CN112199805A - Power transmission line hidden danger identification model evaluation method and device - Google Patents

Power transmission line hidden danger identification model evaluation method and device Download PDF

Info

Publication number
CN112199805A
CN112199805A CN202010952891.6A CN202010952891A CN112199805A CN 112199805 A CN112199805 A CN 112199805A CN 202010952891 A CN202010952891 A CN 202010952891A CN 112199805 A CN112199805 A CN 112199805A
Authority
CN
China
Prior art keywords
power transmission
transmission line
hidden danger
identification model
monitoring device
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202010952891.6A
Other languages
Chinese (zh)
Other versions
CN112199805B (en
Inventor
郭国信
蔡富东
吕昌峰
文刚
陈雷
刘伟
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shandong Senter Electronic Co Ltd
Original Assignee
Shandong Senter Electronic Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shandong Senter Electronic Co Ltd filed Critical Shandong Senter Electronic Co Ltd
Priority to CN202010952891.6A priority Critical patent/CN112199805B/en
Publication of CN112199805A publication Critical patent/CN112199805A/en
Application granted granted Critical
Publication of CN112199805B publication Critical patent/CN112199805B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/10—Geometric CAD
    • G06F30/18—Network design, e.g. design based on topological or interconnect aspects of utility systems, piping, heating ventilation air conditioning [HVAC] or cabling
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/01—Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
    • G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Electricity, gas or water supply
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S10/00—Systems supporting electrical power generation, transmission or distribution
    • Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications

Abstract

The application discloses a method and a device for evaluating a hidden danger identification model of a power transmission line, comprising the following steps of: acquiring analysis indexes of the hidden danger identification models of the power transmission lines corresponding to the monitoring devices in the plurality of power transmission line monitoring devices; the analysis index is related to the accuracy and/or the missing report rate and/or the false report rate of the monitoring device; according to a pre-constructed decision tree model, making a decision on analysis indexes of a power transmission line hidden danger identification model corresponding to each monitoring device in a plurality of power transmission line monitoring devices, and determining a decision value of a power transmission line hidden danger identification model updating task corresponding to any one monitoring device in the plurality of power transmission line monitoring devices; and activating the hidden danger identification model updating task of the power transmission line of the corresponding monitoring device according to the decision value. The method and the device realize that the low-precision hidden danger identification model of the power transmission line corresponding to each monitoring device is updated in time by making a decision on the analysis index of the hidden danger identification model of the power transmission line corresponding to each monitoring device.

Description

Power transmission line hidden danger identification model evaluation method and device
Technical Field
The application relates to the technical field of power transmission lines, in particular to a power transmission line hidden danger identification model evaluation method and device.
Background
The power transmission line is an important component of a power grid and is affected by artificial and natural conditions, various potential safety hazards often appear in the power transmission line, and if the potential safety hazards of the power transmission line cannot be checked in time, the power transmission line not only can endanger the safe operation of the power grid, but also can affect the production and life of users.
Due to the limitations of low computational power and low power consumption of the power transmission line monitoring device, only a lightweight deep neural network model can be deployed. However, the light-weight neural network model has insufficient expression of the characteristics of the image, and the recognition accuracy cannot reach the recognition accuracy of the server, so that a large number of false alarms are caused, and the hidden danger recognition accuracy of the monitoring device on the power transmission line is influenced.
Disclosure of Invention
The embodiment of the application provides a method and a device for evaluating a hidden danger identification model of a power transmission line, and solves the problem that a monitoring device is low in identification precision of hidden danger identification analysis of the power transmission line.
On one hand, the embodiment of the application provides a method for evaluating a hidden danger identification model of a power transmission line, which comprises the following steps: acquiring analysis indexes of the hidden danger identification models of the power transmission lines corresponding to the monitoring devices in the plurality of power transmission line monitoring devices; the analysis index is related to the accuracy and/or the missing report rate and/or the false report rate of the monitoring device; according to a pre-constructed decision tree model, making a decision on analysis indexes of the hidden danger identification models of the power transmission lines corresponding to the monitoring devices in the plurality of power transmission line monitoring devices, and determining a decision value of an update task of the hidden danger identification model of the power transmission line corresponding to any one of the plurality of power transmission line monitoring devices; and activating the hidden danger identification model updating task of the corresponding monitoring device according to the decision value.
According to the embodiment of the application, the decision tree models corresponding to the hidden danger identification and analysis models of the power transmission line of the plurality of monitoring devices are used for deciding the analysis indexes of the hidden danger identification models of the power transmission line corresponding to the monitoring devices, the monitoring devices needing to update the hidden danger identification models of the power transmission line can be screened out from the plurality of monitoring devices, the problem that resources of the power transmission line training identification models are insufficient by a server is solved, the hidden danger identification models of the low-precision power transmission line corresponding to the monitoring devices can be updated in time, and the missing report rate and the false report rate of the hidden danger identification models of the power transmission line of the corresponding monitoring devices are reduced.
In one example, according to a training sample set, discretizing a continuous attribute value corresponding to each continuous attribute of the training sample set through a carry step length to obtain a corresponding discrete value; the continuous attribute is related to the accuracy and/or the missing report rate and/or the false report rate of the monitoring device for identifying the hidden danger of the power transmission line; and respectively carrying out information gain calculation on discrete values corresponding to the continuous attributes to obtain a decision tree model.
According to the embodiment of the application, the continuous attribute values corresponding to the continuous attributes of the training sample set are discretized by setting different step lengths, so that the segmentation granularity of different attributes is realized, the weight of each continuous attribute can be more flexibly controlled, and the decision tree model is more accurate.
In one example, discretizing the continuous attribute values corresponding to each continuous attribute of the training sample set by a carry step specifically includes: determining a discrete interval of each continuous attribute value according to the carry step value of each continuous attribute value; and obtaining discrete values corresponding to the continuous attribute values based on the discrete intervals of the continuous attribute values.
According to the embodiment of the application, each continuous attribute is respectively dispersed into a set of a plurality of intervals through the carry step value, so that the difference between discrete values corresponding to each continuous attribute is small, and the accuracy of the decision tree model is improved.
In one example, the discrete values corresponding to the continuous attributes are respectively subjected to information gain calculation to obtain
The decision tree model specifically comprises: determining the information gain of each continuous attribute according to the information gain value of the discrete value corresponding to each continuous attribute; taking the continuous attribute with the largest information gain in all continuous attributes as the optimal partition attribute of the training sample set; determining the branches of the optimal division attributes according to the optimal splitting points of the optimal division attributes; determining a sample subset of a training sample set corresponding to the branch based on the branch with the optimal division attribute; determining the optimal partition attribute corresponding to the sample subset of the training sample set based on the sample subset of the training sample set corresponding to the branch; the method comprises the following steps of recursively dividing a plurality of optimal attributes of a training sample set from top to bottom until nodes meet growth stopping conditions of a decision tree model; and pruning the grown decision tree model to obtain the decision tree model.
According to the embodiment of the application, the continuous attribute with the largest information gain in the continuous attributes is used as the optimal partition attribute of the training sample set, the carry step value of each continuous attribute is adjusted, the precision of the decision tree model is achieved, and the finally generated decision tree model is small in scale and high in intelligibility while the classification precision is maintained.
In one example, the stopping of growth condition until the node satisfies the decision tree model specifically includes any one or more of the following: the information gain of the discrete value corresponding to each continuous attribute is smaller than a preset threshold value; the discrete value corresponding to each continuous attribute is not divisible; the subset of training samples under each branch of the best partition attribute has the same classification.
In one example, the analysis index is related to the accuracy and/or the false alarm rate of the monitoring device, and specifically includes: the analysis index is the accuracy rate or the rate of missing report or the rate of false report; or a alpha% + b beta% + c gamma%, where a is the accuracy, alpha% is the weight corresponding to the accuracy, b is the false alarm rate, beta% is the weight corresponding to the false alarm rate, c is the false alarm rate, and gamma% is the weight corresponding to the false alarm rate.
In one example, obtaining an analysis index of a power transmission line hidden danger identification model corresponding to each monitoring device in a plurality of power transmission line monitoring devices includes: receiving corresponding power transmission line image information from a monitoring device; the image information comprises hidden danger types and hidden danger coordinates; and determining an accurate alarm power transmission line hidden danger image and/or a false alarm power transmission line hidden danger image and/or a missed alarm power transmission line hidden danger image in the power transmission line image according to the hidden danger type and the hidden danger coordinates in the power transmission line image.
In one example, determining a missed-report potential danger image of the power transmission line in the power transmission line image specifically includes: and performing cascade detection on the power transmission line images of the monitoring devices according to the offline high-precision model, comparing the cascade detection result with the analysis result of the corresponding power transmission line image of each monitoring device, and determining the report missing image.
According to the embodiment of the application, the missed-report picture is determined through the offline high-precision model, and the detection efficiency is improved.
In one example, according to the decision value, activating an update task of the transmission line hidden danger identification model of the corresponding monitoring device specifically includes: and in a preset period, activating the hidden danger identification model updating task of the corresponding monitoring device according to the decision value of the hidden danger identification model updating task of the power transmission line, the number of days exceeding a preset activation threshold value and the number of days exceeding a preset number of days.
On the other hand, the embodiment of the application provides a device for evaluating a hidden danger identification model of a power transmission line, which comprises: the acquisition module is used for acquiring analysis indexes of the hidden danger identification model of the power transmission line corresponding to each monitoring device in the plurality of power transmission line monitoring devices; the analysis index is related to the accuracy and/or the missing report rate and/or the false report rate of the monitoring device; the analysis module is used for deciding analysis indexes of the hidden danger identification models of the power transmission lines corresponding to the monitoring devices in the plurality of power transmission line monitoring devices according to a pre-constructed decision tree model and determining a decision value of an update task of the hidden danger identification model of the power transmission line corresponding to any one of the plurality of power transmission line monitoring devices; and the activation module activates the power transmission line hidden danger identification model updating task of the corresponding monitoring device according to the decision value.
According to the method and the device for evaluating the hidden danger identification model of the power transmission line, under the condition that the customized model of the specific power transmission line monitoring scene is updated, the analysis indexes of the hidden danger identification model of the power transmission line corresponding to each monitoring device are decided through the decision tree models corresponding to the hidden danger identification analysis models of the power transmission lines of the monitoring devices, the monitoring devices needing to be updated of the hidden danger identification model of the power transmission line can be screened out from the monitoring device equipment, the problem that resources of a server for training the power transmission line identification model are insufficient is solved, the hidden danger identification model of the low-precision power transmission line corresponding to the monitoring device can be updated in time, and the missing report rate and the false report rate of the hidden danger identification model of the power transmission line of the corresponding monitoring device are reduced.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the application and together with the description serve to explain the application and not to limit the application. In the drawings:
fig. 1 is a schematic diagram of a power transmission line hidden danger identification model evaluation system provided in an embodiment of the present application;
fig. 2 is a flowchart of a power transmission line hidden danger identification model evaluation method provided in the embodiment of the present application;
fig. 3 is a decision tree training flowchart for evaluating a hidden danger identification model of a power transmission line according to an embodiment of the present disclosure;
fig. 4 is a schematic diagram of an implementation of a power transmission line hidden danger identification model evaluation method according to an embodiment of the present application;
fig. 5 is a structural block diagram of a device for evaluating a hidden danger identification model of a power transmission line according to an embodiment of the present application.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the technical solutions of the present application will be described in detail and completely with reference to the following specific embodiments of the present application and the accompanying drawings. It should be apparent that the described embodiments are only some of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
Fig. 1 is a schematic diagram of a power transmission line hidden danger identification model evaluation system provided in an embodiment of the present application.
As shown in fig. 1, the power transmission line hidden danger identification model evaluation system at least includes: the monitoring platform 110, the server 120, and a plurality of power transmission line monitoring devices, including a monitoring device 130, a monitoring device 140, and a monitoring device 150.
It should be noted that the power transmission line monitoring system according to the embodiment of the present application is provided with a plurality of power transmission line monitoring devices, each of the plurality of power transmission line monitoring devices is connected to the monitoring platform 110, the number of the monitoring devices may be one, or may be multiple, as shown in fig. 1, the monitoring devices 130, 140, and 150 are provided respectively. In the embodiment of the present application, the functions, structures and connections of the monitoring devices are the same, and for convenience of description, the monitoring device 130 is taken as an example for explanation.
The method analysis function according to the embodiment of the present application may be implemented by the server 120 or the monitoring platform 110, which is not limited in this application. For convenience of understanding and description, in the following embodiments, details of the related content of the task of determining whether to activate the monitoring device 130 to identify the hidden danger in the transmission line are described by taking the monitoring platform 110 as an example, and details of training the pre-constructed decision tree model are described by taking the server 120 as an example.
The monitoring platform 110 obtains an analysis index of the power transmission line hidden danger identification model corresponding to the monitoring device 130, and makes a decision on the analysis index of the power transmission line hidden danger identification model corresponding to the monitoring device 130 through a pre-constructed decision tree model to obtain a decision value of an update task of the power transmission line hidden danger identification model corresponding to the monitoring device 130. And in a preset period, judging whether to activate an updating task of the power transmission line hidden danger identification model corresponding to the monitoring device 130. When the training of the neural network model corresponding to the monitoring device 130 is completed, the server 120 stores the neural network model that passes the verification test, and then notifies the monitoring device 130, so that the monitoring device 130 obtains the updated neural network model obtained by the training task from the network model database of the server 120. How to specifically determine whether to activate the update task of the hidden danger identification model of the power transmission line corresponding to the monitoring device 130 in the embodiment of the present application will be specifically explained with reference to fig. 2.
Fig. 2 is a flowchart of a power transmission line hidden danger identification model evaluation method provided in the embodiment of the present application.
S201, the monitoring platform obtains analysis indexes of the hidden danger identification models of the power transmission line corresponding to the monitoring devices in the power transmission line monitoring devices.
In the embodiment of the present application, the monitoring device 130 periodically shoots the corresponding power transmission line to obtain an image of the corresponding power transmission line. For example, the shooting interval of the monitoring device 130 is 30 min/time.
The present application is not limited to the mode of photographing the power transmission line corresponding to the monitoring device 110 in fig. 2.
The monitoring device 130 sends the corresponding transmission line image and the analysis result of the corresponding transmission line image to the monitoring platform 130. Specifically, the corresponding transmission line image includes the hidden danger-free image and the labeled hidden danger image identified by the monitoring device 130, and the analysis result includes image information of the corresponding transmission line image. Further, the image information includes the hidden danger type and the hidden danger coordinate bbox value of the corresponding power transmission line image.
It should be noted that, in the shooting interval, when the corresponding power transmission line fails, the monitoring device 130 automatically triggers to shoot the corresponding failed power transmission line, and immediately sends the image and the analysis result of the corresponding power transmission line to the monitoring platform 110.
The monitoring platform 110 determines an accurate warning transmission line hidden danger image, a false-reported transmission line hidden danger image and a missed-reported transmission line hidden danger image in the transmission line image corresponding to the monitoring device 130 by performing secondary recognition on the corresponding transmission line image. And the analysis index of the power transmission line hidden danger identification model corresponding to the monitoring device 130 is obtained through the corresponding accurate alarm power transmission line hidden danger image, the mistakenly reported power transmission line hidden danger image and the missed reported power transmission line hidden danger image.
In one example, the monitoring platform 130 performs cascade detection on the power transmission line image corresponding to the monitoring device 130 through an offline high-precision model, and compares the cascade detection result with the analysis result of the monitoring device 130 on the corresponding power transmission line image, so as to obtain the power transmission line hidden danger image which is not reported. According to the embodiment of the application, the missed-report picture is determined through the offline high-precision model, and the detection efficiency is improved.
In the embodiment of the application, the analysis index is the accuracy rate or the rate of missing report or the rate of false report; or a α% + b β% + c γ%;
wherein, a is the accuracy, alpha% is the weight corresponding to the accuracy, b is the false alarm rate, beta% is the weight corresponding to the false alarm rate, c is the rate of missing report, and gamma% is the weight corresponding to the rate of missing report.
Further, the accuracy of the power transmission line hidden danger identification model corresponding to the monitoring device 130 is obtained through the corresponding correct alarm power transmission line hidden danger image, and the method is realized through the following formula:
Figure BDA0002677609720000071
wherein, a is the accuracy of the hidden danger identification model of the power transmission line corresponding to the monitoring device 130, r is the total number of hidden danger targets in the hidden danger image of the power transmission line corresponding to the monitoring device 130, and n is the total number of hidden danger targets in the hidden danger image of the power transmission line corresponding to the monitoring device 130.
For example, if the number of the potential danger images of the power transmission line is 2, each image has 5 potential danger targets, the analysis result of the potential danger identification model of the power transmission line corresponding to the monitoring device 130 is that one image includes 3 potential danger targets, and the other image includes 4 potential danger targets, the accuracy is high
Figure BDA0002677609720000072
The false alarm rate of the hidden danger identification model of the power transmission line corresponding to the monitoring device 130 is obtained through the corresponding false alarm hidden danger image of the power transmission line, and is realized through the following formula:
Figure BDA0002677609720000081
wherein, b is a false alarm rate of the hidden danger identification model of the power transmission line corresponding to the monitoring device 130, t is the total number of images in which the hidden danger target does not exist but is identified in the image of the power transmission line, s is the total number of the hidden danger images identified by the hidden danger identification model of the power transmission line, and the identification result is the total number of the hidden danger images.
The false alarm rate of the power transmission line hidden danger identification model corresponding to the monitoring device 130 is obtained through the corresponding false alarm power transmission line hidden danger images, and is realized through the following formula:
Figure BDA0002677609720000082
wherein c is a false negative rate of the hidden danger identification model of the power transmission line corresponding to the monitoring device 130, l is the total number of images in which the hidden danger target exists but the hidden danger target is not identified in the power transmission line image, and m is the total number of the hidden danger images in the power transmission line image, that is, the total number of the hidden danger images actually existing in the power transmission line image.
S202, according to a pre-constructed decision tree model, making a decision on analysis indexes of the hidden danger identification model of the power transmission line corresponding to each monitoring device in the plurality of power transmission line monitoring devices, and determining a decision value of an update task of the hidden danger identification model of the power transmission line corresponding to any one monitoring device in the plurality of power transmission line monitoring devices.
The monitoring platform 110 inputs the analysis index of the power transmission line hidden danger identification model corresponding to the monitoring device 130 into the decision tree model, makes a decision on the analysis index of the power transmission line hidden danger identification model corresponding to the monitoring device 130, and determines whether the power transmission line hidden danger identification model corresponding to the monitoring device 130 needs to be updated. According to the method and the device, analysis indexes of the power transmission line hidden danger identification models corresponding to the monitoring devices in the plurality of monitoring devices are used as training sample sets, and the decision tree model is obtained through training according to the training sample sets. How to obtain the decision tree models corresponding to the plurality of monitoring devices according to the training sample set is specifically described in detail with reference to fig. 3, which is referred to in fig. 3 and related contents.
And S203, activating the power transmission line hidden danger identification model updating task of the corresponding monitoring device according to the decision value.
Specifically, in a preset period, the number of days that the power transmission line hidden danger identification model corresponding to the monitoring device 130 needs to be updated exceeds a preset activation threshold, and an update task of the power transmission line hidden danger identification model of the monitoring device 130 is activated.
For example, taking one week as a calculation cycle, making a decision on the analysis index of the electric transmission line hidden danger identification model corresponding to the monitoring device 130 every day, so that the number of days for which the monitoring device 130 needs to update the corresponding electric transmission line hidden danger identification model reaches or exceeds three days, activating the update task of the monitoring device 130 corresponding to the electric transmission line hidden danger identification model,
in one example, in a preset period, the monitoring platform 110 determines whether to activate an update task of the hidden danger identification model of the corresponding power transmission line of the monitoring device 130 according to the decision value of the total analysis index for identifying the hidden danger of the corresponding power transmission line.
According to the embodiment of the application, the decision tree models corresponding to the hidden danger identification and analysis models of the power transmission line of the plurality of monitoring devices are used for deciding the analysis indexes of the hidden danger identification models of the power transmission line corresponding to the monitoring devices, the monitoring devices needing to update the hidden danger identification models of the power transmission line can be screened out from the plurality of monitoring devices, the problem that resources of the power transmission line training identification models are insufficient by a server is solved, the hidden danger identification models of the low-precision power transmission line corresponding to the monitoring devices can be updated in time, and the missing report rate and the false report rate of the hidden danger identification models of the power transmission line of the corresponding monitoring devices are reduced.
Fig. 3 is a decision tree training flowchart for evaluating a power transmission line hidden danger identification model according to an embodiment of the present disclosure.
S301, discretizing each continuous attribute by the server according to the training sample set, and determining a discrete value corresponding to each continuous attribute.
Specifically, the server 120 divides the continuous attribute value corresponding to each continuous attribute into corresponding discrete intervals by the carry step length, and uses a plurality of split points in the discrete interval corresponding to each continuous attribute value as the discrete value of each continuous attribute.
It should be noted that the monitoring platform 110 stores analysis indexes of the hidden danger identification models of the power transmission line corresponding to the plurality of monitoring devices, and sends the analysis indexes to the server 120 to obtain a training sample set and a test sample set. For example, the monitoring platform 110 performs structured storage on the analysis indexes of the transmission line hidden danger identification models corresponding to the plurality of monitoring devices in units of days, and sends the analysis indexes to the server 120.
Each continuous attribute is an analysis index of the hidden danger identification model of the transmission line corresponding to each monitoring device in the plurality of monitoring devices, and comprises the accuracy of the hidden danger identification model of the transmission line corresponding to each monitoring device, the false alarm rate of the hidden danger identification model of the transmission line corresponding to each monitoring device and the false alarm rate of the hidden danger identification model of the transmission line corresponding to each monitoring device.
For example, the accuracy is 0.8856, 0.6034, and 0.9067, respectively, the carry step value is 0.01, and the discrete intervals are [0-0.01], [ 0.01-0.02....... [0.99-1], and the accuracy is distributed in the discrete intervals for 100 possible values, and the discrete values are 0, 0.01, and 0.02......0.99, and 1, respectively, which are the two latter decimal places. Thus, 0.6034 falls within the interval [0.6-0.61], corresponding discrete values of 0.61, 0.8856 falls within the interval [0.88-0.89], corresponding discrete values of 0.89, 0.9067 falls within the interval [0.9-0.91], corresponding discrete values of 0.91. It should be noted that the discrete value may be an upper critical point or a lower critical point of the interval, but the discrete values of the continuous attributes should be taken in a consistent manner. That is, 0.6034 corresponds to a discrete value of 0.6, 0.8856 corresponds to a discrete value of 0.88, and 0.9067 corresponds to a discrete value of 0.9.
Further, in this embodiment of the present application, the carry step values of the respective continuous attribute values are independent, that is, the carry step value of the accuracy, the carry step value of the false alarm rate, and the carry step value of the false alarm rate may be the same or different.
The embodiment of the application discretizes the continuous attribute values corresponding to the continuous attributes of the training sample set by setting different step lengths, realizes the segmentation granularity of different attributes, can more flexibly control the weight of each continuous attribute, and discretizes each continuous attribute into a set of a plurality of intervals respectively through the carry step length value, so that the difference between the discrete values corresponding to each continuous attribute is smaller, the accuracy of the decision tree model is improved, and the decision tree model is more accurate.
S302, determining the information gain of each continuous attribute according to the information gain value of the discrete value corresponding to each continuous attribute.
Specifically, the information gain of a plurality of discrete values corresponding to each continuous attribute is calculated, and the maximum information gain value corresponding to the plurality of discrete values is used as the information gain of each continuous attribute. The information gain is realized by the following formula:
Gain(D,A)＝Ent(D)-H(D|A)
wherein A represents each continuous attribute, D represents a training sample set, Gain(D, A) represents the information gain of the continuous attribute A to the training sample set D, Ent (D) represents the information entropy of the training sample set, and H (D | A) represents the conditional entropy of the continuous attribute A to the data set D.
Further, the information entropy is realized by the following formula:
Figure BDA0002677609720000111
d is the total number of sample classes, p, divided by the hidden danger identification model of the power transmission line corresponding to the monitoring device according to whether the sample set needs to be updated or not in the training sample set DkRandomly selecting the probability that one sample belongs to the kth sample from the training sample set D,
further, H (D | a) is realized by the following formula:
Figure BDA0002677609720000112
wherein D isiA subset of samples representing the ith value of consecutive attributes in a training sample set D, DikRepresents DiAnd n represents the number of divided sample subsets in the training sample set D.
And S303, taking the continuous attribute with the largest information gain in all continuous attributes as the optimal partition attribute of the training sample set.
Specifically, the information gains of the continuous attributes are compared, and the continuous attribute with the largest information gain is used as the optimal partition attribute of the training sample set. And taking the discrete value corresponding to the maximum information gain value as the optimal splitting point of each continuous attribute.
S304, determining the branches of the optimal division attributes according to the optimal splitting points of the optimal division attributes.
S305, determining a sample subset of the training sample set corresponding to the branch based on the branch with the optimal division attribute.
S306, determining the optimal partition attribute corresponding to the sample subset of the training sample set based on the sample subset of the training sample set corresponding to the branch.
And S307, recursively dividing a plurality of optimal attribute division processes of the training sample set from top to bottom until the nodes meet the growth stopping conditions of the decision tree model.
Specifically, S301-S306 are continued for each node of the decision tree, the continuous attribute with the largest information gain is selected, and the training sample set is continued to be divided, so that each subset can generate a decision sub-tree until the information gain of the discrete value corresponding to each continuous attribute is smaller than a preset threshold value, or the discrete value corresponding to each continuous attribute is not divisible, or the training sample subsets under each branch of the optimal division attribute have the same classification, and the division of the training sample set is stopped, so that the decision tree is obtained.
And S308, pruning the grown decision tree model to obtain the decision tree model.
The embodiment of the present application is not particularly limited to the method for pruning the decision tree model that completes the growth. For example, the decision tree model that completes the growth is pruned by a cost complexity pruning method.
According to the embodiment of the application, the pruned decision tree model is evaluated through the test sample set, the carry step value of each continuous attribute in the S301 is adjusted according to the evaluation result until the precision of the decision tree model is reached, the training process of the decision tree model is stopped, and the decision tree models corresponding to the plurality of monitoring devices are obtained.
According to the embodiment of the application, the continuous attribute with the largest information gain in the continuous attributes is used as the optimal partition attribute of the training sample set, the carry step value of each continuous attribute is adjusted, the precision of the decision tree model is achieved, and the finally generated decision tree model is small in scale and high in intelligibility while the classification precision is maintained.
According to the above description, an embodiment of the present application further provides a schematic diagram of an implementation of the power transmission line hidden danger identification model evaluation method in the scenario of fig. 2, as shown in fig. 4.
In fig. 4, the analysis indexes of the monitoring device 130 are, in units of days: accuracy 0.8856, false alarm rate 0.0214 and missing alarm rate 0.4160. The analysis indexes of the monitoring device 140 are: accuracy 0.6034, false alarm rate 0.1005 and missing report rate 0.0523. The analysis indexes of the monitoring device 150 are: accuracy 0.9067, false alarm rate 0.2545, and false alarm rate 0.1047.
The monitoring platform 110 inputs the analysis index of the monitoring device 130 into a pre-constructed decision tree model, and obtains that the decision value of the power transmission line hidden danger identification model corresponding to the monitoring device 130 is NO, that is, the corresponding power transmission line hidden danger identification model is not updated. And inputting the analysis index of the monitoring device 140 into the pre-constructed decision tree model to obtain a decision value Yes of the power transmission line hidden danger identification model corresponding to the monitoring device 140, namely updating the corresponding power transmission line hidden danger identification model. And inputting the analysis index of the monitoring device 150 into the pre-constructed decision tree model to obtain a Yes decision value of the power transmission line hidden danger identification model corresponding to the monitoring device 150, namely updating the corresponding power transmission line hidden danger identification model.
Taking one week as a statistical period, in the statistical period, the monitoring platform 110 activates a model updating task when the number of days for which the monitoring devices need to update the power transmission new circuit hidden danger identification model reaches or exceeds 3 days. That is to say, if the monitoring platform 110 determines that the number of days that the monitoring device 130 needs to update the new power transmission circuit risk identification model is 2 days, the corresponding new power transmission circuit risk identification model of the monitoring device 130 is not activated. Further, if the monitoring platform 110 determines that the number of days that the monitoring device 130 needs to update the new power transmission circuit hidden danger identification model is 3 days, the monitoring device 130 is activated to activate the corresponding new power transmission circuit hidden danger identification model. Furthermore, if the monitoring platform 110 determines that the number of days that the monitoring device 130 needs to update the new power transmission circuit hidden danger identification model is 4 days, the monitoring device 130 is activated to activate the corresponding new power transmission circuit hidden danger identification model.
In fig. 4, the analysis index corresponding to the monitoring device 130 may be an average value of the total analysis indexes of one week, in units of one week. Similarly, the analysis index corresponding to the monitoring device 140 is an average value of the total analysis indexes of the week, and the analysis index corresponding to the monitoring device 150 is an average value of the total analysis indexes of the week. Further, the monitoring platform 110 determines whether to activate a model updating task according to the decision values of the hidden danger identification models of the power transmission line corresponding to the plurality of monitoring devices. For example, the monitoring platform 110 inputs the analysis index corresponding to the monitoring device 150 into a pre-constructed decision tree model, and activates the update task of the hidden danger identification model of the power transmission line corresponding to the monitoring device 150 if the decision value of the hidden danger identification model of the power transmission line corresponding to the monitoring device 150 is Yes. That is to say, the monitoring platform 110 analyzes the analysis indexes of the hidden danger identification model of the power transmission line corresponding to each of the plurality of monitoring devices, inputs the analysis indexes into the decision tree model, and determines whether to activate the model updating task, and the application is not particularly limited as to the rule for determining whether to activate the model updating task.
Based on the same idea, some embodiments of the present application further provide a device corresponding to the above method.
Fig. 5 is a structural block diagram of a device for evaluating a hidden danger identification model of a power transmission line according to an embodiment of the present application. The device 500 for evaluating the hidden danger identification model of the power transmission line at least comprises an obtaining module 510, an analyzing module 520 and an activating module 530.
An obtaining module 510, configured to obtain an analysis index of a power transmission line hidden danger identification model corresponding to each monitoring device in the plurality of power transmission line monitoring devices; the analysis index is related to the accuracy and/or the missing report rate and/or the false report rate of the monitoring device;
the analysis module 520 is configured to make a decision on an analysis index of the power transmission line hidden danger identification model corresponding to each of the plurality of power transmission line monitoring devices according to a pre-constructed decision tree model, and determine a decision value of an update task of the power transmission line hidden danger identification model corresponding to any one of the plurality of power transmission line monitoring devices;
and an activating module 530, configured to activate, according to the decision value, an update task of the power transmission line hidden danger identification model of the corresponding monitoring device.
The embodiments in the present application are described in a progressive manner, and the same and similar parts among the embodiments can be referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, as for the apparatus embodiment, since it is substantially similar to the method embodiment, the description is relatively simple, and for the relevant points, reference may be made to the partial description of the method embodiment.
The device and the method provided by the embodiment of the application are in one-to-one correspondence, so the device also has the beneficial technical effects similar to the corresponding method, and the beneficial technical effects of the method are explained in detail above, so the beneficial technical effects of the equipment are not described again here.
The above description is only an example of the present application and is not intended to limit the present application. Various modifications and changes may occur to those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims (10)

1. A method for evaluating a hidden danger identification model of a power transmission line is characterized by comprising the following steps:
acquiring analysis indexes of the hidden danger identification models of the power transmission lines corresponding to the monitoring devices in the plurality of power transmission line monitoring devices; the analysis index is related to the accuracy and/or the missing report rate and/or the false report rate of the monitoring device;
according to a pre-constructed decision tree model, making a decision on analysis indexes of a power transmission line hidden danger identification model corresponding to each monitoring device in a plurality of power transmission line monitoring devices, and determining a decision value of a power transmission line hidden danger identification model updating task corresponding to any one monitoring device in the plurality of power transmission line monitoring devices;
and activating the hidden danger identification model updating task of the power transmission line of the corresponding monitoring device according to the decision value.
2. The method for evaluating the hidden danger identification model of the power transmission line according to claim 1, further comprising the following steps of:
discretizing continuous attribute values corresponding to all continuous attributes of a training sample set through a carry step length according to the training sample set to obtain corresponding discrete values; the continuous attribute is related to the accuracy and/or the missing report rate and/or the false report rate of the monitoring device for identifying the hidden danger of the power transmission line;
and respectively carrying out information gain calculation on the discrete values corresponding to the continuous attributes to obtain the decision tree model.
3. The method for evaluating the identification model of the hidden danger of the power transmission line according to claim 2, wherein the discretization of the continuous attribute values corresponding to the continuous attributes of the training sample set by a carry step length specifically comprises:
determining a discrete interval of each continuous attribute value according to the carry step value of each continuous attribute value;
and obtaining a discrete value corresponding to each continuous attribute value based on the discrete interval of each continuous attribute value.
4. The method for evaluating a model for identifying hidden dangers in a power transmission line according to claim 2, wherein the step of performing information gain calculation on the discrete values corresponding to the continuous attributes to obtain the decision tree model specifically comprises the following steps:
determining the information gain of each continuous attribute according to the information gain value of the discrete value corresponding to each continuous attribute;
taking the continuous attribute with the largest information gain in the continuous attributes as the optimal partition attribute of the training sample set;
determining the branches of the optimal partition attribute according to the optimal split point of the optimal partition attribute;
determining a sample subset of the training sample set corresponding to the branch based on the branch with the optimal partition attribute;
determining an optimal partition attribute corresponding to the sample subset of the training sample set based on the sample subset of the training sample set corresponding to the branch;
the process of recursively dividing a plurality of optimal division attributes of the training sample set from top to bottom until the nodes meet the growth stopping condition of the decision tree model;
and pruning the grown decision tree model to obtain the decision tree model.
5. The method for evaluating the identification model of the hidden danger of the power transmission line according to claim 4, wherein the condition that the growth stops until the nodes meet the growth stop condition of the decision tree model specifically comprises any one or more of the following steps:
the information gain of the discrete value corresponding to each continuous attribute is smaller than a preset threshold value;
the discrete value corresponding to each continuous attribute is not divisible;
the subset of training samples under each branch of the best partition attribute has the same classification.
6. The method for evaluating the identification model of the hidden danger of the power transmission line according to claim 1, wherein the analysis index is related to the accuracy and/or the missing report rate and/or the false report rate of the monitoring device, and specifically comprises the following steps:
the analysis index is the accuracy rate or the false alarm rate; or a α% + b β% + c γ%;
wherein, a is the accuracy, alpha% is the weight corresponding to the accuracy, b is the false alarm rate, beta% is the weight corresponding to the false alarm rate, c is the rate of missing report, and gamma% is the weight corresponding to the rate of missing report.
7. The method for evaluating the hidden danger identification model of the power transmission line according to claim 1, wherein the obtaining of the analysis indexes of the hidden danger identification model of the power transmission line corresponding to each monitoring device in the plurality of monitoring devices of the power transmission line comprises the following steps:
receiving corresponding power transmission line image information from the monitoring device; the image information comprises hidden danger types and hidden danger coordinates;
and determining an accurate alarm power transmission line hidden danger image and/or a misinformed power transmission line hidden danger image and/or a missed power transmission line hidden danger image in the power transmission line image according to the hidden danger type and the hidden danger coordinates in the power transmission line image.
8. The method for evaluating the identification model of the hidden danger of the power transmission line according to claim 7, wherein the determining of the missed-reported hidden danger image of the power transmission line image specifically comprises:
and performing cascade detection on the power transmission line images of the monitoring devices according to the offline high-precision model, comparing the cascade detection result with the analysis result of the corresponding power transmission line image of each monitoring device, and determining the report missing image.
9. The method for evaluating the hidden danger identification model of the power transmission line according to claim 1, wherein activating the hidden danger identification model updating task of the corresponding monitoring device according to the decision value specifically comprises:
and in a preset period, activating the hidden danger identification model updating task of the corresponding monitoring device according to the decision value of the hidden danger identification model updating task of the power transmission line, the number of days exceeding a preset activation threshold value and the number of days exceeding a preset number of days.
10. The utility model provides a transmission line hidden danger discernment model evaluation device which characterized in that includes:
the acquisition module is used for acquiring analysis indexes of the hidden danger identification model of the power transmission line corresponding to each monitoring device in the plurality of power transmission line monitoring devices; the analysis index is related to the accuracy and/or the missing report rate and/or the false report rate of the monitoring device;
the analysis module is used for deciding analysis indexes of the hidden danger identification models of the power transmission lines corresponding to the monitoring devices in the plurality of power transmission line monitoring devices according to a pre-constructed decision tree model and determining a decision value of an update task of the hidden danger identification model of the power transmission line corresponding to any one of the plurality of power transmission line monitoring devices;
and the activation module activates the power transmission line hidden danger identification model updating task of the corresponding monitoring device according to the decision value.
CN202010952891.6A 2020-09-11 2020-09-11 Power transmission line hidden danger identification model evaluation method and device Active CN112199805B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010952891.6A CN112199805B (en) 2020-09-11 2020-09-11 Power transmission line hidden danger identification model evaluation method and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010952891.6A CN112199805B (en) 2020-09-11 2020-09-11 Power transmission line hidden danger identification model evaluation method and device

Publications (2)

Publication Number Publication Date
CN112199805A true CN112199805A (en) 2021-01-08
CN112199805B CN112199805B (en) 2022-11-15

Family

ID=74014737

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010952891.6A Active CN112199805B (en) 2020-09-11 2020-09-11 Power transmission line hidden danger identification model evaluation method and device

Country Status (1)

Country Link
CN (1) CN112199805B (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113673634A (en) * 2021-10-22 2021-11-19 智洋创新科技股份有限公司 Method for confirming false alarm of hidden danger of power transmission line channel crane
CN115376074A (en) * 2022-10-25 2022-11-22 济南信通达电气科技有限公司 Method and system for evaluating recognition effect of power transmission line monitoring device

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2011028519A (en) * 2009-07-24 2011-02-10 Nippon Hoso Kyokai <Nhk> Data classification device and program
CN110490121A (en) * 2019-08-14 2019-11-22 北京国网信通埃森哲信息技术有限公司 A kind of transmission line of electricity external force damage prevention method and system for clapping image recognition based on more prisons
CN111583196A (en) * 2020-04-22 2020-08-25 北京智芯微电子科技有限公司 Monitoring system and monitoring method for power transmission line

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2011028519A (en) * 2009-07-24 2011-02-10 Nippon Hoso Kyokai <Nhk> Data classification device and program
CN110490121A (en) * 2019-08-14 2019-11-22 北京国网信通埃森哲信息技术有限公司 A kind of transmission line of electricity external force damage prevention method and system for clapping image recognition based on more prisons
CN111583196A (en) * 2020-04-22 2020-08-25 北京智芯微电子科技有限公司 Monitoring system and monitoring method for power transmission line

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113673634A (en) * 2021-10-22 2021-11-19 智洋创新科技股份有限公司 Method for confirming false alarm of hidden danger of power transmission line channel crane
CN113673634B (en) * 2021-10-22 2022-01-21 智洋创新科技股份有限公司 Method for confirming false alarm of hidden danger of power transmission line channel crane
CN115376074A (en) * 2022-10-25 2022-11-22 济南信通达电气科技有限公司 Method and system for evaluating recognition effect of power transmission line monitoring device

Also Published As

Publication number Publication date
CN112199805B (en) 2022-11-15

Similar Documents

Publication Publication Date Title
CN111158977B (en) Abnormal event root cause positioning method and device
CN106778259B (en) Abnormal behavior discovery method and system based on big data machine learning
CN109189736B (en) Method and device for generating alarm association rule
CN112199805B (en) Power transmission line hidden danger identification model evaluation method and device
CN103699698A (en) Method and system for track traffic failure recognition based on improved Bayesian algorithm
CN110781914A (en) Equipment fault monitoring processing method, device, equipment and storage medium
CN109753499A (en) A kind of O&amp;M monitoring data administering method
CN112579789A (en) Equipment fault diagnosis method and device and equipment
CN109448154A (en) A kind of transmission line of electricity personnel method for inspecting and device
CN112488371B (en) Pipeline intelligent early warning method and system based on big data
CN111078512A (en) Alarm record generation method and device, alarm equipment and storage medium
CN112085869A (en) Civil aircraft flight safety analysis method based on flight parameter data
CN107111609A (en) Lexical analyzer for neural language performance identifying system
CN110489142B (en) Evaluation method and device for equipment software upgrading, storage medium and terminal
CN115102836A (en) Network equipment fault analysis method and device and storage medium
CN111860118A (en) Human behavior analysis method based on artificial intelligence
CN114090393A (en) Method, device and equipment for determining alarm level
CN112906722A (en) Data anomaly detection method, device and equipment
CN116545867A (en) Method and device for monitoring abnormal performance index of network element of communication network
CN110929800B (en) Business body abnormal electricity utilization detection method based on sax algorithm
CN109636943A (en) A kind of transmission line of electricity hand-held set method for inspecting and device
CN111027429A (en) Data preprocessing method and system for intelligent identification of electrical drawings
CN115001941B (en) Method and device for determining communication network management faults
CN111680572B (en) Dynamic judgment method and system for power grid operation scene
CN115514621B (en) Fault monitoring method, electronic device and storage medium

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant