CN111413558A - A Transformer Fault Diagnosis Method Based on DBSCAN - Google Patents

Abstract

The invention discloses a DBSCAN-based transformer fault diagnosis method, comprising the following steps: S1: collecting n typical transformer oil chromatography data sequences X={x ₁ ,x ₂ ,x ₃ ,x ₄ ,x with known fault types ₅ , x ₆ }; S2: Standardize the faulty transformer data, reduce the specific numerical difference of the faulty oil chromatography data column, ensure that each data has continuity and equivalence, and obtain the standardized transformer data matrix Y=(y _ij ) _n×6 ; S3: Calculate the distance between the standardized data of each fault type; Obtain the distance data matrix D=(d _ij ) _n×n ; S4: Use DBSCAN to cluster fault transformer data; S5: Transformer fault diagnosis: Each data to be diagnosed is normalized and added to the known fault data matrix Y, and the cluster of fault types to which the data to be diagnosed belongs is obtained, which is the fault type diagnosed by the method in this paper. The method is applied to transformer fault diagnosis, which can comprehensively reflect the faults inside the transformer.

Description

A Transformer Fault Diagnosis Method Based on DBSCAN

technical field

The invention belongs to the technical field of transformer fault diagnosis and processing, and in particular relates to a DBSCAN-based transformer fault diagnosis method.

Background technique

The power transformer is one of the key equipments in the process of transmitting electric energy in the power system. It is of great significance to ensure the safe and stable operation of the transformer. When the transformer is put into use, it will fail due to excessive load, insulation aging of materials, natural disasters and other reasons. Once the transformer is damaged, it will not be able to transport electricity normally, which will bring huge losses to some national economies and bring serious trouble to people's lives.

At present, the most widely used method for chromatographic analysis and diagnosis of transformer oil is the three-ratio method, but the three-ratio method also has problems such as missing codes and cannot fully reflect the faults that occur inside the transformer.

SUMMARY OF THE INVENTION

The purpose of the present invention is to solve the shortcomings of using the three-ratio method to diagnose the internal faults of the transformer, and to provide a transformer fault diagnosis method based on DBSCAN, which can be applied to the transformer fault diagnosis to more comprehensively reflect the faults that occur inside the transformer. The scientific and technological progress of the electric power industry has important scientific value.

In order to achieve the above-mentioned technical purpose, a technical solution provided by the present invention is a method for diagnosing transformer faults based on DBSCAN, comprising the following steps:

S1: Collect n pieces of typical transformer oil chromatography data sequence X={x ₁ ,x ₂ ,x ₃ ,x ₄ ,x ₅ ,x ₆ }, where {x ₁ ,x ₂ ,x ₃ ,x ₄ , x ₅ } represents the characteristic quantity used for fault diagnosis; x ₆ represents the fault type; constitutes the fault transformer data matrix X=(x _ij ) _n×6 ;

S2: standardize the faulty transformer data, reduce the specific numerical difference of the faulty oil chromatographic data column, ensure that each data has continuity and equivalence, and obtain a standardized transformer data matrix Y=(y _ij ) _n×6 ;

S3: Calculate the distance between the standardized data of each fault type; obtain the distance data matrix D=(d _ij ) _n×n ;

S4: Use DBSCAN to cluster fault transformer data;

S5: Transformer fault diagnosis: standardize each data to be diagnosed and add it to the known fault data matrix Y, and obtain the cluster of fault types to which the data to be diagnosed belongs, which is the fault type diagnosed by the method in this paper.

The characteristic quantities of fault diagnosis in step S1 are: hydrogen, methane, ethane, ethylene, and acetylene.

In step S1, the known fault types are: low temperature overheating, medium temperature overheating, high temperature overheating, partial discharge, low energy discharge, low energy discharge and overheating, high energy discharge and high energy discharge and overheating.

The method of data standardization processing in step S2 is: normalization processing, and the specific formula is as follows:

The distance between the data in step S3 is the Euclidean distance: the Euclidean distance formula for the n-dimensional space is as follows:

The oil chromatographic number is the measured content of H ₂ , CH ₄ , C ₂ H ₆ , C ₂ H ₄ , and C ₂ H ₂ , which are 5-dimensional data in total, so n is 5 here.

Step S4 includes the following steps:

S41. Set DBSCAN parameters, Epslion neighborhood value Eps and MinPts value MinPts;

S42. Find the core points of each category: determine the first data point, and find all the points within the radius Eps of the data point; if the number of points within the radius Eps of the data point is equal to or greater than Minpts, then the data point It is the core point, and the circle with its radius Eps is its neighborhood. All points in the neighborhood belong to the same fault type. These points are boundary points or other core points; find each one in turn. Core point; if the distance between two core points is less than the radius Eps, then they are densely connected, and all points within the intersecting circle formed by the two belong to the same fault type, and these core points can be found All sample sets that the object can reach density are the same fault type;

S43. Find the boundary points of each category: if the number of points within the Eps radius of a certain data (including itself) is less than Minpts, but the data point is within the Eps neighborhood of a certain core point, then the data point is the boundary point; the boundary point and the core point to which it belongs are of the same fault type;

S44. Search for noise points: other points that neither belong to core data points nor boundary points are noise points; the number of points within the Eps radius of noise points is less than Minpts, and also not within the Eps neighborhood of any core point inside, does not belong to any type of failure;

S45, constitute a typical data column and a map of the fault.

Beneficial effects of the present invention: The present invention provides a method for diagnosing transformer faults based on DBSCAN, and applying the method to fault diagnosis of transformers can more comprehensively reflect the faults that occur inside the transformer, and can quickly and accurately determine the type of internal faults in the transformer , has important scientific value to the scientific and technological progress of the electric power industry.

Description of drawings

FIG. 1 is a method flowchart of a DBSCAN-based transformer fault diagnosis method of the present invention.

Fig. 2 is a transformer fault diagnosis effect diagram using a DBSCAN-based transformer fault diagnosis method of the invention.

Detailed ways

In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only the best of the present invention. The embodiments are only used to explain the present invention, and do not limit the protection scope of the present invention. All other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present invention.

Embodiment: a kind of transformer fault diagnosis method based on DBSCAN, its technological process is as shown in Figure 1, and the concrete algorithm is implemented according to the following steps:

S1: Collect 80 typical transformer oil chromatographic data sequences X={x ₁ ,x ₂ ,x ₃ ,x ₄ ,x ₅ ,x ₆ }, where {x ₁ ,x ₂ ,x ₃ ,x ₄ , x ₅ } represents the characteristic quantity used for fault diagnosis; x ₆ represents the fault type; constitutes the fault transformer data matrix X=(x _ij ) _80×6 . Among them, the fault diagnosis characteristic quantities are: hydrogen, methane, ethane, ethylene, acetylene; the known fault types are: low temperature overheating, medium temperature overheating, high temperature overheating, partial discharge, low energy discharge, low energy discharge and overheating, high energy discharge, high energy discharge and combined Eight types of overheating; the corresponding x ₆ values are 1, 2, 3, 4, 5, 6, 7, and 8.

S2: Standardize the faulty transformer data, reduce the specific numerical difference of the faulty oil chromatographic data column, ensure that each data has continuity and equivalence, and obtain a standardized transformer data matrix Y=(y _ij ) _80×6 .

The standardization processing method is: normalization processing, the formula is as follows:

S3: Calculate the distance between the normalized data of each fault type. A distance data matrix D=(d _ij ) _80×80 is obtained.

The distance between the data is Euclidean distance: the Euclidean distance formula for n-dimensional space is as follows:

Since the oil chromatography number is the measured H ₂ , CH ₄ , C ₂ H ₆ , C ₂ H ₄ , C ₂ H ₂ content; A total of 5-dimensional data, so here n is 5.

S4: Use DBSCAN to cluster fault transformer data.

S41. Set the DBSCAN parameter, the Epslion neighborhood value Eps, and the MinPts value MinPts.

S42, find the core points of each category: determine the first data point, and find all the points within the radius Eps of the data point. If the number of points (including itself) within the range of the radius Eps of the data point is equal to or greater than Minpts, the data point is the core point, and the circle with the radius of Eps is its neighborhood. All points belong to the same fault type, and these points are boundary points or other core points. Find each core point in turn. If the distance between two core points is less than the radius Eps, then they are densely connected, and all points within the intersecting circle formed by the two belong to the same fault type, and it can be found that these core objects can be densely All sample sets that are reachable are of the same fault type.

S43. Find the boundary points of each category: if the number of points within the Eps radius of a certain data (including itself) is less than Minpts, but the data point is within the Eps neighborhood of a certain core point, then the data point is the boundary point. The boundary point and the core point to which it belongs are of the same fault type.

S44, looking for noise points: other points that neither belong to the core data points nor the boundary points are noise points. The number of points within the Eps radius of the noise point (including itself) is less than Minpts, and it is not within the Eps neighborhood of any core point, and does not belong to any fault type.

S45, constitute a typical data column and a map of the fault.

S5. Transformer fault diagnosis: standardize each data to be diagnosed and add it to the known fault data matrix Y, and obtain the cluster of fault types to which the data to be diagnosed belongs, which is the fault type diagnosed by the method in this paper.

Specifically, the collected transformer oil chromatographic data sequences of 10 cases of 8 types of faults are shown in Table 1 (two cases for each type):

Table 1: Transformer Oil Chromatographic Data Sequence

The standardized matrix is shown in Table 2:

Table 2: Transformer Oil Chromatographic Data Normalized Sequences

Characteristic gas content H₂ CH₄ C₂H₆ C₂H₄ C₂H₂ low temperature overheating 0.145 0.318 0.471 0.066 0.000 low temperature overheating 0.033 0.336 0.513 0.116 0.001 medium overheat 0.060 0.369 0.185 0.385 0.001 medium overheat 0.035 0.223 0.231 0.511 0.000 high temperature overheating 0.186 0.180 0.067 0.554 0.014 high temperature overheating 0.231 0.164 0.077 0.497 0.031 Partial Discharge 0.873 0.065 0.052 0.011 0.000 Partial Discharge 0.859 0.070 0.045 0.026 0.000 low energy discharge 0.149 0.067 0.138 0.001 0.645 low energy discharge 0.168 0.164 0.037 0.088 0.543 low energy and overheating 0.062 0.201 0.040 0.062 0.636 low energy and overheating 0.034 0.184 0.030 0.048 0.705 high energy discharge 0.586 0.126 0.020 0.108 0.159 high energy discharge 0.533 0.121 0.020 0.141 0.186 High energy and overheating 0.264 0.332 0.025 0.333 0.045 High energy and overheating 0.255 0.269 0.082 0.342 0.051

Compute the Euclidean distance for each data series:

For example, the distance between the first column and the second column is:

The distance between the second and third columns is:

Select DBSCAN parameters, Eps=0.1, MinPts=3 for DBSCAN algorithm analysis.

The algorithm is divided into 13 clusters, the specific types are shown in Table 3

Table 3 Typical clusters of fault types

The 80 columns of known failure data entered will be typical data for judging failures generated by new oil chromatography data. Input the oil chromatography data with diagnostics into the typical data column, and DBSCAN can obtain the cluster to which any data belongs, so as to determine its fault type.

The diagnosis results are shown in Table 4

Table 4 Fault diagnosis results

H₂ CH₄ C₂H₆ C₂H₄ C₂H₂ Fault type diagnostic result 59.26 120.18 185.6 36.84 0.05 low temperature overheating low temperature overheating 22.7 150.3 160.8 400.8 0 medium overheat medium overheat 238 87 189 343 5.2 high temperature overheating noise point 70.4 69.5 28.9 241.2 10.4 high temperature overheating high temperature overheating 260 8 2.5 2 0 Partial Discharge Partial Discharge 30.64 29.84 6.82 16.071 99 low energy discharge low energy discharge 7.85 21.26 4.44 7.85 65.8 low energy and overheating low energy and overheating 410 120 18 71 250 high energy discharge low energy discharge 41 59 8.5 128 270 High energy and overheating High energy and overheating 72 76 twenty two 135 twenty four low energy and overheating High energy and overheating

The specific embodiment described above is a preferred embodiment of a DBSCAN-based transformer fault diagnosis method of the present invention, and is not intended to limit the specific implementation scope of the present invention. The scope of the present invention includes but is not limited to the specific embodiment. Equivalent changes made according to the shape and structure of the present invention are all within the protection scope of the present invention.

Claims (7)

1. a transformer fault diagnosis method based on DBSCAN, is characterized in that: comprise the steps: S1: Collect n pieces of typical transformer oil chromatography data sequence X={x ₁ ,x ₂ ,x ₃ ,x ₄ ,x ₅ ,x ₆ }, where {x ₁ ,x ₂ ,x ₃ ,x ₄ , x ₅ } represents the characteristic quantity used for fault diagnosis; x ₆ represents the fault type; constitutes the fault transformer data matrix

S2: standardize the faulty transformer data, reduce the specific numerical difference of the faulty oil chromatographic data column, ensure that each data has continuity and equivalence, and obtain a standardized transformer data matrix Y=(y _ij ) _n×6 ; S3: Calculate the distance between the standardized data of each fault type; obtain the distance data matrix D=(d _ij ) _n×n ; S4: Use DBSCAN to cluster fault transformer data; S5: Transformer fault diagnosis: standardize each data to be diagnosed and add it to the known fault data matrix Y, and obtain the cluster of fault types to which the data to be diagnosed belongs, which is the fault type diagnosed by the method in this paper. 2. A DBSCAN-based transformer fault diagnosis method according to claim 1, characterized in that: in step S1, the fault diagnosis characteristic quantities are: hydrogen, methane, ethane, ethylene, and acetylene. 3. a kind of transformer fault diagnosis method based on DBSCAN according to claim 1 is characterized in that: In step S1, the known fault types are: low temperature overheating, medium temperature overheating, high temperature overheating, partial discharge, low energy discharge, low energy discharge and overheating, high energy discharge and high energy discharge and overheating. 4. a kind of transformer fault diagnosis method based on DBSCAN according to claim 1 is characterized in that: The method of data standardization processing in step S2 is: normalization processing, and the specific formula is as follows:

5. according to the described a kind of transformer fault diagnosis method based on DBSCAN of claim 1 or 4, it is characterized in that: the distance between data in step S3 is Euclidean distance: the Euclidean distance formula of n-dimensional space is as follows:

6. according to claim 5 or described a kind of transformer fault diagnosis method based on DBSCAN, it is characterized in that: The oil chromatographic number is the measured content of H ₂ , CH ₄ , C ₂ H ₆ , C ₂ H ₄ , and C ₂ H ₂ , which are 5-dimensional data in total, so n is 5 here. 7. a kind of transformer fault diagnosis method based on DBSCAN according to claim 1, is characterized in that: Step S4 includes the following steps: S41. Set DBSCAN parameters, Epslion neighborhood value Eps and MinPts value MinPts; S42. Find the core points of each category: determine the first data point, and find all the points within the radius Eps of the data point; if the number of points within the radius Eps of the data point is equal to or greater than Minpts, then the data point It is the core point, and the circle with its radius Eps is its neighborhood. All points in the neighborhood belong to the same fault type. These points are boundary points or other core points; find each one in turn. Core point; if the distance between two core points is less than the radius Eps, then they are densely connected, and all points within the intersecting circle formed by the two belong to the same fault type, and these core points can be found All sample sets that the object can reach density are the same fault type; S43. Find the boundary points of each category: if the number of points within the Eps radius of a certain data (including itself) is less than Minpts, but the data point is within the Eps neighborhood of a certain core point, then the data point is the boundary point; the boundary point and the core point to which it belongs are of the same fault type; S44. Search for noise points: other points that neither belong to core data points nor boundary points are noise points; the number of points within the Eps radius of noise points is less than Minpts, and also not within the Eps neighborhood of any core point inside, does not belong to any type of failure; S45, constitute a typical data column and a map of the fault.

Images