Transformer insulation condition assessment method and system based on dissolved gas analysis in oil

Abstract

The present invention discloses a method and system for evaluating the insulation state of a transformer based on dissolved gas analysis in oil. The method includes: obtaining the normal historical data of the transformer, preprocessing it, and constructing a corresponding set of grid spaces; calculating the relative entropy thresholds of all grids in each dimension of the set of grid spaces; comparing the relative entropy of each grid with the corresponding relative entropy threshold, performing grid adaptive partitioning, and then comparing the number of normal historical data in each grid with a threshold to determine normal data clusters; calculating the centroids of all normal data clusters as cluster centers; determining the normal state distance range based on the distances between all normal historical data and the cluster centers; obtaining the real-time data of the transformer to be evaluated, preprocessing it, and determining the insulation state according to the relationship between the distance between the real-time data and the cluster centers and the normal state distance range. The present invention is not affected by manually set thresholds, has high adaptability, and does not require fault sample data.

Description

Technical Field

[0001] The present invention relates to transformer condition assessment technology, and specifically relates to a method and system for assessing the insulation condition of a transformer based on dissolved gas analysis in oil. Background Art

[0002] A transformer is an essential electrical equipment in the power system, and its operating reliability directly affects the safety of the power system. Dissolved Gas Analysis (DGA) technology is the main method for identifying the condition of a transformer. DGA technology can monitor in real time the multi-component gases dissolved in the insulating oil of a transformer, such as hydrogen (H 2 ), carbon monoxide (CO), carbon dioxide (CO 2 ), methane (CH 4 ), ethane (C 2 H 6 ), ethylene (C 2 H 4 ), acetylene (C 2 H 2 ), etc. According to these real-time monitoring values, the assessment system determines the operating condition of the transformer by comparing the real-time gas concentration and its growth rate with a preset alarm threshold. Therefore, setting a reasonable warning threshold is crucial for ensuring the reliability of the assessment.

[0003] The current IEC 60599 standard has collected more than 20,000 transformers from 25 power grids around the world. According to the curve drawn from the dissolved gas data, the gas concentration threshold is obtained according to a certain percentage (i.e., 90%) of the total cumulative analysis. This standard does not take into account the differences between transformers, and the given "90%" percentage is relatively subjective. In this regard, domestic and foreign research institutions have proposed their own standards on the basis of the IEC60599 standard. For example, in the GB / T 7252 standard, transformers are divided into "above 330 kV" and "below 220 kV", and warning values are given respectively. This standard takes into account the different voltage levels of transformers, but the classification standard is relatively vague and cannot provide the assessment threshold for all dissolved gases. Another research has proposed new criteria such as the new three-ratio method based on DGA to obtain diagnostic results, overcoming the conflicts of traditional interpretation techniques. The graphical technology based on characteristic gases has also been used for transformer fault diagnosis. The above methods based on graph theory and new criteria can overcome some limitations, but it is difficult to completely solve the problem of fuzzy boundaries in assessment. In addition, some artificial intelligence algorithms such as artificial neural networks, support vector machines, fuzzy logic, etc. have also been widely used in the assessment of transformer insulation condition, which can achieve high-accuracy assessment, but problems such as the need for a large amount of sample data, large computational volume, and failure to consider transformer differentiation limit the application of the above methods in practical engineering.

[0004] In summary, the existing methods for evaluating the insulation state of transformers have the following technical problems:

[0005] 1. The existing DGA-based method for evaluating the insulation state of transformers does not consider the differences of transformers. For example, the gas concentration and the threshold of gas growth rate for evaluating the insulation state of transformers of different grades and different types of insulating oil should be different. Using the evaluation threshold under the same standard to diagnose the insulation state of different transformers is likely to result in low evaluation accuracy, easy misjudgment and missed judgment, and reduce the reliability of the evaluation system;

[0006] 2. Artificial intelligence algorithms require a large amount of sample data and have high computational complexity, making it difficult to apply them to engineering practice;

[0007] 3. The existing evaluation algorithms require fault sample data, and it is difficult to obtain fault sample data during the actual operation of transformers. Summary of the Invention

[0008] The technical problem to be solved by the present invention: In view of the above problems of the prior art, a method and system for evaluating the insulation state of transformers based on dissolved gas analysis in oil are provided, which have high reliability and scalability, are not affected by the thresholds set manually, have high self-adaptability, and do not require fault sample data.

[0009] To solve the above technical problems, the technical solution adopted by the present invention is as follows:

[0010] A method for evaluating the insulation state of transformers based on dissolved gas analysis in oil, comprising the following steps:

[0011] Obtain the normal historical data of the dissolved gas detection in the insulating oil of transformers of a specified category under normal operating conditions;

[0012] Preprocess the normal historical data, then establish a coordinate system and construct a corresponding grid space set, calculate the relative entropy threshold of all grids on each dimension of the grid space set, and finally perform grid adaptive partitioning according to the comparison results of the relative entropy of each grid and the corresponding relative entropy threshold;

[0013] Compare the number of normal historical data in each grid in the adaptively partitioned grid space set with the threshold to determine the normal data clusters, calculate the centroid of all normal data clusters as the cluster center, and calculate the distance between all normal historical data in the coordinate system and the cluster center, and determine the normal state distance range according to the distance between all normal historical data and the cluster center;

[0014] Obtain the real-time data of the dissolved gas in the insulating oil of the transformer under operation to be evaluated, preprocess the real-time data, then establish a coordinate system and calculate the distance between the real-time data in the coordinate system and the cluster center, and determine the insulation status of the transformer to be evaluated according to the relationship between the distance between the real-time data and the cluster center and the normal state distance range.

[0015] Further, the specified type of transformer specifically refers to the transformer of the same type as the transformer to be evaluated. Before obtaining the normal historical data of the dissolved gas in the insulating oil of the specified type of transformer under normal operation, it also includes: classifying the transformer according to one or more parameters among voltage level, insulating oil type, volume, operation years, manufacturer, actual load, and equipment type.

[0016] Further, the normal historical data is an X 1 ×N 1 -dimensional data set, and the real-time data is an X 1 -dimensional data set, where X 1 represents the number of gas types, and N 1 represents the dimension of the historical data under normal operation. When preprocessing the normal historical data and preprocessing the real-time data, it both includes: calculating the feature quantity in the normal historical data or real-time data, and using the data reduction method to reduce the dimension of the feature quantity, so that the dimension-reduced normal historical data is an X 2 ×N 1 -dimensional data set, and the dimension-reduced real-time data is an X 2 -dimensional data set, where X 2 represents the dimension of the eigenvalue calculated according to the gas concentration.

[0017] Further, when establishing a coordinate system and constructing the corresponding grid space set, it includes: setting the coordinate origin and establishing an X 2 -dimensional coordinate system, setting the step size L 1 according to the value range of the N i data in the i-th dimension coordinate, i = 1, 2,..., X 2 , equally dividing the data set according to the step size L i , and constructing the grid space set D = {d ij}, where d ij represents the number of normal historical data in the j-th grid in the i-th dimension.

[0018] Further, when calculating the relative entropy threshold of all grids in each dimension of the grid space set, it includes:

[0019] Calculate the relative entropy of each grid in the i-th dimension of the grid space set respectively, and the expression is as follows:

[0020]

[0021] where d ij represents the j-th grid in the i-th dimension of the grid space set, and s(d ij ) is the proportion of the number of elements in this grid to the data of the entire grid space set. is the number of grids in the i-th dimension;

[0022] Calculate the median of the relative entropy of all grids in the i-th dimension, and determine the relative entropy threshold according to the relative entropy of all grids in the i-th dimension and the median. The expression is as follows:

[0023]

[0024] where D i is the set of all grids in the i-th dimension, and M i is the median of the relative entropy of the grids in the i-th dimension.

[0025] Furthermore, when performing grid adaptive partitioning according to the comparison result of the relative entropy of each grid and the corresponding relative entropy threshold, it includes: comparing the relative entropy of the current grid with the corresponding relative entropy threshold. If the relative entropy of the current grid is greater than the corresponding relative entropy threshold, then merge the current grid with the previous grid in the same row, and then compare the relative entropy of the next grid in the same row of the current grid with the corresponding relative entropy threshold; if the relative entropy of the current grid is less than the corresponding relative entropy threshold, then directly compare the relative entropy of the next grid in the same row of the current grid with the corresponding relative entropy threshold.

[0026] Furthermore, when comparing the number of normal historical data in each grid in the adaptively partitioned grid space set with the threshold to determine the normal data cluster, it includes:

[0027] Sort the number of normal historical data in each grid in the adaptively partitioned grid space set where the number of normal historical data is greater than 0 in a specified order, and then select the number of normal historical data in the grid at the specified rank as the threshold;

[0028] Traverse the adaptively partitioned grid space set. If the number of normal historical data in the current grid is greater than or equal to the threshold, then the normal historical data in the current grid is the normal data cluster, and then compare the number of normal historical data in other grids centered on the current grid with the threshold; if the number of normal historical data in the current grid is less than the threshold, then the normal historical data in the current grid is noise.

[0029] Furthermore, when calculating the centroid of all normal data clusters as the cluster center, it includes:

[0030] Calculate the centroid of the grid corresponding to each normal data cluster, and then determine the cluster center grid according to the centroid of the grid. The expression of the cluster center grid is as follows:

[0031]

[0032] where, represents the coordinate of the cluster center grid in the i-th dimension, and h s represents the starting grid coordinate of the normal data cluster in the i-th dimension, and h e represents the ending grid coordinate of the normal data cluster in the i-th dimension, represents the grid coordinate of the h-th centroid in the i-th dimension, represents the number of normal historical data within the h-th grid coordinate;

[0033] Take the center position of the cluster center grid as the cluster center.

[0034] Furthermore, when determining the insulation state of the transformer to be evaluated according to the relationship between the distance between the real-time data and the cluster center and the normal state distance range, it includes:

[0035] If the distance between the current real-time data and the cluster center is within the normal state distance range, the insulation state of the transformer to be evaluated is normal;

[0036] If the distance between the current real-time data and the cluster center is outside the normal state distance range, reduce the time interval for obtaining the real-time data and obtain the next real-time data, preprocess the next real-time data, and then establish a coordinate system and calculate the distance between the next real-time data in the coordinate system and the cluster center;

[0037] If the distance between the next real-time data and the cluster center is within the normal state distance range, the insulation state of the transformer to be evaluated is normal; if the distance between the next real-time data and the cluster center is outside the normal state distance range, the insulation state of the transformer to be evaluated is abnormal.

[0038] The present invention also proposes a transformer insulation state evaluation system based on dissolved gas analysis in oil, including:

[0039] A dissolved gas detection module in oil, used to detect the dissolved gas in the insulating oil;

[0040] A monitoring period module, used to enable the dissolved gas detection module in oil according to a preset period;

[0041] An insulation state evaluation module, used to execute any one of the transformer insulation state evaluation methods based on dissolved gas analysis in oil to evaluate whether the insulation state of the transformer is abnormal.

[0042] Compared with the prior art, the advantages of the present invention are:

[0043] The present invention calculates the cluster centers based on normal historical data, further calculates the range of the distances between the normal data and the cluster centers as the normal state range, and then determines whether there is an abnormality in the transformer insulation state by calculating the distance between the real-time detection data and the cluster centers and judging the relationship between the calculation result and the normal state range. The evaluation of the transformer insulation state can be realized by judging the real-time measurement data, without the need for fault sample data.

[0044] When calculating the cluster centers, the present invention constructs a grid space set for the data set of normal historical data, then calculates the relative entropy threshold of all grids in each dimension, and then merges the grids by comparing the relative entropy of each grid with the relative entropy threshold, so as to realize adaptive grid division to effectively reduce the calculation amount. Finally, the number of samples in the divided grid is compared with the threshold to realize data cleaning, and the centroid of the cleaned data is calculated as the cluster center, without the need for manual threshold setting and with high adaptability.

[0045] The present invention adopts the same steps for each type of classified transformer to realize adaptive threshold evaluation, which can be applied to the insulation state evaluation of different voltage levels, different insulation oil types, and even the same type of transformer under different operation years, and can effectively improve the reliability and scalability of the evaluation system. Brief Description of the Drawings

[0046] Figure 1 It is a flowchart of the method of the embodiment of the present invention.

[0047] Figure 2 It is a schematic diagram of the coordinate system established in the embodiment of the present invention.

[0048] Figure 3 It is a schematic diagram of the grid space set constructed in the embodiment of the present invention.

[0049] Figure 4 It is a schematic diagram of the adaptively divided grid in the embodiment of the present invention.

[0050] Figure 5 It is a schematic diagram of the cluster center grid in the embodiment of the present invention. Detailed Embodiment

[0051] The following further describes the present invention in conjunction with the accompanying drawings of the specification and specific preferred embodiments, but does not limit the protection scope of the present invention thereby.

[0052] Embodiment 1

[0053] This embodiment proposes a method for evaluating the insulation state of a transformer based on the analysis of dissolved gases in oil, which can be applied to the insulation state evaluation of different voltage levels, different insulation oil types, and even the same type of transformer under different operation years, and can effectively improve the reliability and scalability of the evaluation system. AsFigure 1 As shown in the figure, the method of this embodiment includes the following steps:

[0054] S1) Classify the transformers;

[0055] S2) Obtain the normal historical data of the dissolved gas detection in the insulating oil under the normal operation state of the transformers of the specified category. In this embodiment, the transformers of the specified category specifically refer to the transformers with the same category as the transformers to be evaluated;

[0056] S3) Preprocess the normal historical data;

[0057] S4) Establish a coordinate system for the preprocessed normal historical data and construct a corresponding grid space set, calculate the relative entropy threshold of all grids in each dimension of the grid space set, and perform grid adaptive division according to the comparison results of the relative entropy of each grid and the corresponding relative entropy threshold;

[0058] S5) Compare the number of normal historical data in each grid in the adaptively divided grid space set with the threshold to determine the normal data clusters, calculate the centroid of all normal data clusters as the cluster center, and calculate the distance between all normal historical data in the coordinate system and the cluster center, and determine the normal state distance range according to the distance between all normal historical data and the cluster center;

[0059] S6) Obtain the real-time data of the dissolved gas detection in the insulating oil under the operation state of the transformers to be evaluated;

[0060] S7) Preprocess the real-time data, establish a coordinate system for the preprocessed real-time data and calculate the distance between the real-time data in the coordinate system and the cluster center, and determine the insulation state of the transformers to be evaluated according to the relationship between the distance between the real-time data and the cluster center and the normal state distance range.

[0061] The following is a specific description of each main step.

[0062] In step S1 of this embodiment, the basis for classifying the transformers includes, but is not limited to, one or more parameters such as voltage level, insulating oil type, volume, operation years, manufacturer, actual load, and equipment type. Based on these parameters, the parameter intervals corresponding to different categories can be divided through expert discussion to classify the transformers with different parameter values.

[0063] In steps S2 and S6 of this embodiment, the dissolved gas detection includes, but is not limited to, detecting carbon monoxide, carbon dioxide, methane, ethane, ethylene, acetylene, hydrogen, etc. Correspondingly, the normal historical data is an X 1 ×N 1 -dimensional data set, and the real-time data is an X 1 -dimensional data set, where X1 Denotes the number of gas types for dissolved gas detection, N 1 Denotes the dimension of historical data under normal operating conditions.

[0064] In step S3 and step S7 of this embodiment, the specific process of preprocessing is to first calculate characteristic quantities such as the carbon dioxide / carbon monoxide ratio, acetylene / hydrogen ratio, and total hydrocarbon content, and then use data dimensionality reduction methods such as principal component analysis, streamline learning, and support vector machines to reduce the dimensionality of the aforementioned characteristic quantities, so that the normal historical data after dimensionality reduction is X 2 ×N 1 -dimensional data set, and the real-time data after dimensionality reduction is X 2 -dimensional data set, where X 2 Denotes the dimension of the eigenvalue calculated according to the gas concentration, and X 2 <X 1 .

[0065] In step S4 of this embodiment, when establishing a coordinate system and constructing a corresponding grid space set, it includes:

[0066] Establishing a coordinate system: Set the coordinate origin and establish an X 2 -dimensional coordinate system, as Figure 2 shown. When X 2 =2, take the first dimension as the horizontal axis and the second dimension as the vertical axis to establish a two-dimensional coordinate system, and then plot each normal historical data as a normal state sample in the coordinate system according to the value of each dimension of the normal historical data;

[0067] Constructing a grid space set: The preprocessed normal historical data is X 2 ×N 1 -dimensional data set. Set the step size L 1 according to the value range of the N i data samples on the i-th dimension coordinate, i = 1, 2,..., X 2 . Divide the data set equally according to the step size L i to construct a grid space set D = {d ij}, where d ij represents the number of normal historical data in the j-th grid on the i-th dimension, as Figure 3 shown. For example, if there are two normal state samples in the fifth grid of the i = 1 dimension, then d 15 = 2.

[0068] In step S4 of this embodiment, when calculating the relative entropy threshold of all grids on each dimension of the grid space set, it includes:

[0069] S41) Calculate the relative entropy of each grid on the i-th dimension of the grid space set respectively, and the expression is as follows:

[0070]

[0071] Among them, d ij represents the j-th grid in the i-th dimension of the grid space set, and s(d ij ) is the proportion of the number of elements in this grid in the data of the entire grid space set. is the number of grids in the i-th dimension;

[0072] S42) Calculate the relative entropy of all grids in the i-th dimension, and the expression is as follows:

[0073]

[0074] Among them, D i is the set of all grids in the i-th dimension;

[0075] S43) Calculate the median of the relative entropy of all grids in the i-th dimension, and determine the relative entropy threshold according to the relative entropy of all grids in the i-th dimension and the median, and the expression is as follows:

[0076]

[0077] Among them, M i is the median of the relative entropy of the grids in the i-th dimension.

[0078] In step S4 of this embodiment, when performing grid adaptive partitioning according to the comparison result of the relative entropy of each grid and the corresponding relative entropy threshold, it includes:

[0079] Traverse the relative entropy e(d ij ) of all grids in the i-th dimension of the grid space, compare the relative entropy e(d ij ) of each grid with the corresponding relative entropy threshold T. If e(d ij ) > T, that is, the relative entropy of the current grid is greater than the corresponding relative entropy threshold, then merge the current grid with the previous grid in the same row, and then compare the relative entropy of the next grid in the same row of the current grid with the corresponding relative entropy threshold; if e(d ij ) < T, that is, the relative entropy of the current grid is less than the corresponding relative entropy threshold, then directly compare the relative entropy of the next grid in the same row of the current grid with the corresponding relative entropy threshold.

[0080] For each dimension of the grid space set, after performing steps S41 to S44, perform grid adaptive partitioning according to the comparison result of the relative entropy of each grid and the corresponding relative entropy threshold in this dimension. Finally, obtain the grid space set D' = {d′ ij} after adaptive partitioning, where d′ ij represents the number of normal state samples in the j-th grid in the i-th dimension, as Figure 4 shown.

[0081] In step S5 of this embodiment, when comparing the number of normal historical data in each grid in the adaptively partitioned grid space set with a threshold to determine the normal data clusters, it includes:

[0082] S51) Sort the number of normal historical data in each grid in the adaptively partitioned grid space set where the number of normal historical data is greater than 0 in a specified order, and then select the number of normal historical data in the grid at the specified ranking as the threshold; in this embodiment, sort the number d′ of non-zero normal state samples in the grid space set D' ij in descending order, and take the number corresponding to the ranking of 95% * N 2 as the density threshold T 2 , where N 2 is the number of the number d′ of non-zero normal state samples in the grid space set D' ij ;

[0083] S52) Traverse the adaptively partitioned grid space set. If the number of normal historical data in the current grid, that is, the number d′ of normal state samples ij is greater than or equal to the density threshold T 2 , then the normal historical data in the current grid is a normal data cluster, and then compare the number of normal historical data in other grids centered on the current grid with the threshold; if the number of normal historical data in the current grid is less than the threshold, the normal historical data in the current grid is noise; mark the grids that have been visited. After traversing all grids, the normal data clusters and noise in the normal historical data can be obtained, as Figure 5 shown, where the black dots represent normal data clusters and the white dots represent noise.

[0084] In step S5 of this embodiment, when calculating the centroid of all normal data clusters as the cluster center, it includes:

[0085] S501) Calculate the centroid of the grid corresponding to each normal data cluster, and then determine the centroid grid, that is, the grid coordinates where the cluster center is located, according to the centroid of the grid. The expression of the centroid grid is as follows:

[0086]

[0087] where, represents the centroid grid coordinate of the i-th dimension, h s represents the starting grid coordinate of the normal data cluster in the i-th dimension, h e represents the ending grid coordinate of the normal data cluster in the i-th dimension, represents the h-th grid coordinate in the i-th dimension, represents the number of normal historical data in the h-th grid coordinate;

[0088] S502) Use the center position of the cluster center grid as the cluster center C normal , such as Figure 5 shown. The cluster center grid obtained by step S501 is grid A, and its coordinates are a, b, c, and d respectively. Then the coordinates of the cluster center are [(a + b) / 2, (c + d) / 2].

[0089] In step S5 of this embodiment, calculating the distance between all normal historical data in the coordinate system and the cluster center specifically calculates the Euclidean distance between all normal historical data in the coordinate system and the cluster center. Determining the normal state distance range according to the distance between all normal historical data and the cluster center specifically selects the maximum value d max and the minimum value d min of the Euclidean distances between all normal historical data and the cluster center as the upper and lower limit values of the normal state distance range.

[0090] In step S7 of this embodiment, when establishing a coordinate system for the preprocessed real-time data and calculating the distance between the real-time data in the coordinate system and the cluster center, it includes:

[0091] Establishing a coordinate system: Set the coordinate origin and establish an X 2 -dimensional coordinate system, and then plot the real-time data in the coordinate system according to the value of each dimension of the preprocessed real-time data;

[0092] Calculating the distance between the real-time data in the coordinate system and the cluster center: Calculate the Euclidean distance d r between the real-time data and the cluster center according to the real-time data and the coordinates of the cluster center in the coordinate system.

[0093] In step S7 of this embodiment, when determining the insulation state of the transformer to be evaluated according to the relationship between the distance between the real-time data and the cluster center and the normal state distance range, it includes:

[0094] Calculating

[0095] If ρ ≥ 0 and d r ≠ d max , that is, the distance d r between the current real-time data and the cluster center is within the normal state distance range, then the insulation state of the transformer to be evaluated is normal;

[0096] If ρ < 0 and d r ≠ d max , that is, the distance between the current real-time data and the cluster center is outside the normal state distance range, reduce the time interval for obtaining the real-time data and obtain the next real-time data for rapid measurement, preprocess the next real-time data, and then establish a coordinate system and calculate the distance d' between the next real-time data in the coordinate system and the cluster centerr ;

[0097] Calculate

[0098] If ρ≥0 and d’ r ≠d max , that is, the distance between the next real-time data and the cluster center is within the normal state distance range, then the insulation state of the transformer to be evaluated is normal; if ρ < 0 and d’ r ≠d max , that is, the distance between the next real-time data and the cluster center is outside the normal state distance range, indicating that the measurement result of the rapid measurement does not meet the judgment condition either, then it is judged that the insulation state of the transformer to be evaluated is abnormal.

[0099] Embodiment 2

[0100] This embodiment proposes a transformer insulation state evaluation system based on dissolved gas analysis in oil, including:

[0101] A central processing module for system control, data processing, display and storage, evaluation information sending, etc.;

[0102] A dissolved gas detection module in oil for detecting dissolved gases in insulating oil, including but not limited to a gas precipitation unit and a gas detection unit in transformer insulating oil, etc., and the detection principle of the gas detection unit includes but not limited to gas chromatography, photoacoustic spectroscopy, tunable diode laser absorption spectroscopy, etc.;

[0103] A monitoring period module for enabling the dissolved gas detection module in oil according to a preset period. The monitoring period module includes a periodic monitoring mode and a rapid monitoring mode. The periodic monitoring mode is the monitoring period set by the user, and the default is to measure once every 24 hours. The rapid detection mode is to immediately start the system for the second measurement after the current measurement ends;

[0104] An insulation state evaluation module for executing the transformer insulation state evaluation method based on dissolved gas analysis in oil described in Embodiment 1 to evaluate whether the insulation state of the transformer is abnormal.

[0105] In summary, the present invention discloses a transformer insulation state evaluation method and system based on dissolved gas analysis in oil, having the following advantages:

[0106] The present invention provides adaptive threshold evaluation, which can be applied to the insulation state evaluation of different voltage levels, different types of insulating oil, and even the same type of transformer under different operating years, and can effectively improve the reliability and scalability of the evaluation system.

[0107] Based on the historical data of the normal operation state of the transformer, first, a data preprocessing method is adopted to reduce the data dimension; then, an adaptive grid division is performed on the data samples, and the number of samples in the grid is compared with a threshold to achieve data cleaning. Finally, the centroid of the cleaned data is calculated as the cluster center. The implementation process of this method does not require manual setting of the threshold, has high adaptability, and effectively reduces the calculation amount through data dimensionality reduction and adaptive grid division.

[0108] Based on the existing normal operation sample set, the present invention realizes the insulation state evaluation of the transformer by judging the subordinate relationship between the real-time measurement data and the normal sample set, without the need for fault sample data.

[0109] The above is only the preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions falling within the idea of the present invention belong to the protection scope of the present invention. It should be pointed out that for those of ordinary skill in the art, several improvements and refinements made without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A transformer insulation condition assessment method based on dissolved gas analysis in oil, characterized in that: The following steps are involved: Obtain normal historical data of dissolved gas detection in insulating oil of a specified type of transformer under normal operating conditions; Preprocessing the normal historical data, then establishing a coordinate system and constructing a corresponding grid space set, calculating the relative entropy threshold of all grids in each dimension of the grid space set, and finally performing grid adaptive division according to the comparison result of the relative entropy of each grid with the corresponding relative entropy threshold; The number of normal historical data in each grid in the adaptively divided grid space set is compared with the threshold to determine the normal data cluster, the centroid of all normal data clusters is calculated as the cluster center, and the distance between all normal historical data and the cluster center in the coordinate system is calculated, and the normal state distance range is determined according to the distance between all normal historical data and the cluster center; Real-time data of dissolved gas detection in insulating oil of a transformer to be evaluated under operating conditions is obtained, the real-time data is preprocessed, a coordinate system is established, and the distance between the real-time data and the cluster center in the coordinate system is calculated. The insulation state of the transformer to be evaluated is determined according to the relationship between the distance between the real-time data and the cluster center and the distance range of the normal state.

2. The transformer insulation condition assessment method based on dissolved gas analysis in oil according to claim 1, characterized in that: The transformer of the specified category specifically refers to a transformer of the same category as the transformer to be evaluated. Before obtaining the normal historical data of the detection of dissolved gas in the insulating oil of the transformer of the specified category under normal operating conditions, it also includes: classifying the transformer according to one or more parameters of voltage level, insulating oil type, volume, operating years, manufacturer, actual load, and equipment type.

3. The transformer insulation condition assessment method based on dissolved gas analysis in oil according to claim 1, characterized in that: The normal historical data is a data set of X1×N1 dimensions, and the real-time data is a data set of X1 dimensions, wherein X1 represents the number of gas types, and N1 represents the dimension of historical data under normal operating conditions. Preprocessing the normal historical data and preprocessing the real-time data both include: calculating characteristic quantities in the normal historical data or real-time data, and reducing the dimension of the characteristic quantities using a data reduction method, so that the normal historical data after dimensionality reduction is a data set of X2×N1 dimensions, and the real-time data after dimensionality reduction is a data set of X2 dimensions, wherein X2 represents the dimension of characteristic values ​​calculated based on gas concentration.

4. The transformer insulation condition assessment method based on dissolved gas analysis in oil according to claim 3 is characterized in that: When establishing a coordinate system and constructing the corresponding grid space set, it includes: setting the coordinate origin, establishing an X2-dimensional coordinate system, setting the step size L according to the value range of the N1 data in the i-th dimension coordinate i , i=1,2,…,X2, according to the step length L i Divide the data set equally and construct a grid space set D = {d ij }, d ij Represents the number of normal historical data in the jth grid on the i-th dimension.

5. The transformer insulation condition assessment method based on dissolved gas analysis in oil according to claim 4 is characterized in that: When calculating the relative entropy threshold of all grids in each dimension of the grid space set, it includes: Calculate the relative entropy of each grid on the i-th dimension of the grid space set respectively, the expression is as follows: Among them, d ij represents the jth grid in the i-th dimension of the grid space set, s(d ij ) is the ratio of the number of elements in the grid to the total grid space data set, is the number of grids in the i-th dimension; Calculate the median of the relative entropy of all grids in the i-th dimension, and determine the relative entropy threshold based on the relative entropy and median of all grids in the i-th dimension. The expression is as follows: Among them, D i is the set of all grids in the i-th dimension, M i is the median of the relative entropy of the i-th grid.

6. The transformer insulation condition assessment method based on dissolved gas analysis in oil according to claim 1, characterized in that: When adaptively dividing the grid according to the comparison result between the relative entropy of each grid and the corresponding relative entropy threshold, it includes: comparing the relative entropy of the current grid with the corresponding relative entropy threshold; if the relative entropy of the current grid is greater than the corresponding relative entropy threshold, merging the current grid with the previous grid in the same row, and then comparing the relative entropy of the next grid in the same row of the current grid with the corresponding relative entropy threshold; if the relative entropy of the current grid is less than the corresponding relative entropy threshold, directly comparing the relative entropy of the next grid in the same row of the current grid with the corresponding relative entropy threshold.

7. The transformer insulation condition assessment method based on dissolved gas analysis in oil according to claim 1, characterized in that: When comparing the number of normal historical data in each grid in the adaptively divided grid space set with the threshold to determine the normal data cluster, it includes: The number of normal historical data in each grid whose number of normal historical data is greater than 0 in the adaptively divided grid space set is sorted in a specified order, and then the number of normal historical data in the grid with the specified ranking is selected as the threshold; Traverse the adaptively divided grid space set. If the number of normal historical data of the current grid is greater than or equal to the threshold, the normal historical data of the current grid is a normal data cluster. Then compare the number of normal historical data in other grids centered on the current grid with the threshold. If the number of normal historical data of the current grid is less than the threshold, the normal historical data of the current grid is noise.

8. The transformer insulation condition assessment method based on dissolved gas analysis in oil according to claim 1, characterized in that: When calculating the centroid of all normal data clusters as cluster centers, it includes: Calculate the centroid of the grid corresponding to each normal data cluster, and then determine the cluster center grid according to the centroid of the grid. The cluster center grid expression is as follows: in, represents the grid coordinates of the cluster center in the i-th dimension, h s represents the starting grid coordinates of the normal data cluster in the i-th dimension, h e represents the ending grid coordinates of the normal data cluster in the i-th dimension, represents the hth grid coordinate in the i-th dimension, Indicates the number of normal historical data in the hth grid coordinate; The center position of the cluster center grid is taken as the cluster center.

9. The transformer insulation condition assessment method based on dissolved gas analysis in oil according to claim 1, characterized in that: When determining the insulation state of the transformer to be evaluated based on the relationship between the distance between the real-time data and the cluster center and the normal state distance range, it includes: If the distance between the current real-time data and the cluster center is within the normal state distance range, the insulation state of the transformer to be evaluated is normal; If the distance between the current real-time data and the cluster center is outside the normal distance range, reduce the time interval for acquiring real-time data and acquire the next real-time data, pre-process the next real-time data, and then establish a coordinate system and calculate the distance between the next real-time data and the cluster center in the coordinate system; If the distance between the next real-time data and the cluster center is within the normal state distance range, the insulation state of the transformer to be evaluated is normal; if the distance between the next real-time data and the cluster center is outside the normal state distance range, the insulation state of the transformer to be evaluated is abnormal.

10. A transformer insulation status assessment system based on dissolved gas analysis in oil, characterized in that: include: Dissolved gas detection module in oil, used to detect dissolved gas in insulating oil; A monitoring cycle module, used to enable the dissolved gas detection module in oil according to a preset cycle; The insulation state evaluation module is used to execute the transformer insulation state evaluation method based on dissolved gas analysis in oil as described in any one of claims 1 to 9 to evaluate whether the insulation state of the transformer is abnormal.

Images