A method and system for identifying and locating abnormalities of substation equipment based on MaskRCNN

Through the MarkRCNN method, combined with iForest and K-means clustering and Mask RCNN and LSSVM models, the problems of low image recognition accuracy and low fault positioning efficiency of traditional power equipment are solved, and higher recognition accuracy and positioning accuracy are achieved. .

CN112418202BActive Publication Date: 2024-11-08SHANGHAI HENGNENGTAI ENTERPRISE MANAGEMENT CO LTD PUNENG ELECTRIC POWER TECH BRANCH +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202011060711.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-01-11
Publication Date
2024-11-08
Estimated Expiration
2041-01-11

AI Technical Summary

Technical Problem

Traditional image recognition methods perform poorly in power equipment, have low accuracy, cannot reach the industrial application level, and have weak fault tolerance and low positioning efficiency.

Method used

The abnormal identification and positioning method of substation equipment based on MarkRCNN is adopted, and real-time data of power equipment is collected and preprocessed, outliers are identified using iForest algorithm and K-means clustering strategy, and combined with Mask RCNN target recognition network model and LSSVM positioning model. Perform preliminary identification and precise positioning.

Benefits of technology

It significantly improves the accuracy of image recognition of substation equipment and the fault tolerance, efficiency and accuracy of fault positioning, and achieves higher recognition accuracy and positioning accuracy.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN112418202B_ABST
    Figure CN112418202B_ABST
Patent Text Reader

Abstract

The present invention discloses a method and system for identifying and locating abnormalities of substation equipment based on MarkRCNN, including collecting real-time data of power equipment for preprocessing; using the iForest algorithm to identify abnormal values ​​of preprocessed operating data, and combining the abnormal values ​​marked by the K-means clustering strategy; constructing a Mask RCNN target recognition network model based on a convolutional neural network; inputting the marked abnormal values ​​into the Mask RCNN target recognition network model for preliminary identification, and outputting target identification results; training and parameter optimization of LSSVM, setting accuracy requirements and thresholds, and outputting a positioning model after training; importing the target identification results into the positioning model to obtain abnormal location information of the power equipment. The present invention greatly improves the accuracy of substation image recognition, while improving the positioning and identification of abnormal locations of equipment, and improves the fault tolerance, positioning efficiency and accuracy of fault location information.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of power systems and image recognition, and in particular to a method and system for identifying and locating abnormalities of substation equipment based on MarkRCNN. Background Art

[0002] Traditional image recognition methods perform well on large sample data sets, but currently there are few image samples of power equipment, and the accuracy of traditional recognition methods is not high, and they cannot reach the level of industrial application. However, substation equipment has rich correlations, and the probability of specific types of equipment appearing at the same time is high. If this correlation is introduced during image recognition, the recognition accuracy can be greatly improved, making industrial application possible.

[0003] There are two main methods for locating fault information based on the collection of field equipment: a. Matrix method: A unified matrix algorithm has been proposed in some literature. The basic process is to first construct a network description matrix according to the topological structure of the distribution network, and generate a fault information matrix according to the overcurrent information, thereby deriving a fault judgment matrix to accurately determine the fault interval; b. Artificial intelligence method: This type of method can still accurately locate faults when the network structure changes, the uploaded real-time information is distorted or incomplete, etc. There are mainly artificial neural networks, genetic algorithms, rough set theory, data mining, Petri nets, simulated electromagnetics and other algorithms. Summary of the invention

[0004] The purpose of this section is to summarize some aspects of embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and the invention title of this application to avoid blurring the purpose of this section, the specification abstract and the invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0005] In view of the above existing problems, the present invention is proposed.

[0006] Therefore, the present invention provides a method and system for identifying and locating abnormalities of substation equipment based on MarkRCNN, which can solve the problems of weak fault tolerance, low positioning efficiency and low accuracy of fault location information.

[0007] To solve the above technical problems, the present invention provides the following technical solutions: including collecting real-time data of power equipment for preprocessing; using the iForest algorithm to identify the outliers of the preprocessed operating data, and marking the outliers in combination with the K-means clustering strategy; constructing a Mask RCNN target recognition network model based on a convolutional neural network; inputting the marked outliers into the Mask RCNN target recognition network model for preliminary identification, and outputting the target identification results; training and parameter optimization of LSSVM, setting accuracy requirements and thresholds, and outputting a positioning model after training is completed; importing the target identification results into the positioning model to obtain the abnormal location information of the power equipment.

[0008] As a preferred solution of the MarkRCNN-based method for identifying and locating abnormalities of substation equipment described in the present invention, it includes: selecting a radial basis function as the objective function of the positioning model, as follows:

[0009]

[0010] Where x={x1;x2;…;x 14}: frequency characteristic matrix composed of frequency characteristic vectors of real-time operation data of substation equipment, y: frequency characteristic vector of historical operation related data of substation equipment, σ: kernel width, reflecting the distribution and range characteristics of training.

[0011] As a preferred solution of the method for identifying and locating abnormalities of substation equipment based on MarkRCNN described in the present invention, wherein: outputting the positioning model includes initializing penalty parameters C and σ, training and testing the LSSVM using a preprocessed data set; setting the accuracy requirement, if the accuracy of the LSSVM model does not meet the requirement, optimizing the C and σ according to the error until the accuracy of the test data meets the accuracy requirement; setting the threshold, and outputting the positioning model.

[0012] As a preferred solution of the MarkRCNN-based method for identifying and locating abnormalities of substation equipment described in the present invention, the positioning model is used for positioning, including importing the pre-processed operating data into the positioning model; if the operating data of a certain device in the substation exceeds the threshold, the position of the device component is abnormal.

[0013] As a preferred solution of the abnormal identification and positioning method of substation equipment based on MarkRCNN described in the present invention, wherein: constructing the Mask RCNN target recognition network model includes superimposing multiple residual networks ResNet, y = F(x) + x; establishing a regional generation network, P i =FC2[FC1[Pooling(f,Ri )]], set the threshold to 0.5, if P i If P is greater than 0.5, the candidate region is retained. i If it is lower than 0.5, it will be discarded; generate classification branches, P c i =FC4[FC3[Pooling(f,R i ′)]]; Generate mask branch, M i =FC6[FC5[Pooling(f,R i ′)]]; where y: output of the residual network, x: input of the residual network, F: convolution operation function, f: image feature of the residual network output, R i : Candidate region, Pooling: Pooling operation, FC1 and FC2 are the first and second fully connected layer operations respectively, P i : Candidate region R i The probability of belonging to the foreground (i.e., containing the object to be identified), R i ′: reserved candidate region, FC3 and FC4 are the third and fourth fully connected layer operations respectively, P c i : Candidate region R i ′The probability of the object c to be identified, FC5 and FC6 are the operations of the fifth and sixth fully connected layers respectively, and the matrix M i With the candidate region R i ′Pixel size is consistent, M i Each position in represents the probability that the pixel in the candidate area belongs to the recognized object.

[0014] As a preferred solution of the MarkRCNN-based method for identifying and locating abnormalities of substation equipment described in the present invention, the preliminary identification includes extracting image features using the residual network; the region generation network locates the object to be identified using the image features, and sends the features of the located area to the classification branch and the mask branch respectively; the classification branch identifies the type of the object to be identified; and the mask branch locates the pixel points of the object to be identified in the image.

[0015] As a preferred solution of the MarkRCNN-based method for identifying and locating abnormalities in substation equipment described in the present invention, the target identification result includes type, orientation and size.

[0016] As a preferred solution of the MarkRCNN-based substation abnormality identification and positioning method described in the present invention, the identification and marking of the abnormal value includes: randomly sampling the operating data; defining a partitioning dimension and placing the operating data smaller than the partitioning point in the dimension on the left side of the current node, and placing the operating data larger than the partitioning point on the right side; looping and iterating until the operating data can no longer be time-divided; using the K-means clustering strategy to select k points as the initial centroid and calculate the Euclidean distance between all remaining points and the centroid; dividing all points whose distance values ​​from the centroid point are less than the threshold into a cluster; recalculating the center point of the cluster and defining the label.

[0017] As a preferred solution of the MarkRCNN-based substation abnormality identification and positioning system described in the present invention, it includes: an identification and acquisition module, which is used to collect the image information and the relevant relationship information between the substation equipment, and obtain the historical operation data and real-time operation data of the substation equipment; a data processing center module is connected to the acquisition module, which is used to receive, calculate, store, and output the data information to be processed, which includes a calculation unit, a database, and an input and output management unit. The calculation unit is connected to the acquisition module, which is used to receive the data information obtained by the information acquisition module to perform identification and positioning calculation processing and normalization processing, and calculate the type, size, and position data. The database is connected to each module, which is used to store all received data information and provide allocation and supply services for the data processing center module. The input and output management unit is used to receive information from each module and output the calculation result of the calculation unit; the positioning module is connected to the data processing center module, which is used to receive the calculation result of the calculation unit, and through calling the decoded body analysis to determine whether the size exceeds the threshold and whether the position is within the area, comprehensively determine whether the target recognition and data matching correspond to each other, so as to locate the abnormal position.

[0018] Beneficial effects of the present invention: The present invention identifies the outliers of the preprocessed operating data through the iForest algorithm, combines the outliers marked by the K-means clustering strategy, uses the Mask RCNN target recognition network model for preliminary recognition, and adds the LSSVM positioning model, which greatly improves the image recognition accuracy of the substation equipment, improves the positioning and recognition of the abnormal position of the equipment, and improves the fault tolerance, positioning efficiency and accuracy of the fault location information. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:

[0020] Figure 1 A flow chart of a method for identifying and locating abnormalities of substation equipment based on MarkRCNN according to the first embodiment of the method of the present invention;

[0021] Figure 2 A schematic diagram of a MarkRCNN algorithm framework of a MarkRCNN-based method for identifying and locating abnormalities of substation equipment described in the first embodiment of the method of the present invention;

[0022] Figure 3 This is a schematic diagram of the module structure distribution of a MarkRCNN-based substation abnormality identification and positioning system described in the second embodiment of the method of the present invention. DETAILED DESCRIPTION

[0023] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0024] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0025] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0026] The present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

[0027] At the same time, in the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the directions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0028] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0029] Example 1

[0030] Reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, provides a method for identifying and locating abnormalities of substation equipment based on MarkRCNN, comprising:

[0031] S1: Collect real-time data of power equipment for preprocessing.

[0032] S2: Use the iForest algorithm to identify the outliers of the preprocessed running data, and combine it with the K-means clustering strategy to mark the outliers. It should be noted that the identification and marking of outliers in this step includes:

[0033] Random sampling of running data;

[0034] Define the partitioning dimension and place the running data in the dimension that is smaller than the partition point on the left side of the current node, and the running data that is larger than the partition point on the right side;

[0035] The loop is repeated until the running data can no longer be divided into different time periods;

[0036] Use K-means clustering strategy to select k points as the initial centroid and calculate the Euclidean distance between all remaining points and the centroid;

[0037] All points whose distance to the centroid is less than the threshold are divided into a cluster;

[0038] Recalculate the cluster centers and define the labels.

[0039] S3: Constructing the Mask RCNN target recognition network model based on the convolutional neural network. It should also be noted that constructing the Mask RCNN target recognition network model includes:

[0040] Superimpose multiple residual networks ResNet, y = F (x) + x;

[0041] Establishing a region generation network, P i =FC2[FC1[Pooling(f,R i )]], set the threshold to 0.5, if P i If P is greater than 0.5, the candidate region is retained. i If it is less than 0.5, it will be discarded;

[0042] Generate classification branch, P c i =FC4[FC3[Pooling(f,R i ′)]];

[0043] Generate mask branch, M i =FC6[FC5[Pooling(f,R i ′)]];

[0044] Among them, y: output of residual network, x: input of residual network, F: convolution operation function, f: image feature of residual network output, R i : Candidate region, Pooling: Pooling operation, FC1 and FC2 are the first and second fully connected layer operations respectively, P i : Candidate region R i The probability of belonging to the foreground (i.e., containing the object to be identified), R i ′: reserved candidate region, FC3 and FC4 are the third and fourth fully connected layer operations respectively, P c i : Candidate region R i ′The probability of the object c to be identified, FC5 and FC6 are the operations of the fifth and sixth fully connected layers respectively, and the matrix M i With the candidate region R i ′Pixel size is consistent, M i Each position in represents the probability that the pixel in the candidate area belongs to the recognized object.

[0045] S4: Input the marked outliers into the Mask RCNN target recognition network model for preliminary recognition, and output the target recognition result. It should also be noted that the preliminary recognition includes:

[0046] Extract image features using residual network;

[0047] The region generation network uses image features to locate the object to be identified, and sends the features of the located area to the classification branch and the mask branch respectively;

[0048] The classification branch identifies the type of object to be identified;

[0049] The mask branch locates the pixels of the object to be identified in the image.

[0050] Specifically, the target recognition results include:

[0051] Type, location and size.

[0052] S5: Train and optimize the parameters of LSSVM, set the accuracy requirements and thresholds, and output the positioning model after the training is completed. It is necessary to explain again that:

[0053] The radial basis function is selected as the objective function of the positioning model, as follows:

[0054]

[0055] Where x={x1;x2;…;x 14}: frequency characteristic matrix composed of frequency characteristic vectors of real-time operation data of substation equipment, y: frequency characteristic vector of historical operation related data of substation equipment, σ: kernel width, reflecting the distribution and range characteristics of training.

[0056] Furthermore, the output positioning model includes:

[0057] Initialize the penalty parameters C and σ, and use the preprocessed data set to train and test the LSSVM;

[0058] Set the accuracy requirement. If the accuracy of the LSSVM model does not meet the requirement, optimize C and σ according to the error until the accuracy of the test data meets the accuracy requirement.

[0059] Set the threshold and output the positioning model.

[0060] S6: Import the target recognition result into the positioning model to obtain the abnormal location information of the power equipment. It should be explained again in this step that positioning using the positioning model includes:

[0061] Import the preprocessed operation data into the positioning model;

[0062] If the operating data of a device in the substation exceeds the threshold, the position of the device component is abnormal.

[0063] Preferably, the present embodiment also needs to explain that the traditional method for identifying and locating abnormalities in power equipment is to use an infrared thermometer to measure the hot spot temperature of the power equipment and the temperature of the thermometer environment reference body of the non-fault phase, calculate the relative temperature difference through the temperature difference, and combine with a PC to identify different abnormalities according to different power equipment. The main problem it solves is how to conveniently determine the abnormal location of the heating equipment for troubleshooting; compared with the traditional method, the present invention identifies the abnormal values ​​of the pre-processed operating data through the iForest algorithm, combines the abnormal values ​​marked by the K-means clustering strategy, uses the Mask RCNN target recognition network model for preliminary identification, and adds the LSSVM positioning model, which greatly improves the image recognition accuracy of the substation equipment while improving the positioning and identification of the abnormal location of the equipment, and improves the fault tolerance, positioning efficiency and accuracy of the fault location information.

[0064] Preferably, in order to better verify and illustrate the technical effects adopted in the method of the present invention, this embodiment selects a traditional method for identifying and locating abnormalities in power equipment and adopts this method for comparative testing, and compares the test results by means of scientific demonstration to verify the real effect of this method.

[0065] Traditional methods for identifying and locating abnormalities in power equipment have limitations in use, weak fault tolerance, low positioning efficiency, and accuracy is only higher for abnormal hot spots. In order to verify that the method of the present invention has higher fault tolerance, more comprehensive applicability, higher positioning efficiency and accuracy than traditional methods, this embodiment will use traditional methods and the method of the present invention to perform positioning and identification tests on the abnormal positions of substation equipment respectively.

[0066] Test environment: (1) Take 100 sets of pictures of the equipment to be tested, collect historical operation data of the equipment and real-time operation data of the entire period, and input them into the simulation software after processing;

[0067] (2) Types of substation equipment include transformers, GIS, bushings, insulators, switches, circuit breakers, transformers, and capacitors;

[0068] (3) The method of the present invention and the traditional method are trained on the training set and tested on the test set respectively, and the automatic test equipment is turned on and the MATLB simulation is used.

[0069] The test results are as follows:

[0070] Table 1: Results of device abnormal positioning tolerance rate comparison between the two methods.

[0071]

[0072]

[0073] Referring to Table 1, it can be seen intuitively that, in the face of the same test data and test equipment, the recognition and positioning results processed by the method of the present invention have a higher fault tolerance rate than the recognition and positioning results processed by the traditional method, that is, higher accuracy, which verifies the real effect of the method of the present invention.

[0074] Example 2

[0075] Reference Figure 3 , which is the second embodiment of the present invention. This embodiment is different from the first embodiment in that it provides a substation abnormality identification and positioning system based on MarkRCNN, including:

[0076] The identification and collection module 100 is used to collect image information and relevant relationship information between substation equipment, and obtain historical operation data and real-time operation data of substation equipment.

[0077] The data processing center module 200 is connected to the acquisition module 100, and is used to receive, calculate, store, and output data information to be processed. It includes a calculation unit 201, a database 202, and an input-output management unit 203. The calculation unit 201 is connected to the acquisition module 100, and is used to receive the data information acquired by the information acquisition module 100 to perform identification and positioning calculation processing and normalization processing, and calculate the type, size, and location data. The database 202 is connected to each module, and is used to store all received data information and provide allocation and supply services for the data processing center module 200. The input-output management unit 203 is used to receive information from each module and output the calculation results of the calculation unit 201.

[0078] The positioning module 300 is connected to the data processing center module 200, and is used to receive the calculation results of the calculation unit 201, and to retrieve the decoded body for analysis to determine whether the size exceeds the threshold, whether the position is within the area, and to comprehensively determine whether the target identification and data matching correspond to each other, so as to locate the abnormal position.

[0079] Generally speaking, the data processing center module 200 is mainly divided into three levels, including a control layer, an operation layer and a storage layer. The control layer is the command and control center of the data processing center module 200, and is composed of an instruction register IR, an instruction decoder ID and an operation controller OC. The control layer can take out each instruction from the memory in sequence according to the program pre-compiled by the user, put it in the instruction register IR, analyze and determine it through the instruction decoder, notify the operation controller OC to operate, and send micro-operation control signals to the corresponding components according to the determined timing; the operation layer is the core of the computing unit 201, and can perform arithmetic operations (such as addition, subtraction, multiplication, division and additional operations) and logical operations (such as shifting, logical testing or comparison of two values). It is connected to the control layer and performs operations by receiving control signals from the control layer; the storage layer is the database of the data processing center module 200, and can store data (data to be processed and data that has been processed).

[0080] It should be appreciated that embodiments of the present invention may be implemented or enforced by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-transitory computer-readable memory. The method may be implemented in a computer program using standard programming techniques - including a non-transitory computer-readable storage medium configured with a computer program, wherein the storage medium so configured causes the computer to operate in a specific and predefined manner - according to the methods and drawings described in the specific embodiments. Each program may be implemented in a high-level procedural or object-oriented programming language to communicate with a computer system. However, if desired, the program may be implemented in an assembly or machine language. In any case, the language may be a compiled or interpreted language. In addition, the program may be run on a programmed ASIC for this purpose.

[0081] Furthermore, the operations of the processes described herein may be performed in any suitable order unless otherwise indicated herein or otherwise clearly contradicted by context. The processes described herein (or variations and / or combinations thereof) may be performed under the control of one or more computer systems configured with executable instructions, and may be implemented as code (e.g., executable instructions, one or more computer programs, or one or more applications) that is executed collectively on one or more processors, by hardware, or a combination thereof. The computer program includes a plurality of instructions that may be executed by one or more processors.

[0082] Further, the method can be implemented in any type of computing platform that is operably connected to a suitable computer, including but not limited to a personal computer, a minicomputer, a mainframe, a workstation, a network or distributed computing environment, a separate or integrated computer platform, or in communication with a charged particle tool or other imaging device, etc. Aspects of the present invention can be implemented in machine-readable code stored on a non-transitory storage medium or device, whether removable or integrated into a computing platform, such as a hard disk, an optical read and / or write storage medium, a RAM, a ROM, etc., so that it can be read by a programmable computer, and when the storage medium or device is read by the computer, it can be used to configure and operate the computer to perform the process described herein. In addition, the machine-readable code, or part thereof, can be transmitted via a wired or wireless network. When such media includes instructions or programs that implement the steps described above in conjunction with a microprocessor or other data processor, the invention described herein includes these and other different types of non-transitory computer-readable storage media. When programmed according to the methods and techniques described in the present invention, the present invention also includes the computer itself. The computer program can be applied to input data to perform the functions described herein, thereby converting the input data to generate output data stored in a non-volatile memory. The output information can also be applied to one or more output devices such as a display. In a preferred embodiment of the present invention, the converted data represents a physical and tangible object, including a specific visual depiction of the physical and tangible object produced on a display.

[0083] As used in this application, the terms "component", "module", "system", etc. are intended to refer to a computer-related entity, which can be hardware, firmware, a combination of hardware and software, software, or software in operation. For example, a component can be, but is not limited to: a process running on a processor, a processor, an object, an executable file, a thread in execution, a program, and / or a computer. As an example, an application running on a computing device and the computing device can both be components. One or more components can exist in a process and / or thread in execution, and a component can be located in a computer and / or distributed between two or more computers. In addition, these components can be executed from various computer-readable media having various data structures thereon. These components can communicate in a local and / or remote process manner, such as according to a signal having one or more data packets (e.g., data from a component that interacts with another component in a local system, a distributed system, and / or interacts with other systems in a signal manner through a network such as the Internet).

[0084] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for identifying and locating abnormalities of substation equipment based on MarkRCNN, characterized in that: include, Collect real-time data of substation equipment for preprocessing; The iForest algorithm is used to identify the outliers of the preprocessed running data, and the outliers are marked by combining the K-means clustering strategy; the outliers are identified and marked, including: randomly sampling the operating data; Define a partitioning dimension and place the running data smaller than the partitioning point in the dimension on the left side of the current node, and place the running data larger than the partitioning point on the right side; The loop is iterated until the operation data can no longer be divided into time periods; Using the K-means clustering strategy, k points are selected as initial centroids and the Euclidean distances between all remaining points and the centroids are calculated; All points whose distance to the centroid is less than the threshold are divided into a cluster; Recalculate the center point of the cluster and define the label; Construct the Mask RCNN target recognition network model based on convolutional neural network; include, Superimpose multiple residual networks ResNet, y = F (x) + x; Establishing a region generation network, P i =FC2[FC1[Pooling(f,R i )]], set the threshold to 0.5, if P i If P is greater than 0.5, the candidate region is retained. i If it is less than 0.5, it will be discarded; Generate classification branches, Generate mask branches, Among them, y: output of residual network, x: input of residual network, F: convolution operation function, f: image feature of residual network output, Pooling: pooling operation, FC1 and FC2 are the first and second fully connected layer operations respectively, R i : Candidate region, P i : Candidate region R i The probability of containing the object to be identified, R i ′: the candidate region to be retained, FC3 and FC4 are the third and fourth fully connected layer operations respectively, The retained candidate region R i ′The probability of the object c to be identified, FC5 and FC6 are the operations of the fifth and sixth fully connected layers respectively, and the matrix M i and the retained candidate region R i ′Pixel size is consistent, M i Each position in represents the probability that the pixel point in the candidate area belongs to the identified object; The marked outliers are input into the Mask RCNN target recognition network model for preliminary recognition, and the target recognition result is output; the preliminary recognition includes: Extracting image features using the residual network; The region generation network locates the object to be identified using the image features, and sends the features of the located region to the classification branch and the mask branch respectively; The classification branch identifies the type of the object to be identified; The mask branch locates the pixel points of the object to be identified in the image; The target recognition result includes type, orientation and size; Train and optimize the parameters of LSSVM, set the accuracy requirements and thresholds, and output the positioning model after training. Select the radial basis function as the objective function of the positioning model, as follows: Where x={x1; x2; ...; x 14 }: frequency characteristic matrix composed of frequency characteristic vectors of real-time operation data of substation equipment, y: frequency characteristic vector of historical operation data of substation equipment, σ: kernel width, reflecting the distribution and range characteristics of training; Outputting the positioning model includes, Initialize penalty parameters C and σ, and train and test the LSSVM using the preprocessed data set; The accuracy requirement is set. If the accuracy of the LSSVM model does not meet the requirement, the C and σ are assigned and optimized according to the error until the accuracy of the test data meets the accuracy requirement; Setting the threshold and outputting the positioning model; Importing the target recognition result into the positioning model to obtain abnormal position information of the substation equipment; using the positioning model to perform positioning, including: Importing the preprocessed operation data into the positioning model; If the operating data of a certain device in the power transformation equipment exceeds the threshold, the location of the device is abnormal.

2. A MarkRCNN-based power substation abnormality identification and positioning system application The MarkRCNN-based power substation abnormality identification and positioning method as claimed in claim 1 is characterized by: include, An identification and collection module (100) is used to collect image information and related relationship information between substation equipment, and obtain historical operation data and real-time operation data of substation equipment; The data processing center module (200) is connected to the acquisition module (100) and is used to receive, calculate, store and output data information to be processed. It includes a calculation unit (201), a database (202) and an input / output management unit (203). The calculation unit (201) is connected to the acquisition module (100) and is used to receive the data information acquired by the information acquisition module (100) to perform identification and positioning calculation processing and normalization processing, and calculate the type, size and position data. The database (202) is connected to each module and is used to store all received data information and provide allocation and supply services for the data processing center module (200). The input / output management unit (203) is used to receive information from each module and output the calculation result of the calculation unit (201). The positioning module (300) is connected to the data processing center module (200) and is used to receive the operation result of the operation unit (201), retrieve the decoded body for analysis to determine whether the size exceeds the threshold, whether the position is within the area, and comprehensively determine whether the target identification and data matching correspond, so as to locate the abnormal position.