Converter station fault matching classification method and system based on DTW and clustering method

Abstract

The invention relates to a DTW and clustering method-based converter station fault matching classification method and system, and the method comprises the steps: collecting the historical fault recording data of a converter station, and recognizing the features of the fault recording data; based on historical fault recording data, a DTW algorithm is adopted to calculate a distance matrix between fault recording data features; classifying historical fault data by adopting a clustering method based on the distance matrix between the fault recording data features to obtain a basic typical category; when new fault recording data is triggered, new fault recording data features are obtained; calculating a distance matrix between the new fault recording data and the typical category by using a DTW algorithm; and using a clustering method to match the new fault recording data with the basic typical category based on the distance matrix to obtain a fault classification result. Compared with the prior art, the method achieves the accurate matching and classification of the fault type of the converter station.

Description

Technical Field

[0001] The present invention relates to the technical field of operation and maintenance of ultra-high voltage converter stations, and in particular to a converter station fault matching classification method based on DTW and clustering method. Background Art

[0002] The development of smart grids and the integration of physical information systems in power systems have led to a sharp increase in power data. Converter stations are equipped with infrared temperature measurement, integrated monitoring and robot inspection systems. These systems are characterized by numerous devices, wide distribution and huge amounts of data. The complexity of converter stations and the intelligence level of equipment are constantly improving, which puts higher requirements on the intelligent operation and maintenance of converter stations. Traditional manual analysis methods have timeliness issues and the limitation of difficulty in accurately predicting equipment failures. It is difficult for the operation and maintenance personnel of converter stations to quickly grasp the operating status of the equipment in the station.

[0003] In the process of building smart grids, UHV converter stations, as key nodes in the power grid system, play an important role in converting AC and DC when transmitting electricity over long distances. With the increasing requirements for the reliability of equipment in UHV converter stations, the workload of operation and maintenance is also gradually increasing. In order to improve the inspection and maintenance capabilities of UHV DC converter stations, it is necessary to implement strict monitoring plans for the equipment in converter stations. Traditional equipment maintenance methods usually include regular inspections and manual data recording, or temperature measurement using infrared monitoring equipment. However, these maintenance methods are not efficient, and as the amount of data required for operation and maintenance increases, the difficulty of fully discovering faults also increases. At present, converter stations generally adopt a planned maintenance method, which is to perform annual maintenance by arranging a power outage for the entire station. However, this planned maintenance requires a long power outage time, which may have a negative impact on the economic benefits of the DC transmission system. Summary of the invention

[0004] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and provide a converter station identification method based on DTW and clustering method.

[0005] The purpose of the present invention can be achieved by the following technical solutions:

[0006] As a first aspect of the present invention, a converter station fault matching classification method based on DTW and clustering method is provided, the steps comprising:

[0007] Collect historical fault recording data of converter stations and identify the characteristics of fault recording data;

[0008] Based on the historical fault recording data, the DTW algorithm is used to calculate the distance matrix between the fault recording data features;

[0009] Based on the distance matrix between the characteristics of the fault recording data, the historical fault data is classified by clustering method to obtain the basic typical categories;

[0010] When new fault recording data is triggered, the new fault recording data features are obtained;

[0011] Use the DTW algorithm to calculate the distance matrix between the new fault recording data and the typical categories;

[0012] The clustering method is used to match the new fault recording data with the basic typical categories based on the distance matrix to obtain the fault classification results.

[0013] As a preferred technical solution, the DTW algorithm is used to calculate the distance matrix between the fault recording data features, specifically for the fault recording data feature β of the jth fault, the following dynamic programming is solved:

[0014] α(i,j)=d(β i ,β j )+min{α(i-1,j-1),α(i-1,j),α(i,j-1)}

[0015] In the formula, d(β i ,β j ) represents the Euclidean distance between two faults; α(i,j) represents the cumulative distance.

[0016] As a preferred technical solution, the DTW algorithm is used to calculate the distance matrix between the features of the fault recording data, specifically solving the following plan for the historical fault data:

[0017]

[0018] In the formula, x, y represent the faults to be clustered; C i , C j Indicates the category of clustering.

[0019] As a preferred technical solution, the DTW algorithm is used to calculate the distance matrix between the new fault recording data and the typical categories, specifically, for the fault recording data feature δ of the basic typical category i With the new fault recording data feature θ j , solve the following dynamic programming:

[0020] α(i,j)=d(δ i ,θ j )+min{α(i-1,j-1),α(i-1,j),α(i,j-1)}

[0021] In the formula, d(δ i ,θ j) represents the fault recording data characteristics of the basic typical category δ i With the new fault recording data feature θ j The Euclidean distance of , α(i,j) represents the cumulative distance.

[0022] As a preferred technical solution, the distance matrix is ​​calculated based on the DTW algorithm, specifically to solve the following planning:

[0023]

[0024] In the formula, s represents a new fault, t represents the basic typical category, and C i ,C j The category of the cluster.

[0025] As a preferred technical solution, the new fault recording data is matched based on the clustering method, as follows:

[0026] If the clustering result shows that the new fault data belongs to a basic typical category, the specific fault category will be fed back;

[0027] If the new fault data cannot be classified into an existing typical category, the new fault data is determined to be a new typical category.

[0028] As a second aspect of the present invention, a converter station fault matching classification system based on DTW and clustering method is provided, the system comprising:

[0029] Basic typical category generation module: collect historical fault recording data of converter stations and identify the characteristics of fault recording data; calculate the distance matrix between the characteristics of fault recording data based on the DTW algorithm; use clustering method to classify the existing fault data based on the distance matrix to obtain basic typical categories;

[0030] New fault data classification and matching module: When new fault recording data is generated, the new fault recording data features are obtained; the DTW algorithm is used to calculate the distance matrix between the new fault recording data and the typical categories; based on the distance matrix, the clustering method is used to match the new fault recording data to obtain the fault classification.

[0031] As a preferred technical solution, the basic typical category generation module uses the DTW algorithm to calculate the distance matrix between the fault recording data features, specifically for the fault recording data feature β of the jth fault, to solve the following dynamic programming:

[0032] α(i,j)=d(β i ,β j )+min{α(i-1,j-1),α(i-1,j),α(i,j-1)}

[0033] In the formula, d(βi ,β j ) represents the Euclidean distance between two faults; α(i,j) represents the cumulative distance;

[0034] The DTW algorithm is used to calculate the distance matrix between the features of the fault recording data. Specifically, the following plan is solved for the historical fault data:

[0035]

[0036] In the formula, x, y represent the faults to be clustered; C i , C j Indicates the category of clustering.

[0037] As a preferred technical solution, the new fault data classification matching module uses the DTW algorithm to calculate the distance matrix between the new fault recording data and the typical categories, specifically for the fault recording data feature δ of the basic typical category. i With the new fault recording data feature θ j , solve the following dynamic programming:

[0038] α(i,j)=d(δ i ,θ j )+min{α(i-1,j-1),α(i-1,j),α(i,j-1)}

[0039] In the formula, d(δ i ,θ j ) represents the fault recording data characteristics of the basic typical category δ i With the new fault recording data feature θ j The Euclidean distance of , α(i,j) represents the cumulative distance;

[0040] The distance matrix is ​​calculated based on the DTW algorithm, specifically to solve the following plan:

[0041]

[0042] In the formula, s represents a new fault, t represents the basic typical category, and C i ,C j The category of the cluster.

[0043] As a preferred technical solution, when the new fault data classification matching module matches the new fault recording data:

[0044] If the clustering result shows that the new fault data belongs to a basic typical category, the specific fault category will be fed back;

[0045] If the new fault data cannot be classified into an existing typical category, the new fault data is determined to be a new typical category.

[0046] Compared with the prior art, the present invention has the following beneficial effects:

[0047] The present invention obtains basic typical fault categories by classifying and identifying the historical fault recording data of the converter station; when new fault recording data is triggered, the new fault recording data can be matched with the basic typical fault categories based on the DTW algorithm and clustering method to obtain the fault detection result. In the subsequent fault matching and classification process, the typical categories can also be expanded, and the accurate matching and classification of the converter station fault types can be achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 This is a flow chart of a converter station fault matching classification method based on DTW and clustering method of the present invention;

[0049] Figure 2 It is a schematic diagram of clustering results in one embodiment of the present invention. DETAILED DESCRIPTION

[0050] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0051] Example 1

[0052] The purpose of the present invention is to provide a converter station fault matching classification method based on DTW and clustering method, which can solve some of the technical problems mentioned above. The specific scheme is:

[0053] Step S1, collecting historical fault recording data of the converter station and identifying the characteristics of the fault recording data;

[0054] Step S2, using the DTW algorithm to calculate the distance matrix between the features of the fault recording data based on the historical fault recording data;

[0055] Step S3, classifying the historical fault data using a clustering method based on the distance matrix between the fault recording data features to obtain basic typical categories;

[0056] Step S4: when new fault recording data is triggered, collect and obtain new fault recording data features;

[0057] Step S5, using the DTW algorithm to calculate the distance matrix between the new fault recording data and the typical categories;

[0058] Step S6: Use the clustering method to match the new fault recording data with the basic typical categories based on the distance matrix to obtain the fault classification result.

[0059] Specifically, in step S1, the converter station fault recording data is collected and the fault recording data features are identified. The steps are as follows:

[0060] The collector converter station has m faults, and the current of the i-th fault is recorded as β i =(β i1 ,β i2 ,…,β in );

[0061] Specifically, in step S2, the distance matrix is ​​calculated based on the DTW algorithm, and the steps are as follows:

[0062] Assume that the current of the jth fault is Solve the dynamic program:

[0063] α(i,j)=d(β i ,β j )+min{α(i-1,j-1),α(i-1,j),α(i,j-1)}

[0064] In the formula, d(β i ,β j ) represents the Euclidean distance between two faults; α(i,j) represents the cumulative distance.

[0065] Specifically, in step S3, the existing fault data is classified based on the clustering method as the basic typical category, and the steps are as follows:

[0066] Solution planning:

[0067]

[0068] In the formula, x, y represent the faults to be clustered; C i , C j Represents the category of clustering. Get k basic typical categories, with k≤m.

[0069] Specifically, in step S5, the DTW algorithm is used to calculate the distance matrix between the new fault recording data and the typical categories, which is specifically:

[0070] For k basic categories, the current of the ith category is recorded as

[0071] The distance matrix is ​​calculated based on the DTW algorithm, which is the current flow for the new fault. Solve the dynamic program:

[0072] α(i,j)=d(δ i ,θ j )+min{α(i-1,j-1),α(i-1,j),α(i,j-1)}

[0073] In the formula, d(δ i ,θ j ) represents the Euclidean distance between the current of the new fault and the basic typical category current; α(i,j) represents the cumulative distance.

[0074] Specifically, in step S6, the new fault recording data is matched based on the clustering method, as follows:

[0075] Use clustering method to match new fault recording data with basic typical categories based on distance matrix

[0076] Solution planning:

[0077]

[0078] In the formula, s represents a new fault, t represents the basic typical category, and C i ,C j The category of the cluster.

[0079] If the clustering result shows that the new fault data is classified into a typical category, the specific category will be fed back. If the new fault data cannot be classified into an existing typical category, the new fault data is determined to be a typical category, that is, one new typical category is added.

[0080] Example 2

[0081] As one of the specific implementation examples of the present invention, in this embodiment, historical fault recording data of a UHV converter station in a certain city in the past year is collected, and the new fault recording data generated in the recent two months is subjected to the converter station fault classification matching method based on DTW and clustering method as described above for fault diagnosis. The specific process is as follows:

[0082] Collect historical fault recording data of converter stations and identify the characteristics of fault recording data;

[0083] The distance matrix between the fault recording data features is calculated based on the DTW algorithm; the existing fault data is classified based on the distance matrix using the clustering method as the basic typical category;

[0084] When new fault recording data is generated, collect new fault recording data features;

[0085] Use the DTW algorithm to calculate the distance matrix between the new fault recording data and the typical categories;

[0086] Match the new fault recording data based on clustering method.

[0087] The clustering results are as follows Figure 2 The specific classification results are shown in Table 1:

[0088] Table 1 Fault classification table

[0089] category Typical categories Failure Date Class 1 20240712 20240712 Class 2 20240720 20240720 Class 3 20240612 20240612、20240614、20240621、20240622、20240717

[0090] Example 3

[0091] As another embodiment of the present invention, the present invention also provides a converter station fault classification matching and identification system based on DTW and clustering method, which specifically includes:

[0092] Basic typical category generation module: collect historical fault recording data of converter stations and identify the characteristics of fault recording data; calculate the distance matrix between the characteristics of fault recording data based on the DTW algorithm; use clustering method to classify the existing fault data based on the distance matrix as the basic typical category;

[0093] New fault data matching module: When new fault recording data is generated, the new fault recording data features are collected; the DTW algorithm is used to calculate the distance matrix between the new fault recording data and the typical categories; based on the distance matrix, the clustering method is used to match the new fault recording data to obtain the fault classification.

[0094] Specifically, the basic typical category generation module uses the DTW algorithm to calculate the distance matrix between the fault recording data features. Specifically, for the fault recording data feature β of the jth fault, the following dynamic programming is solved:

[0095] α(i,j)=d(β i ,β j )+min{α(i-1,j-1),α(i-1,j),α(i,j-1)}

[0096] In the formula, d(β i ,β j ) represents the Euclidean distance between two faults; α(i,j) represents the cumulative distance;

[0097] The DTW algorithm is used to calculate the distance matrix between the features of the fault recording data. Specifically, the following plan is solved for the historical fault data:

[0098]

[0099] In the formula, x, y represent the faults to be clustered; C i , C j Indicates the category of clustering.

[0100] Specifically, the new fault data classification matching module uses the DTW algorithm to calculate the distance matrix between the new fault recording data and the typical categories, specifically for the fault recording data feature δ of the basic typical category i With the new fault recording data feature θ j , solve the following dynamic programming:

[0101] α(i,j)=d(δ i ,θ j )+min{α(i-1,j-1),α(i-1,j),α(i,j-1)}

[0102] In the formula, d(δ i ,θ j ) represents the fault recording data characteristics of the basic typical category δ i With the new fault recording data feature θ j The Euclidean distance of , α(i,j) represents the cumulative distance;

[0103] The distance matrix is ​​calculated based on the DTW algorithm, specifically to solve the following plan:

[0104]

[0105] In the formula, s represents a new fault, t represents the basic typical category, and C i ,C j The category of the cluster.

[0106] When the new fault data classification matching module matches the new fault recording data:

[0107] If the clustering result shows that the new fault data belongs to a basic typical category, the specific fault category will be fed back;

[0108] If the new fault data cannot be classified into an existing typical category, the new fault data is determined to be a new typical category.

[0109] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A converter station fault matching classification method based on DTW and clustering method, characterized in that the steps include: Collect historical fault recording data of converter stations and identify the characteristics of fault recording data; Based on the historical fault recording data, the DTW algorithm is used to calculate the distance matrix between the fault recording data features; Based on the distance matrix between the characteristics of the fault recording data, the historical fault data is classified by clustering method to obtain the basic typical categories; When new fault recording data is triggered, the new fault recording data features are obtained; Use the DTW algorithm to calculate the distance matrix between the new fault recording data and the typical categories; The clustering method is used to match the new fault recording data with the basic typical categories based on the distance matrix to obtain the fault classification results.

2. The converter station fault matching classification method based on DTW and clustering method according to claim 1 is characterized in that: The DTW algorithm is used to calculate the distance matrix between the fault recording data features, specifically for the fault recording data feature β of the jth fault, the following dynamic programming is solved: α(i,j)=d(β i ,b j )+min{α(i-1,j-1),α(i-1,j),α(i,j-1)} In the formula, d(β i ,β j ) represents the Euclidean distance between two faults; α(i,j) represents the cumulative distance.

3. The converter station fault matching classification method based on DTW and clustering method according to claim 2 is characterized in that: The DTW algorithm is used to calculate the distance matrix between the features of the fault recording data, specifically to solve the following plan for the historical fault data: In the formula, x, y represent the faults to be clustered; C i , C j Indicates the category of clustering.

4. The converter station fault matching classification method based on DTW and clustering method according to claim 1 is characterized in that: The DTW algorithm is used to calculate the distance matrix between the new fault recording data and the typical categories, specifically, for the fault recording data feature δ of the basic typical category i With the new fault recording data feature θ j , solve the following dynamic programming: α(i,j)=d(δ i ,i j )+min{α(i-1,j-1),α(i-1,j),α(i,j-1)} In the formula, d(δ i ,θ j ) represents the fault recording data characteristics of the basic typical category δ i With the new fault recording data feature θ j The Euclidean distance of , α(i,j) represents the cumulative distance.

5. The converter station fault matching classification method based on DTW and clustering method according to claim 4 is characterized in that: The distance matrix is ​​calculated based on the DTW algorithm, specifically to solve the following plan: In the formula, s represents a new fault, t represents the basic typical category, and C i ,C j The category of the cluster.

6. The converter station fault matching classification method based on DTW and clustering method according to claim 5 is characterized in that: The new fault recording data is matched based on the clustering method, as follows: If the clustering result shows that the new fault data belongs to a basic typical category, the specific fault category will be fed back; If the new fault data cannot be classified into an existing typical category, the new fault data is determined to be a new typical category.

7. A converter station fault matching classification system based on DTW and clustering method, characterized in that: The system comprises: Basic typical category generation module: collect historical fault recording data of converter stations and identify the characteristics of fault recording data; calculate the distance matrix between the characteristics of fault recording data based on the DTW algorithm; use clustering method to classify the existing fault data based on the distance matrix to obtain basic typical categories; New fault data classification and matching module: When new fault recording data is generated, the new fault recording data features are obtained; the DTW algorithm is used to calculate the distance matrix between the new fault recording data and the typical categories; based on the distance matrix, the clustering method is used to match the new fault recording data to obtain the fault classification.

8. The converter station fault matching classification system based on DTW and clustering method according to claim 7, characterized in that: The basic typical category generation module uses the DTW algorithm to calculate the distance matrix between the fault recording data features. Specifically, for the fault recording data feature β of the jth fault, the following dynamic programming is solved: α(i,j)=d(β i ,b j )+min{α(i-1,j-1),α(i-1,j),α(i,j-1)} In the formula, d(β i ,β j ) represents the Euclidean distance between two faults; α(i,j) represents the cumulative distance; The DTW algorithm is used to calculate the distance matrix between the features of the fault recording data. Specifically, the following plan is solved for the historical fault data: In the formula, x, y represent the faults to be clustered; C i , C j Indicates the category of clustering.

9. The converter station fault matching classification system based on DTW and clustering method according to claim 7, characterized in that: The new fault data classification matching module uses the DTW algorithm to calculate the distance matrix between the new fault recording data and the typical categories, specifically for the fault recording data feature δ of the basic typical category i With the new fault recording data feature θ j , solve the following dynamic programming: α(i,j)=d(δ i ,i j )+min{α(i-1,j-1),α(i-1,j),α(i,j-1)} In the formula, d(δ i ,θ j ) represents the fault recording data characteristics of the basic typical category δ i With the new fault recording data feature θ j The Euclidean distance of , α(i,j) represents the cumulative distance; The distance matrix is ​​calculated based on the DTW algorithm, specifically to solve the following plan: In the formula, s represents a new fault, t represents the basic typical category, and C i ,C j The category of the cluster.

10. The converter station fault matching classification system based on DTW and clustering method according to claim 7, characterized in that: When the new fault data classification matching module matches the new fault recording data: If the clustering result shows that the new fault data belongs to a basic typical category, the specific fault category will be fed back; If the new fault data cannot be classified into an existing typical category, the new fault data is determined to be a new typical category.

Images