top , the first post-peak valley V _v1 , the post-peak initial peak V _p1 , the post-peak second valley V _v2 and the post-peak second peak V The five eigenvalues of _p2 ;

（5）用上述五个特征值构建特征量集合在特征量集合

中取区分度最高的三个特征量构建一三维特征空间{x，y,z}；(5) Use the above five eigenvalues to construct a feature set in the set of features

Take the three feature quantities with the highest discrimination to construct a three-dimensional feature space {x, y, z};

为了获得区分效果最佳的三个特征量，需要计算上述四个特征量的区分度D_ab：In order to obtain the three feature quantities with the best distinguishing effect, it is necessary to calculate the discrimination degree D _ab of the above four feature quantities:

${D D.}_{ab ab} = = | | \frac{{\overset{&OverBar; &OverBar;}{λ λ}}_{a a} - - {\overset{&OverBar; &OverBar;}{λ λ}}_{b b}}{{σ σ}_{a a} + + {σ σ}_{b b}} | | - - - - - - ((11))$

式中，D_ab为同一特征量上随机变量a、b的区分度，

分别代表了任意两种不同的放电类型的某一相同特征量的随机变量的样本平均值，σ_a、σ_b分别代表了任意两种不同的放电类型的某一相同特征量的随机变量的无偏样本标准差，（例如，要计算特征量

对沿面放电和悬浮电位这两种放电信号的区分度，则沿面放电信号的

和悬浮电位放电信号的

分别对应公式中的a、b）其中，通过采集到的n个a的样本为{a₁,a₂,a₃,a₄…a_n}，m个b的样本为{b₁,b₂,b₃,b₄…b_m}，可以求得：In the formula, D _ab is the discrimination degree of random variables a and b on the same feature quantity,

Represent the sample mean value of a random variable with the same characteristic quantity of any two different discharge types, σ _a and σ _b respectively represent the random variables of a certain characteristic quantity of any two different discharge types Partial sample standard deviation, (for example, to calculate the feature quantity

For the discrimination of the two kinds of discharge signals, the surface discharge and the floating potential, the surface discharge signal

and floating potential discharge signal of the

Corresponding to a and b in the formula respectively) Among them, the collected n samples of a are {a ₁ , a ₂ , a ₃ , a ₄ …a _n }, and m samples of b are {b ₁ , b ₂ ,b ₃ ,b ₄ …b _m }, we can get:

a的样本平均值为：The sample mean of a is:

${\overset{&OverBar; &OverBar;}{λ λ}}_{a a} = = \frac{11}{n no} {Σ Σ}_{i i = = 11}^{n no} {a a}_{i i},,$

b的样本平均值为：The sample mean of b is:

${\overset{&OverBar; &OverBar;}{λ λ}}_{b b} = = \frac{11}{m m} {Σ Σ}_{i i = = 11}^{m m} {b b}_{i i},,$

a的无偏样本标准差为：The unbiased sample standard deviation of a is:

${σ σ}_{a a} = = \sqrt{\frac{{Σ Σ}_{i i = = 11}^{n no} {(({a a}_{i i} - - {\overset{&OverBar; &OverBar;}{λ λ}}_{a a}))}^{22}}{n no - - 11}},,$

b的无偏样本标准差为：The unbiased sample standard deviation of b is:

${σ σ}_{b b} = = \sqrt{\frac{{Σ Σ}_{i i = = 11}^{m m} {(({b b}_{i i} - - {\overset{&OverBar; &OverBar;}{λ λ}}_{b b}))}^{22}}{m m - - 11}},,$

将求得的

σ_a及σ_b代入式（1）以获得区分度D_ab值。will seek

σ _a and σ _b are substituted into formula (1) to obtain the value of discrimination D _ab .

区分度D_ab的取值越大，则说明同一特征量上的两种不同放电类型出现混淆的可能性就越小。The larger the value of the discrimination degree D _ab is, the less likely it is that there will be confusion between two different discharge types on the same feature quantity.

也就是说，本技术方案中，分别将不同局放类型的相同特征量两两组合，并求取该相同的特征量对于这选取的两种局部放电类型的区分度D_ab。为了构造出能区别不同放电类型的三维特征空间，需要在选取特征量时遵循“最劣不入”的原则，即尽量避免采用对区分某一个或几个局放类型表现最差的特征量，基于该原则将最差特征量剔除的步骤为：先把不同放电类型两两成对，再就各对中的某一相同特征量进行区分度D_ab计算，特征量为四个，即为及

接着从四个特征量的区分度中任意选取三个进行求和以构成不同的求和组合，然后将不同求和组合的区分度对应于相应的局放类型对形成列表，最后对于同一局放类型对中不同求和组合进行对比以将每一局放类型对中“区分最劣”的求和组合予以剔除，保留求和组合数值高的其他三个特征量以构成三维特征空间的参量{x，y，z}。由于区分度D_ab求和数值越大越能区别不同放电类型，因此，当某一局放类型对中的某一求和组合的数值越小时，说明该求和组合中所剔除的特征量越是能够区分不同的放电类型，反之也亦然。That is to say, in this technical solution, the same characteristic quantities of different PD types are combined in pairs, and the degree of discrimination D _ab of the same characteristic quantity for the two selected partial discharge types is calculated. In order to construct a three-dimensional feature space that can distinguish different discharge types, it is necessary to follow the principle of "the worst does not enter" when selecting feature quantities, that is, try to avoid using the feature quantity that is the worst for distinguishing one or several PD types. Based on this principle, the steps to eliminate the worst feature quantity are: first pair different discharge types in pairs, and then calculate the discrimination degree D _ab of a certain same feature quantity in each pair. There are four feature quantities, that is, and

Then, three of the discrimination degrees of the four feature quantities are arbitrarily selected to be summed to form different summation combinations, and then the discrimination degrees of different summation combinations correspond to the corresponding PD type pairs to form a list, and finally for the same PD Different summation combinations in the type pairs are compared to eliminate the summation combination that "distinguishes the worst" in each PD type pair, and the other three feature quantities with high summation combination values are retained to form the parameters of the three-dimensional feature space{ x,y,z}. Since the greater the value of the summation value of the discrimination degree D _ab , the more it can distinguish different discharge types. Therefore, when the value of a certain summation combination in a certain PD type pair is smaller, it means that the feature quantity eliminated in the summation combination is more Ability to distinguish between different discharge types and vice versa.

区分度概念和计算是本领域内的技术人员所熟知的，故本文不在此进行进一步详细的介绍。The concept and calculation of discrimination are well known to those skilled in the art, so no further detailed introduction will be made here.

（6）采用对应典型局部放电类型的局部放电信号的样本数据对样本空间训练分类器进行训练；(6) Using sample data of partial discharge signals corresponding to typical partial discharge types to train the sample space training classifier;

（7）采用经过训练的样本空间训练分类器对映射到三维特征空间{x，y,z}内的局部放电类型进行分类和识别。(7) Use the trained sample space to train the classifier to classify and identify the PD types mapped into the three-dimensional feature space {x, y, z}.

进一步地，在上述步骤（2）中，采用通带为300MHz～850MHz的带通滤波器进行滤波降噪。在滤波过程中采用通带为300MHz～850MHz的带通滤波器的原因在于：1）不同类型的局放信号在300MHz以下的频率成分相似，不具备区分不同局放类型的操作性；2）某些局部放电的信号频谱中在幅值最大处（约870MHz左右）实质上为噪声频带；3）大约在900MHz以上的频率分量对信号时域波形包络影响不大，不能为获得局部放电超高频信号的包络线提供有效的数值依据。Further, in the above step (2), a band-pass filter with a pass band of 300 MHz to 850 MHz is used for filtering and noise reduction. The reasons for using a band-pass filter with a passband of 300MHz to 850MHz in the filtering process are: 1) Different types of PD signals have similar frequency components below 300MHz, which does not have the operability to distinguish different types of PD; In some partial discharge signal spectrums, the maximum amplitude (about 870MHz) is essentially a noise frequency band; 3) the frequency components above about 900MHz have little effect on the time-domain waveform envelope of the signal, which cannot be used to obtain ultra-high partial discharge The envelope curve of the frequency signal provides an effective numerical basis.

进一步地，在步骤（2）中，采用12阶巴特沃斯带通滤波器进行滤波降噪。Further, in step (2), a 12-order Butterworth bandpass filter is used for filtering and noise reduction.

进一步地，在步骤（3）中，采用希尔伯特变换法获得局部放电超高频信号的包络线。Further, in step (3), the envelope of the partial discharge ultra-high frequency signal is obtained by using the Hilbert transform method.

当然，也可以采用检波—滤波法得到包络线，但是采用这种方法得到的是信号正半周中线的包络线，其没有采用希尔伯特变换法得到的包络线准确性高。Of course, the detection-filtering method can also be used to obtain the envelope curve, but the envelope curve obtained by this method is the envelope curve of the center line of the positive half cycle of the signal, which is not as accurate as the envelope curve obtained by the Hilbert transform method.

另外，也可以采用高通绝对值法得到包络线，但是采用这种方法得到的包络线也是信号中线的包络线，其准确性也没有采用希尔伯特变换法得到的包络线准确性高。In addition, the high-pass absolute value method can also be used to obtain the envelope, but the envelope obtained by this method is also the envelope of the signal center line, and its accuracy is not as accurate as the envelope obtained by the Hilbert transform method. high sex.

此外，还可以采用样条曲线法来得到包络线，但是样条曲线法的插值点选取原则比较难以确定，且该算法对于不同信号的适应性不强。In addition, the spline method can also be used to obtain the envelope, but the selection principle of the interpolation point of the spline method is difficult to determine, and the adaptability of the algorithm to different signals is not strong.

希尔伯特变换法也是本领域内技术人员所熟知的，故本文在此仅做简单介绍：The Hilbert transform method is also well known to those skilled in the art, so this article only briefly introduces it here:

对于一个连续时域信号x(t)，其与的卷积为For a continuous time-domain signal x(t), it is the same as The convolution of

$\overset{^^}{x x} ((t t)) = = H h [[x x ((t t))]] = = x x ((t t)) * * h h ((t t)) = = \frac{11}{π π} {&Integral; &Integral;}_{- - \infty \infty}^{+ + \infty \infty} x x ((τ τ)) \cdot \cdot \frac{11}{t t - - τ τ} dτ dτ$

连续时域信号x(t)的解析信号为The analytical signal of the continuous time domain signal x(t) is

$a a ((t t)) = = x x ((t t)) + + j j \overset{^^}{x x} ((t t))$

其中，j表示虚数单位；Among them, j represents the imaginary unit;

那么该解析信号的模为Then the modulus of the analytical signal is

$E E. ((t t)) = = | | a a ((t t)) | | = = \sqrt{{x x}^{22} ((t t)) + + {\overset{^^}{x x}}^{22} ((t t))},,$

该模就是连续时域信号x(t)的包络线。This modulus is the envelope of the continuous time-domain signal x(t).

对于长度为m的离散信号序列x(n)及其FFT（快速傅里叶变换）序列X(k)之间的关系可表示为：For a discrete signal sequence x(n) of length m and its FFT (Fast Fourier Transform) sequence X(k), the relationship can be expressed as:

$A A ((k k)) = = \{\begin{matrix} X x ((k k)),, & k k = = 00 \\ 22 X x ((k k)),, & k k = = 1,2,3 1,2,3,, . . . . . .,, \frac{m m}{22} - - 11 \\ 00 & k k = = \frac{m m}{22},, \frac{m m}{22} + + 11,, . . . . . .,, m m - - 11 \end{matrix}$

其中，A(k)为x(n)的离散解析信号a(n)对应的FFT序列，将式中所得的A(k)经过IFFT（逆快速傅里叶变换）处理后代入式

中求模值可以得到包络线E(n)：Among them, A(k) is the FFT sequence corresponding to the discrete analytical signal a(n) of x(n), and the A(k) obtained in the formula is processed by IFFT (inverse fast Fourier transform) and then substituted into the formula

The envelope E(n) can be obtained by calculating the modulus value in:

E(n)＝|a(n)|＝|IFFT[A(k)]|。E(n)=|a(n)|=|IFFT[A(k)]|.

进一步地，在步骤（6）中，样本空间训练分类器采用神经网络或支持向量机。Further, in step (6), the classifier is trained in the sample space using a neural network or a support vector machine.

更进一步地，在步骤（6）中，典型局部放电类型包括：沿面放电信号、悬浮电位信号和金属微粒信号的至少其中之一。Furthermore, in step (6), typical partial discharge types include: at least one of creeping discharge signals, floating potential signals and metal particle signals.

本发明所述的气体绝缘开关设备绝缘缺陷种类的识别方法，基于局放脉冲的波形且不需要工频相位同步信息，该识别方法所具有优点如下：The method for identifying insulation defect types of gas insulated switchgear according to the present invention is based on the waveform of the partial discharge pulse and does not require power frequency phase synchronization information. The advantages of the identification method are as follows:

（1）能准确地分类和区别气体绝缘开关设备的不同放电类型；(1) Can accurately classify and distinguish different discharge types of gas insulated switchgear;

（2）极大地提升了判断气体绝缘开关设备绝缘缺陷的效率；(2) Greatly improved the efficiency of judging insulation defects of gas insulated switchgear;

（3）为气体绝缘开关设备故障维修策略提供重要依据；(3) Provide an important basis for the fault maintenance strategy of gas insulated switchgear;

（4）及时获得气体绝缘开关设备绝缘缺陷的诊断结果以避免重大安全事故的发生。(4) Obtain the diagnosis results of insulation defects of gas-insulated switchgear in time to avoid major safety accidents.

附图说明Description of drawings

图1显示了沿面放电绝缘故障的局部放电超高频信号的频谱分析。Fig. 1 shows the spectrum analysis of partial discharge UHF signal for creeping discharge insulation fault.

图2显示了悬浮电位绝缘故障的局部放电超高频信号的频谱分析。Figure 2 shows the spectral analysis of the partial discharge UHF signal for a floating potential insulation fault.

图3显示了金属微粒绝缘故障的局部放电超高频信号的频谱分析。Fig. 3 shows the spectrum analysis of the partial discharge UHF signal of the metal particle insulation fault.

图4显示了经过预处理后的沿面放电的局部放电超高频信号。Figure 4 shows the partial discharge UHF signal of creeping discharge after preprocessing.

图5显示了经过预处理后的悬浮电位的局部放电超高频信号。Figure 5 shows the PD UHF signal of the levitating potential after preprocessing.

图6显示了经过预处理后的金属微粒的局部放电超高频信号。Figure 6 shows the partial discharge UHF signal of the pretreated metal particles.

图7显示了局部放电超高频信号的包络线中脉冲峰值V_top，峰后初现波谷V_v1，峰后初现波峰V_p1，峰后次现波谷V_v2和峰后次现波峰V_p2等五个特征值。Figure 7 shows the pulse peak value V _top , the first post-peak valley V _v1 , the post-peak initial peak V _p1 , the post-peak secondary peak V _v2 and the post-peak secondary peak V in the envelope of the partial discharge UHF signal Five eigenvalues such as _p2 .

图8显示了三种局部放电超高频信号的特征量

的取值。Figure 8 shows the characteristic quantities of three partial discharge UHF signals

value of .

图9显示了三种局部放电超高频信号的特征量的取值。Figure 9 shows the characteristic quantities of three partial discharge UHF signals value of .

图10显示了三种局部放电超高频信号的特征量

的取值。Figure 10 shows the characteristic quantities of three partial discharge UHF signals

value of .

图11显示了三种局部放电超高频信号的特征量的取值。Figure 11 shows the characteristic quantities of three partial discharge UHF signals value of .

图12显示了三种局部放电类型的局部放电信号的样本数据三维特征空间内的分布。Fig. 12 shows the distribution in the three-dimensional feature space of the sample data of the partial discharge signals of the three partial discharge types.

具体实施方式Detailed ways

以下结合具体实施例和说明书附图来对本发明所述的气体绝缘开关设备绝缘缺陷种类的识别方法做进一步的解释说明，但是该解释说明并不构成对本发明所涉及的技术方案的不当限定。The method for identifying the type of insulation defect of the gas insulated switchgear according to the present invention will be further explained below in combination with specific embodiments and accompanying drawings, but the explanation does not constitute an improper limitation of the technical solution involved in the present invention.

在本实施例中涉及了三种典型的气体绝缘开关设备的绝缘故障，分别为沿面放电、悬浮电位和金属微粒，这三种典型绝缘故障的局部放电信号的频谱分析如图1至图3所示。通过图1至图3可以看出，三种典型的局部放电类型信号在300MHz以下的频率成分相似，不具备区分不同局部放电类型的操作性；由金属微粒和沿面放电所导致的局部放电的信号频谱中幅值最大处（约870MHz左右）实质上为噪声频带；三种典型的局放类型信号大约在900MHz以上的频率分量对信号时域波形包络影响不大，不能为获得局部放电超高频信号的包络线提供有效的数值依据。In this embodiment, three typical insulation faults of gas-insulated switchgear are involved, namely creeping discharge, floating potential and metal particles. The frequency spectrum analysis of partial discharge signals of these three typical insulation faults is shown in Figures 1 to 3 Show. It can be seen from Figures 1 to 3 that the three typical partial discharge types have similar frequency components below 300MHz, and do not have the operability to distinguish different partial discharge types; the signals of partial discharge caused by metal particles and surface discharge The maximum amplitude in the frequency spectrum (about 870MHz) is essentially a noise frequency band; the frequency components of the three typical PD types above about 900MHz have little effect on the time-domain waveform envelope of the signal, and cannot be used to obtain ultra-high partial discharge. The envelope curve of the frequency signal provides an effective numerical basis.

采用本发明所述的技术方案用来对气体绝缘开关设备的上述三种绝缘缺陷种类进行识别的步骤如下：The steps for identifying the above three types of insulation defects of the gas insulated switchgear by adopting the technical solution described in the present invention are as follows:

（1）通过超高频电磁波信号传感器采集气体绝缘开关设备的局部放电超高频信号。(1) The partial discharge ultra-high frequency signal of the gas insulated switchgear is collected by the ultra-high frequency electromagnetic wave signal sensor.

（2）为了获得高信噪比的局部放电超高频信号，对采集到的局部放电超高频信号采用通带为300MHz～850MHz的12阶巴特沃斯带通滤波器进行滤波降噪处理；经过滤波降噪后的沿面放电，悬浮电位及金属微粒的局部放电超高频信号如图4至6所示。(2) In order to obtain partial discharge ultra-high frequency signals with high signal-to-noise ratio, the collected partial discharge ultra-high frequency signals are filtered and noise-reduced using a 12-order Butterworth band-pass filter with a passband of 300MHz to 850MHz; After filtering and noise reduction, the UHF signals of surface discharge, suspended potential and partial discharge of metal particles are shown in Figures 4 to 6.

（3）为了能够获得局部放电超高频信号在时域中的总体趋势，采用希尔伯特变换法来获得局部放电超高频信号的包络线。(3) In order to obtain the overall trend of the partial discharge UHF signal in the time domain, the Hilbert transform method is used to obtain the envelope of the partial discharge UHF signal.

（4）分别从沿面放电，悬浮电位及金属微粒的局部放电超高频信号的包络线中提取脉冲峰值V_top，峰后初现波谷V_v1，峰后初现波峰V_p1，峰后次现波谷V_v2和峰后次现波峰V_p2五个特征值，如图7所示。(4) Extract the pulse peak value V _top , initial post-peak trough V _v1 , post-peak initial post-peak V _p1 , and post-peak sub-peak from the envelopes of creeping discharge, suspension potential and partial discharge UHF signals of metal particles respectively. There are five eigenvalues of V _v2 of the current trough and V _p2 of the current peak after the peak, as shown in Figure 7.

（5）用步骤（4）中的五个特征值构建特征量集合

图8至图11显示了集合中四个特征量的分布图，图8至图11中的1，2和3分别代表沿面放电，悬浮电位和金属微粒三种缺陷的局部放电信号。(5) Use the five eigenvalues in step (4) to construct a set of feature quantities

Figures 8 to 11 show the distribution diagrams of the four characteristic quantities in the set, and 1, 2 and 3 in Figures 8 to 11 represent the partial discharge signals of surface discharge, suspension potential and metal particle defects, respectively.

(6)在特征量集合中取区分度最高的三个特征量构建三维特征空间{x，y,z}的参量，区分度

其中，分别代表了任意两种不同的放电类型的某一相同特征量的随机变量的样本平均值，σ_a、σ_b分别代表了任意两种不同的放电类型的某一相同特征量的随机变量的无偏样本标准差，特征量集合

中各特征量的随机变量的样本平均值和无偏样本标准差如表1所示。通过采集到的n个a的样本为{a₁,a₂,a₃,a-₄…a_n}，m个b的样本为{b₁,b₂,b₃,b₄…b_m}，可以求得a的样本平均值为

b的样本平均值为

并且a的无偏样本标准差为

σ_{a} = \sqrt{\frac{Σ_{i = 1}^{n} {(a_{i} - {\overset{&OverBar;}{λ}}_{a})}^{2}}{n - 1}},

b的无偏样本标准差为

σ_{b} = \sqrt{\frac{Σ_{i = 1}^{m} {(b_{i} - {\overset{&OverBar;}{λ}}_{b})}^{2}}{m - 1}} .

将三种局部放电类型两两成对，即沿面-悬浮，悬浮-金属和金属-沿面，将其对应各特征量计算相应的区分度，得到表2。接着从四个特征量的区分度D_ab值中任意选取三个进行求和以构成不同的求和组合，然后将不同求和组合的区分度D_ab值对应于相应的每一对局放类型，得到表3。表3中的粗斜体数据表示区分度最劣的数据。因此，在本实施例中，能够区分这三种局部放电类型的特征量为

及

采用这三个特征量构成三维特征空间{x，y,z}。(6) In the feature set Take the three feature quantities with the highest discrimination to construct the parameters of the three-dimensional feature space {x, y, z}, the discrimination

in, Represent the sample mean value of a random variable with the same characteristic quantity of any two different discharge types, σ _a and σ _b respectively represent the random variables of a certain characteristic quantity of any two different discharge types Partial sample standard deviation, feature set

Table 1 shows the sample mean and unbiased sample standard deviation of the random variables of each characteristic quantity in . The collected n samples of a are {a ₁ , a ₂ , a ₃ , a- ₄ …a _n }, and m samples of b are {b ₁ , b ₂ , b ₃ , b ₄ …b _m } , the sample mean value of a can be obtained as

The sample mean of b is

And the unbiased sample standard deviation of a is

σ_{a} = \sqrt{\frac{Σ_{i = 1}^{no} {(a_{i} - {\overset{&OverBar;}{λ}}_{a})}^{2}}{no - 1}},

The unbiased sample standard deviation of b is

σ_{b} = \sqrt{\frac{Σ_{i = 1}^{m} {(b_{i} - {\overset{&OverBar;}{λ}}_{b})}^{2}}{m - 1}} .

The three types of partial discharge are paired in pairs, that is, surface-suspension, suspension-metal and metal-surface, and the corresponding discrimination degree is calculated according to each characteristic quantity, and Table 2 is obtained. Then select three arbitrarily summed values _of the degree of discrimination D _ab of the four feature quantities to form different summation combinations, and then correspond to each corresponding pair of PD types , get Table 3. Data in bold italics in Table 3 represent the data with the worst discrimination. Therefore, in this embodiment, the characteristic quantities that can distinguish these three types of partial discharge are

and

The three-dimensional feature space {x, y, z} is formed by using these three feature quantities.

（7）采用沿面放电，悬浮电位和金属微粒三种局部放电类型的局部放电信号的样本数据对一三层BP（Back Propagation）神经网络，即误差反传神经网络进行训练。(7) A three-layer BP (Back Propagation) neural network, that is, an error backpropagation neural network, is trained using the sample data of partial discharge signals of three types of partial discharge, creeping discharge, floating potential and metal particles.

神经网络的训练过程是本领域内技术人员所熟知的，故本文在此不再赘述具体的训练过程。The training process of the neural network is well known to those skilled in the art, so the specific training process will not be repeated here.

（8）采用经过训练的三层BP神经网络对映射到三维特征空间{x，y,z}内的三种局部放电类型，即沿面放电，悬浮电位和金属微粒的局部放电类型进行分类，如图12所示，发现特征量的随机数据已明显地聚为三个区域，即可对于上述三种局部放电类型进行识别，识别结果如表4所示。(8) Use the trained three-layer BP neural network to classify the three types of partial discharges mapped into the three-dimensional feature space {x, y, z}, namely, creeping discharges, suspended potentials and partial discharges of metal particles, such as As shown in Fig. 12, it is found that the random data of the characteristic quantity has been obviously clustered into three regions, and the above three types of partial discharge can be identified, and the identification results are shown in Table 4.

表1三种局部放电的特征量集合中各特征量的随机变量的样本平均值和无偏样本标准差Table 1 The sample mean and unbiased sample standard deviation of the random variables of each characteristic quantity in the three kinds of partial discharge characteristic quantity sets

表2三种局部放电成对中的各特征量的区分度D_ab数值Table 2 Discrimination degree D _ab value of each characteristic quantity in three kinds of partial discharge pairs

特征量Feature amount 沿面-悬浮surface-suspension 悬浮-金属suspension - metal 金属-沿面metal-surface V_v1/V_top V _v1 /V _top 1.22851.2285 0.65280.6528 1.69871.6987 V_p1/V_top V _p1 /V _top 0.17160.1716 1.01491.0149 1.47251.4725 V_v2/V_top V _v2 /V _top 1.00821.0082 0.80890.8089 1.63891.6389 V_p2/V_top V _p2 /V _top 1.49321.4932 0.73100.7310 0.06970.0697

表3各局放类型对中不同区分度求和组合的数值Table 3 Values of sum combinations of different discrimination degrees in each PD type pair

表4三种局部放电类型在三层BP神经网络的混淆矩阵中的识别率Table 4 The recognition rates of the three types of partial discharge in the confusion matrix of the three-layer BP neural network

由表4可知，采用本发明所述的气体绝缘开关设备绝缘缺陷种类的识别方法效果较好，能快速准确地识别不同的放电类型。It can be seen from Table 4 that the method for identifying insulation defect types of gas insulated switchgear according to the present invention has a better effect, and can quickly and accurately identify different discharge types.

要注意的是，以上列举的仅为本发明的具体实施例，显然本发明不限于以上实施例，随之有着许多的类似变化。本领域的技术人员如果从本发明公开的内容直接导出或联想到的所有变形，均应属于本发明的保护范围。It should be noted that the above examples are only specific embodiments of the present invention, and obviously the present invention is not limited to the above embodiments, and there are many similar changes accordingly. All modifications directly derived or associated by those skilled in the art from the content disclosed in the present invention shall belong to the protection scope of the present invention.

Claims

1. A method for identifying the type of insulation defect of gas insulated switchgear is characterized by comprising the following steps:

(1) collecting a local discharge ultrahigh frequency signal of the gas insulated switchgear;

(2) filtering and denoising the collected partial discharge ultrahigh frequency signal;

(3) obtaining an envelope line of a partial discharge ultrahigh frequency signal;

(4) extracting the pulse peak value V from the envelope curve_topInitial trough V after peak_v1First appearance of wave crest after peakV_p1Second occurrence of trough V after peak_v2Secondary peak V after peak summation_p2Five eigenvalues;

(5) constructing a feature quantity set by using the five feature values

In feature quantity set

Three characteristic quantities with highest distinguishing degrees are taken to construct a three-dimensional characteristic space { x, y, z };

(6) training a sample space training classifier by adopting sample data of a partial discharge signal corresponding to a typical partial discharge type;

(7) and (3) adopting a trained sample space to train a classifier to classify and identify the partial discharge type mapped into the three-dimensional characteristic space { x, y, z }.

2. The method for identifying the kind of insulation defect of the gas insulated switchgear according to claim 1, wherein: in the step (2), a band-pass filter with a passband of 300 MHz-850 MHz is adopted for filtering and noise reduction.

3. The method for identifying the kind of insulation defect of the gas insulated switchgear according to claim 2, wherein: in the step (2), a 12-order Butterworth band-pass filter is adopted for filtering and denoising.

4. The method for identifying the type of the insulation defect of the gas insulated switchgear as claimed in claim 1, wherein in the step (3), the envelope of the partial discharge UHF signal is obtained by using a Hilbert transform method.

5. The method for identifying the type of the insulation defect of the gas insulated switchgear according to claim 1, wherein in the step (6), the sample space training classifier uses a neural network or a support vector machine.

6. The method for identifying the kind of insulation defect of gas insulated switchgear according to claim 1, wherein in the step (6), the typical partial discharge type includes: at least one of a creeping discharge signal, a floating potential signal, and a metal particle signal.