WO2015070513A1 - 一种三相共筒式超高压gis局部放电的模式识别方法 - Google Patents
一种三相共筒式超高压gis局部放电的模式识别方法 Download PDFInfo
- Publication number
- WO2015070513A1 WO2015070513A1 PCT/CN2014/000766 CN2014000766W WO2015070513A1 WO 2015070513 A1 WO2015070513 A1 WO 2015070513A1 CN 2014000766 W CN2014000766 W CN 2014000766W WO 2015070513 A1 WO2015070513 A1 WO 2015070513A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- partial discharge
- phase
- signal
- spectrum
- pattern recognition
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/12—Testing dielectric strength or breakdown voltage ; Testing or monitoring effectiveness or level of insulation, e.g. of a cable or of an apparatus, for example using partial discharge measurements; Electrostatic testing
- G01R31/1227—Testing dielectric strength or breakdown voltage ; Testing or monitoring effectiveness or level of insulation, e.g. of a cable or of an apparatus, for example using partial discharge measurements; Electrostatic testing of components, parts or materials
- G01R31/1254—Testing dielectric strength or breakdown voltage ; Testing or monitoring effectiveness or level of insulation, e.g. of a cable or of an apparatus, for example using partial discharge measurements; Electrostatic testing of components, parts or materials of gas-insulated power appliances or vacuum gaps
Definitions
- the invention relates to the technical field of high-voltage electric discharge identification, in particular to a pattern recognition method for partial discharge of a three-phase common-tube ultra-high pressure GIS.
- GIS Gas lnsulated Switchgear
- the main defects affecting the performance of insulating media in GIS are: serious installation errors, poor contact between conductors, high-voltage conductor protrusions, fixed particles, insulator defects, steam, etc.
- GIS Global System for Mobile Communications
- the three-phase common-tube GIS has obvious differences in internal structure and electric field distribution from the coaxial GIS.
- the prior art research mainly focuses on the coaxial GIS, but the three-phase common-tube ultra-high pressure GIS partial discharge detection pattern recognition Less research.
- the object of the present invention is to provide a pattern recognition method for three-phase common-tube ultra-high pressure GIS partial discharge in order to achieve the above problems, so as to realize the advantages of multi-function, wide application range and good accuracy.
- the technical solution adopted by the present invention is: a pattern recognition method for partial discharge of a three-phase common-tube ultra-high pressure GIS, comprising the following steps:
- Step 1 Using UHF to detect three-phase common-tube GIS partial discharge, using UHF sensor to sample the partial discharge signal;
- Step 2 Denoising the collected partial discharge signal by using the improved wavelet threshold filtering method to obtain a true partial discharge signal
- Step 3 extracting characteristic parameters of the sampling signal by using a phase analysis mode algorithm
- Step 4 Using the improved kernel principal component analysis method to reduce the dimension of the feature space composed of the feature parameters, and obtain the feature parameter matrix after dimension reduction;
- Step 5 Pattern recognition of GIS insulation defect types using cluster-based K-nearest neighbor classification.
- step 2 the improved wavelet threshold filtering method is used to perform the denoising processing on the collected partial discharge signal, and an adaptive threshold calculation method is specifically adopted; the adaptive threshold calculation method is as follows:
- N j is the number of wavelet coefficients on the scale
- ) is the median of all wavelet coefficients on the scale
- ⁇ is called the signal-to-noise ratio factor, which is the signal The signal-to-noise ratio is reflected in the threshold calculation.
- ⁇ j is called the scale factor. It is the maximum value of the wavelet coefficient on the scale to correct the estimation error caused by the different lengths of the sample sequence.
- T j is the calculated threshold.
- the characteristic parameters include a skewness Sk, a steepness Ku, a local peak number Pe, a cross-correlation coefficient Cc, and a discharge factor Q.
- skewness Sk is specifically:
- w is the number of phase windows in a half cycle
- x i is the phase of the i th phase window
- y i is the ordinate of the spectrum, representing the apparent discharge amount q or the number of discharges n;
- the parameter ⁇ represents the position at the center of the collected partial discharge map, and ⁇ represents the steep condition represented by the symmetry axis at the center of the spectrum, ⁇ x is a parameter related to the average distribution of the partial discharge map. The phase corresponding to a point in the map;
- w is the number of phase windows in a half cycle
- x i is the phase of the i th phase window
- y i is the ordinate of the spectrum, representing the apparent discharge amount q or the number of discharges n;
- the parameter ⁇ represents the position at the center of the collected partial discharge map, and ⁇ represents the steep condition represented by the symmetry axis at the center of the spectrum, ⁇ x is a parameter related to the average distribution of the partial discharge map. The phase corresponding to a point in the map;
- the steepness Ku is used to describe the degree of protrusion of a shape compared to the shape of a normal distribution: the steepness Ku of the normal distribution is 0; if Ku>0, the profile is sharper than the normal profile; If Ku ⁇ 0, the spectral profile is flatter than the normal distribution profile.
- the local peak point number Pe the number of local peak points is used to describe the number of local peaks on the contour of the spectrum; If there is a local peak, it needs to be judged by the following difference equation:
- phase windows the phase axis is equally divided the larger the number of local peaks.
- cross-correlation coefficient Cc is specifically:
- discharge factor Q is specifically:
- the improved kernel principal component analysis method is an improved kernel principal component analysis method
- the sampled kernel function is:
- step 5 the algorithm of the K-nearest neighbor classification method specifically includes:
- Step1 In the training set, firstly map all the partial discharge data to a space vector
- Step2 Starting from the first class, perform two-two similarity calculation on all signal data belonging to this category, set a minimum threshold, and obtain a cluster with similar similarity according to statistics;
- Step3 For each cluster, combine all the signal data, and then calculate its center vector; in addition, calculate the number of clusters/category total, which represents the contribution coefficient of this cluster to this class;
- Step4 After the new text arrives, preprocess it to obtain its vector space
- Step 5 Calculate the distance between the space vector of the new text and the center vector of each cluster generated by Step3, multiply the distances by the contribution coefficients of the corresponding clusters, and add the results of the cluster calculations belonging to the same category, and compare the largest ones.
- One category is the category of the typical defect partial discharge to be classified.
- the pattern recognition method for the partial discharge of the three-phase common-tube ultra-high pressure GIS includes the steps; the ultra-high frequency detection of the three-phase common-tube GIS partial discharge, and the use of the UHF sensor for the partial discharge signal Using the improved wavelet threshold filtering method to denoise the collected local discharge signal to obtain the real partial discharge signal; extract the characteristic parameters of the sampled signal based on the phase analysis mode algorithm; use the improved kernel principal component analysis method Dimensionality processing is performed on the feature space composed of feature parameters to obtain the feature parameter matrix after dimension reduction.
- the K-nearest neighbor classification method based on cluster idea is used to identify the GIS insulation defect type. It can overcome the defects of the prior art and improve the three-phase.
- the accuracy of the common-tube ultra-high pressure GIS partial discharge detection mode recognition can overcome the defects of the prior art, such as less function, small application range and poor accuracy, so as to realize the advantages of multiple functions, wide application range and good accuracy.
- FIG. 1 is a schematic structural view of a partial discharge test device for a gas insulated combined electric appliance in a mode recognition method for a partial discharge of a three-phase common cylinder type ultrahigh pressure GIS according to the present invention
- N and K are both natural numbers
- FIG. 3 is a schematic diagram showing the relationship between the statistical distribution and the Sk and Ku in the pattern recognition method for the partial discharge of the three-phase common-tube ultra-high pressure GIS according to the present invention, wherein (a) is a positive offset, (b) is a non-offset, and (c) is a negative Offset, Sk is the offset of the shape of the map relative to the normal distribution; (d) is a positive offset, (e) is no offset, (f) is a negative offset, and steepness Ku is used to describe a certain shape The distribution is compared to the extent of the protrusion of the normal distribution shape;
- FIG. 4 is a schematic diagram showing the effect of a kernel function in a pattern recognition method for a partial discharge of a three-phase common-tube ultra-high pressure GIS according to the present invention, wherein (a) is a polynomial function, (b) is a Gaussian kernel function, and (c) is a novel kernel function;
- FIG. 5 is a comparison diagram of waveforms before (a) and after (b) before filtering in a pattern recognition method for partial discharge of a three-phase common-tube ultra-high pressure GIS according to the present invention.
- the horizontal axis is 0-360, angle, and vertical axis. Is the signal amplitude (Q).
- a pattern recognition method for partial discharge of a three-phase common-tube ultra-high pressure GIS is provided.
- the partial discharge test device for a gas-insulated combination electric appliance using the three-phase common-tube ultra-high pressure GIS partial discharge pattern recognition method includes a disk insulator 5 and a high voltage mounted on the disk insulator 5.
- the common end of the resistor 1 and the high voltage bushing 4 is grounded through a voltage divider composed of a capacitor.
- the gas-insulated combined electric appliance partial discharge test device mainly comprises a transformer 3, a voltage divider composed of a capacitor, an oscilloscope, a three-phase common-tube gas insulated combined electrical appliance (GIS), a sensor and a partial discharge signal detector (PDSG);
- GIS three-phase common-tube gas insulated combined electrical appliance
- PDSG partial discharge signal detector
- step 1) the three-phase common-tube GIS partial discharge is detected by ultra-high frequency method (UHF).
- UHF ultra-high frequency method
- different fault types can be identified according to the spectral characteristics of the measured signal and the position of the discharge occurring on the power frequency voltage waveform.
- step 1) the partial discharge pattern under typical defect conditions is measured using an actual three-phase common-tube GIS.
- three-phase common-tube GIS should be provided by professional high-voltage switchgear enterprises.
- the three-phase conductor is pressurized in two phases, one connected to a high voltage, and only one high voltage bushing is provided in the model.
- a typical insulation defect with contrast is set.
- step 2) the improved wavelet threshold filtering algorithm is visible by the calculation method of median Median (Cj, k), and the choice of threshold is closely related to the length of the analyzed signal. In practical applications, there is no guarantee that the proportion and position of the effective signal in the entire sampling sequence in the entire sampling sequence is constant. Because the length of the signal determines the number of wavelet coefficients Nj at each scale after wavelet transform, it affects the median, Median (Cj, k), thus affecting the threshold. This effect will lead to the result that the same one has The signal sequence u is included in a sample sequence s. If the s lengths are different (that is, the width of the time observation window is different), the wavelet threshold filtering result will have a large difference.
- the result of the soft threshold filtering algorithm is applied.
- the original signal sampling frequency is 20 GHz
- the long sampling sequence with a wide time window is 50000 points
- the short sampling sequence with narrow time window is 16000 points. Both sequences completely contain useful UHF PD signals.
- the correlation coefficient between the two is 0.6677.
- N j is the number of wavelet coefficients on the scale
- ) is the median of all wavelet coefficients on the scale
- ⁇ is called the signal-to-noise ratio factor, which is the signal The signal-to-noise ratio is reflected in the threshold calculation.
- ⁇ j is called the scale factor. It is the maximum value of the wavelet coefficient on the scale to correct the estimation error caused by the different lengths of the sample sequence.
- T j is the calculated threshold.
- the effectiveness of the wavelet denoising method mainly depends on wavelet basis function, wavelet decomposition scale, threshold function and threshold selection.
- a large number of simulation experiments, laboratory simulations and field measured data analysis are used to verify the effectiveness of the method used.
- the results show that compared with the denoising algorithms of other threshold rules, the wavelet denoising method significantly improves the noise cancellation capability in partial discharge signal processing, and also has less distortion of the processed signal waveform, more accurate extraction and influence.
- the advantages are few factors.
- the characteristic parameters include: skewness Sk, steepness Ku, local peak number Pe, cross-correlation coefficient Cc, and discharge factor Q;
- step 3 for the sampled signal, a phase analysis mode (PRPD) based extraction feature parameter is used, wherein: the skewness Sk, the definition of Sk:
- PRPD phase analysis mode
- w is the number of phase windows in a half cycle
- xi is the phase of the i th phase window
- y i is the ordinate of the spectrum, representing the apparent discharge amount q or the number of discharges n;
- the parameter ⁇ represents the position at the center of the collected partial discharge map, and ⁇ represents the steep condition represented by the symmetry axis at the center of the spectrum,
- ⁇ x is a parameter related to the average distribution of the partial discharge map. The phase corresponding to a point in the map.
- phase window Construction method of ⁇ -q-n spatial surface: The power frequency phase is followed. 0-360° is divided into 256 cells, and the discharge pulse amplitude q is divided into 128 cells according to the maximum amplitude, so the ⁇ -q plane is divided into 128 ⁇ 256 cells; the statistics are on the ⁇ -q plane. The number of discharges in the small interval is the middle ⁇ -qn spatial surface.
- each variable is the same as the definition of the variable in the skewness.
- the steepness Ku is used to describe the degree of protrusion of a shape compared to the shape of a normal distribution: the steepness Ku of the normal distribution is 0; if Ku>0, the profile is sharper than the normal profile; If Ku ⁇ 0, the spectral profile is flatter than the normal distribution profile.
- the number of local peaks Pe the number of local peaks is used to describe the number of local peaks on the contour of the spectrum. At the outline point Whether there is a local peak at the place can be determined by:
- the number of local peaks is closely related to the number of phase windows of the spectrum. In general, the more phase windows the phase axis is equally divided, the larger the number of local peaks.
- I the discharge repetition rate in the phase window i
- the superscript "+", "-" corresponds to The positive and negative half of the spectrum.
- the discharge amount factor Q reacted The difference in average discharge between the positive and negative half cycles of the spectrum.
- the unified calculation is calculated by analyzing the spectrum and the algorithm, and the feature parameter offset (SK), steepness (Ku), local peak number (Pe), discharge factor (Q), The correlation coefficient (Cc) is used for pattern recognition.
- SK feature parameter offset
- Ku steepness
- Pe local peak number
- Q discharge factor
- Cc correlation coefficient
- step 4 since it is impossible to know in advance which feature parameters can construct the simplest feature space of the UHF PD signal, that is, the redundant feature parameter matrix, the constructed feature space dimension is high, and there may be dimension redundancy. I am not good for running and identifying results. Therefore, there is a need to perform dimensionality reduction processing of the feature space.
- kernel function In the application of KPCA, the choice of nonlinear transformation (ie kernel function) is very important.
- kernel functions are Polynomial, Gaussian, and Sigmoid kernel functions, as follows:
- ⁇ x i , x j > is the sample vector
- are the Euclidean norms of the two.
- the polynomial kernel function is the power of b from the distance ( ⁇ x i , x j >+a) and is a monotonically increasing function of ⁇ x i , x j >.
- the distance ⁇ x i , x j > will be compressed. It can be seen that the function of the polynomial kernel function is to compress the small distance and further expand the large distance.
- the Gaussian kernel function also known as the radial basis kernel function, is also generally defined as an exponential monotonically decreasing function of two vector Euclidean distances, which is a radially symmetric scalar function.
- ⁇ is called the width parameter and is used to control the radial range of the function, ie the width of the Gaussian pulse.
- the Gaussian kernel function has a small scope of action, and the effect is exactly opposite to the polynomial kernel function, that is, the small distance is expanded to compress the large distance.
- this embodiment combines the advantages and disadvantages of polynomial kernel function and Gaussian kernel function, and proposes a new kernel function, the effect of which is shown in Figure 4.
- step 4 the kernel function sampled by the improved kernel principal component analysis is:
- step 5 the basic idea of the K nearest neighbor method is: to give a test document, the system finds the nearest K neighbors in the already classified training set, and obtains the category of the test document according to the category distribution of these neighbors. Among them, these neighbors can be weighted with the similarity of the test documents to obtain a better classification effect.
- the so-called cluster means a set of texts with similar properties.
- the present invention considers those subsets of partial discharge signal data with the largest distance between texts belonging to the same category in the training set as a cluster. Therefore, the algorithm of the K-nearest neighbor classification method Can be described as follows:
- Step1 In the training set, firstly map all the partial discharge data to a space vector
- Step2 Starting from the first class, perform two-two similarity calculation on all signal data belonging to this category, set a minimum threshold, and obtain a cluster with similar similarity according to statistics;
- Step3 For each cluster, combine all the signal data, and then calculate its center vector, in addition, calculate the number of clusters / total number of categories, this value represents the contribution coefficient of this cluster to this class, denoted as C;
- Step4 After the new text arrives, preprocess it to obtain its vector space
- Step 5 Calculate the distance between the space vector of the new text and the center vector of each cluster generated by Step3, multiply the distances by the contribution coefficients of the corresponding clusters, and add the results of the cluster calculations belonging to the same category, and compare the largest ones.
- One category is the category of the typical defect partial discharge to be classified.
- a clustering algorithm for generating clusters hypothesis category:
- Step1 setting a threshold of similarity a
- Step3 Start with di
- Step4 Calculate the similarity with the first text in Tn to obtain the value s;
- the nearest neighbor is extended to the K nearest neighbor.
- the K nearest neighbor method does not select a nearest neighbor for classification, but selects the K representative points closest to the classified text, and then according to this K
- the category information of the representative points determines the category of the text to be classified.
- the K-nearest neighbor classifier is used to identify the GIS insulation defect type by applying the KPCA, RST and CCMDR algorithm to reduce the dimensionality of the feature parameters.
- a program file is written in a C language software environment, and the design, training, and classification identification test of the classifier are implemented.
- the output of the classifier designed in this embodiment is not distributed like a BP neural network in a distributed manner around a certain point, but corresponds to the type 4 GIS defect type, the output value includes only four kinds of results [1, 2, 3, 4], so the pattern recognition result is only expressed by the recognition accuracy rate, as shown in Table 1.
- the pattern recognition method for the partial discharge of the three-phase common-tube ultra-high pressure GIS includes the steps of: detecting ultra-high frequency (UHF) three-phase common-tube GIS partial discharge, using UHF transmission sense
- the partial discharge (PD) signal is sampled by the device; the localized signal is denoised by the improved wavelet threshold filtering method; the characteristic parameters of the sampled signal are extracted based on the phase analysis mode, and the characteristic parameters include: skewness Sk, steepness Ku , local peak point number Pe, cross-correlation coefficient Cc, and discharge factor Q; using the improved kernel principal component analysis method to reduce the dimensionality of the feature space composed of characteristic parameters, and obtain the feature parameter matrix after dimensionality reduction; using K-nearest neighbor classification method Pattern recognition of GIS insulation defect types.
- At least the beneficial effects of the three-phase common-tube ultra-high pressure GIS partial discharge pattern recognition method include: overcoming the defects of the prior art, and improving the accuracy of the three-phase common-tube ultra-
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Testing Relating To Insulation (AREA)
Abstract
一种三相共筒式超高压GIS局部放电的模式识别方法,包括步骤:采用超高频检测三相共筒式GIS局部放电,利用UHF传感器对局部放电信号采样;利用改进的小波阀值滤波方法对采集到的局部放信号进行消噪处理,得到真实的局部放电信号;通过基于相位分析模式算法提取采样信号的特征参数;利用改进的核主成分分析法对特征参数组成的特征空间进行降维处理,得到降维后的特征参数矩阵;利用基于簇思想的K近邻分类法对GIS绝缘缺陷类型进行模式识别。所述三相共筒式超高压GIS局部放电的模式识别方法,可以克服现有技术中功能少、适用范围小和准确性差等缺陷,以实现功能多、适用范围广和准确性好的优点。
Description
本发明涉及高压电放电识别技术领域,具体地,涉及一种三相共筒式超高压GIS局部放电的模式识别方法。
气体绝缘开关设备(Gas lnsulated Switchgear,简称GIS)是特高压电网中的重要组成设备之一,它将一座变电站中的断路器、电流互感器、电压互感器、避雷器、隔离开关、接地开关、母线、电缆终端、进出线套管等优化设计后分别装在各自密封间中最后集中组装在一个充以SF6作为绝缘介质的整体外壳中。
GIS内部影响绝缘介质性能的缺陷主要有:严重的安装错误、导体之间接触不良、高压导体突出物、固定微粒、绝缘子缺陷、蒸气等。
GIS的发展趋于三相共筒化、复合化和智能化,由于实现了小型化,可在工厂内进行整机装配和试验合格后以间隔的形式运达现场,因此可缩短现场安装工期,同时可靠性又得到了提高。
三相共筒式GIS在内部结构,电场分布等方面与共轴式GIS有着明显的不同,现有技术研究主要集在共轴式GIS,但对于三相共筒式超高压GIS局部放电检测模式识别的研究较少。
在实现本发明的过程中,发明人发现现有技术中至少存在功能少、适用范围小和准确性差等缺陷。
发明内容
本发明的目的在于,针对上述问题,提出一种三相共筒式超高压GIS局部放电的模式识别方法,以实现功能多、适用范围广和准确性好的优点。
为实现上述目的,本发明采用的技术方案是:一种三相共筒式超高压GIS局部放电的模式识别方法,包括以下步骤:
步骤1:采用超高频检测三相共筒式GIS局部放电,利用UHF传感器对局部放电信号采样;
步骤2:利用改进的小波阀值滤波方法对采集到的局部放信号进行消噪处理,得到真实的局部放电信号;
步骤3:通过基于相位分析模式算法提取采样信号的特征参数;
步骤4:利用改进的核主成分分析法对特征参数组成的特征空间进行降维处理,得到降维后的特征参数矩阵;
步骤5:利用基于簇思想的K近邻分类法对GIS绝缘缺陷类型进行模式识别。
进一步地,在步骤2中,所述利用改进的小波阀值滤波方法对采集到的局部放信号进行消噪处理的操作,具体采用自适应阈值计算方法;该自适应阈值计算方法如下:
其中,j为尺度,Nj为该尺度上小波系数的个数,Median(|Cj,k|)为该尺度上所有小波系数的中位数,α称为信噪比因子,是信号的信噪比在阀值计算中的体现,βj称为尺度因子,是尺度上小波系数的最大值纠正采样序列长度不同引起的估计误差,Tj为计算的阀值。
进一步地,在步骤3中,所述特征参数包括偏斜度Sk、陡峭度Ku、局部峰点数Pe、互相关系数Cc和放电因数Q。
进一步地,所述偏斜度Sk具体为:
上式中,w是半周期内的相窗数;xi是第i个相窗的相位;
其中,yi是谱图的纵坐标,代表视在放电量q或放电次数n;参数μ代表采集到的局部放电图谱中心处的位置,σ代表图谱的中心处对称轴所体现的陡峭情况,Δx则是关于局部放电图谱的一个与平均分布相关的参数,图谱中的某点所对应的相位;
偏斜度Sk反映谱图形状相对于正态分布的左右偏斜情况:Sk=0说明该谱图形状左右对称;Sk>0说明该谱图相对于正态分布形状向左偏;Sk<0说明该谱图相对于正态分布形状向右偏。
进一步地,所述陡峭度Ku具体为:
上式中,w是半周期内的相窗数;xi是第i个相窗的相位;
其中,yi是谱图的纵坐标,代表视在放电量q或放电次数n;参数μ代表采集到的局部放电图谱中心处的位置,σ代表图谱的中心处对称轴所体现的陡峭情况,Δx则是关于局部放电图谱的一个与平均分布相关的参数,图谱中的某点所对应的相位;
陡峭度Ku用于描述某种形状的分布对比于正态分布形状的突起程度:正态分布的陡峭度Ku为0;如果Ku>0,则说明该谱图轮廓比正态分布轮廓尖锐陡峭;如果Ku<0,则说明该谱图轮廓比正态分布轮廓平坦。
(yi-yi-1)>0,(yi+1-yi)<0;
相位轴被等分的相窗越多,局部峰点数越大。
进一步地,所述互相关系数Cc具体为:
式中,是相窗i内的放电量,上标“+”、“-”对应于谱图正负半轴;c反应了正负半周期内的放电强弱和相位分布的相关性,互相关系数Cc接近于1意味着谱图正负半周的轮廓十分相似;Cc接近于0,谱图轮廓差异巨大。
进一步地,所述放电因数Q具体为:
进一步地,在步骤4中,所述改进的核主成分分析法即改进的核主成分分析法,所采样的核函数为:
其中,(a∈R,b∈N,σ>0),参数a、b与σ的选择是根据特征矩阵中元素的数值大小确定的,参数σ用于控制核函数的径向作用范围;xi和xj代表不同的样本向量,(xi,xj)代表样本向量的矢量积,R代表向量的取值范围在实数集,N代表整数集,k(xi,xj)代表结合多项式核函数与高斯核函数的优点而得到的新的核函数。
进一步地,在步骤5中,所述K近邻分类法的算法具体包括:
Step1:在训练集中,首先将所有局部放电数据进行预处理映射成为空间向量;
Step2:从第一个类开始,对属于这个类别的所有信号数据进行两两相似度计算,设定一个最小阈值,根据统计获得相似度接近的一个个簇;
Step3:对于每一个簇,将其中的所有信号数据合并,然后计算它的中心向量;此外,计算簇个数/类别总数,这个值代表此簇对这个类的贡献系数;
Step4:当新文本到来后,进行预处理取得它的向量空间;
Step5:将新文本的空间向量与Step3所生成的每一簇的中心向量计算距离,将这些距离与对应簇的贡献系数相乘,属于同一类别的簇计算的结果相加,比较得到最大的那一类别就是待分类典型缺陷局部放电所属类别。
本发明各实施例的三相共筒式超高压GIS局部放电的模式识别方法,由于包括步骤;采用超高频检测三相共筒式GIS局部放电,利用UHF传感器对局部放电信号采
样;利用改进的小波阀值滤波方法对采集到的局部放信号进行消噪处理,得到真实的局部放电信号;通过基于相位分析模式算法提取采样信号的特征参数;利用改进的核主成分分析法对特征参数组成的特征空间进行降维处理,得到降维后的特征参数矩阵;利用基于簇思想的K近邻分类法对GIS绝缘缺陷类型进行模式识别;可以克服现有技术的缺陷,提高三相共筒式超高压GIS局部放电检测模式识别的准确性;从而可以克服现有技术中功能少、适用范围小和准确性差的缺陷,以实现功能多、适用范围广和准确性好的优点。
本发明的其它特征和优点将在随后的说明书中阐述,并且,部分地从说明书中变得显而易见,或者通过实施本发明而了解。
下面通过附图和实施例,对本发明的技术方案做进一步的详细描述。
附图用来提供对本发明的进一步理解,并且构成说明书的一部分,与本发明的实施例一起用于解释本发明,并不构成对本发明的限制。在附图中:
图1为本发明三相共筒式超高压GIS局部放电的模式识别方法中气体绝缘组合电器局部放电试验装置的结构示意图;
图2为本发明三相共筒式超高压GIS局部放电的模式识别方法的流程示意图,在图2中,N和K均为自然数;
图3为本发明三相共筒式超高压GIS局部放电的模式识别方法中统计分布与Sk、Ku关系示意图,(a)为正偏移,(b)为无偏移,(c)为负偏移,Sk为图谱形状相对于正态分布的偏移情况;(d)为正偏移,(e)为无偏移,(f)为负偏移,陡峭度Ku用于描述某种形状的分布对比于正态分布形状的突起程度;
图4为本发明三相共筒式超高压GIS局部放电的模式识别方法中核函数的作用效果示意图,(a)为多项式函数,(b)为高斯核函数,(c)为新型核函数;
图5为本发明三相共筒式超高压GIS局部放电的模式识别方法中滤波前(a)、后(b)波形对比图,在图5中,横轴为0~360,角度,纵轴为信号幅值(Q)。
结合附图,本发明实施例中附图标记如下:
1-水电阻;2-高压试验变压器;3-变压器;4-高压套管;5-盘式绝缘子。
以下结合附图对本发明的优选实施例进行说明,应当理解,此处所描述的优选实施例仅用于说明和解释本发明,并不用于限定本发明。
针对现有技术中存在的缺陷,根据本发明实施例,如图1-图5所示,提供了一种三相共筒式超高压GIS局部放电的模式识别方法。
如图1所示,本实施例的三相共筒式超高压GIS局部放电的模式识别方法使用的气体绝缘组合电器局部放电试验装置,包括盘式绝缘子5,安装在盘式绝缘子5上的高压套管4,依次与高压套管4连接的水电阻1、高压试验变压器2、变压器3和局部放电信号检测仪(PDSG);局部放电信号检测仪(PDSG)还与盘式绝缘子5连接,水电阻1和高压套管4的公共端通过由电容组成的分压器后接地。
该气体绝缘组合电器局部放电试验装置,主要包括变压器3、由电容组成的分压器、示波器、三相共筒式气体绝缘组合电器(GIS)、传感器和局部放电信号检测仪(PDSG);通过在GIS内分别设置高压导体金属突出物、自由金属微粒、绝缘子表面固定金属、绝缘子气隙等缺陷,检测相应的局部放电信号,进行模式识别。
本实施例的三相共筒式超高压GIS局部放电的模式识别方法,包括以下步骤:
1)采用超高频(UHF)检测三相共筒式GIS局部放电,利用UHF传感器对局部放电(PD)信号采样;
在步骤1)中,采用超高频法(UHF)对三相共筒式GIS局部放电进行检测。用UHF法检测GIS局部放电时,可以根据测得信号的频谱特性和放电发生在工频电压波形上的位置来识别不同故障类型。
在步骤1)中,采用实际三相共筒式GIS,对典型缺陷条件下局部放电图谱进行测量。其中三相共筒式GIS应有专业高压开关设备企业提供。三相导体的加压方式为两相接地,一相接高压,模型中只设有一支高压套管。根据同轴型GIS试验模型测试结果,设置具有对比性的典型绝缘缺陷。三相共筒式GIS腔体内存在自由金属微粒,绝缘子表面存在固定金属颗粒,绝缘子上存在气隙缺陷,并将局部放电物理模型置于GIS模拟装置中,进行局部放电的测量。
2)利用改进的小波阀值滤波方法对采集到的局部放信号进行消噪处理,得到真实的局部放电信号;
在步骤2)中,改进的小波阀值滤波算法,由中位数Median(Cj,k)的计算方法可见,阈值的选择与被分析信号的长度关系十分密切。在实际应用当中,并不能保证有效信号在整段采样点数在整段采样序列中所占的比例及位置是不变的。因为信号的长度决定了小波变换后在每个尺度上小波系数的个数Nj,它影响中位数,Median(Cj,k)的取值,从而影响了阈值的大小。这种影响会导致这样的结果:同一个有
用信号序列u被包含在一段采样序列s当中,如果s长度不同(即时间观察窗的宽度不同),小波阈值滤波后的结果会有较大差异。对于包含同一个UHF PD信号的不同长度的s,应用软阈值滤波算法的结果。原始信号采样频率为20GHz,时间窗较宽的长采样序列为50000点,时间窗较窄的短采样序列为16000点,两个序列都完整的包含了有用UHF PD信号。显而易见,应用软阈值算法滤波后的UHF PD信号,尤其是振荡衰减将要结束的部分,存在明显的差异,两者的相关系数为0.6677。
产生这种后果的根本原因在于阈值的计算公式中完全忽略了有用信号的幅值与信噪比的因素。鉴于此,本实施例经过多次分析,考虑了有用信号幅值与信噪比的因素,提出了一种新的自适应阈值计算方法,该自适应阈值计算方法大大降低了小波阈值滤波结果对采样点的敏感程度,虑波前后的图形对比如图5所示,图5中(a)为滤波前图形,(b)为滤波后图形:
其中,j为尺度,Nj为该尺度上小波系数的个数,Median(|Cj,k|)为该尺度上所有小波系数的中位数,α称为信噪比因子,是信号的信噪比在阀值计算中的体现,βj称为尺度因子,是尺度上小波系数的最大值纠正采样序列长度不同引起的估计误差,Tj为计算的阀值。
小波消噪方法的有效性主要取决于小波基函数、小波分解尺度、阀值函数、阀值选取等几个方面。本实施例中用大量的仿真实验、实验室模拟和现场实测数据的分析验证了所用方法的有效性。结果表明,该小波消噪方法同其它阀值规则的消噪算法相比,明显提高了局部放电信号处理中的消噪能力,而且还具有处理后信号波形失真小、提取更准确、所受影响因素少等优点。
3)通过基于相位分析模式算法提取采样信号的特征参数,优选的,特征参数包括:偏斜度Sk、陡峭度Ku、局部峰点数Pe、互相关系数Cc和放电因数Q;
在步骤3)中,对于采样后的信号,采用基于相位分析模式(PRPD)的提取特征参数,其中:偏斜度Sk,Sk的定义:
上式中,w是半周期内的相窗数;xi是第i个相窗的相位;
其中,yi是谱图的纵坐标,代表视在放电量q或放电次数n;参数μ代表采集到的局部放电图谱中心处的位置,σ代表图谱的中心处对称轴所体现的陡峭情况,Δx则是关于局部放电图谱的一个与平均分布相关的参数,图谱中的某点所对应的相位。
偏斜度Sk反映谱图形状相对于正态分布的左右偏斜情况:Sk=0说明该谱图形状左右对称;Sk>0说明该谱图相对于正态分布形状向左偏;Sk<0说明该谱图相对于正态分布形状向右偏。
相窗的定义:Φ-q-n空间曲面的构造方法:将工频相位按照。0-360°划分为256个小区间,将放电脉冲幅值q按最大幅值划分为128个小区间,因而Φ-q平面被划分成128×256个小区间;统计Φ-q平面上各小区间内的放电次数,即得到中Φ-q-n空间曲面。
陡峭度Ku,Ku的定义:
其中的各变量的定义和偏斜度中变量定义相同。陡峭度Ku用于描述某种形状的分布对比于正态分布形状的突起程度:正态分布的陡峭度Ku为0;如果Ku>0,则说明该谱图轮廓比正态分布轮廓尖锐陡峭;如果Ku<0,则说明该谱图轮廓比正态分布轮廓平坦。
上式变为差分方程即:
差分方程可以简化为:
(yi-yi-1)>0,(yi+1-yi)<0;
在实际计算中局部峰点数与谱图的相窗数目密切相关。一般而言,相位轴被等分的相窗越多,局部峰点数越大。
互相关系数Cc,互相关系数Cc的定义:
式中,是相窗i内的放电量,上标“+”、“-”对应于谱图正负半轴。Cc反应了正负半周期内的放电强弱和相位分布的相关性,互相关系数Cc接近于1意味着谱图正负半周的轮廓十分相似;Cc接近于0,谱图轮廓差异巨大。
放电因数Q,
按照上面各个统计算子的公式,通过分析谱图和算法计算出统计算子,提取特征参数偏移度(SK)、陡峭度(Ku)、局部峰点数(Pe)、放电因数(Q)、互相关系数(Cc)来进行模式识别。
4)利用改进的核主成分分析法对特征参数组成的特征空间进行降维处理,得到
降维后的特征参数矩阵;
在步骤4)中,由于事先无法知道哪些特征参数可以构造UHF PD信号的最简特征空间,即无冗余的满轶特征参数矩阵,构造的特征空间维数较高,而且可能存在维数冗余,对运行和识别结果不利。因此,有需要进行特征空间的降维处理。
在KPCA的应用当中,非线性变换(即核函数)的选择非常重要。常用的核函数有多项式核函数(Polynomial)、高斯核函数(Gauss)与Sigmoid核函数,分别如下:
k(xi,xj)=(<xi,xj>+a)b,(a∈R,b∈N);
k(xi,xj)=tanh(<xi,xj>+a),(a∈R);
其中,<xi,xj>为样本向量,xi和xj的矢量积,||xi-xj||为两者的欧氏范数。多项式核函数是距离(<xi,xj>+a)的b次幂,是<xi,xj>的单调增函数。经过变换,如果(<xi,xj>+a)>1,原始距离<xi,xj>会被放大;相反,如果(<xi,xj>+a)<1,则原始距离<xi,xj>会被压缩。可见多项式核函数的作用是把小距离压缩而把大距离进一步扩大。高斯核函数也称作径向基核函数,通常也被定义为两个向量欧氏距离的指数单调下降函数,它是一种径向对称的标量函数。其中σ称为宽度参数,用于控制函数的径向作用范围,即高斯脉冲的宽度。但通常高斯核函数的作用范围较小,作用效果则正好与多项式核函数相反,即把小距离扩大而把大距离压缩。
其实核函数的作用应该是将原始距离进一步扩大,或者将同类样本间的距离压缩而将非同类样本间的距离扩大,以便于进行分类识别。鉴于此,本实施例结合多项式核函数与高斯核函数的优缺点,提出了一种新的核函数,其作用效果如图4所示。
在步骤4)中,改进的核主成分分析法所采样的核函数为:
其中,(a∈R,b∈N,σ>0),参数a、b与σ的选择是根据特征矩阵中元素的数值大小确定的,参数σ用于控制核函数的径向作用范围;xi和xj代表不同的样本向量,(xi,xj)代表样本向量的矢量积,R代表向量的取值范围在实数集,N代表整数集,k(xi,xj)代表结合多项式核函数与高斯核函数的优点而得到的新的核函数。此处取a=5,b=1是为了使变换后的距离随原距离作一定比例的变化。参数σ用于控制核函数的径向作用范围,由于原始特征矩阵中两个向量的距离一般不超过7,所以取σ=7。由图4可见,本发明所提出的新型核函数将原始的小距离变大,将大距离适当缩小。
5)利用基于簇思想的K近邻分类法对GIS绝缘缺陷类型进行模式识别;
在步骤5)中,K最近邻方法其基本思想是:给出测试文档,系统在已经分类好的训练集中查找与其最近的K个邻居,根据这些邻居的类别分布情况获得测试文档的类别。其中可以用这些邻居与测试文档的相似度进行加权,从而获得较好的分类效果。所谓簇,意思就是一类具有相似性质的文本的集合,本发明把训练集中属于同一类别的文本之间距离最大的那些局部放电信号数据子集合认为是一个簇,因此,K近邻分类法的算法可以描述如下:
Step1:在训练集中,首先将所有局部放电数据进行预处理映射成为空间向量;
Step2:从第一个类开始,对属于这个类别的所有信号数据进行两两相似度计算,设定一个最小阈值,根据统计获得相似度接近的一个个簇;
Step3:对于每一个簇,将其中的所有信号数据合并,然后计算它的中心向量,此外,计算簇个数/类别总数,这个值代表此簇对这个类的贡献系数,记作C;
Step4:当新文本到来后,进行预处理取得它的向量空间;
Step5:将新文本的空间向量与Step3所生成的每一簇的中心向量计算距离,将这些距离与对应簇的贡献系数相乘,属于同一类别的簇计算的结果相加,比较得到最大的那一类别就是待分类典型缺陷局部放电所属类别。
这个算法的基础是如何找出同一类别中的哪些文本属于同一簇,以下给出找出同
一类别簇的生成簇算法思想:假设类别:
c={d1,d2,...................,dm}
Step1:设定一个相似度的阈值a;
Step2:首先创建一个簇,记作T0,用Ki记录簇内所包含的文档数量,total记录创建的簇数量,初始化已处理文档i=2;
Step3:从di开始;
Step4:与Tn中的第一个文本进行相似度计算得到值s;
Step5:如果s>=a,且Tn中还有未与此样本进行比较的样本,那么继续进行相似度计算并更新s;如果没有未比较样本,那么将此数据加入到簇Tn中去;如果s<a,如果有其它未比较的簇,则n++,返回step4;如果没有未比较的簇,那么创建新簇,记为T++total;将此文档归为T++total簇中;
Step6:如果i!=m,那么i++;返回Step3;否则,结束。
为了克服最近邻法错判率较高的缺陷,将最近邻推广到K近邻,K近邻法不是选取一个最近邻进行分类,而是选取离待分类文本最近的K个代表点,然后根据这K个代表点的类别信息来确定待分类文本的类别。
对于降维后的特征参数矩阵,一半样本用于训练K近邻分类器,另一半用于测试分类器的性能。对于分别应用KPCA、RST以及CCMDR算法降维后的特征参数矩阵,应用K近邻分类器对GIS绝缘缺陷类型进行识别。本实施例在C语言软件环境下编写了程序文件,实现分类器的设计、训练及分类识别测试。由于本实施例设计的分类器的输出并不象BP神经网络一样以某点为中心呈分散状分布,而是对应于4类GIS缺陷类型,输出取值仅包括4种结果[1,2,3,4],所以模式识别结果仅以识别正确率表示,如表1所示。
表1:K近邻算法模式识别正确率
缺陷类型 | K近邻法识别正确率 |
高压导体金属突出物 | 92% |
自由金属微粒 | 91.5% |
绝缘子表面固定金属 | 88% |
绝缘子气隙缺陷 | 90% |
综上所述,本发明上述各实施例的三相共筒式超高压GIS局部放电的模式识别方法,包括步骤:采用超高频(UHF)检测三相共筒式GIS局部放电,利用UHF传感
器对局部放电(PD)信号采样;利用改进的小波阀值滤波方法对局部放信号进行消噪处理;基于相位分析模式提取采样信号的特征参数,特征参数包括:偏斜度Sk、陡峭度Ku、局部峰点数Pe、互相关系数Cc、和放电因数Q;利用改进的核主成分分析法对特征参数组成的特征空间进行降维处理,得到降维后的特征参数矩阵;利用K近邻分类法对GIS绝缘缺陷类型进行模式识别。该三相共筒式超高压GIS局部放电的模式识别方法至少可以达到的有益效果包括:克服了现有技术的缺陷,提高了三相共筒式超高压GIS局部放电检测模式识别的准确性。
最后应说明的是:以上所述仅为本发明的优选实施例而已,并不用于限制本发明,尽管参照前述实施例对本发明进行了详细的说明,对于本领域的技术人员来说,其依然可以对前述各实施例所记载的技术方案进行修改,或者对其中部分技术特征进行等同替换。凡在本发明的精神和原则之内,所作的任何修改、等同替换、改进等,均应包含在本发明的保护范围之内。
Claims (10)
- 一种三相共筒式超高压GIS局部放电的模式识别方法,其特征在于,包括以下步骤:步骤1:采用超高频检测三相共筒式GIS局部放电,利用UHF传感器对局部放电信号采样;步骤2:利用改进的小波阀值滤波方法对采集到的局部放信号进行消噪处理,得到真实的局部放电信号;步骤3:通过基于相位分析模式算法提取采样信号的特征参数;步骤4:利用改进的核主成分分析法对特征参数组成的特征空间进行降维处理,得到降维后的特征参数矩阵;步骤5:利用基于簇思想的K近邻分类法对GIS绝缘缺陷类型进行模式识别。
- 根据权利要求1或2所述的三相共筒式超高压GIS局部放电的模式识别方法,其特征在于,在步骤3中,所述特征参数包括偏斜度Sk、陡峭度Ku、局部峰点数Pe、互相关系数Cc和放电因数Q。
- 根据权利要求1或2所述的三相共筒式超高压GIS局部放电的模式识别方法,其特征在于,在步骤5中,所述K近邻分类法的算法具体包括:Step1:在训练集中,首先将所有局部放电数据进行预处理映射成为空间向量;Step2:从第一个类开始,对属于这个类别的所有信号数据进行两两相似度计算,设定一个最小阈值,根据统计获得相似度接近的一个个簇;Step3:对于每一个簇,将其中的所有信号数据合并,然后计算它的中心向量;此外,计算簇个数/类别总数,这个值代表此簇对这个类的贡献系数;Step4:当新文本到来后,进行预处理取得它的向量空间;Step5:将新文本的空间向量与Step3所生成的每一簇的中心向量计算距离,将这些距离与对应簇的贡献系数相乘,属于同一类别的簇计算的结果相加,比较得到最大的那一类别就是待分类典型缺陷局部放电所属类别。
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CA2918679A CA2918679C (en) | 2013-11-14 | 2014-08-13 | Pattern recognition method for partial discharge of three-phase cylinder type ultrahigh voltage gis |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310566573.6 | 2013-11-14 | ||
CN201310566573.6A CN103558529B (zh) | 2013-11-14 | 2013-11-14 | 一种三相共筒式超高压gis局部放电的模式识别方法 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2015070513A1 true WO2015070513A1 (zh) | 2015-05-21 |
Family
ID=50012831
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2014/000766 WO2015070513A1 (zh) | 2013-11-14 | 2014-08-13 | 一种三相共筒式超高压gis局部放电的模式识别方法 |
Country Status (3)
Country | Link |
---|---|
CN (1) | CN103558529B (zh) |
CA (1) | CA2918679C (zh) |
WO (1) | WO2015070513A1 (zh) |
Cited By (29)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105842588A (zh) * | 2016-03-18 | 2016-08-10 | 深圳供电局有限公司 | 一种修正超声波局部放电检测的方法和系统 |
CN107907807A (zh) * | 2017-12-25 | 2018-04-13 | 国网湖北省电力公司信息通信公司 | 一种气体绝缘组合电器局部放电模式识别方法 |
CN108896878A (zh) * | 2018-05-10 | 2018-11-27 | 国家电网公司 | 一种基于超声波的局部放电检测方法 |
CN109829412A (zh) * | 2019-01-24 | 2019-05-31 | 三峡大学 | 基于动态模式分解分形特征的局部放电模式识别方法 |
CN110287514A (zh) * | 2019-05-10 | 2019-09-27 | 杭州电子科技大学 | 基于振动信号处理的超高速碰撞源智能定位方法 |
CN110412431A (zh) * | 2019-08-05 | 2019-11-05 | 国网湖南省电力有限公司 | 一种电力设备的局部放电缺陷类型的诊断方法及诊断系统 |
CN110514976A (zh) * | 2019-09-29 | 2019-11-29 | 国网江苏省电力有限公司 | 一种gis绝缘缺陷监测装置、系统和检测方法 |
CN110533064A (zh) * | 2019-07-17 | 2019-12-03 | 西安西电开关电气有限公司 | 一种gis设备的局部放电图谱模式识别方法和系统 |
CN110991376A (zh) * | 2019-12-10 | 2020-04-10 | 上海欧秒电力监测设备有限公司 | 一种局放类型识别的特征提取方法 |
CN111160185A (zh) * | 2019-12-20 | 2020-05-15 | 中国农业大学 | 多尺度时间序列遥感影像趋势和断点检测方法 |
CN111610417A (zh) * | 2020-05-28 | 2020-09-01 | 华乘电气科技股份有限公司 | 一种基于社区发现的放电信号源分离方法 |
CN112730654A (zh) * | 2020-12-18 | 2021-04-30 | 国网河北省电力有限公司电力科学研究院 | 六氟化硫电气设备故障检测方法、装置及终端设备 |
CN113253066A (zh) * | 2021-04-27 | 2021-08-13 | 国网山东省电力公司烟台供电公司 | 一种局部放电特高频信号prpd/prps图谱相位同步方法 |
CN113343550A (zh) * | 2021-06-09 | 2021-09-03 | 上海交通大学 | 基于局部图像特征的局部放电故障的诊断方法 |
CN113655348A (zh) * | 2021-07-28 | 2021-11-16 | 国网湖南省电力有限公司 | 一种基于深度孪生网络的电力设备局部放电故障诊断方法、系统终端及可读存储介质 |
CN113985318A (zh) * | 2021-10-26 | 2022-01-28 | 国网山东省电力公司检修公司 | 变压器运行状态的多类型局放信号实时提取方法 |
CN114035007A (zh) * | 2021-11-22 | 2022-02-11 | 国家电网有限公司 | 一种非接触式变压器套管局部放电检测系统 |
CN114200020A (zh) * | 2021-11-04 | 2022-03-18 | 广西电网有限责任公司南宁供电局 | 基于耦合超声的二次电缆线芯绝缘损伤识别方法 |
CN114384383A (zh) * | 2022-03-22 | 2022-04-22 | 东华理工大学南昌校区 | 一种用于定位特高频局部放电点的电路及方法 |
CN114839255A (zh) * | 2022-04-29 | 2022-08-02 | 西安热工研究院有限公司 | 一种基于XGBoost算法的六氟化硫电气设备微水检测方法 |
CN115186772A (zh) * | 2022-09-13 | 2022-10-14 | 云智慧(北京)科技有限公司 | 一种电力设备的局部放电的检测方法、装置及设备 |
CN116383602A (zh) * | 2022-12-14 | 2023-07-04 | 国网安徽省电力有限公司电力科学研究院 | 考虑噪声与样本量的gis隔离开关机械缺陷辨识方法 |
WO2023213332A1 (zh) * | 2022-06-27 | 2023-11-09 | 上海格鲁布科技有限公司 | 一种多源混合型特高频局部放电图谱的分离识别方法 |
CN117092458A (zh) * | 2023-02-07 | 2023-11-21 | 特变电工山东鲁能泰山电缆有限公司 | 一种电缆交流耐压局放脉冲波形的确定方法及系统 |
CN117708760A (zh) * | 2024-02-05 | 2024-03-15 | 国网江西省电力有限公司电力科学研究院 | 基于多模型融合的开关柜多源局放模式识别方法及系统 |
CN117786466A (zh) * | 2024-02-23 | 2024-03-29 | 兰州交通大学 | 基于智能频谱感知的信号盲识别方法 |
CN117892067A (zh) * | 2024-03-15 | 2024-04-16 | 国网上海市电力公司 | 一种低频电流局放监测抗干扰方法、装置、设备及介质 |
CN117969958A (zh) * | 2024-04-02 | 2024-05-03 | 杭州永德电气有限公司 | 一种电阻片配组产品检测方法及系统 |
CN118608876A (zh) * | 2024-08-06 | 2024-09-06 | 国网浙江省电力有限公司电力科学研究院 | Gis局部放电模式识别模型及其训练方法 |
Families Citing this family (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103558529B (zh) * | 2013-11-14 | 2016-11-23 | 国家电网公司 | 一种三相共筒式超高压gis局部放电的模式识别方法 |
CN104198901A (zh) * | 2014-08-13 | 2014-12-10 | 广东电网公司电力科学研究院 | 变电站局部放电信号的定位方法和系统 |
CN104977515B (zh) * | 2015-07-17 | 2017-10-31 | 许继集团有限公司 | 一种三相共筒gis局部放电监测中放电类型的识别方法 |
CN105223475B (zh) * | 2015-08-25 | 2018-05-15 | 国家电网公司 | 基于高斯参数拟合的局部放电谱图特征模式识别算法 |
EP3698388A4 (en) | 2017-10-16 | 2021-05-05 | ABB Power Grids Switzerland AG | METHOD OF MONITORING A CIRCUIT BREAKER AND DEVICE AND INTERNET OF THINGS USING THEREOF |
US10401420B2 (en) * | 2017-12-08 | 2019-09-03 | Hamilton Sundstrand Corporation | Voltage suppressor test circuit and method of testing a voltage suppressor |
CN108169643A (zh) * | 2018-01-30 | 2018-06-15 | 西南石油大学 | 一种用于电缆局部放电模式识别的方法和系统 |
CN109063780B (zh) * | 2018-08-10 | 2020-09-15 | 国网上海市电力公司 | 变压器局部放电识别方法 |
ES2780448A1 (es) * | 2019-02-22 | 2020-08-25 | Ormazabal Corp Technology Aie | Método y sistema de reconocimiento de descargas parciales para diagnóstico de redes eléctricas |
EP3913383B1 (en) * | 2020-05-22 | 2023-10-04 | Rohde & Schwarz GmbH & Co. KG | Method and system for detecting anomalies in a spectrogram, spectrum or signal |
CN112698160A (zh) * | 2020-12-01 | 2021-04-23 | 广东电网有限责任公司广州供电局 | 开关柜局部放电故障识别方法、装置、计算机设备及存储介质 |
CN114548845B (zh) * | 2022-04-27 | 2022-07-12 | 北京智芯微电子科技有限公司 | 一种配网管理方法、装置及系统 |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN202975247U (zh) * | 2012-11-23 | 2013-06-05 | 山东兴驰高压开关有限公司 | Gis局放在线监测装置 |
CN103558529A (zh) * | 2013-11-14 | 2014-02-05 | 国家电网公司 | 一种三相共筒式超高压gis局部放电的模式识别方法 |
CN103558528A (zh) * | 2013-11-14 | 2014-02-05 | 国家电网公司 | 一种局部放电超高频检测系统及方法 |
-
2013
- 2013-11-14 CN CN201310566573.6A patent/CN103558529B/zh active Active
-
2014
- 2014-08-13 WO PCT/CN2014/000766 patent/WO2015070513A1/zh active Application Filing
- 2014-08-13 CA CA2918679A patent/CA2918679C/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN202975247U (zh) * | 2012-11-23 | 2013-06-05 | 山东兴驰高压开关有限公司 | Gis局放在线监测装置 |
CN103558529A (zh) * | 2013-11-14 | 2014-02-05 | 国家电网公司 | 一种三相共筒式超高压gis局部放电的模式识别方法 |
CN103558528A (zh) * | 2013-11-14 | 2014-02-05 | 国家电网公司 | 一种局部放电超高频检测系统及方法 |
Non-Patent Citations (3)
Title |
---|
DUAN, DAPENG: "GIS Partial Discharge Detection and Biomimetic Pattern Recognition Based on UHF Method", PH .D. DISSERTATION SUBMITTED TO SHANGHAI JIAO TONG UNIVERSITY, 28 February 2011 (2011-02-28), pages 1 - 17 , 35, AND 55-82 * |
LI, GUOWEI ET AL.: "Judgment of Partial Discharge Types in GIS Based on UHF Method", HIGH VOLTAGE APPARATUS, vol. 49, no. 1, 31 January 2013 (2013-01-31), pages 63 - 68 * |
LIU, FAN ET AL.: "Recognition of PD mode based on KNN algorithm for converter transformer", ELECTRIC POWER AUTOMATION EQUIPMENT, vol. 33, 31 May 2013 (2013-05-31), pages 89 - 93 * |
Cited By (46)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105842588A (zh) * | 2016-03-18 | 2016-08-10 | 深圳供电局有限公司 | 一种修正超声波局部放电检测的方法和系统 |
CN105842588B (zh) * | 2016-03-18 | 2018-09-28 | 深圳供电局有限公司 | 一种修正超声波局部放电检测的方法和系统 |
CN107907807A (zh) * | 2017-12-25 | 2018-04-13 | 国网湖北省电力公司信息通信公司 | 一种气体绝缘组合电器局部放电模式识别方法 |
CN108896878A (zh) * | 2018-05-10 | 2018-11-27 | 国家电网公司 | 一种基于超声波的局部放电检测方法 |
CN109829412A (zh) * | 2019-01-24 | 2019-05-31 | 三峡大学 | 基于动态模式分解分形特征的局部放电模式识别方法 |
CN109829412B (zh) * | 2019-01-24 | 2023-03-24 | 三峡大学 | 基于动态模式分解分形特征的局部放电模式识别方法 |
CN110287514A (zh) * | 2019-05-10 | 2019-09-27 | 杭州电子科技大学 | 基于振动信号处理的超高速碰撞源智能定位方法 |
CN110287514B (zh) * | 2019-05-10 | 2023-03-28 | 杭州电子科技大学 | 基于振动信号处理的超高速碰撞源智能定位方法 |
CN110533064A (zh) * | 2019-07-17 | 2019-12-03 | 西安西电开关电气有限公司 | 一种gis设备的局部放电图谱模式识别方法和系统 |
CN110533064B (zh) * | 2019-07-17 | 2022-11-22 | 西安西电开关电气有限公司 | 一种gis设备的局部放电图谱模式识别方法和系统 |
CN110412431A (zh) * | 2019-08-05 | 2019-11-05 | 国网湖南省电力有限公司 | 一种电力设备的局部放电缺陷类型的诊断方法及诊断系统 |
CN110514976A (zh) * | 2019-09-29 | 2019-11-29 | 国网江苏省电力有限公司 | 一种gis绝缘缺陷监测装置、系统和检测方法 |
CN110514976B (zh) * | 2019-09-29 | 2024-02-13 | 国网江苏省电力有限公司 | 一种gis绝缘缺陷监测装置、系统和检测方法 |
CN110991376A (zh) * | 2019-12-10 | 2020-04-10 | 上海欧秒电力监测设备有限公司 | 一种局放类型识别的特征提取方法 |
CN110991376B (zh) * | 2019-12-10 | 2024-03-19 | 上海欧秒电力监测设备有限公司 | 一种局放类型识别的特征提取方法 |
CN111160185A (zh) * | 2019-12-20 | 2020-05-15 | 中国农业大学 | 多尺度时间序列遥感影像趋势和断点检测方法 |
CN111160185B (zh) * | 2019-12-20 | 2023-11-10 | 中国农业大学 | 多尺度时间序列遥感影像趋势和断点检测方法 |
CN111610417B (zh) * | 2020-05-28 | 2022-03-15 | 华乘电气科技股份有限公司 | 一种基于社区发现的放电信号源分离方法 |
CN111610417A (zh) * | 2020-05-28 | 2020-09-01 | 华乘电气科技股份有限公司 | 一种基于社区发现的放电信号源分离方法 |
CN112730654A (zh) * | 2020-12-18 | 2021-04-30 | 国网河北省电力有限公司电力科学研究院 | 六氟化硫电气设备故障检测方法、装置及终端设备 |
CN112730654B (zh) * | 2020-12-18 | 2023-02-03 | 国网河北省电力有限公司电力科学研究院 | 六氟化硫电气设备故障检测方法、装置及终端设备 |
CN113253066B (zh) * | 2021-04-27 | 2024-01-26 | 国网山东省电力公司烟台供电公司 | 一种局部放电特高频信号prpd/prps图谱相位同步方法 |
CN113253066A (zh) * | 2021-04-27 | 2021-08-13 | 国网山东省电力公司烟台供电公司 | 一种局部放电特高频信号prpd/prps图谱相位同步方法 |
CN113343550A (zh) * | 2021-06-09 | 2021-09-03 | 上海交通大学 | 基于局部图像特征的局部放电故障的诊断方法 |
CN113655348A (zh) * | 2021-07-28 | 2021-11-16 | 国网湖南省电力有限公司 | 一种基于深度孪生网络的电力设备局部放电故障诊断方法、系统终端及可读存储介质 |
CN113655348B (zh) * | 2021-07-28 | 2023-12-08 | 国网湖南省电力有限公司 | 一种基于深度孪生网络的电力设备局部放电故障诊断方法、系统终端及可读存储介质 |
CN113985318A (zh) * | 2021-10-26 | 2022-01-28 | 国网山东省电力公司检修公司 | 变压器运行状态的多类型局放信号实时提取方法 |
CN114200020A (zh) * | 2021-11-04 | 2022-03-18 | 广西电网有限责任公司南宁供电局 | 基于耦合超声的二次电缆线芯绝缘损伤识别方法 |
CN114035007A (zh) * | 2021-11-22 | 2022-02-11 | 国家电网有限公司 | 一种非接触式变压器套管局部放电检测系统 |
CN114384383A (zh) * | 2022-03-22 | 2022-04-22 | 东华理工大学南昌校区 | 一种用于定位特高频局部放电点的电路及方法 |
CN114839255A (zh) * | 2022-04-29 | 2022-08-02 | 西安热工研究院有限公司 | 一种基于XGBoost算法的六氟化硫电气设备微水检测方法 |
CN114839255B (zh) * | 2022-04-29 | 2024-04-23 | 西安热工研究院有限公司 | 一种基于XGBoost算法的六氟化硫电气设备微水检测方法 |
WO2023213332A1 (zh) * | 2022-06-27 | 2023-11-09 | 上海格鲁布科技有限公司 | 一种多源混合型特高频局部放电图谱的分离识别方法 |
CN115186772B (zh) * | 2022-09-13 | 2023-02-07 | 云智慧(北京)科技有限公司 | 一种电力设备的局部放电的检测方法、装置及设备 |
CN115186772A (zh) * | 2022-09-13 | 2022-10-14 | 云智慧(北京)科技有限公司 | 一种电力设备的局部放电的检测方法、装置及设备 |
CN116383602A (zh) * | 2022-12-14 | 2023-07-04 | 国网安徽省电力有限公司电力科学研究院 | 考虑噪声与样本量的gis隔离开关机械缺陷辨识方法 |
CN116383602B (zh) * | 2022-12-14 | 2024-05-03 | 国网安徽省电力有限公司电力科学研究院 | 考虑噪声与样本量的gis隔离开关机械缺陷辨识方法 |
CN117092458A (zh) * | 2023-02-07 | 2023-11-21 | 特变电工山东鲁能泰山电缆有限公司 | 一种电缆交流耐压局放脉冲波形的确定方法及系统 |
CN117708760A (zh) * | 2024-02-05 | 2024-03-15 | 国网江西省电力有限公司电力科学研究院 | 基于多模型融合的开关柜多源局放模式识别方法及系统 |
CN117786466A (zh) * | 2024-02-23 | 2024-03-29 | 兰州交通大学 | 基于智能频谱感知的信号盲识别方法 |
CN117786466B (zh) * | 2024-02-23 | 2024-04-26 | 兰州交通大学 | 基于智能频谱感知的信号盲识别方法 |
CN117892067A (zh) * | 2024-03-15 | 2024-04-16 | 国网上海市电力公司 | 一种低频电流局放监测抗干扰方法、装置、设备及介质 |
CN117892067B (zh) * | 2024-03-15 | 2024-05-28 | 国网上海市电力公司 | 一种低频电流局放监测抗干扰方法、装置、设备及介质 |
CN117969958A (zh) * | 2024-04-02 | 2024-05-03 | 杭州永德电气有限公司 | 一种电阻片配组产品检测方法及系统 |
CN117969958B (zh) * | 2024-04-02 | 2024-06-07 | 杭州永德电气有限公司 | 一种电阻片配组产品检测方法及系统 |
CN118608876A (zh) * | 2024-08-06 | 2024-09-06 | 国网浙江省电力有限公司电力科学研究院 | Gis局部放电模式识别模型及其训练方法 |
Also Published As
Publication number | Publication date |
---|---|
CA2918679C (en) | 2020-08-18 |
CN103558529A (zh) | 2014-02-05 |
CA2918679A1 (en) | 2015-05-21 |
CN103558529B (zh) | 2016-11-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2015070513A1 (zh) | 一种三相共筒式超高压gis局部放电的模式识别方法 | |
CN112014700B (zh) | 基于局部放电多信息融合的gis绝缘子缺陷识别方法及系统 | |
EP3699614B1 (en) | Method and system of partial discharge recognition for diagnosing electrical networks | |
Mondal et al. | Detection, measurement, and classification of partial discharge in a power transformer: Methods, trends, and future research | |
Chan et al. | Time-frequency sparsity map on automatic partial discharge sources separation for power transformer condition assessment | |
Wu et al. | Defect recognition and condition assessment of epoxy insulators in gas insulated switchgear based on multi-information fusion | |
Seo et al. | Probabilistic wavelet transform for partial discharge measurement of transformer | |
CN109596955A (zh) | 局部放电状态确定方法及装置 | |
CN113325277A (zh) | 一种局部放电处理方法 | |
CN103558528A (zh) | 一种局部放电超高频检测系统及方法 | |
Bajwa et al. | An investigation into partial discharge pulse extraction methods | |
Zhao et al. | Interpretation of transformer winding deformation fault by the spectral clustering of FRA signature | |
CN114397569A (zh) | 基于vmd参数优化、样本熵的断路器故障电弧检测方法 | |
CN105785236A (zh) | 一种gis局放检测外部干扰信号排除方法 | |
Chan et al. | Automatic blind equalization and thresholding for partial discharge measurement in power transformer | |
Chan et al. | Hybrid method on signal de‐noising and representation for online partial discharge monitoring of power transformers at substations | |
CN113745049B (zh) | 一种真空灭弧室内真空度监测方法及系统 | |
CN111157853A (zh) | 一种输电线路的放电状态识别方法及系统 | |
Faisal et al. | Prediction of incipient faults in underground power cables utilizing S-transform and support vector regression | |
CN116910470A (zh) | 一种gis组合电器局部放电故障模式识别方法 | |
Seo et al. | A novel signal extraction technique for online partial discharge (PD) measurement of transformers | |
Chang et al. | Online source recognition of partial discharge for gas insulated substations using independent component analysis | |
CN115130516A (zh) | 一种基于图谱功率谱熵的gis设备状态辨识方法及装置 | |
Evagorou et al. | Evaluation of partial discharge denoising using the wavelet packets transform as a preprocessing step for classification | |
Rajendran et al. | Simulation of partial discharges and implementation of noise elimination techniques |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
ENP | Entry into the national phase |
Ref document number: 2918679 Country of ref document: CA |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 14862009 Country of ref document: EP Kind code of ref document: A1 |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 14862009 Country of ref document: EP Kind code of ref document: A1 |