An online early warning system and method for intelligent electric energy meter faults

Abstract

The present invention relates to the technical field of smart electricity meters, and discloses an online early warning system and method for smart electricity meter faults, including a connection module, a transmission module, a metering detection module, a data hierarchical module, an autonomous processing module, a matching module, and an online early warning module. The system can obtain the power environment information, fault information, and operation quality information of the smart electricity meter in real time, carry out long-term operation quality monitoring of the equipment, and perform hierarchical division of the metering data; construct a deep learning network with self-evolution capability, introduce disturbance factors and matching algorithms, and realize online early warning of smart electricity meter faults and self-evolution of early warning technology. The present invention solves the problems in the prior art of the smart electricity meter metering anomaly monitoring method, such as relatively simple functions, large amount of calculation data and complex process, low degree of intelligence, inaccurate early warning results caused by disturbances of factors that cannot be simulated and emulated under complex working conditions, untimely fault warnings, and inability to realize online early warning and intelligent evolution.

Description

Technical Field

[0001] The present invention relates to the technical field of smart electric energy meters, and in particular to an online early warning system and method for smart electric energy meter faults. Background Art

[0002] Smart energy meters are essential data acquisition devices for smart grids. They collect, measure, and transmit raw energy data, forming the foundation for information integration, analysis, optimization, and presentation. Smart energy meters are subject to a wide variety of fault types, requiring rapid maintenance by maintenance personnel. However, in practice, the inability of maintenance systems to quickly and accurately identify the specific fault type leads to delayed repairs. Therefore, rapid and accurate fault diagnosis for smart energy meters is crucial for improving maintenance efficiency.

[0003] Chinese patent application number: CN202210100621.1, publication date: June 3, 2022, discloses a method, apparatus, computer device, and storage medium for determining faults in a smart electricity meter. The method comprises: obtaining data curves measured by the smart electricity meter over a preset time period, the data curves comprising multiple cycles of first voltage curves and current curves corresponding to each first voltage curve. A preset number of first voltage curves are extracted from the data curves as target voltage curves, where the target voltage curves correspond to current curves with current amplitudes less than a preset value. Based on the target voltage curves, a target voltage measurement error value for the smart electricity meter is determined. This determines the voltage measurement error value caused by the fault in the smart electricity meter. This error value quantifies the voltage fault condition, thereby determining the fault in the electricity meter. This allows personnel to promptly repair or replace the smart electricity meter, preventing power outages caused by incorrect metering or meter failure.

[0004] However, in the process of implementing the technical solution of the above invention, it was found that the above technology has at least the following problems: the current smart electricity meter measurement anomaly monitoring method has relatively simple functions, a large amount of calculated data and a complex process, and a low degree of intelligence. Disturbances caused by factors that cannot be simulated and emulated under complex working conditions cause inaccurate early warning results, untimely fault warnings, and inability to achieve online early warning and intelligent evolution. Summary of the Invention

[0005] The present invention provides an online early warning system and method for smart electricity meter faults, which solves the problems in the existing technology of smart electricity meter measurement anomaly monitoring methods, such as relatively simple functions, large amount of calculation data and complex processes, low degree of intelligence, inaccurate early warning results caused by disturbances of factors that cannot be simulated and emulated under complex working conditions, untimely fault warnings, and inability to achieve online early warning and intelligent evolution. The present invention realizes online early warning of smart electricity meter faults and self-evolution of early warning technology.

[0006] The present invention specifically includes the following technical solutions:

[0007] An online early warning system for smart electric energy meter faults includes the following parts:

[0008] Connection module, transmission module, measurement and detection module, data layering module, autonomous processing module, matching module and online early warning module;

[0009] The connection module is used to connect the smart energy meter with the single-chip microcomputer and the transmission module, and to establish a number of sensor interfaces for data acquisition. The connection module is located on the smart energy meter and is connected to the transmission module and the metering detection module through data transmission.

[0010] The transmission module is used to receive the metering data collected by the smart electric energy meter and the sensor, and transmit it to the metering detection module. The transmission module is connected to the metering detection module through data transmission;

[0011] The metering detection module is used to perform periodic round-robin readings on the smart electric energy meter and the sensor to sense metering data. The metering detection module is located on the single-chip microcomputer and is connected to the data layering module through data transmission.

[0012] The data hierarchical module is used to hierarchically divide the metering data according to the transmission relationship between the metering data. The data hierarchical module is located on the single chip microcomputer and is connected to the autonomous processing module through data transmission;

[0013] The autonomous processing module is used to build a deep learning network with self-evolution capability. Through deep training, it learns from a complex combination of functions from layer to layer to find a mapping function that defines the mapping from input to output, optimizes the parameters of each layer, reduces errors, and ultimately outputs the level of abnormal data of the smart electricity meter and its abnormal parameters. The autonomous processing module is connected to the matching module via data transmission.

[0014] The matching module is used to match the corresponding fault type according to the level and abnormal parameters, and the matching module is connected to the online warning module through data transmission.

[0015] A method for online early warning of smart electric energy meter faults includes the following steps:

[0016] S1. Obtain real-time information on the power environment, faults, and operating quality of smart energy meters, conduct long-term equipment operating quality monitoring, and classify metering data into different levels;

[0017] S2. Build a deep learning network with self-evolution capabilities, introduce disturbance factors and matching algorithms, and realize online early warning of smart electricity meter failures and the self-evolution of early warning technology.

[0018] Furthermore, the step S1 specifically includes:

[0019] The metering data is divided into levels according to the transmission relationship between the metering data, and the stored standard metering data of the smart electricity meter is used as the reference value for calculating the correlation relationship. The reference value is in a dynamic adjustment process; the metering data directly read from the smart electricity meter belongs to the first level; for the metering data collected by the sensor, if the metering data is related to the smart electricity meter or related to the occurrence of a smart electricity meter failure, then these data belong to the second level; if the metering data is the parameters of the upstream and downstream devices of the smart electricity meter collected by the sensor, then these data belong to the third level, and the data of the third level needs to be related to the smart electricity meter through the data of the second level; other data belong to the fourth level.

[0020] Furthermore, the step S2 specifically includes:

[0021] A deep learning network with self-evolution capability is constructed, and the collected metering data is input into the deep learning network. Through deep training, the network finds the mapping function that defines the input to output from the complex combination of functions from layer to layer, optimizes the parameters of each layer, reduces the error, and finally outputs the abnormal level and abnormal parameters of the smart electricity meter data, so as to match the corresponding fault type according to the level and abnormal parameters and issue an early warning.

[0022] Furthermore, the step S2 specifically includes:

[0023] A deep learning network consists of an input layer, a mapping layer, a correlation layer, a perturbation layer, a recurrent layer, and an output layer.

[0024] Furthermore, the step S2 specifically includes:

[0025] The mapping layer maps the input metering data to the calculation space. The association layer calculates the correlation between data at different levels based on the divided data hierarchy. The perturbation layer calculates the perturbation factor according to the preset active correlation coefficient and passive correlation coefficient, perturbs the data, and optimizes the problem of inaccurate fault warning results of smart electricity meters caused by disturbances of factors that cannot be simulated and emulated under complex working conditions. The circulation layer selects the learning function in reinforcement learning based on the correlation of the metering data of smart electricity meters in time series.

[0026] Furthermore, the step S2 specifically includes:

[0027] Correlations include the correlation between the metering data collected by the sensor and the smart energy meter, as well as the correlation between the metering data collected by the sensor from upstream and downstream devices of the smart energy meter and the metering data collected by the sensor from the smart energy meter. Correlations between data at different levels are calculated using a combination of active and passive correlations. Active correlations are based on expert settings or common knowledge, while passive correlations are analyzed and calculated using the correlation analysis method constructed by this invention.

[0028] Furthermore, the step S2 specifically includes:

[0029] Taking action a strategy from the current loop state s to reach the next state will produce impact benefits during the execution of this action. The relationship function between the impact benefit and the action strategy is the learning function.

[0030] The present invention has at least the following technical effects or advantages:

[0031] 1. By obtaining the power environment information, fault information, and operation quality information of the smart electricity meter in real time, long-term operation quality monitoring of the equipment is carried out. The metering data is hierarchically divided according to the transmission relationship between the metering data, which facilitates the mining of the association between the data and achieves the retention of data relevance. The smart electricity meter and sensor are collected in proportion using the data relevance method, which will not generate too much duplicate data and reduce the calculation complexity and cost.

[0032] 2. A deep learning network with self-evolution capability was established. Using big data modeling and analysis as a means, a fault mapping model of smart electricity meter measurement data was fitted through the input layer, mapping layer, association layer, disturbance layer, loop layer and output layer. A reasonable arrangement was made according to the correlation between the measurement data of smart electricity meters. Disturbance factor matching and algorithms were introduced to optimize the problem of inaccurate warning results caused by disturbances of factors that cannot be simulated and emulated under complex working conditions, thereby realizing online warning of smart electricity meter faults and the self-evolution of warning technology.

[0033] 3. The technical solution of the present invention can effectively solve the problems in the existing technology of the abnormal measurement monitoring method of smart electricity meters, such as relatively simple functions, large amount of calculation data and complex process, low degree of intelligence, inaccurate warning results caused by factors that cannot be simulated and emulated under complex working conditions, untimely fault warning, and inability to realize online warning and intelligent evolution. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 This is a structural diagram of an online early warning system for smart electric energy meter faults according to the present invention;

[0035] Figure 2This is a flow chart of an online early warning method for smart electric energy meter faults according to the present invention;

[0036] Figure 3 This is a structural diagram of a deep learning network with self-evolution capability described in the present invention. DETAILED DESCRIPTION

[0037] The embodiments of the present application provide an online early warning system and method for smart electricity meter faults, which solves the problems in the prior art of the smart electricity meter measurement abnormality monitoring method, such as relatively simple functions, large amount of calculation data and complex process, low degree of intelligence, inaccurate early warning results caused by disturbances of factors that cannot be simulated and emulated under complex working conditions, untimely fault warnings, and inability to achieve online early warning and intelligent evolution.

[0038] The technical solution in the embodiments of the present application is to solve the above problems, and the overall idea is as follows:

[0039] By obtaining the power environment information, fault information, and operation quality information of smart electricity meters in real time, long-term equipment operation quality monitoring is carried out. The metering data is hierarchically divided according to the transmission relationship between the metering data, which facilitates the mining of the associations between the data and achieves the retention of data correlation. The smart electricity meters and sensors are collected in proportion using data correlation, which does not generate excessive duplicate data and reduces the computational complexity and cost. A deep learning network with self-evolution capability is established. Using big data modeling and analysis as a means, a fault mapping model of the metering data of smart electricity meters is fitted through the input layer, mapping layer, association layer, perturbation layer, loop layer and output layer. According to the correlation between the metering data of smart electricity meters, a reasonable arrangement is made, and disturbance factor matching and algorithm are introduced to optimize the problem of inaccurate warning results caused by disturbances of factors that cannot be simulated and emulated under complex working conditions, thereby realizing online warning of smart electricity meter faults and the self-evolution of warning technology.

[0040] In order to better understand the above technical solution, the above technical solution will be described in detail below with reference to the accompanying drawings and specific implementation methods.

[0041] Refer to the attached Figure 1 The present invention provides an online early warning system for smart electric energy meter faults, comprising the following parts:

[0042] Connection module 10, transmission module 20, measurement and detection module 30, data layering module 40, autonomous processing module 50, matching module 60 and online warning module 70;

[0043] The connection module 10 is used to connect the smart energy meter with the single-chip microcomputer and the transmission module 20, and to establish several sensor interfaces for data acquisition. The connection module 10 is located on the smart energy meter and is connected to the transmission module 20 and the metering detection module 30 through data transmission;

[0044] The transmission module 20 is used to receive the metering data collected by the smart energy meter and the sensor and transmit it to the metering detection module 30. The transmission module 20 is connected to the metering detection module 30 through data transmission;

[0045] The metering detection module 30 is used to periodically collect data from smart energy meters and sensors and sense metering data. The metering detection module 30 is located on the single-chip microcomputer and is connected to the data layering module 40 through data transmission.

[0046] The data layering module 40 is used to divide the metering data into different levels according to the transmission relationship between the metering data. The data layering module 40 is located on the single chip microcomputer and is connected to the autonomous processing module 50 through data transmission;

[0047] The autonomous processing module 50 is used to build a deep learning network with self-evolution capabilities. Through deep training, it learns from a complex combination of functions from layer to layer to find the mapping function that defines the input to output, optimizes the parameters of each layer, reduces errors, and ultimately outputs the level of abnormal data in the smart electricity meter and its abnormal parameters. The autonomous processing module 50 is connected to the matching module 60 via data transmission.

[0048] A matching module 60 is used to match the corresponding fault type according to the level and abnormal parameters. The matching module 60 is connected to the online warning module 70 through data transmission;

[0049] The online warning module 70 is used to issue a warning signal of the fault type of the smart energy meter.

[0050] Refer to the attached Figure 2 The present invention provides an online early warning method for smart electric energy meter faults, comprising the following steps:

[0051] S1. Obtain the power environment information, fault information, and operation quality information of the smart electricity meter in real time, carry out long-term operation quality monitoring of the equipment, and divide the metering data into levels.

[0052] The smart energy meter is equipped with a connection module 10. This module, connected to the microcontroller and the transmission module 20, has several sensor interfaces for data collection. The microcontroller receives metering data from sensors via the sensor interface and transmits this data to the metering detection module 30 via the transmission module 20. The metering detection module 30 on the microcontroller also receives metering data generated by the smart energy meter via the transmission module 20. The metering detection module 30 periodically reads the smart energy meter and its sensors. Metering data flows from the local smart energy meter and sensors to the metering detection module 30 on the remote microcontroller. The collected metering data includes power environment information, fault information, and operational quality information. By acquiring real-time power environment information, fault information, and operational quality information from the smart energy meter, long-term operational quality monitoring of the equipment can be performed.

[0053] Power environment information includes electrical power, current, voltage, active power, reactive power, apparent power, frequency, power factor, and electricity data; fault information includes the manufacturer, equipment type, service life, equipment status, fault time, and fault cause of the smart electricity meter; operation quality information includes the power supply of the smart electricity meter, actual power consumption, fixed loss, line loss, and the relationship between the measurement errors of each smart electricity meter.

[0054] The data stratification module 40 stratifies metering data based on the transfer relationships between them. It uses the stored standard metering data of smart meters as a reference value for calculating correlations, with the reference value undergoing a dynamic adjustment process. Metering data directly read from smart meters belongs to the first tier. Metering data collected by sensors, if relevant to the smart meter or related to a smart meter failure, belongs to the second tier. Metering data collected by sensors for parameters of upstream and downstream devices of the smart meter belongs to the third tier, and data in the third tier must be linked to the smart meter through data in the second tier. All other data belongs to the fourth tier.

[0055] The MCU processes the received metering data in layers and transmits it to the autonomous processing module 50 via the communication module. The autonomous processing module 50 establishes a self-evolving deep learning network. By introducing disturbance factors and matching algorithms, it optimizes the problem of inaccurate smart meter fault warning results caused by disturbances that cannot be simulated or emulated under complex working conditions. This enables online warning of smart meter faults and the self-evolution of warning technology.

[0056] The beneficial effects of step S1 are: by obtaining the power environment information, fault information, and operation quality information of the smart electricity meter in real time, long-term operation quality monitoring of the equipment is carried out, and the metering data is hierarchically divided according to the transmission relationship between the metering data, which is convenient for mining the association between the data and retaining the data correlation. The smart electricity meter and sensor are collected in proportion using the data correlation method, which will not generate too much duplicate data and reduce the calculation complexity and cost.

[0057] S2. Build a deep learning network with self-evolution capabilities, introduce disturbance factors and matching algorithms, and realize online early warning of smart electricity meter failures and the self-evolution of early warning technology.

[0058] A self-evolving deep learning network is constructed, consisting of an input layer, a mapping layer, an association layer, a perturbation layer, a recurrent layer, and an output layer. The collected metering data is fed into the deep learning network, which then learns through deep training to find the mapping function that defines the input to output from a complex combination of layer-by-layer functions. The network then optimizes the parameters of each layer, reducing errors. Ultimately, the network outputs the level and parameters of abnormal smart meter data, matching the level and parameters to the corresponding fault type and issuing an early warning.

[0059] By comparing the sampled data set with the standard measurement data, the reference value is adaptively adjusted and the initial parameters of the deep learning network are defined according to the reference value. 12 ,ω 23 ,ω 34 ,ω 45 ,ω 56}, the input data of the input layer is Ip 1 ={X1,X2,…,X i ,…,X t}, Among them, X i Represents the measurement data collected at the i-th moment, i∈[1,t], t is the total number of collection moments in the time series segment, represents the mth measurement data collected at the i-th moment, where m is the total number of data collected at each moment. The collected data at each moment is put into a neuron in time series, and the input layer has a total of t neurons. The input layer passes the data to the mapping layer:

[0060] IP 1 =Op 1

[0061] IP 2 =ω 12 Op 1

[0062] Among them, Op1 is the output of the input layer, Ip 2 is the input of the mapping layer, ω 12 is the connection weight between the input layer and the mapping layer.

[0063] The mapping layer maps the input measurement data into the calculation space to facilitate subsequent calculations. The mapping process is as follows:

[0064]

[0065]

[0066] Where c is the coupling coefficient, Op 2 is the output of the mapping layer. The mapping layer passes the data to the association layer:

[0067] IP 3 =ω 23 Op 2

[0068] Among them, IP 3 is the input of the association layer, ω 23 is the connection weight between the mapping layer and the association layer.

[0069] The correlation layer calculates the correlation between data at different levels based on the divided data hierarchy. This correlation includes the correlation between the metering data collected by the sensor and the smart energy meter, as well as the correlation between the metering data collected by the sensor from the upstream and downstream devices of the smart energy meter and the metering data related to the smart energy meter collected by the sensor. The present invention uses a combination of active and passive correlations to calculate the correlation between data at different levels. Active correlation is the setting of correlations for data at different levels based on expert settings or common knowledge; passive correlation is analyzed and calculated using the correlation analysis method constructed by the present invention.

[0070] In each level of measurement data, by analyzing whether each measurement data of the current level and the measurement data to be analyzed show correlation in any time series, the passive correlation coefficient is calculated. The measurement data to be analyzed The correlation is calculated as:

[0071]

[0072] Among them, γ is the correlation analysis evaluation factor. If γ is greater than the preset correlation threshold, it means that there is no correlation between the data. Otherwise, there is correlation. n and n' both represent different levels of measurement data. If there is a correlation between the data, the passive correlation coefficient is calculated:

[0073]

[0074] Among them, δ represents the passive correlation coefficient. The association layer passes the preset active correlation coefficient and passive correlation coefficient to the perturbation layer:

[0075] Op 3 ={Ip 3 ,δ',δ}

[0076] IP 4 =ω 34 Op 3

[0077] Among them, Op 3 is the output of the correlation layer, including management layer input, active correlation coefficient and passive correlation coefficient, Ip 4 is the input of the perturbation layer, ω 34 is the connection weight between the association layer and the perturbation layer.

[0078] The disturbance layer calculates the disturbance factor based on the preset active correlation coefficient and passive correlation coefficient, perturbs the data, and optimizes the problem of inaccurate fault warning results of smart electricity meters caused by disturbances that cannot be simulated and emulated under complex working conditions. The calculation formula of the disturbance factor is:

[0079]

[0080] in, is the disturbance factor, from Randomly selected from . Then the output of the perturbation layer is:

[0081]

[0082] Among them, Op 4 is the output of the perturbation layer, and b is the bias. The perturbation layer passes the data to the recurrent layer:

[0083] IP 5 =ω 45 Op 4

[0084] Among them, IP 5 is the input of the recurrent layer.

[0085] The loop layer selects the learning function in reinforcement learning based on the correlation of the smart meter data in the time series. Taking action a strategy from the current loop state s to reach the next state will produce an impact benefit during the execution of this action. The relationship between the impact benefit and the action strategy is the learning function. The formula of the learning function is:

[0086]

[0087] Among them, Q(s,a) is the learning function, E is the learning process, D is the maximum number of cycles, d∈[1,D], γ is the decay factor used for self-learning control. The closer γ is to 1, the lower the current impact benefit is, and the more emphasis is placed on the value of the subsequent cycle state. The closer γ is to 0, the higher the current impact benefit is. For any action strategy A i Denote by a, for any cycle state S i Indicated by s. The loop layer passes the results of loop learning to the output layer:

[0088] IP 6 =ω 56 Op 5 =ω 56 Q(s,a)

[0089] Among them, IP 6 Represents the input of the output layer, Op 5 represents the output of the recurrent layer, ω 56 Represents the connection weight between the recurrent layer and the output layer.

[0090] After deep learning, a fault mapping model of the smart electricity meter measurement data is fitted. After comparison with the standard value, the real-time error is obtained. The initial parameters are adjusted based on the real-time error so that the final real-time error is within an acceptable range, automatically realizing a dynamic adjustment process.

[0091] Based on the output metering data fault mapping model, the data anomaly level and its anomaly parameters are obtained. The matching module 60 can preset the fault types corresponding to different data anomaly levels and anomaly parameters according to the actual situation, thereby matching the fault type of the current data, and the online warning module 70 issues a warning signal.

[0092] The beneficial effects of step S2 are: establishing a deep learning network with self-evolution capability, using big data modeling and analysis as a means, fitting a smart electricity meter metering data fault mapping model through the input layer, mapping layer, association layer, disturbance layer, loop layer and output layer, making a reasonable arrangement based on the correlation between the smart electricity meter metering data, introducing disturbance factor matching and algorithms, optimizing the problem of inaccurate warning results caused by disturbances of factors that cannot be simulated and emulated under complex working conditions, and realizing online warning of smart electricity meter faults and the self-evolution of warning technology.

[0093] In summary, the smart energy meter fault online early warning system and method described in the present invention are completed.

[0094] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0095] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0096] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0097] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. An online early warning system for smart electric energy meter faults, characterized in that: Includes the following sections: Connection module, transmission module, measurement and detection module, data layering module, autonomous processing module, matching module and online early warning module; The connection module is used to connect the smart energy meter with the single-chip microcomputer and the transmission module, and to establish a number of sensor interfaces for data acquisition. The connection module is located on the smart energy meter and is connected to the transmission module and the metering detection module through data transmission. The transmission module is used to receive the metering data collected by the smart electric energy meter and the sensor, and transmit it to the metering detection module. The transmission module is connected to the metering detection module through data transmission; The metering detection module is used to perform periodic round-robin readings on the smart electric energy meter and the sensor to sense metering data. The metering detection module is located on the single-chip microcomputer and is connected to the data layering module through data transmission. The data stratification module is used to hierarchically divide metering data based on the transmission relationship between metering data, using the stored standard metering data of smart electric energy meters as a reference value for calculating the correlation relationship, and the reference value is in a dynamic adjustment process; metering data directly read from smart electric energy meters belongs to the first level; metering data collected by sensors, if the metering data is related to the smart electric energy meter or is related to the occurrence of smart electric energy meter failures, this data belongs to the second level; If the metering data is the parameters of the upstream and downstream devices of the smart energy meter collected by sensors, then these data belong to the third level, and the third level data needs to be related to the smart energy meter through the second level data; All other data belong to the fourth level. The data stratification module is located on the single-chip microcomputer and is connected to the autonomous processing module via data transmission. The autonomous processing module is used to build a deep learning network with self-evolution capabilities. Through deep training, it learns from the complex combination of functions from layer to layer to find the mapping function that defines the mapping from input to output, optimizes the parameters of each layer, reduces errors, and ultimately outputs the level of abnormal data of the smart electricity meter and its abnormal parameters. The autonomous processing module is connected to the matching module via data transmission. The matching module is used to match the corresponding fault type according to the level and abnormal parameters, and the matching module is connected to the online warning module through data transmission.

2. A smart energy meter fault online early warning method, characterized in that: The following steps are involved: S1. Real-time acquisition of smart meter power environment information, fault information, and operational quality information, conducting long-term equipment operational quality monitoring. Metered data is hierarchically divided based on the transmission relationships between them. Stored standard metered data from smart meters is used as a reference value for calculating correlations, with the reference value undergoing a dynamic adjustment process. Metered data directly read from smart meters belongs to the first tier. Metered data collected by sensors, if relevant to the smart meter or related to smart meter faults, belongs to the second tier. If the metering data is the parameters of the upstream and downstream devices of the smart energy meter collected by sensors, then these data belong to the third level, and the third-level data must be related to the smart energy meter through the second-level data; other data belong to the fourth level; S2. Construct a deep learning network with self-evolution capabilities, introduce perturbation factors and matching algorithms, input the collected metering data into the deep learning network, and use deep training to learn the complex combination of functions from layer to layer to find the mapping function that defines the input to output. Optimize the parameters of each layer to reduce errors, and ultimately output the level of abnormal smart electricity meter data and its abnormal parameters. Based on the level and abnormal parameters, the corresponding fault type is matched and an early warning is issued, realizing online early warning of smart electricity meter faults and the self-evolution of early warning technology.

3. The online early warning method for smart electric energy meter fault according to claim 2, characterized in that: The step S2 specifically includes: A deep learning network consists of an input layer, a mapping layer, a correlation layer, a perturbation layer, a recurrent layer, and an output layer.

Images