Online monitoring method and system for power equipment based on embedded intelligent sensors

Abstract

The present invention discloses an online monitoring method and system for power equipment based on embedded intelligent sensors, which mainly include: collecting data information of each power equipment, dividing the data information, and obtaining several data groups; judging the fault status of each data group based on a prediction model, and marking the data group with faults with labels; analyzing the marked data group to obtain the fault information of the power equipment. The present invention uses embedded intelligent sensors to collect and pre-process data information of terminal power equipment, and uses edge computing and distributed computing collaborative operation technology to calculate terminal data, diagnose potential safety hazards and the service life of power equipment, reduce the failure rate of power equipment, and realize reliable full life cycle prediction. By placing computing, storage, and application resources on the edge side of the network to reduce transmission delay and bandwidth consumption, perceptible services can be provided based on real-time network information.

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment failure, and in particular to an online monitoring method and system for power equipment based on embedded intelligent sensors. Background Art

[0002] With the continuous development of China's ultra-high voltage AC and DC systems, the stability and reliability of the power grid are becoming increasingly important. Consequently, the power grid is placing increasingly stringent requirements on the safety and reliability of power equipment. As a crucial component of power equipment safety and reliability, the reliability of power equipment itself should be given due attention. The power system is an electricity production and consumption system consisting of power generation, transmission, transformation, distribution, and consumption. It converts natural primary energy into electricity through power generation devices, and then supplies electricity to users through transmission, transformation, and distribution. It primarily includes two categories: power generation equipment and power supply equipment. Power generation equipment primarily includes power station boilers, steam turbines, gas turbines, hydro turbines, generators, point motors, transformers, and other equipment. Power supply equipment primarily includes transmission lines, transformers, contactors, and other components of various voltage levels.

[0003] Existing power equipment is monitored manually at irregular intervals, which not only requires a lot of labor but also greatly increases the monitoring cost. Due to the large human error, the monitoring efficiency is greatly reduced, which cannot meet the detection requirements of power equipment. The use is limited, or a lot of time and energy is often spent on finding the fault points of power equipment. Fault points in many remote areas are even more difficult to find, resulting in the inability to timely inspect and maintain the fault points, affecting the power quality of users. Summary of the Invention

[0004] (1) Purpose of the invention

[0005] In view of the above problems, the purpose of the present invention is to propose an online monitoring method and system for power equipment based on embedded intelligent sensors, which can realize real-time online monitoring of power equipment patrols, reduce the number of patrols, and also realize minute-level patrols through the monitoring system, thereby reducing the impact of power equipment failures and quickly finding the fault point, reducing the impact of power equipment failures on users. The present invention discloses the following technical solutions.

[0006] (2) Technical solution

[0007] As a first aspect of the present invention, the present invention discloses an online monitoring method for power equipment based on an embedded intelligent sensor, comprising:

[0008] Collecting data information of each power device and dividing the data information to obtain several data groups;

[0009] Based on the prediction model, determine the fault status of each data group and label the data group with faults;

[0010] The data group marked with the tag is analyzed to obtain the fault information of the power equipment.

[0011] In a possible implementation manner, the obtaining of fault information of the power equipment further includes:

[0012] The fault location is determined according to the fault information of the electric power equipment.

[0013] In a possible implementation, dividing the data information to obtain a plurality of data groups specifically includes:

[0014] Using artificial intelligence attribute feature analysis algorithms to label the data information with descriptive labels;

[0015] Segmentation and clustering based on descriptive labels;

[0016] According to the subdivision clustering, several data groups are obtained.

[0017] In a possible implementation, labeling the data group with faults specifically includes:

[0018] If there is a fault in the data group, marking the data group with a fault label;

[0019] If there is no fault in the data set, the information of the data set is stored.

[0020] In one possible implementation, based on a preset data threshold of each power device, the data value in the data group is compared with the data threshold. If it is greater than the threshold, the data value has a fault; if it is not less than the threshold, the data value has no fault.

[0021] As a second aspect of the present invention, the present invention further discloses an online monitoring system for power equipment based on an embedded intelligent sensor, comprising:

[0022] A data acquisition module, comprising a sensor and an intelligent control unit, wherein the sensor is used to collect data information of each power device, and the intelligent control unit is used to divide the data information to obtain several data groups;

[0023] An edge computing module, comprising a prediction unit and a marking unit, wherein the prediction unit determines the fault status of each data group based on a prediction model, and marks the data group with a fault label through the marking unit;

[0024] The cloud platform includes an analysis unit, which analyzes the data group marked with tags to obtain fault information of the power equipment.

[0025] In a possible implementation, the system further includes a positioning module, which determines a fault location based on the fault information of the power equipment.

[0026] In one possible implementation, the intelligent control unit includes a labeling subunit, a classification subunit, and a grouping subunit;

[0027] The labeling subunit uses an artificial intelligence attribute feature analysis algorithm to label the data information with a descriptive label;

[0028] The classification subunits are subdivided and clustered according to descriptive labels;

[0029] The grouping subunit obtains several data groups based on subdivision clustering.

[0030] In a possible implementation manner, the marking unit marks a fault label on a data group having a fault, specifically including:

[0031] If there is a fault in the data group, a fault label is marked on the data group; if there is no fault in the data group, information of the data group is stored.

[0032] In one possible implementation, the analysis unit includes a threshold database and a comparison subunit;

[0033] The threshold database is used to store preset data thresholds of various power devices;

[0034] The comparison subunit compares the data value in the data group with the data threshold based on the preset data threshold. If the data value is greater than the threshold, the data value has a fault; if the data value is not less than the threshold, the data value does not have a fault.

[0035] (3) Beneficial effects

[0036] The present invention discloses an online monitoring method and system for power equipment based on embedded intelligent sensors, which have the following beneficial effects: Embedded intelligent sensors are used to collect and preprocess data from terminal power equipment, and edge computing and distributed computing collaboratively operate to calculate terminal data, diagnose potential safety hazards and power equipment service life, reduce power equipment failure rates, and achieve reliable full life cycle prediction. By placing computing, storage, and application resources at the edge of the network, transmission delays and bandwidth consumption are reduced, and perceptible services can be provided based on real-time network information. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The embodiments described below with reference to the accompanying drawings are exemplary and are intended to explain and illustrate the present invention, but should not be construed as limiting the scope of protection of the present invention.

[0038] Figure 1 This is a schematic flow chart of the method for online monitoring of power equipment based on embedded intelligent sensors disclosed in the present invention;

[0039] Figure 2 It is a schematic diagram of the process of dividing data information disclosed in the present invention;

[0040] Figure 3 This is an architectural diagram of the edge computing technology disclosed in the present invention;

[0041] Figure 4 The present invention discloses an online monitoring system for electric power equipment based on embedded intelligent sensors.

[0042] Figure numerals: 500, data acquisition module; 510, sensor; 520, intelligent control unit; 521, labeling subunit; 522, classification subunit; 523, grouping subunit; 600, edge computing module; 610, prediction unit; 620, labeling unit; 700, cloud platform; 710, analysis unit; 711, comparison subunit; 720, threshold database; 730, resource library; 800, positioning module. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical solutions and advantages of the present invention more clear, the technical solutions in the embodiments of the present invention will be described in more detail below with reference to the accompanying drawings in the embodiments of the present invention.

[0044] It should be noted that: in the drawings, the same or similar reference numerals throughout represent the same or similar elements or elements with the same or similar functions. The described embodiments are part of the embodiments of the present invention, not all of them. In the absence of conflict, the embodiments and features in the embodiments of this application can be combined with each other. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0045] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like to indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying 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 orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the scope of protection of the present invention.

[0046] Reference below Figure 1-3 This article describes in detail the first embodiment of an online monitoring method for power equipment based on embedded intelligent sensors. This embodiment is primarily applicable to power equipment failures. By using embedded intelligent sensors to collect and preprocess data from terminal power equipment, and utilizing edge computing and distributed computing collaborative technologies to calculate terminal data, potential safety hazards and power equipment service life can be diagnosed, reducing power equipment failures and enabling reliable full lifecycle prediction. By placing computing, storage, and application resources at the edge of the network to reduce transmission delays and bandwidth consumption, perceptible services can be provided based on real-time network information.

[0047] like Figure 1 As shown, this embodiment mainly includes the following steps:

[0048] S100: Collect data information of each power device and divide the data information to obtain several data groups.

[0049] In step S100, smart sensors embedded in each grid terminal are used to collect data information from each power device connected to the grid, and the data information is divided according to certain attributes or rules, and data information with the same attributes or the same category are divided into one group to obtain several data groups with different attributes or different categories.

[0050] Furthermore, the data information includes operating parameters and equipment parameters, wherein the operating parameters include voltage, current, noise, etc. of the equipment operation, and the equipment parameters include the unique number of the equipment, the rated voltage, rated current, etc. of the equipment.

[0051] Furthermore, when collecting data information of each power device, one smart sensor can be used to collect data for one power device, or multiple power devices can be collected through one smart sensor. When one smart sensor collects data for multiple power devices, each power device is collected in turn, and the data information of the same device is marked with the device number and collection time.

[0052] Furthermore, power equipment mainly includes two categories: power generation equipment and power supply equipment. Among them, power generation equipment mainly includes power station boilers, steam turbines, gas turbines, water turbines, generators, point motors, transformers, etc., and power supply equipment mainly includes transmission lines of various voltage levels, mutual inductors, contactors, etc.

[0053] like Figure 2 As shown, in step S100, the data information is divided to obtain several data groups, which specifically includes the following steps:

[0054] S110. Using artificial intelligence attribute feature analysis algorithms, label the data information with descriptive labels;

[0055] S120, performing segmentation and clustering according to descriptive labels;

[0056] S130. Obtain several data groups based on subdivision and clustering.

[0057] The data information of the power equipment is obtained through the intelligent sensor, and the artificial intelligence attribute feature analysis algorithm is used to mark the data information with descriptive labels. For example, the generator in the power generation equipment is marked with descriptive labels, such as equipment attributes, equipment number, equipment location, etc. For example, the descriptive label content of the generator is: generator; NO02145 (the generator number); 8:30 (data collection time); Tianjin Power Plant (the location of the generator). According to the descriptive labels, segmentation and clustering are performed. Through segmentation and clustering, the collected data is divided into several data groups. Specifically: when an intelligent sensor collects multiple power When an intelligent sensor collects data from an electric power device, it is classified according to the type of data, such as classifying the generator's operating data into one category, and the equipment parameter data into another category, etc., so that several data groups can be obtained. The data in the data group has the same attributes or the same source, which is convenient for subsequent analysis and judgment of the data. The data can be grouped so that it can be processed in parallel, saving data processing time.

[0058] In one embodiment, an intelligent sensor includes an intelligent control module and a sensor. The intelligent control module includes a knowledge base, an inference engine, a knowledge acquisition program, and a comprehensive database. The knowledge base stores the expert knowledge, experience values, and basic sensor parameters required for the operation of the intelligent sensor. The knowledge in the knowledge base serves as the basis for the inference engine to issue commands. The comprehensive data module stores the raw data of the intelligent sensor, various commonly used data, and various parameters. The inference engine uses the knowledge in the knowledge base to think, judge, and reason based on the data from the sensor and the comprehensive data, and then makes a judgment. The knowledge base stores artificial intelligence attribute feature analysis algorithms and other technologies. The comprehensive data module stores data information collected from various power equipment. The inference engine integrates the data information in the data module, selects an algorithm from the knowledge base through the knowledge acquisition program, and uses this algorithm to process the data information, ultimately obtaining multiple data sets.

[0059] S200: Based on the prediction model, determine the fault status of each data group and label the data group with faults;

[0060] like Figure 3 As shown, in step S200, edge computing is used to configure the data of each data group with the prediction model to obtain whether there is fault data in the data group. If there is fault data, the data group is marked; if there is no fault data, the data group is stored for later tracing.

[0061] In step S200, the data of each data group is configured with the prediction model, specifically including:

[0062] The data group is represented as a data entity combination x = {..., l′2, l′1, m′1, m′2, h′1, h′2, ...} according to the time sequence of data collection, where m = {m′1, m′2, ...} represents the current data entity of the data group according to the time sequence, l′ = {..., l′2, l′1} and h′ = {h′1, h′2, ...} respectively represent the previous and subsequent adjacent data entities of the current data entity of the data group according to the time sequence.

[0063] Then, the BiLSTM network is used to extract features from the data set to obtain the entity feature vector of the data set. The data entity combination x = {..., l'2, l'1, m'1, m'2, h'1, h'2, ...} of the data set is input into the BiLSTM network, and the entity feature vector of the data set can be obtained, which is expressed as X m ,Right now

[0064] X m =f BiLSTM (..., l′2, l′1, m′1, m′2, h′1, h′2,...)

[0065] where f BiLSTM (*) indicates that the data entity combination x = {..., l′2, l′1, m′1, m′2, h′1, h′2, ...} is input into BiLSTM for feature extraction.

[0066] Furthermore, a prototype network prediction model is adopted. After the prototype network prediction model is trained using the support set, the fault state category is obtained according to the entity feature vector of the data group.

[0067] The prototype network prediction model is an important model in the field of artificial intelligence. After training with support set samples, feature vectors of the same category have greater similarity in features after being processed by the model, and can achieve category prediction for the input feature vector. The support set of the prototype network prediction model is expressed as:

[0068] S={(x1, y1), (x2, y2),...(x N ,yN)}

[0069] where x1, x2...x N Represents the entity features of the sample data group in the support set, y1, y2...y N Represents the fault status category corresponding to each support set sample data group. Assuming there are K fault status categories, each fault status category is represented by k, then k∈{1, 2..., K}, then for each fault status category k, the support set S can be set k , support set S k If the fault state categories of the sample data groups in are all k, then the category prototype of each category k is calculated through the support set:

[0070]

[0071] where f θ (*) represents the data entity feature vector obtained by extracting the entity features of the sample data group. In this invention, it is the entity feature vector of the data group extracted by the BiLSTM network. k The category prototype represents the average representation of the feature vectors of the data entities in the kth category.

[0072] Furthermore, the prototype network can calculate the entity feature vector X of the input data set m The distribution of each fault status category k relative to the K fault status categories:

[0073]

[0074] Where k′ represents the other categories in the K categories that do not belong to category k. Therefore, the prototype network prediction model can be used to predict the feature vector X m With respect to the distribution of each category k among the K categories, a fault state category of the data set is determined.

[0075] Furthermore, the edge computing node stores a local distribution network model, deploys fault handling applications, and configures event processing rules. Upon receiving data, the edge computing node immediately performs on-site data processing, fault monitoring, and analysis, and sends the processed information to the cloud platform. In this mode, fault analysis and alarm tasks can be completed within milliseconds. Based on edge computing technology, the grouped data sets are input into the prediction model, and fault data is obtained through data processing. For example, if the generator service life in the prediction model is 30 years, the generator number data and acquisition time are compared with the initial number data and the time when the initial number data was established in the prediction model. If it is found that the generator has been in operation for 29 years, then the data set is judged to have a fault. If a data set is found to have a fault, the processing of other data in the group is stopped, and the data set is directly judged to have a fault. Through the above method, faults can be quickly determined, improving monitoring efficiency.

[0076] Furthermore, the data group with faults is marked with a fault label, which can be divided into two types: faults that have occurred and faults that are about to occur. For example, the fault of the operating data can be marked as a fault that has occurred, and the life fault of the equipment can be marked as a fault that is about to occur.

[0077] S300: Analyze the data group marked with tags to obtain fault information of the power equipment.

[0078] In step S300, based on the preset data threshold of each power device, the data value in the data group is compared with the data threshold. If it is greater than the threshold, the data value has a fault; if it is not less than the threshold, the data value has no fault.

[0079] Furthermore, in the cloud platform, the basic parameters and rated operating parameters of each power device are stored in the resource library, and the basic parameter threshold and rated operating parameter threshold of each power device are preset in the database, wherein the preset threshold is less than the parameter value, and the preset threshold is the same as the data in the prediction model. When the cloud platform receives a data group marked with a fault, it first judges the marking label of the data group, first processes the data group labeled as having a fault, compares the data values, basic parameters / rated parameters and preset basic parameter thresholds / rated parameter thresholds in the data group, filters out data values ​​that are simultaneously greater than the preset data threshold and parameter value, and obtains fault information based on the data value. After processing all data values ​​in the data group that are simultaneously greater than the preset data threshold and parameter value, it filters out data values ​​that are greater than the preset data threshold but less than the parameter value, and obtains fault information based on the data value, and then sorts the acquired fault information, so that the most serious fault can be intuitively obtained.

[0080] Furthermore, the fault information includes the device number where the fault occurred, the specific fault content and the fault time, etc.

[0081] In step S300, fault information of the power equipment is obtained, and then the following steps are further included:

[0082] S400: Determine the fault location based on the fault information of the power equipment.

[0083] In step S400, based on the device number and fault time in the fault information, combined with the device number in the resource library, the device installation location can be obtained, thereby determining the location of the fault. The fault point can be inspected and maintained in time, reducing the steps and energy required to find the fault and reducing the impact of the fault on the user's electricity consumption.

[0084] Reference below Figure 4Detailed description, based on the same inventive concept, an embodiment of the present invention also provides a first embodiment of an online monitoring system for electric power equipment based on embedded intelligent sensors. Since the principle of the problem solved by this method is similar to the aforementioned online monitoring method for electric power equipment based on embedded intelligent sensors, the problem solved by this method can be referred to the aforementioned, and the repeated parts will not be repeated. This embodiment is mainly used for electric power equipment failures. It uses embedded intelligent sensors to collect and pre-process data information of terminal electric power equipment, and uses edge computing and distributed computing collaborative operation technology to calculate terminal data, diagnose potential safety hazards and the service life of electric power equipment, reduce electric power equipment failures, and achieve reliable full life cycle prediction. By placing computing, storage, and application resources on the edge side of the network to reduce transmission delay and bandwidth consumption, perceptible services can be provided based on real-time network information.

[0085] like Figure 4 As shown, this embodiment mainly includes: a data acquisition module 500, an edge computing module 600 and a cloud platform 700.

[0086] The data acquisition module 500 includes a sensor 510 and an intelligent control unit 520. The sensor 510 is used to collect data information of each power device, and the intelligent control unit 520 is used to divide the data information to obtain several data groups.

[0087] The edge computing module 600 includes a prediction unit 610 and a marking unit 620. The prediction unit 610 determines the fault status of each data group based on the prediction model, and marks the data group with faults with a fault label through the marking unit 620.

[0088] The cloud platform 700 includes an analysis unit 710 and a threshold database 720. The analysis unit 710 analyzes the data group with the tag to obtain the fault information of the power equipment; the threshold database 720 is used to store the preset data threshold of each power equipment.

[0089] Smart sensors embedded in each power grid terminal are used to collect data information from each power equipment connected to the power grid, and the data information is divided according to certain attributes or rules. Data information with the same attributes or categories are grouped together to obtain several data groups with different attributes or categories.

[0090] Furthermore, the data information includes operating parameters and equipment parameters, wherein the operating parameters include voltage, current, noise, etc. of the equipment operation, and the equipment parameters include the unique number of the equipment, the rated voltage, rated current, etc. of the equipment.

[0091] Furthermore, when collecting data information of each power device, one smart sensor can be used to collect data for one power device, or multiple power devices can be collected through one smart sensor. When one smart sensor collects data for multiple power devices, each power device is collected in turn, and the data information of the same device is marked with the device number and collection time.

[0092] Furthermore, power equipment mainly includes two categories: power generation equipment and power supply equipment. Among them, power generation equipment mainly includes power station boilers, steam turbines, gas turbines, water turbines, generators, point motors, transformers, etc., and power supply equipment mainly includes transmission lines of various voltage levels, mutual inductors, contactors, etc.

[0093] In one embodiment, the intelligent control unit 520 includes a labeling subunit 521, a classification subunit 522 and a grouping subunit 523, wherein the labeling subunit 521 uses an artificial intelligence attribute feature analysis algorithm to label the data information with descriptive labels; the classification subunit 522 performs subdivision clustering based on the descriptive labels; and the grouping subunit 523 obtains several data groups based on the subdivision clustering.

[0094] The data information of the power equipment is obtained through the intelligent sensor, and the artificial intelligence attribute feature analysis algorithm is used to mark the data information with descriptive labels. For example, the generator in the power generation equipment is marked with descriptive labels, such as equipment attributes, equipment number, equipment location, etc. For example, the descriptive label content of the generator is: generator; NO02145 (the generator number); 8:30 (data collection time); Tianjin Power Plant (the location of the generator). According to the descriptive labels, segmentation and clustering are performed. Through segmentation and clustering, the collected data is divided into several data groups. Specifically: when an intelligent sensor collects multiple power When an intelligent sensor collects data from an electric power device, it is classified according to the type of data, such as classifying the generator's operating data into one category, and the equipment parameter data into another category, etc., so that several data groups can be obtained. The data in the data group has the same attributes or the same source, which is convenient for subsequent analysis and judgment of the data. The data can be grouped so that it can be processed in parallel, saving data processing time.

[0095] In one embodiment, the smart sensor includes an intelligent control module and a sensor 510. The intelligent control module includes a knowledge base, an inference engine, a knowledge acquisition program, and a comprehensive database. The knowledge base stores the expert knowledge, experience values, and basic parameters of the sensor 510 required for the operation of the smart sensor. The knowledge in the knowledge base serves as the basis for the inference engine to issue commands. The comprehensive data module stores the raw data of the smart sensor, various commonly used data, and various parameters. The inference engine uses the knowledge in the knowledge base to think, judge, and reason based on the data from the sensor 510 and the comprehensive data, and then makes a judgment. The knowledge base stores artificial intelligence attribute feature analysis algorithms and other technologies. The comprehensive data module stores data information collected from various power equipment. The inference engine integrates the data information in the data module, selects an algorithm from the knowledge base through the knowledge acquisition program, and uses this algorithm to process the data information, ultimately obtaining multiple data sets.

[0096] In one embodiment, edge computing is used to configure the data of each data group with a prediction model to determine whether there is fault data in the data group. If there is fault data, the data group is marked; if there is no fault data, the data group is stored for later tracing.

[0097] Furthermore, the data of each data group is configured with the prediction model, specifically including:

[0098] The data group is represented as a data entity combination x = {..., l′2, l′1, m′1, m′2, h′1, h′2, ...} according to the time sequence of data collection, where m = {m′1, m′2, ...} represents the current data entity of the data group according to the time sequence, l′ = {..., l′2, l′1} and h′ = {h′1, h′2, ...} respectively represent the previous and subsequent adjacent data entities of the current data entity of the data group according to the time sequence.

[0099] Then, the BiLSTM network is used to extract features from the data set to obtain the entity feature vector of the data set. The data entity combination x = {..., l'2, l'1, m'1, m'2, h'1, h'2, ...} of the data set is input into the BiLSTM network, and the entity feature vector of the data set can be obtained, which is expressed as X m ,Right now

[0100] X m =f BiLSTM (..., l′2, l′1, m′1, m′2, h′1, h′2,...)

[0101] where f BiLSTM(*) indicates that the data entity combination x = {..., l′2, l′1, m′1, m′2, h′1, h′2, ...} is input into BiLSTM for feature extraction.

[0102] Furthermore, a prototype network prediction model is adopted. After the prototype network prediction model is trained using the support set, the fault state category is obtained according to the entity feature vector of the data group.

[0103] The prototype network prediction model is an important model in the field of artificial intelligence. After training with support set samples, feature vectors of the same category have greater similarity in features after being processed by the model, and can achieve category prediction for the input feature vector. The support set of the prototype network prediction model is expressed as:

[0104] S={(x1, y1), (x2, y2),...(x N ,y N )}

[0105] where x1, x2...x N Represents the entity features of the sample data group in the support set, y1, y2...y N Represents the fault status category corresponding to each support set sample data group. Assuming there are K fault status categories, each fault status category is represented by k, then k∈{1, 2..., K}, then for each fault status category k, the support set S can be set k , support set S k If the fault state categories of the sample data groups in are all k, then the category prototype of each category k is calculated through the support set:

[0106]

[0107] where f θ (*) represents the data entity feature vector obtained by extracting the entity features of the sample data group. In this invention, it is the entity feature vector of the data group extracted by the BiLSTM network. k The category prototype represents the average representation of the feature vectors of the data entities in the kth category.

[0108] Furthermore, the prototype network can calculate the entity feature vector X of the input data set m The distribution of each fault status category k relative to the K fault status categories:

[0109]

[0110] Where k′ represents the other categories in the K categories that do not belong to category k. Therefore, the prototype network prediction model can be used to predict the feature vector X mWith respect to the distribution of each category k among the K categories, a fault state category of the data set is determined.

[0111] Furthermore, the edge computing node stores a local distribution network model, deploys a fault handling application, and configures event handling rules. Upon receiving data, the edge computing node immediately performs on-site data processing, fault monitoring, and analysis, and sends the processed information to the cloud platform 700. In this mode, fault analysis and alarm tasks can be completed within milliseconds. Based on edge computing technology, the grouped data sets are input into the prediction model, and fault data is obtained through data processing. For example, if the service life of a generator in the prediction model is 30 years, the generator number data and acquisition time are compared with the initial number data and the time when the initial number data was established in the prediction model. If it is found that the generator has been in operation for 29 years, then the data set is judged to have a fault. If a data set in the data set is found to have a fault, the processing of other data in the group is stopped, and the data set is directly judged to have a fault. In this way, faults can be quickly determined, improving monitoring efficiency.

[0112] Furthermore, the marking unit 620 marks the data group with a fault with a fault label. The label can be divided into two types: a fault has occurred and a fault is about to occur. For example, a fault in the operating data can be marked as a fault has occurred, and a life fault of the equipment can be marked as a fault is about to occur.

[0113] In one embodiment, the analysis unit 710 includes a comparison subunit 711, which compares the data value in the data group with the data threshold based on a preset data threshold. If the data value is greater than the threshold, the data value has a fault; if the data value is not less than the threshold, the data value does not have a fault.

[0114] Based on the preset data threshold of each power device, the data value in the data group is compared with the data threshold. If it is greater than the threshold, the data value has a fault; if it is not less than the threshold, the data value has no fault.

[0115] Furthermore, in the cloud platform 700, the basic parameters and rated operating parameters of each power device are stored in the resource library 730, and the basic parameter threshold and rated operating parameter threshold of each power device are preset in the threshold database 720, wherein the preset threshold is less than the parameter value, and the preset threshold is the same as the data in the prediction model. When the cloud platform 700 receives a data group marked with a fault, it first judges the marking label of the data group, first processes the data group marked as having a fault, compares each data value, basic parameter / rated parameter and preset basic parameter threshold / rated parameter threshold in the data group, filters out data values ​​that are simultaneously greater than the preset data threshold and parameter value, and obtains fault information based on the data value. After processing all data values ​​in the data group that are simultaneously greater than the preset data threshold and parameter value, it filters out data values ​​that are greater than the preset data threshold but less than the parameter value, and obtains fault information based on the data value, and then sorts the acquired fault information, so that the most serious fault can be intuitively obtained.

[0116] Furthermore, the fault information includes the device number where the fault occurred, the specific fault content and the fault time, etc.

[0117] In one embodiment, the system further includes a positioning module 800 , which determines the fault location based on the fault information of the power equipment.

[0118] According to the equipment number and fault time in the fault information, combined with the equipment number in the resource library 730, the equipment installation location can be obtained, thereby determining the location of the fault. The fault point can be inspected and maintained in time, reducing the steps and energy required to find the fault and reducing the impact of the fault on the user's electricity consumption.

[0119] In one embodiment, the system further includes a network unit, which can be a dedicated power network (fiber), a dedicated IoT network, the Internet (carrier broadband), or a mobile Internet (4G). The edge computing node is connected to the cloud platform via a network adapter via the network layer.

[0120] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A method for online monitoring of power equipment based on embedded intelligent sensors, characterized in that: include: Collecting data information of each power device and dividing the data information to obtain several data groups, dividing the data information to obtain several data groups, specifically including: Using artificial intelligence attribute feature analysis algorithms to label the data information with descriptive labels; Segmentation and clustering based on descriptive labels; According to the subdivision clustering, several data groups are obtained; Based on the prediction model, the fault status of each data group is determined, and the data group with faults is labeled, and the data of each data group is configured with the prediction model, specifically including: Represent the data group as a combination of data entities according to the time sequence of data collection ,in, Indicates the current data entity of the data group in time sequence, and Respectively represent the preceding and succeeding adjacent data entities of the current data entity of the data group in time sequence; Then, the BiLSTM network is used to extract features from the data group, obtain the entity feature vector of the data group, and combine the data entities of the data group By inputting the BiLSTM network as input, the entity feature vector of the data group can be obtained, which is expressed as ,Right now in Indicates combining data entities Input BiLSTM for feature extraction; Then, a prototype network prediction model is adopted. After the prototype network prediction model is trained using the support set, the fault state category is obtained according to the entity feature vector of the data group. The prototype network prediction model is an important model in the field of artificial intelligence. After training with support set samples, feature vectors of the same category have greater similarity in features after being processed by the model. It can realize category prediction for the input feature vector. The support set of the prototype network prediction model is expressed as: in Represents the entity features of the sample data group in the support set, Represents the fault status category corresponding to each support set sample data group. Assuming there are K fault status categories, each fault status category is represented by k, then , then for each fault state category k, the support set can be set , support set If the fault state categories of the sample data groups in are all k, then the category prototype of each category k is calculated through the support set: in Represents the data entity feature vector obtained by extracting the entity features of the sample data group, that is, the entity feature vector of the data group extracted by the BiLSTM network. As the category prototype, it represents the average representation of the feature vectors of the data entities in the kth category; Furthermore, the prototype network can calculate the entity feature vector of the input data set The distribution of each fault status category k relative to the K fault status categories: in Represents other categories in K categories that do not belong to category k, so the prototype network prediction model can feature vector determining a fault state category of the data set with respect to a distribution for each category k of the K categories; The data group marked with the tag is analyzed to obtain the fault information of the power equipment.

2. The method for online monitoring of electric power equipment based on embedded intelligent sensors according to claim 1, characterized in that: After obtaining the fault information of the electric power equipment, the method further includes: The fault location is determined according to the fault information of the electric power equipment.

3. The method for online monitoring of electric power equipment based on embedded intelligent sensors according to claim 1, characterized in that: The step of labeling the data group with faults specifically includes: If there is a fault in the data group, marking the data group with a fault label; If there is no fault in the data set, the information of the data set is stored.

4. The method for online monitoring of electric power equipment based on embedded intelligent sensors according to claim 1, characterized in that: The analyzing of the data group of the marked tags to obtain the fault information of the power equipment specifically includes: Based on the preset data threshold of each power device, the data value in the data group is compared with the data threshold. If it is greater than the threshold, the data value has a fault; if it is not less than the threshold, the data value has no fault.

5. An online monitoring system for power equipment based on embedded intelligent sensors, characterized in that: include: A data acquisition module, comprising a sensor and an intelligent control unit. The sensor is used to collect data information of each power device. The intelligent control unit is used to divide the data information to obtain several data groups. The intelligent control unit includes a label subunit, a classification subunit, and a grouping subunit. The labeling subunit uses an artificial intelligence attribute feature analysis algorithm to label the data information with a descriptive label; The classification subunits are subdivided and clustered according to descriptive labels; The grouping subunit obtains a number of data groups based on subdivision clustering; The edge computing module includes a prediction unit and a marking unit. The prediction unit determines the fault status of each data group based on the prediction model, and marks the data group with a fault label through the marking unit. The data of each data group is configured with the prediction model, specifically including: Represent the data group as a combination of data entities according to the time sequence of data collection ,in, Indicates the current data entity of the data group in time sequence, and Respectively represent the preceding and succeeding adjacent data entities of the current data entity of the data group in time sequence; Then, the BiLSTM network is used to extract features from the data group, obtain the entity feature vector of the data group, and combine the data entities of the data group By inputting the BiLSTM network as input, the entity feature vector of the data group can be obtained, which is expressed as ,Right now in Indicates combining data entities Input BiLSTM for feature extraction; Then, a prototype network prediction model is adopted. After the prototype network prediction model is trained using the support set, the fault state category is obtained according to the entity feature vector of the data group. The prototype network prediction model is an important model in the field of artificial intelligence. After training with support set samples, feature vectors of the same category have greater similarity in features after being processed by the model. It can realize category prediction for the input feature vector. The support set of the prototype network prediction model is expressed as: in Represents the entity features of the sample data group in the support set, Represents the fault status category corresponding to each support set sample data group. Assuming there are K fault status categories, each fault status category is represented by k, then , then for each fault state category k, the support set can be set , support set If the fault state categories of the sample data groups in are all k, then the category prototype of each category k is calculated through the support set: in Represents the data entity feature vector obtained by extracting the entity features of the sample data group, that is, the entity feature vector of the data group extracted by the BiLSTM network. As the category prototype, it represents the average representation of the feature vectors of the data entities in the kth category; Furthermore, the prototype network can calculate the entity feature vector of the input data set The distribution of each fault status category k relative to the K fault status categories: in Represents other categories in K categories that do not belong to category k, so the prototype network prediction model can feature vector determining a fault state category of the data set with respect to a distribution for each category k of the K categories; A cloud platform, comprising an analysis unit and a threshold database, wherein the analysis unit analyzes the data group marked with the tag to obtain fault information of the power equipment; The threshold database is used to store preset data thresholds of various power devices.

6. The online monitoring system for power equipment based on embedded intelligent sensors according to claim 5 is characterized in that: The system further includes a positioning module, which determines a fault location based on the fault information of the electric equipment.

7. The online monitoring system for power equipment based on embedded intelligent sensors according to claim 5 is characterized in that: The marking unit marks a fault label on a data group with a fault, specifically including: If there is a fault in the data group, marking the data group with a fault label; If there is no fault in the data set, the information of the data set is stored.

8. The online monitoring system for power equipment based on embedded intelligent sensors according to claim 5 is characterized in that: The analysis unit includes a comparison subunit; The comparison subunit compares the data value in the data group with the data threshold based on the preset data threshold. If the data value is greater than the threshold, the data value has a fault; if the data value is not less than the threshold, the data value does not have a fault.

Images