Power grid line loss grade classification method, system and equipment

Abstract

The invention discloses a power grid line loss grade classification method, system and device, and relates to the technical field of line loss detection, and the method comprises the steps: obtaining the mixed data of the line loss detection of a power grid; analyzing the mixed data through a fuzzy monotonicity discrimination criterion, and dividing all attributes A of the mixed data into fuzzy monotonic rule attributes B and non-monotonic conventional attributes C; using fuzzy mutual information FMI and fuzzy ordered mutual information FRMI to measure the importance degree of the conventional attribute and the rule, obtaining the optimal segmentation point of the rule attribute B and the optimal segmentation point of the non-monotonic conventional attribute C through the size relationship between the importance degree of the conventional attribute and a threshold value, and carrying out bifurcation according to the optimal segmentation point and constructing a binary decision tree; and processing the mixed data through a binary decision tree to obtain a power grid line loss grade classification result. According to the method, excellent classification precision and high interpretability are shown when the stage classification task of the power grid line loss is processed through the process.

Description

Technical Field

[0001] The present invention relates to the technical field of line loss detection, and in particular to a method, system and device for classifying power grid line loss levels. Background Art

[0002] In recent years, with the continuous evolution of power systems, traditional line loss detection methods have gradually become inadequate to meet the demands of modern power networks. Against this backdrop, the introduction of big data technology has become an inevitable trend in the development of the power industry, bringing profound changes to the industry. Line loss classification, as a core component of line loss detection, has garnered widespread attention within the industry.

[0003] Line loss classification plays a crucial role in power systems. It not only quickly reveals the extent of line losses but also provides a key basis for power companies to manage and assess line losses. Line loss classification is typically categorized into five levels: severe line loss, significant line loss, moderate line loss, minor line loss, and no line loss. In machine learning, this classification task falls under the umbrella of fuzzy monotonic ordered classification, which requires explicit preferences and fuzzy monotonicity constraints between feature values ​​and decision values.

[0004] However, in the actual process of power grid line loss classification, the acquired datasets often contain both fuzzy monotonic and non-monotonic attributes, both of which are crucial for accurately determining line loss levels. Unfortunately, current mainstream classification algorithms often ignore this fuzzy monotonic ordering prior knowledge or focus solely on processing fuzzy monotonic features, which to some extent limits the accuracy and interpretability of line loss classification. Summary of the Invention

[0005] The purpose of the present invention is to address the deficiencies of the above-mentioned prior art and provide a method, system and device for classifying power grid line loss levels, so as to solve the problem in the prior art that the current mainstream classification algorithms often ignore this fuzzy monotonous and orderly prior knowledge, or only focus on processing fuzzy monotonous features, which to a certain extent restricts the accuracy of line loss level classification.

[0006] The present invention specifically provides the following technical solution: a method for classifying power grid line loss levels, comprising the following steps:

[0007] Acquire mixed data during power grid line damage detection; the mixed data includes power loss of power grid lines;

[0008] Analyzing the mixed data by using the fuzzy monotonicity discrimination criterion, dividing all attributes A of the mixed data into fuzzy monotonic regular attributes B and non-monotonic regular attributes C;

[0009] Fuzzy mutual information FMI and fuzzy ordered mutual information FRMI are used to measure the importance of regular attributes and rule attributes in mixed data respectively;

[0010] According to the relationship between the importance of the regular attribute and the threshold value, the optimal split point of the rule attribute B and the optimal split point of the non-monotonic regular attribute C are obtained, and bifurcations are performed according to the two optimal split points, and a binary decision tree is constructed through all the bifurcations;

[0011] Fuzzy monotonic consistency judgment is performed on the mixed data of the power grid line loss detection to be classified by the binary decision tree, and the power grid line loss grade classification result is obtained according to the fuzzy monotonic consistency judgment result.

[0012] Preferably, the method of analyzing the mixed data by using the fuzzy monotonicity criterion to divide all attributes A of the mixed data into fuzzy monotonic regular attributes B and non-monotonic regular attributes C comprises the following steps:

[0013] The possibility degree is used as the fuzzy membership degree, and all attributes A of the mixed data are judged by the order of the attribute values ​​according to the possibility degree;

[0014] Based on the sorting result, a fuzzy sorting inconsistency rate is obtained, and the fuzzy monotonicity of the mixed data attributes is evaluated according to the fuzzy sorting inconsistency rate. The mixed data attributes are divided into fuzzy monotonic regular attributes B and non-monotonic regular attributes C.

[0015] Preferably, the possibility degree is used as the fuzzy membership degree, and all attributes A of the mixed data are judged by the possibility degree according to the order of attribute values, specifically:

[0016] Get fuzzy ordered set The specific expression is:

[0017]

[0018] Obtain the possibility of each condition attribute through the fuzzy ordered set Among them, the possibility is x i Better than x j The possibility of v(x i ,a) and v(x j ,a) are two samples x i and x j The attribute value on the conditional attribute a, g is a constant, i,j∈[1,n], n is a constant.

[0019] Preferably, the fuzzy monotonicity of all attributes of the mixed data is evaluated according to the fuzzy sorting inconsistency rate, and the attributes of the mixed data are divided into fuzzy monotonic regular attributes B and non-monotonic regular attributes C, including the following steps:

[0020] Obtain the forward fuzzy sorting inconsistent value through the possibility The specific expression is:

[0021]

[0022] Among them, a∈A is the conditional attribute, D is the decision attribute, U is the sample set of all data, N is the number of samples, For the attribute value on A to be greater than x i The set of all samples taking values ​​on A, is greater than x on the decision attribute D i The set of all examples of the decision value of ;

[0023] Obtain the inconsistent values ​​of the backward fuzzy sorting by the possibility The specific expression is:

[0024]

[0025] in, The value of the attribute on A is less than x i The set of all samples taking values ​​on A, is less than x on the decision attribute D i The set of all examples of the decision value of ;

[0026] Obtaining a fuzzy sorting inconsistency difference value by using the forward fuzzy sorting inconsistency value and the backward fuzzy sorting inconsistency value The specific expression is:

[0027]

[0028] Inconsistent differences are sorted by the fuzzy The size of the data is used to judge the fuzzy monotonicity and non-monotonicity of the attributes, that is, it is divided into fuzzy monotonic regular attributes B and non-monotonic regular attributes C.

[0029] Preferably, the fuzzy sorting inconsistent difference The size of determines the fuzzy monotonicity and non-monotonicity of data attributes, including the following steps:

[0030] like If the data is fuzzy monotonically consistent on the conditional attribute a, it is a fuzzy monotonic regular attribute, otherwise it is a non-monotonic regular attribute; where δ is the fuzzy monotonicity threshold;

[0031] like When , the data is fuzzy monotonically decreasing on the conditional attribute a; When , the data is fuzzy monotonically increasing on the conditional attribute a, and the fuzzy monotonic rule attribute is obtained.

[0032] Preferably, the step of obtaining the optimal splitting point of the rule attribute B and the optimal splitting point of the non-monotonic regular attribute C through the relationship between the importance of the regular attribute and the threshold value, bifurcating according to the two optimal splitting points, and constructing a binary decision tree through all bifurcations includes the following steps:

[0033] Set the general attribute a i The values ​​are binarized to obtain the maximized fuzzy mutual information FMI value, and the maximized fuzzy mutual information FMI value is used as the importance of the conventional attribute;

[0034] If the importance of the general attribute is greater than the threshold t, the value of the general attribute is selected as the split point, and the optimal split point c is used as the split point. * Perform a fork;

[0035] If the importance of the regular attribute is less than the threshold t, the maximum value of the maximized fuzzy ordered mutual information FRM I is selected for rule attribute selection and used as the optimal split point b* for bifurcation;

[0036] According to the optimal segmentation point c * Or b* forks, and iterates the process of obtaining the best split point, using all the forks to build a binary decision tree.

[0037] Preferably, the optimal segmentation point c * Fork, specifically:

[0038] Get the best split point c * ; The specific expression is:

[0039]

[0040] The non-monotonic regular attribute C is selected as the extended attribute, and the specific expression is:

[0041]

[0042] Among them, i,j∈[1,n], n is a constant, D is the decision attribute, and a is the conditional attribute.

[0043] Preferably, the method of selecting the maximum value of the maximized fuzzy ordered mutual information FRMI for rule attribute selection and bifurcation of the corresponding optimal segmentation point b* comprises the following steps:

[0044] Select the optimal cut point b corresponding to the rule attribute j * , the specific expression is:

[0045]

[0046] Select rule attribute B as the extended attribute. The specific expression is:

[0047]

[0048] Among them, b j is the split point.

[0049] The present invention provides a power grid line loss classification system, comprising:

[0050] A data acquisition module is used to acquire mixed data during online power grid damage detection; the mixed data includes power loss of power grid lines;

[0051] A judgment module is used to analyze the mixed data by using a fuzzy monotonicity judgment criterion, and divide all attributes A of the mixed data into fuzzy monotonic regular attributes B and non-monotonic regular attributes C;

[0052] Importance acquisition module, used to measure the importance of regular attributes and rule attributes in mixed data using fuzzy mutual information FMI and fuzzy ordered mutual information FRMI respectively;

[0053] A decision tree construction module is used to obtain the optimal split point of the rule attribute B and the optimal split point of the non-monotonic regular attribute C based on the relationship between the importance of the regular attribute and the threshold value, and to bifurcate according to the two optimal split points, and to construct a binary decision tree through all the bifurcations;

[0054] The classification module is used to perform fuzzy monotonic consistency judgment on the mixed data of the power grid line loss detection to be classified through the binary decision tree, and obtain the power grid line loss grade classification result through the fuzzy monotonic consistency judgment result.

[0055] The present invention provides a computer device, comprising a memory and a processor, wherein a program is stored in the memory, and when the program is executed by the processor, the processor executes the steps of a power grid line loss level classification method.

[0056] Compared with the prior art, the present invention has the following significant advantages:

[0057] This paper, based on a fuzzy monotonicity discrimination criterion algorithm, quantitatively analyzes the fuzzy monotonic and non-monotonic features in line loss big data. Based on this evaluation framework, a hybrid decision tree model is designed that can simultaneously process fuzzy monotonic and non-monotonic attributes, thereby achieving efficient and accurate classification of power grid line loss levels. When handling the task of classifying substation power grid line loss levels, this paper not only demonstrates excellent classification accuracy but also possesses strong interpretability, providing strong technical support for line loss management in the power industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 This is an overall flow chart of a method for classifying power grid line loss levels according to the present invention;

[0059] Figure 2 This is a flow chart of constructing a binary decision tree through bifurcation of the optimal split point in the present invention. DETAILED DESCRIPTION

[0060] The following is a clear and complete description of the technical solutions of the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0061] Fuzzy monotone classification, a core research area in machine learning, has attracted the attention of numerous researchers. Constructing fuzzy monotone decision trees is one of the most effective approaches to solving fuzzy monotone classification tasks. Ben-David et al. innovatively introduced the concepts of ordered decision trees and fuzzy monotone decision trees, providing new approaches to addressing this problem. Later, in 2010, Hu et al. proposed the concepts of rank entropy and rank mutual information and, based on these concepts, designed a novel algorithm—the rank entropy-based fuzzy monotone decision tree (REMDT). Due to its excellent generalization capabilities, REMDT quickly became a research hotspot in fuzzy monotone classification and has been further improved and extended by researchers. Pei et al. designed a partially fuzzy monotone tree for the real-valued partially fuzzy monotone classification problem. These decision-making methods primarily focus on real-valued data. However, their application in real-world scenarios is limited because multi-attribute decision-making problems often involve multiple types of fuzzy data rather than single values.

[0062] For example, when determining the level of line loss in power systems, line loss is generally categorized into five levels: severe line loss, large line loss, moderate line loss, minor line loss, and no line loss. Severe line loss occurs due to overload, tripping, or force majeure, resulting in significant energy dissipation and losses, requiring urgent remediation measures. Major line loss refers to significant energy loss, with noticeable losses, but not reaching the level of a trip or force majeure. Moderate line loss refers to minimal energy loss, requiring immediate remediation measures. Minor line loss refers to minimal energy loss, but requires regular maintenance. No line loss refers to low-voltage power station measurements and commissioning, where no significant losses are detected. Big data plays a crucial role in classifying line loss levels. First, big data analysis can deeply analyze the factors influencing line loss, fully exploring previously unknown influencing patterns, enabling better control. Furthermore, big data analysis can help companies determine the appropriate measures for different levels of line loss, enabling more accurate control. This paper presents a method for identifying power grid line loss levels. In the actual application scenarios of power grid line loss, fuzzy monotonic features and non-monotonic features often coexist, and non-monotonic features also play an important role in improving classifier performance, which cannot be ignored.

[0063] To effectively diagnose and evaluate the coexistence of fuzzy monotonic and non-monotonic attributes in complex uncertain data classification tasks, an innovative adaptive hybrid model was designed. This model aims to construct a hybrid fuzzy monotonic decision tree that can flexibly handle data containing both fuzzy monotonic and non-monotonic attributes. In the implementation process, the likelihood degree is first used as the fuzzy membership degree to compare the size of different types of data. Next, a new metric called the fuzzy rank inconsistency rate is introduced to accurately distinguish fuzzy monotonic and non-monotonic attributes in the data. When selecting the extended attributes and split points, the strategy of maximizing the fuzzy mutual information (FMI) and the fuzzy ranking mutual information (FRMI) is adopted. The proposed method provides new ideas and methods for handling uncertain complex data classification tasks.

[0064] Firstly, a data sorting discriminant model is constructed, which is used to distinguish the fuzzy monotonicity of line loss characteristics.

[0065] Definition: Given a mixed data attribute information system, U is the sample set of all data.

[0066] like Figure 1 As shown, the present invention proposes a method for classifying power grid line loss levels, which specifically includes the following steps:

[0067] Step S1: Acquire mixed data during power grid line damage detection; the mixed data includes power loss of power grid lines.

[0068] Step S2: Analyze the mixed data using the fuzzy monotonicity criterion, and divide all attributes A of the mixed data into fuzzy monotonic regular attributes B and non-monotonic regular attributes C.

[0069] Specifically:

[0070] The mixed data is analyzed by using the fuzzy monotonicity discrimination criterion, and all attributes A of the mixed data are divided into fuzzy monotonic regular attributes B and non-monotonic regular attributes C, including the following steps:

[0071] The possibility degree is used as the fuzzy membership degree, and all attributes A in the mixed data are judged by the possibility degree according to the order of attribute values.

[0072] Based on the sorting results, the fuzzy sorting inconsistency rate is obtained, and the fuzzy monotonicity of the mixed data attributes is evaluated according to the fuzzy sorting inconsistency rate. The mixed data attributes are divided into fuzzy monotonic regular attributes B and non-monotonic regular attributes C.

[0073] Step S21: Using the possibility degree as the fuzzy membership degree, all attributes A of the mixed data are judged by the possibility degree according to the order of the attribute values, specifically:

[0074] Get fuzzy ordered set The specific expression is:

[0075]

[0076] Obtain the possibility of each condition attribute through fuzzy ordered sets Among them, the possibility For example x i Better than example x j The possibility of v(x i ,a) and v(x j ,a) are two samples x i and x j The attribute value on the regular attribute a, g is a constant, i,j∈[1,n], n is a constant.

[0077] Step S22: Evaluate the fuzzy monotonicity of all attributes of the mixed data based on the fuzzy sorting inconsistency rate, and divide the mixed data attributes into fuzzy monotonic regular attributes B and non-monotonic regular attributes C. Specifically:

[0078] Get the inconsistent values ​​of forward fuzzy sorting by possibility The specific expression is:

[0079]

[0080] in, is the conditional attribute, D is the decision attribute, U is the sample set of all data, N is the number of samples, For the attribute value on A to be greater than x i The set of all samples taking values ​​on A, is greater than x on the decision attribute D i The set of all examples of the decision value of .

[0081] Get inconsistent values ​​of backward fuzzy sorting by possibility The specific expression is:

[0082]

[0083] in, The value of the attribute on A is less than x i The set of all samples taking values ​​on A, is less than x on the decision attribute D i The set of all examples of the decision value of .

[0084] Obtain the fuzzy sorting inconsistency difference value through the forward fuzzy sorting inconsistency value and the backward fuzzy sorting inconsistency value. The specific expression is:

[0085]

[0086] Inconsistent differences are sorted by the fuzzy The size of is used to judge the fuzzy monotonicity and non-monotonicity of data attributes, and divide them into fuzzy monotonic rule attributes B and non-monotonic regular attributes C; the steps are as follows:

[0087] like If the data is fuzzy monotonically consistent on the conditional attribute a, it is a fuzzy monotonic regular attribute, otherwise it is a non-monotonic regular attribute. Where δ is the fuzzy monotonicity threshold.

[0088] like When , the data is fuzzy monotonically decreasing on the conditional attribute a; When , the data is fuzzy monotonically increasing on the conditional attribute a, and the fuzzy monotonic rule attribute is obtained.

[0089] Step S3: Use fuzzy mutual information FMI and fuzzy ordered mutual information FRMI to measure the importance of regular attributes and rule attributes in mixed data respectively.

[0090] Step S4: Obtain the optimal split point of the rule attribute B and the optimal split point of the non-monotonic regular attribute C through the relationship between the importance of the regular attribute and the threshold value, and bifurcate according to the two optimal split points, and construct a binary decision tree through all the bifurcations. Figure 2 shown.

[0091] Specifically:

[0092] Step S41: Set attribute a i The values ​​are binarized to obtain the maximized fuzzy mutual information FMI value, and the maximized fuzzy mutual information FMI value is used as the importance of the conventional attribute.

[0093] Step S42: If the importance of the general attribute is greater than the threshold t, the general attribute is selected as the segmentation attribute and the optimal segmentation point c is used. * Perform a fork.

[0094] Step S43: If the importance of the regular attribute is less than the threshold t, the maximum value of the maximized fuzzy ordered mutual information FRMI is selected for rule attribute selection, and the corresponding optimal segmentation point b* is bifurcated.

[0095] Step S44: According to the optimal segmentation point c * Or b* forks, and iterates the process of obtaining the best split point, using all the forks to build a binary decision tree.

[0096] Among them, the general attribute a i Binarization is performed to obtain the maximum mutual information FMI, including the following steps:

[0097] Judge v(x i ,a) and c j The size of v(x i ,a)<c j , then let v(x i ,a)=1; otherwise v(x i ,a)=2.

[0098] Through a i The value of (x) is obtained to maximize the fuzzy mutual information FMI.

[0099] Get the best split point c * ; The specific expression is:

[0100]

[0101] The non-monotonic regular attribute C is selected as the extended attribute, and the specific expression is:

[0102]

[0103] Among them, i,j∈[1,n], n is a constant, D is the decision attribute, and a is the conditional attribute.

[0104] The maximum value of the maximized fuzzy ordered mutual information FRMI is selected for rule attribute selection and used as the optimal split point b* for bifurcation, including the following steps:

[0105] Select the optimal cut point b corresponding to the rule attribute j * , the specific expression is:

[0106]

[0107] Select rule attribute B as the extended attribute. The specific expression is:

[0108]

[0109] Among them, b j is the split point.

[0110] According to the best split point b j * Fork and get the best split point b j * The process is iterated, all bifurcations are used to build an ordered decision tree, and the fuzzy sorting inconsistency difference is output through the ordered decision tree.

[0111] Step S5: Perform fuzzy monotonic consistency judgment on the mixed data of the power grid line loss detection to be classified through a binary decision tree, and obtain the power grid line loss grade classification result through the fuzzy monotonic consistency judgment result.

[0112] Specifically:

[0113] for It is considered that there is a global fuzzy monotonic relationship between attributes a and D, otherwise it is considered that there is a global non-monotonic relationship between attributes a and D.

[0114] based on Fuzzy monotonicity judgment of:

[0115] Input: S = (x i ,y i ): object set; δ: fuzzy monotonicity threshold.

[0116] Output: (Difference in inconsistency rates between upward and downward sorting).

[0117] 1: Read in the object set (x i ,y i ).

[0118] 2: for each attribute of the object do:

[0119] 3: Application calculate

[0120] 4: Application calculate

[0121] 5: Application

[0122] 6: end for:

[0123] 7: if then:

[0124] 8: Returns the fuzzy monotonic rule attributes.

[0125] 9: else:

[0126] 10: Returns non-monotonic regular properties.

[0127] 11: end if.

[0128] According to the above criteria, the fuzzy monotonic consistency of mixed data can be determined. In addition, with the help of experiments, the present invention has drawn the following conclusions: (1) When When , the samples in the data set show a strict fuzzy monotonically increasing or decreasing trend between the condition attribute a and the decision attribute D; (2) As the data is no longer fuzzy monotonically consistent, and The absolute value of will become smaller. The absolute value of will become larger. (3) As the fuzzy monotonically increasing degree of a increases, It approaches 0 from the left. Similarly, as the degree of a blur decreases monotonically, Approaching 0 from the right.

[0129] Through the above analysis, it is necessary to adjust the threshold parameter to determine whether there is a global fuzzy monotonic relationship between a and D. According to the previous training, the rules for judging fuzzy monotonicity are as follows: It is believed that there is a global fuzzy monotonic relationship between a and D, otherwise it is believed that there is a global non-monotonic relationship between a and D. The above fuzzy sorting inconsistency judgment criterion is used to judge the fuzzy monotonic consistency of the regional power grid line loss data. The judgment process is based on The fuzzy monotonicity judgment of .

[0130] Based on the above method, the present invention provides a power grid line loss grade classification system, comprising: a data acquisition module, a judgment module, an importance acquisition module, a decision tree construction module and a classification module.

[0131] Among them, the data acquisition module is used to obtain mixed data during power grid line loss detection; the mixed data includes the power loss of the power grid line; the judgment module is used to analyze the mixed data through the fuzzy monotonicity discrimination criterion, and divide all attributes A of the mixed data into fuzzy monotonic rule attributes B and non-monotonic conventional attributes C; the importance acquisition module is used to use fuzzy mutual information FMI and fuzzy ordered mutual information FRMI to measure the importance of conventional attributes and rule attributes in the mixed data respectively; the decision tree construction module is used to obtain the optimal segmentation point of the rule attribute B and the optimal segmentation point of the non-monotonic conventional attribute C through the relationship between the importance of the conventional attribute and the threshold size, and to bifurcate according to the two optimal segmentation points, and to construct a binary decision tree through all bifurcations; the classification module is used to perform fuzzy monotonic consistency judgment on the mixed data of the power grid line loss detection to be classified through the binary decision tree, and obtain the power grid line loss level classification result through the fuzzy monotonic consistency judgment result.

[0132] The present invention also provides a computer device, including a memory and a processor. The memory stores a program, and when the program is executed by the processor, the processor executes the steps of a power grid line loss level classification method.

[0133] According to the disclosed embodiments, a computing device may communicate with one or more external devices (e.g., a keyboard, a pointing device, Bluetooth communications, etc.), or with any device that enables a computing device to communicate with one or more other computing devices (e.g., a router, a modem, etc.).

[0134] The present invention also provides a storage medium on which a computer program is stored. When the computer program is executed by a processor, a method for predicting the remaining time of an aircraft is implemented.

[0135] According to the disclosed embodiments, the storage medium may be a non-volatile computer-readable storage medium, such as, but not limited to, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, the storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0136] The various embodiments of the present invention are described in a progressive manner. Each embodiment focuses on the differences from other embodiments, and the same or similar parts between the various embodiments can be referred to in detail.

[0137] The present invention uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only intended to help understand the method and core concept of the present invention. At the same time, those skilled in the art will find that the specific implementation methods and application scopes may vary based on the concept of the present invention. In summary, the contents of this specification should not be construed as limiting the present invention.

Claims

1. A method for classifying power grid line loss levels, characterized in that: The steps include: Acquire mixed data during power grid line damage detection; the mixed data includes power loss of power grid lines; Analyze the mixed data by using the fuzzy monotonicity discrimination criterion, and divide all attributes A of the mixed data into fuzzy monotonic regular attributes B and non-monotonic regular attributes C; Fuzzy mutual information FMI and fuzzy ordered mutual information FRMI are used to measure the importance of regular attributes and regular attributes in mixed data respectively; Through the relationship between the importance of the conventional attribute and the threshold value, the optimal segmentation point of the rule attribute B and the optimal segmentation point of the non-monotonic conventional attribute C are obtained, and bifurcation is performed according to the two optimal segmentation points, and a binary decision tree is constructed through all the bifurcations; The binary decision tree is used to perform fuzzy monotonic consistency judgment on the mixed data of the power grid line loss detection to be classified, and the power grid line loss grade classification result is obtained through the fuzzy monotonic consistency judgment result.

2. A method for classifying power grid line loss levels according to claim 1, characterized in that: The mixed data is analyzed by using a fuzzy monotonicity discrimination criterion to divide all attributes A of the mixed data into fuzzy monotonic regular attributes B and non-monotonic regular attributes C, including the following steps: The possibility degree is used as the fuzzy membership degree, and all attributes A of the mixed data are judged according to the order of attribute values ​​by the possibility degree; Based on the sorting result, a fuzzy sorting inconsistency rate is obtained, and the fuzzy monotonicity of the mixed data attribute is evaluated according to the fuzzy sorting inconsistency rate, and the mixed data attribute is divided into a fuzzy monotonic regular attribute B and a non-monotonic regular attribute C.

3. A method for classifying power grid line loss levels according to claim 2, characterized in that: The possibility degree is used as the fuzzy membership degree, and all attributes A of the mixed data are judged according to the order of attribute values ​​by the possibility degree, specifically: Get fuzzy ordered set The specific expression is: The possibility of obtaining each condition attribute through the fuzzy ordered set Among them, the possibility For x i Better than x j The possibility of v(x i ,a) and v(x j ,a) are two samples x i and x j The attribute value on the conditional attribute a, g is a constant, i,j∈[1,n], n is a constant.

4. A method for classifying power grid line loss levels according to claim 2, characterized in that: The fuzzy monotonicity of the mixed data attributes is evaluated according to the fuzzy sorting inconsistency rate, and the mixed data attributes are divided into fuzzy monotonic regular attributes B and non-monotonic regular attributes C, including the following steps: Obtain the inconsistent values ​​of the forward fuzzy sorting by the possibility The specific expression is: Among them, a∈A is the conditional attribute, D is the decision attribute, U is the sample set of all data, N is the number of samples, For the attribute value on A to be greater than x i The set of all examples taking values ​​on A, is greater than x on the decision attribute D i The set of all examples of decision values ​​of ; Obtain the inconsistent values ​​of the backward fuzzy sorting by the possibility The specific expression is: in, For the attribute value on A to be less than x i The set of all examples taking values ​​on A, is less than x on the decision attribute D i The set of all examples of decision values ​​of ; Obtain the fuzzy sorting inconsistency difference value by using the forward fuzzy sorting inconsistency value and the backward fuzzy sorting inconsistency value. The specific expression is: Inconsistent differences by fuzzy sorting The size of is used to judge the fuzzy monotonicity and non-monotonicity of data attributes, and divide them into fuzzy monotonic rule attributes B and non-monotonic regular attributes C.

5. A method for classifying power grid line loss levels according to claim 4, characterized in that: Inconsistent differences by fuzzy sorting The size of determines the fuzzy monotonicity and non-monotonicity of data attributes, including the following steps: like Then the judgment data is fuzzy monotonically consistent on the conditional attribute a, that is, it is a fuzzy monotonic regular attribute, otherwise it is a non-monotonic regular attribute; Where δ is the fuzzy monotonicity threshold; like When , the data is fuzzy monotonically decreasing on the conditional attribute a; When , the data is fuzzy monotonically increasing on the conditional attribute a, and the fuzzy monotonic rule attribute is obtained.

6. A method for classifying power grid line loss levels according to claim 1, characterized in that: The optimal segmentation point of the rule attribute B and the optimal segmentation point of the non-monotonic regular attribute C are obtained through the relationship between the importance of the regular attribute and the threshold value, and bifurcation is performed according to the two optimal segmentation points, and a binary decision tree is constructed through all the bifurcations, including the following steps: Set the general attribute a i The values ​​are binarized to obtain the maximized fuzzy mutual information FMI value, and the maximized fuzzy mutual information FMI value is used as the importance of the conventional attribute; If the importance of the general attribute is greater than the threshold t, the value of the general attribute is selected as the split point, and the optimal split point c is used as the split point. * Make a fork; If the importance of the regular attribute is less than the threshold t, the maximum value of the maximized fuzzy ordered mutual information FRMI is selected for rule attribute selection and used as the optimal split point b* for bifurcation; According to the optimal segmentation point c * Or b* forks, and iterates the process of obtaining the best split point, using all the forks to build a binary decision tree.

7. A method for classifying power grid line loss levels according to claim 6, characterized in that: The optimal segmentation point c * Fork, specifically: Get the best split point c * ; The specific expression is: Select the non-monotonic regular attribute C as the extended attribute, and the specific expression is: Among them, i,j∈[1,n], n is a constant, D is the decision attribute, and a is the conditional attribute.

8. A method for classifying power grid line loss levels according to claim 7, characterized in that: The method of selecting the maximum value of the maximized fuzzy ordered mutual information FRMI for rule attribute selection and using it as the best segmentation point b* for bifurcation includes the following steps: Select the optimal cut point b corresponding to the rule attribute j * , the specific expression is: Select rule attribute B as the extended attribute, the specific expression is: Among them, b j is the split point.

9. A power grid line loss classification system, characterized in that: include: A data acquisition module is used to acquire mixed data during the online power grid damage detection; the mixed data includes the power loss of the power grid line; A judgment module, used for analyzing the mixed data by using a fuzzy monotonicity judgment criterion, and dividing all attributes A of the mixed data into fuzzy monotonic regular attributes B and non-monotonic regular attributes C; An importance acquisition module is used to measure the importance of conventional attributes and rule attributes in mixed data using fuzzy mutual information FMI and fuzzy ordered mutual information FRMI respectively; A decision tree construction module is used to obtain the optimal segmentation point of the rule attribute B and the optimal segmentation point of the non-monotonic regular attribute C through the relationship between the importance of the regular attribute and the threshold value, and to bifurcate according to the two optimal segmentation points, and to construct a binary decision tree through all the bifurcations; The classification module is used to perform fuzzy monotonic consistency judgment on the mixed data of the power grid line loss detection to be classified through the binary decision tree, and obtain the power grid line loss grade classification result through the fuzzy monotonic consistency judgment result.

10. A computer device, characterized in that: It comprises a memory and a processor, wherein a program is stored in the memory, and when the program is executed by the processor, the processor executes the steps of a method for classifying power grid line loss levels as claimed in any one of claims 1 to 8.

Images