Oil-paper insulation PRPD spectrum recognition method

Abstract

A method for identifying PRPD patterns of oil-paper insulation based on the 3-sigma principle includes determining identification criteria: determining the total number of patterns to be classified; obtaining a PRPD pattern corresponding to each pattern from a partial discharge detector; extracting fingerprint features from the PRPD pattern; calculating the mean and variance of each fingerprint feature corresponding to each pattern; determining each fingerprint criterion for pattern i based on the 3-sigma principle; and obtaining various fingerprint criteria corresponding to various patterns. Identifying the PRPD pattern: obtaining the PRPD pattern to be identified from a commercial partial discharge detector; generating and initializing a scoring variable; and calculating the fingerprint feature x of the PRPD pattern to be identified. j The number of fingerprints that meet the criteria of each pattern is added to the scoring variable; the maximum value and its corresponding pattern are extracted from the scoring variable, and this pattern is the pattern corresponding to the PRPD spectrum to be identified.

Description

Technical Field

[0001] The present invention belongs to the technical field of power transmission and transformation equipment status monitoring, and particularly relates to an oil-paper insulation PRPD spectrum recognition method based on the 3-sigma principle. Background Art

[0002] Accurately and efficiently assessing the operating status of power transformers and comprehensively improving the intelligent operation and maintenance of power transmission and transformation equipment are fundamental to ensuring power supply reliability and are crucial for promoting the transformation of power systems toward a clean, low-carbon future. However, due to the complex production and assembly processes of power transformers, insulation defects such as metal particles and moisture are inevitably introduced into the oil-paper insulation. These defects trigger various types of partial discharge (PD) and may even indirectly lead to insulation degradation and failure. Because PD is a key indicator for quantitatively describing the level of insulation degradation, PD detection has become a key component of factory and preventive testing for power transformers. Because the PD phase and discharge volume caused by different defects vary significantly, identifying insulation defect types through PD detection provides an important reference for condition-based maintenance of power transformers.

[0003] The phase resolved partial discharge (PRPD) spectrum obtained by partial discharge detection can objectively and comprehensively describe the quantitative characteristics of partial discharge, including the discharge phase. The apparent discharge quantity q and the number of discharges n are known. Therefore, pattern recognition research using PRPD patterns as the research object is an important research branch in the field of power transformer condition monitoring. Existing research typically addresses this problem using hard thresholding or model recognition methods. Hard thresholding methods offer a low computational burden, but their generalization and adaptability struggle to meet the requirements of on-site condition monitoring. Model recognition methods offer strong generalization capabilities, but the computational resources required to train the model are prohibitively large, making implementation difficult on inexpensive hardware. Therefore, there is an urgent need for computationally lightweight, accurate, and efficient PRPD pattern recognition methods for oil-paper insulation in engineering applications. Summary of the Invention

[0004] The present invention provides a method for identifying PRPD patterns in oil-paper insulation based on the 3-sigma principle. This method extracts fingerprint features from PRPD patterns corresponding to various patterns and sets adaptive criteria. The adaptive criteria are then used to determine the pattern corresponding to the PRPD pattern to be identified. This method is simple and easy to implement, while offering significant technical advantages: high recognition accuracy, strong adaptability, and good robustness.

[0005] The present invention provides a PRPD spectrum recognition method for oil-paper insulation based on the 3-sigma principle, comprising the following steps:

[0006] S1, determine the fingerprint criterion, where:

[0007] S1.1, determine the total number of modes to be classified, s, where the mode is the cause of the oil-paper partial discharge;

[0008] S1.2, initialize the counting variable i = 1, the counting variable i represents the mode number of the current calculation;

[0009] S1.3, obtain the PRPD spectrum corresponding to mode i from the partial discharge detector;

[0010] S1.4, extract fingerprint features from PRPD maps, extract from each PRPD map corresponding to pattern i and Three two-dimensional vectors, calculate the positive skewness S of each two-dimensional vector k + , negative skewness S k - , positive kurtosis K u + , negative kurtosis K u - , asymmetry Asy and correlation coefficient CC, a total of 18 fingerprint features x are extracted from each PRPD map ij , where the subscript j represents the fingerprint feature number, j = 1, 2, ..., 18, is the discharge phase, q max is the maximum discharge capacity, q ave is the average discharge capacity, n is the number of discharges;

[0011] S1.5, calculate the average value μ of each fingerprint feature corresponding to pattern i ij and variance σ ij ;

[0012] S1.6, fingerprint criteria based on 3-sigma principle mode i: μ ij -3σ ij ≤λ ij ≤μ ij +3σ ij ;

[0013] S1.7, count variable i increments by 1 in a loop. If count variable i is greater than s, the loop is exited; otherwise, it returns to S1.3.

[0014] S1.8, obtain various fingerprint criteria λ corresponding to various patterns ij , where i = 1, 2, ..., s; j = 1, 2, ..., 18;

[0015] S2, identify PRPD patterns, where

[0016] S2.1, obtaining a PRPD spectrum to be identified from a partial discharge detector;

[0017] S2.2, generate and initialize s scoring variables, k1 = k2 = ... = k s =0, the scoring variable is used to quantitatively characterize the similarity between the PRPD spectrum to be identified and various patterns;

[0018] S2.3, initialize the counting variable i=1;

[0019] S2.4, calculate fingerprint feature x j The fingerprint criterion λ is satisfied ij The number of a i , k i =k i +a i ;

[0020] S2.5, count variable i increments by 1 in a loop. If count variable i is greater than s, exit the loop; otherwise, return to S2.4.

[0021] S2.6, extracting the maximum value and its corresponding pattern b from the scoring variable, where pattern b is the pattern corresponding to the PRPD spectrum to be identified.

[0022] In the oil-paper insulation PRPD spectrum identification method based on the 3-sigma principle, the patterns include metal particles, moisture and no defects.

[0023] The beneficial effects of the present invention are:

[0024] (1) Strong adaptability. This method first extracts fingerprint features from the PRPD spectra corresponding to various patterns, and then designs fingerprint feature judgment criteria corresponding to each pattern based on the 3-sigma principle, avoiding the deviation caused by subjective threshold setting. At the same time, the judgment criteria designed using this method can be dynamically adjusted according to the PRPD spectrum characteristics output by commercial partial discharge detectors, showing strong adaptability.

[0025] (2) Simple calculation and high recognition accuracy. Compared with pattern recognition methods such as machine learning or deep learning, this method avoids the complex model training process and has extremely low requirements on the computing power of the hardware. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a schematic diagram of the overall structure of the present invention;

[0027] Figure 2 This is a flow chart of step S1;

[0028] Figure 3 This is a flow chart of step S2;

[0029] Figure 4 This is the PRPD spectrum of the metal particle mode;

[0030] Figure 5 This is the PRPD map of the severe dampening mode;

[0031] Figure 6 This is the PRPD spectrum of the defect-free mode. DETAILED DESCRIPTION

[0032] In order to make the purpose and technical solution of the present invention clearer and easier to understand, Figures 1 to 6 The present invention is further described in detail with reference to the accompanying drawings and embodiments. The specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0033] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first" and "second" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "multiple" means two or more. In the description of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed, detachable, or integral connections; mechanical or electrical connections; direct or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on the specific circumstances.

[0034] Example 1

[0035] The following describes the preferred embodiments of the present invention in detail with reference to the accompanying drawings, so that the advantages and features of the present invention can be more easily understood by those skilled in the art, thereby making a clearer and more precise definition of the scope of protection of the present invention. Figure 1 、 Figure 2 and Figure 3 The embodiment of the present invention includes: a method for identifying PRPD patterns of oil-paper insulation based on the 3-sigma principle, comprising the following steps:

[0036] S1. Determine the recognition criteria.

[0037] S1.1 Determine the total number of patterns to be classified, s. In this example, there are three types of patterns: metal particles, severe moisture, and no defects, so s = 3.

[0038] S1.2 Initialize the counting variable i = 1. The counting variable i represents the mode number currently being calculated.

[0039] S1.3 Obtain the PRPD spectrum corresponding to mode i from the partial discharge detector. In this example, a commercial partial discharge detector (HAEFELY DDX 9121b) is used. 100 PRPD spectra are collected from various modes to determine the fingerprint criteria for each mode. The typical PRPD spectra of the three modes used in this example are as follows: metal particles, severe moisture, and no defects. Figure 4 、 Figure 5 and Figure 6 shown.

[0040] S1.4 Extract fingerprint features from PRPD maps. Extract fingerprint features from each PRPD map corresponding to pattern i. and Three two-dimensional vectors, calculate the positive skewness S of each two-dimensional vector k + , negative skewness S k - , positive kurtosis K u + , negative kurtosis K u - , asymmetry Asy and correlation coefficient CC, so a total of 18 fingerprint features x are extracted from each PRPD map ij The subscript j represents the fingerprint feature number, j = 1, 2, ..., 18. Since PRPD patterns of different modes have certain similarities, the more fingerprint features are selected, the higher the reliability of the established recognition method.

[0041] S1.5 Calculate the average value μ of each fingerprint feature corresponding to pattern i ij and variance σ ij .

[0042] S1.6 gives the fingerprint criteria for pattern i: μ ij -3σ ij ≤λ ij ≤μ ij +3σ ij .

[0043] In step S1.7, the counting variable i is incremented by 1. If the counting variable i is greater than 3, the loop is exited; otherwise, the loop is returned to step S1.3.

[0044] S1.8 Obtain various fingerprint criteria λ corresponding to various patterns ij , where i = 1, 2, 3; j = 1, 2, ..., 18. The fingerprint features of various patterns are shown in Table 1.

[0045] Table 1 Fingerprint criteria for various patterns

[0046]

[0047] S2. Identify PRPD patterns.

[0048] S2.1 obtains the PRPD pattern to be identified from the partial discharge detector and extracts 18 characteristic fingerprints from the PRPD pattern. In this example, a PRPD pattern corresponding to a metal particle defect pattern is selected. The fingerprint features extracted from the PRPD pattern are shown in Table 2.

[0049] Table 2 PRPD pattern fingerprint features to be identified

[0050]

[0051] S2.2 generates and initializes three scoring variables, k1 = k2 = k3 = 0. The scoring variables are used to quantitatively represent the similarity between the PRPD spectrum to be identified and various patterns.

[0052] S2.3 Initialize the counting variable i=1.

[0053] S2.4 Calculate fingerprint feature x j The fingerprint criterion λ is satisfied ij The number of a i , k i =k i +a i ;

[0054] In step S2.5, the counting variable i is incremented by 1. If the counting variable i is greater than 3, the loop is exited; otherwise, the loop is returned to step S2.4.

[0055] S2.6 selects the pattern corresponding to the maximum value from each scoring variable. This pattern is the pattern corresponding to the PRPD spectrum to be identified. In this case, the scoring variables corresponding to the metal particle, severe moisture and defect-free modes are 10, 7 and 8 respectively, so it is determined that the spectrum corresponds to the metal particle mode. Correct identification. Since there is a high similarity between the PRPD spectra of various modes and the partial discharge has a strong randomness, the fingerprint characteristics of a certain mode PRPD spectrum may also meet the criteria of the fingerprint characteristics of other modes. Therefore, in this case, the scoring variables k2 and k3 of the severe defect and defect-free modes are not 0. At the same time, the operation of selecting the maximum value from the scoring variables can also avoid the interference caused by the self-similarity of the PRPD spectrum and the randomness of the partial discharge to the greatest extent.

[0056] The criterion determined using this method can be dynamically adjusted according to the PRPD spectrum characteristics output by commercial partial discharge detectors, and has strong adaptability. At the same time, it is simple to calculate and has high recognition accuracy, and is expected to be widely used in power transformer operation sites.

[0057] The above content is only for explaining the technical idea of the present invention and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution in accordance with the technical idea proposed by the present invention shall fall within the protection scope of the claims of the present invention.

Claims

1. A PRPD spectrum recognition method for oil-paper insulation based on the 3-sigma principle, characterized in that: include: S1, determine the fingerprint criterion, where: S1.1, determine the total number of modes to be classified, s, where the mode is the cause of the oil-paper partial discharge; S1.2, initialize the counting variable i = 1, the counting variable i represents the mode number of the current calculation; S1.3, obtain the PRPD spectrum corresponding to mode i from the partial discharge detector; S1.4, extract fingerprint features from PRPD maps, extract from each PRPD map corresponding to pattern i and Three two-dimensional vectors, calculate the positive skewness S of each two-dimensional vector k + , negative skewness S k - , positive kurtosis K u + , negative kurtosis K u - , asymmetry Asy and correlation coefficient CC, a total of 18 fingerprint features x are extracted from each PRPD map ij , where the subscript j represents the fingerprint feature number, j = 1, 2, ..., 18, is the discharge phase, q max is the maximum discharge capacity, q ave is the average discharge capacity, n is the number of discharges; S1.5, calculate the average value μ of each fingerprint feature corresponding to pattern i ij and variance σ ij ; S1.6, fingerprint criteria based on 3-sigma principle mode i: μ ij -3σ ij ≤λ ij ≤μ ij +3σ ij ; S1.7, count variable i increments by 1 in a loop. If count variable i is greater than s, the loop is exited; otherwise, it returns to S1.

3. S1.8, obtain various fingerprint criteria λ corresponding to various patterns ij , where i = 1, 2, ..., s; j = 1, 2, ..., 18; S2, identify PRPD patterns, where S2.1, obtaining a PRPD spectrum to be identified from a partial discharge detector; S2.2, generate and initialize s scoring variables, k1 = k2 = ... = k s =0, the scoring variable is used to quantitatively characterize the similarity between the PRPD spectrum to be identified and various patterns; S2.3, initialize the counting variable i=1; S2.4, calculate fingerprint feature x j The fingerprint criterion λ is satisfied ij The number of a i , k i =k i +a i ; S2.5, count variable i increments by 1 in a loop. If count variable i is greater than s, exit the loop; otherwise, return to S2.

4. S2.6, extracting the maximum value and its corresponding pattern b from the scoring variable, where pattern b is the pattern corresponding to the PRPD spectrum to be identified.

2. A PRPD spectrum recognition method for oil-paper insulation based on the 3-sigma principle according to claim 1, characterized in that: Preferably, the pattern includes metal particles, moisture and no defects.

Images