A Multi-classification Strategy of Support Vector Machine for Transformer Fault Diagnosis

Abstract

The present invention discloses a multi-classification strategy for a support vector machine used in transformer fault diagnosis, belonging to the technical field of artificial intelligence. A multi-classification strategy for a support vector machine used in transformer fault diagnosis, the steps of obtaining the multi-classification strategy include: calculating the between-class scatter matrix and the within-class scatter matrix after merging two classes for any two classes of samples in the training set, and adding the sum of the traces of the between-class scatter matrix and the within-class scatter matrix as the separability index; finding a pair of combinations with the smallest separability index, merging the pair of combinations into one class, and recording the synthesis order of the classes; generating a classification decision tree in reverse according to the recorded class merging order before, and using it as the multi-classification strategy of the support vector machine. The beneficial effects of the present invention: It can improve the training efficiency and classification accuracy of the transformer fault diagnosis model based on the support vector machine, enabling it to be practically applied to the power system diagnosis site.

Description

Technical Field

[0001] The present invention belongs to the technical field of artificial intelligence, and particularly relates to a multi-classification strategy for a support vector machine used for transformer fault diagnosis. Background Art

[0002] Power transformers are key large-scale equipment in the power transmission and distribution process, directly affecting the safety and stability of the power system. Therefore, it is very necessary to monitor the operating status of transformers and adopt fault diagnosis strategies to timely detect potential insulation problems and repair them in time to avoid serious accidents. At present, the most widely used power transformer fault diagnosis tool is the three-ratio method based on the analysis of dissolved gases in oil proposed by the International Electrotechnical Commission. However, due to the relatively rigid diagnostic criteria of this method, misdiagnosis is likely to occur when diagnosing fault samples near the classification threshold, resulting in serious consequences for the power system.

[0003] With the increasing maturity of artificial intelligence technology in recent years, scholars have tried to apply these new technologies to establish complex non-linear mapping relationships between dissolved gas data in oil and transformer faults. Among these new technologies, support vector machines have attracted the attention of scholars due to their excellent learning ability and perfect theoretical basis. There are already many cases proving that transformer fault diagnosis based on support vector machines can improve the diagnostic accuracy to a certain extent.

[0004] However, at the present stage, there is no targeted design for the multi-classification strategy of support vector machines, and traditional multi-classification methods are not applicable to transformer fault diagnosis problems. Summary of the Invention

[0005] The purpose of the present invention is to provide a multi-classification strategy for a support vector machine used for transformer fault diagnosis.

[0006] The purpose of the present invention can be achieved through the following technical solutions:

[0007] A multi-classification strategy for a support vector machine used for transformer fault diagnosis, and the steps to obtain the multi-classification strategy include:

[0008] Step 1: Calculate the between-class scatter matrix for any two classes in the training set and the within-class scatter matrix after merging the two classes, and add the sum of the traces of the between-class scatter matrix and the within-class scatter matrix as the separability index;

[0009] Step 2: Specify any two classes as a pair of combinations. After the calculation in Step 1, find the pair of combinations with the smallest separability index, merge the pair of combinations into one class, and record the synthesis order of the class; repeat the operation in Step 1 until all classes are merged into one class, and terminate the merging operation.

[0010] Step 3: According to the merging order of the classes recorded previously, generate a classification decision tree in reverse, and use it as the multi-classification strategy of the support vector machine.

[0011] Further, in Step 1, when calculating the between-class scatter matrix and the within-class scatter matrix after merging any two classes in the training set, assume that there are a total of C class samples in the training set, and the number of samples contained in each class is N i . Let x i k be the k-th dimensional sample vector of class ω i , and there are a total of training samples, then the within-class scatter matrix S W , and the between-class scatter matrix S B are respectively defined as:

[0012]

[0013]

[0014] where μ i and μ 0 are the mean vectors of class ω i and the whole data respectively.

[0015]

[0016]

[0017] Further, the tree structure of the classification decision tree in Step 3 is of the binary tree type.

[0018] Advantages of the present invention:

[0019] The present invention can improve the training efficiency and classification accuracy of the transformer fault diagnosis model based on the support vector machine, enabling it to be practically applied to the power system diagnosis site. Description of the Drawings

[0020] The present invention will be further described below with reference to the accompanying drawings.

[0021] Figure 1 is the implementation flowchart of the multi-classification strategy of the support vector machine for transformer fault diagnosis. Specific Embodiments

[0022] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0023] As Figure 1 shown, a multi-classification strategy for a support vector machine for transformer fault diagnosis is characterized in that the steps of obtaining the multi-classification strategy include:

[0024] Step 1: Calculate the between-class scatter matrix and the within-class scatter matrix after combining any two classes in the training set. Add the sum of the traces of the between-class scatter matrix and the within-class scatter matrix as the separability index. The larger the trace of the between-class scatter matrix, the more conducive to classifying the two classes, while the smaller the trace of the within-class scatter matrix indicates the better integrity after combining the two classes. Therefore, the smaller the separability index, the higher the similarity of the features of these two classes of samples and the more difficult they are to separate.

[0025] Step 2: Specify any two classes as a pair of combinations. After the calculation in Step 1, find the pair of combinations with the smallest separability index, merge the pair of combinations into one class, and record the synthesis order of the class; repeat the operation in Step 1 until all classes are merged into one class and terminate the merging operation. Among them, the two classes merged earlier will be located in the lower layer in the classification decision tree.

[0026] Step 3: Generate a classification decision tree in reverse according to the previously recorded class merging order and use it as the multi-classification strategy of the support vector machine. Since the support vector machine to be trained is the least, and except for the top support vector machine that needs to use all the samples during training, the lower support vector machines only need to use a subset of the total samples during training. Therefore, the training efficiency is higher than the traditional multi-classification strategy, and the decision tree structure designed based on the sample distribution can ensure a higher classification accuracy.

[0027] Further, in Step 1, when calculating the between-class scatter matrix and the within-class scatter matrix after combining any two classes in the training set, assume that there are a total of C class samples in the training set, and the number of samples contained in each class is N i . Let x i k be the k-th dimensional sample vector of class ω i , and there are a total of training samples. Then the within-class scatter matrix S W , and the between-class scatter matrix S B are respectively defined as:

[0028]

[0029]

[0030] where μ i and μ 0 are the mean vectors of class ω i and the entire data, respectively.

[0031]

[0032]

[0033] Furthermore, the tree structure of the classification decision tree in step three is of the binary tree type. Binary trees can utilize the advantages of linked lists and ordered arrays, and at the same time have the advantage of being able to merge ordered arrays and linked lists.

[0034] In the description of this specification, the descriptions referring to terms such as "one embodiment", "example", "specific example", etc. mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.

[0035] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art of this industry should understand that the present invention is not limited by the above embodiments. The above embodiments and the descriptions in the specification only illustrate the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed.

Claims

1. A multi-classification strategy of support vector machine for transformer fault diagnosis, characterized in that, the steps of obtaining the multi-classification strategy include: Step 1: Calculate the between-class scatter matrix for any two classes of samples in the training set and the within-class scatter matrix after merging the two classes, and add the sum of the traces of the between-class scatter matrix and the within-class scatter matrix as the separability index; Step 2: Specify any two classes as a pair of combinations. After the calculation in Step 1, find the pair of combinations with the smallest separability index, merge the pair of combinations into one class, and record the synthesis order of the class; repeat the operation in Step 1 until all classes are merged into one class, and terminate the merging operation; Step 3: According to the recorded merging order of the classes before, reversely generate a classification decision tree, and use it as the multi-classification strategy of the support vector machine; In the first step, when calculating the between-class scatter matrix for any two classes in the training set and the within-class scatter matrix after merging the two classes, assume that there are a total of C class samples in the training set, and the number of samples contained in each class is ; Assume is the k-th dimensional sample vector of class , and there are a total of W training samples. Then the within-class scatter matrix S B are respectively defined as: wherein and are respectively the mean vectors of the classes and the entire data; 。 2. The multi-classification strategy of support vector machine for transformer fault diagnosis according to claim 1, characterized in that, the tree structure of the classification decision tree in Step 3 is a binary tree type.

Images