Multiplexing management method and system for dismounting and replacing fault ammeter
By constructing a feature vector and root cause localization model for dismantling and replacing faulty meters, the sorting and detection of faulty meters is automated, solving the problems of low efficiency and unstable accuracy in existing technologies, and realizing efficient automation of meter fault diagnosis and health management.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- GANSU SHINING SCI & TECH
- Filing Date
- 2025-11-11
- Publication Date
- 2026-04-21
AI Technical Summary
In existing technologies, the sorting, detection, and processing of faulty meters rely on manual labor, which is inefficient, has unstable accuracy, and is overly dependent on human experience.
By acquiring the work order text, usage year, and environmental data, a return feature vector is constructed. The root cause localization model is used to locate the root cause of the failure. Combined with the health score, reuse management is carried out, including reusability judgment and sorting suggestions.
It has automated the diagnosis and health management of electricity meters, improved the efficiency and accuracy of replacing faulty meters, and reduced reliance on human experience.
Smart Images
Figure CN121903586A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of electricity meter reuse management technology, specifically a reuse management method and system for replacing faulty electricity meters. Background Technology
[0002] Replacement of electricity meters refers to meters that have been used for a certain period of time and have been removed due to malfunctions or replaced with updated versions. This mainly targets all returned smart meters except those removed for operational spot checks or user-requested calibration. The reuse and disposal of removed meters is an essential stage for putting them back into use and improving energy efficiency. Specifically, it consists of three important stages: equipment sorting, equipment re-inspection, and classification and disposal. Sorting includes three business stages: post-removal equipment sorting, sorting testing, and sorting and disposal analysis. The relationships between these stages are as follows: Figure 1 As shown.
[0003] In existing technologies, the sorting, inspection, and processing of replaced meters all rely on manual labor. During inspection, human experience is needed to determine the type of fault before sorting and classifying the meters. This reliance on manual labor is inefficient and, due to its excessive dependence on human experience, results in poor accuracy and stability. Summary of the Invention
[0004] In view of this, the purpose of the present invention is to provide a reuse management method and system for replacing faulty electricity meters, so as to solve the above-mentioned technical problems.
[0005] To achieve the above objectives, the present invention adopts the following technical solution:
[0006] The present invention provides a reuse management method for replacing faulty electricity meters, comprising the following steps:
[0007] The system obtains the work order text for replacing faulty meters, the year of use, the operating environment sampling data corresponding to multiple sampling time points, and the metering error data at multiple sampling time points. The operating environment sampling data includes temperature, humidity, and load.
[0008] Extract the key entities from the returned work order text, wherein the key entities include the manufacturer entity, batch entity, fault phenomenon entity, and reason for return entity;
[0009] Based on the key entities and the year of use, a repurchase feature vector is constructed, and the repurchase feature vector is matched with a pre-constructed typical repurchase scenario information library to obtain the current scenario label. The typical repurchase scenario information library includes feature vector templates for typical repurchase scenarios.
[0010] An input vector is constructed based on the operating environment sampling data corresponding to multiple sampling time points, the measurement error data of multiple sampling time points, the year of use, the manufacturer entity, and the batch entity. The current scene label is then fused into the input vector and input into a pre-constructed root cause localization model to obtain the root cause of the fault. The root cause localization model represents the probability distribution of multiple root cause labels corresponding to the input vector.
[0011] A health assessment is performed based on the root cause of the failure to obtain a health score; and reuse management is performed based on the root cause of the failure and the health score, wherein the reuse management includes reusability judgment and sorting suggestions.
[0012] In one embodiment of this application, the method for constructing the typical recall scenario information database includes:
[0013] Obtain the dismantled sample dataset, wherein the dismantled sample dataset includes multiple dismantled sample data, including samples of service life, samples of manufacturers, samples of batches, samples of fault phenomena, and samples of reasons for dismantling;
[0014] The extracted sample data in the extracted sample dataset is vectorized to obtain the extracted sample vector.
[0015] Density clustering is performed on the reclaimed sample vectors in the reclaimed sample dataset to obtain multiple sample clusters;
[0016] Sample clusters with a number of samples within a cluster greater than a set threshold are designated as target sample clusters. Typical values of multiple parameter values corresponding to each feature dimension of multiple decomposed sample vectors are extracted from the target sample clusters. The typical values of text-type feature dimensions are obtained through annotation, while the typical values of numeric-type feature dimensions are obtained by calculating the average value.
[0017] Based on the typical values corresponding to the feature dimensions, scene labels and feature vector templates for typical split-off scenarios are constructed.
[0018] In one embodiment of this application, the retraction feature vector is matched with a pre-built database of typical retraction scenarios to obtain scenario labels, including:
[0019] Calculate the cosine similarity between the re-retrieval feature vector and multiple feature vector templates in the typical re-retrieval scenario information database;
[0020] The scene label corresponding to the feature vector template with the highest similarity is used as the current scene label.
[0021] In one embodiment of this application, the method for constructing the root cause localization model includes:
[0022] Obtain a fault sample dataset of electricity meters, wherein the fault sample dataset includes multiple fault sample data, each fault sample data includes trend features extracted from metering error data at sampling time points, typical operating environment features extracted from operating environment sampling data corresponding to multiple sampling time points, manufacturer features, batch features, root cause labels, and scenario labels.
[0023] A feature vector is constructed based on the aforementioned trend characteristics, typical operating environment characteristics, manufacturer characteristics, and batch characteristics. And based on root cause labels and scene tags Building training data The training data The mathematical expression is:
[0024]
[0025] In the formula, Indicates the sample number. Indicates the total sample size;
[0026] For the scene tags Scene embedding, feature crossing, and feature fusion are performed to obtain the input feature vector. , wherein the input feature vector The mathematical expression is:
[0027]
[0028] In the formula, Represents the scene embedding vector. Represents the feature cross vector;
[0029] Construct a loss function based on the input feature vector. The input feature vector Root cause label The loss function is used to train the random forest model to obtain the root cause localization model and a set of model parameters representing the probability distribution of multiple root cause labels corresponding to the input vector. The mathematical expression of the loss function is:
[0030]
[0031] The mathematical expression for the output module of the root cause localization model is:
[0032]
[0033]
[0034] In the formula, Indicates the root cause category number. Indicates the total number of root causes. Indicates the first The weight of the root cause, Indicates the use of the sample for reference Is the root cause... Indicator functions, Samples predicted by the model The root cause Confidence level, Represents the model parameter set, The root of the input feature vector is because The probability, This indicates the index of the decision tree in the random forest. This represents the total number of decision trees in the random forest. Indicates the first Tree root cause The predicted probability, Indicates the root cause of the target. This represents the maximum probability.
[0035] In one embodiment of this application, the scene label Scene embedding, feature crossing, and feature fusion are performed to obtain the input feature vector. ,include:
[0036] For the scene tags Perform scene embedding to obtain scene embedding vectors. , where the scene embedding vector The mathematical expression is:
[0037]
[0038] In the formula, Represents the scene embedding matrix. Representing scene tags One-hot encoding;
[0039] Embed the scene vector With the feature vector Perform feature crossing to obtain the feature crossing vector. , where the feature cross vector The mathematical expression is:
[0040]
[0041] Embed the scene vector The feature cross vector With the feature vector The input feature vector is obtained by fusion. .
[0042] In one embodiment of this application, the method for constructing the input vector includes:
[0043] The average value of the operating environment sampling data corresponding to multiple sampling time points is calculated to obtain the average value of various environmental parameters; and the average value of the various environmental parameters is quantized based on a preset quantization relationship table to obtain the operating environment characteristics. The preset quantization relationship table includes multiple value ranges and corresponding quantization values.
[0044] Linear fitting is performed on the measurement error data at multiple sampling time points to obtain the slope of the linear fitting equation; and the slope of the linear fitting equation is used as a feature of the error change trend. ;
[0045] The usage year, the manufacturer entity, and the batch entity are vectorized to obtain a year vector. Manufacturer Vector and batch vector ;
[0046] Based on the aforementioned operating environment characteristics Error change trend characteristics Year vector Manufacturer Vector and batch vector Construct the input vector.
[0047] In one embodiment of this application, the current scene label is fused into the input vector and then input into a pre-built root cause localization model to obtain the root cause of the fault, including:
[0048] The current scene label fusion value is added to the input vector to obtain the fused input vector. ;
[0049] The fused input vector Substituting these values into the root cause localization model yields the root cause of the fault.
[0050] In one embodiment of this application, a health assessment is performed based on the root cause of the failure to obtain a health score, including:
[0051] The recurrence rate of the root cause of the failure under the current scenario label is determined from a pre-built multi-scenario recurrence probability table. The multi-scenario recurrence probability table is pre-constructed based on historical reuse data, and includes the failure recurrence rate corresponding to multiple root cause types under multiple scenarios.
[0052] Based on the recurrence rate Calculate health score ,in, .
[0053] In one embodiment of this application, reuse management based on the root cause of the failure and the health score includes:
[0054] Based on the root cause of the failure and the pre-built suggestion rule table, the corresponding sorting method is selected, wherein the suggestion rule table includes processing and sorting methods corresponding to various root causes of failure.
[0055] The reusability of the health score and the preset health threshold is determined. When the health score is greater than or equal to the preset health threshold, the meter is determined to be reusable. When the health score is less than the preset health threshold, the meter is determined to be unreusable.
[0056] This application also provides a reuse management system for replacing faulty electricity meters, including:
[0057] The acquisition module is used to acquire the work order text of the replacement of the faulty meter, the year of use, the operating environment sampling data corresponding to multiple sampling time points, and the metering error data of multiple sampling time points. The operating environment sampling data includes temperature value, humidity value, and load.
[0058] The entity extraction module is used to extract key entities from the returned work order text, wherein the key entities include the manufacturer entity, batch entity, fault phenomenon entity, and reason for return entity;
[0059] The scene matching module is used to construct a repurchase feature vector based on the key entity and the year of use, and match the repurchase feature vector with a pre-built typical repurchase scene information library to obtain scene tags, wherein the typical repurchase scene information library includes feature vectors of typical repurchase scenes.
[0060] The root cause localization module is used to construct an input vector based on the operating environment sampling data corresponding to multiple sampling time points, the measurement error data of multiple sampling time points, the year of use, the manufacturer entity, and the batch entity, and then input the scene label into the input vector and input it into the pre-constructed root cause localization model to obtain the root cause of the fault. The root cause localization model represents the probability distribution of multiple root cause labels corresponding to the input vector.
[0061] The reuse management module is used to evaluate the health status based on the root cause of the failure and obtain a health status score; and to perform reuse management based on the root cause of the failure and the health status score, wherein the reuse management includes reusability judgment and sorting suggestions.
[0062] The beneficial effects of this invention are as follows: This invention provides a reuse management method and system for replacing faulty electricity meters. This application extracts key entities from the replacement work order of the faulty electricity meter and extracts the usage year of the faulty electricity meter, operating environment sampling data corresponding to multiple sampling time points, and metering error data from multiple inspection time points from the historical data recording system. It extracts replacement feature vectors from the basic data and obtains the scene label of the faulty electricity meter through feature matching. Then, it uses a root cause localization model to locate the root cause of the faulty electricity meter. Reuse management is based on the root cause of the fault and health score. This application, through a root cause localization modeling method that integrates scene labels, achieves several key technological breakthroughs and significant business value in the task of electricity meter fault diagnosis and health management. Attached Figure Description
[0063] The present invention will be further described below with reference to the accompanying drawings and embodiments:
[0064] Figure 1 This is a schematic diagram of the sorting process.
[0065] Figure 2 This is a flowchart illustrating a reuse management method for replacing faulty electricity meters in one embodiment of this application;
[0066] Figure 3 This is a flowchart illustrating the construction of a root cause model in one embodiment of this application.
[0067] Figure 4 This is a diagram illustrating the significance of the reuse management process in one embodiment of this application;
[0068] Figure 5 This is a flowchart of a reuse management system for replacing faulty electricity meters, shown in one embodiment of this application;
[0069] Figure 6 A schematic diagram of the structure of a computer system suitable for implementing the electronic device of the present application is shown. Detailed Implementation
[0070] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, unless otherwise specified, the following embodiments and features described therein can be combined with each other.
[0071] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Therefore, the drawings only show the layers related to the present invention and are not drawn according to the actual number, shape and size ratio of the layers in the actual implementation. In the actual implementation, the shape and number of each layer can be arbitrarily changed, and the layer layout may also be more complex.
[0072] Numerous details are explored in the following description to provide a more thorough explanation of embodiments of the invention; however, it will be apparent to those skilled in the art that embodiments of the invention may be practiced without these specific details.
[0073] Figure 2 This is a flowchart illustrating a reuse management method for replacing faulty electricity meters in one embodiment of this application, as shown below. Figure 2 As shown: A reuse management method for replacing faulty electricity meters according to this embodiment may include steps S210 to S250:
[0074] S210, obtain the work order text for replacing the faulty meter, the year of use, the operating environment sampling data corresponding to multiple sampling time points, and the metering error data of multiple sampling time points, wherein the operating environment sampling data includes temperature value, humidity value and load;
[0075] This application is based on the operational processes of actual data management systems in the power industry (such as smart meter management systems, work order systems, and metering automation systems). The data acquisition process strictly follows power industry standards (such as "DL / T 645-2007 Electricity Meter Communication Protocol") to ensure data traceability and verifiability.
[0076] The table below shows the data acquisition method, source, and time range in this embodiment:
[0077] Table 1. Data Acquisition Information Table
[0078] Data types Get System How to obtain Time range frequency Retrieve work order text Power work order management system (such as PMS) Linking the work order to the meter ID One year before dismantling Automatic recording when work orders are generated Year of use Electricity Meter Record Management System Extract directly from the meter installation file Electricity meter activation time One-time use Operating environment sampling data Smart meter sensor + weather API Real-time data collection via built-in sensors in the electricity meter + supplementary data from the meteorological bureau's API. Six months before dismantling Once daily (optional) Measurement error data Automated metering systems (such as MDS) The system automatically records during regular spot checks. Six months before dismantling Once every 2 months
[0079] As shown in the table above, the "Fault Description" field can be extracted from the power work order management system (PMS) by querying the returned work order record through the meter ID (e.g., E-2023-001).
[0080] From the electricity meter record management system, query the "Installation Date" field using the meter ID (usually the date the meter was first put into operation). For example:
[0081] Meter ID: E-2023-001
[0082] Installation date: 2018-03-12
[0083] Year of use: 2018 (meaning the meter has been in operation for 5.7 years)
[0084] Methods for obtaining runtime environment data include:
[0085] Temperature / humidity: The smart meter's built-in sensor collects data every 10 minutes and stores it in the metering system;
[0086] Load: Real-time load (unit: kW) is obtained through the distribution transformer monitoring terminal (DTU) once per hour;
[0087] Data integration: The system merges environmental data by timestamp (if a sensor is missing, the meteorological bureau's API is called to complete it).
[0088] Example (6 months before dismantling, sampling points: 2023-07-01, 2023-08-01, 2023-09-01)
[0089] Metering error data is obtained by querying the "sampling record" through the meter ID in the metering automation system (MDS) and extracting the error value (unit: %).
[0090] S220, Extract the key entities from the dismantling work order text, wherein the key entities include the manufacturer entity, batch entity, fault phenomenon entity, and dismantling reason entity;
[0091] This application uses a large language model to extract key entities from the returned work order text, and adopts a three-part prompt of "instruction + example + text to be processed" to guide LLM to extract accurately.
[0092] S230, construct a repurchase feature vector based on the key entity and the year of use, and match the repurchase feature vector with a pre-built typical repurchase scenario information library to obtain the current scenario label, wherein the typical repurchase scenario information library includes feature vector templates for typical repurchase scenarios;
[0093] The dismantling feature vector is used to compare with feature vector templates in the typical dismantling scenario information database to find the scenario label for the dismantled faulty meter. In this embodiment, the scenario label is a high-level semantic category that integrates prior information such as operating environment, usage conditions, and maintenance behavior, used to characterize the typical combination of external conditions when the meter fails. For example, "high temperature + high frequency + manufacturer A + batch X" indicates that batch X meters from manufacturer A experienced frequent dismantling under high temperature conditions. This phenomenon may suggest that relevant components of batch X meters from manufacturer A may malfunction under high temperature conditions.
[0094] The following is a method for constructing a typical reunification scenario information database according to an embodiment of this application, specifically including:
[0095] (1) Obtain the sample dataset of the dismantled sample, wherein the sample dataset of the dismantled sample includes multiple sample data of the dismantled sample, including samples of the years of use, samples of the manufacturer, samples of the batch, samples of the fault phenomenon and samples of the reason for dismantling.
[0096] Extract historical data of dismantled meters from power system data warehouses (such as MDS metering automation system, PMS work order system, and meter archive).
[0097] (2) Vectorize the withdrawn sample data in the withdrawn sample dataset to obtain the withdrawn sample vector;
[0098] Extract semantic embedding vectors using pre-trained language models (such as BERT);
[0099] The numerical features and textual features are embedded into a vector and then concatenated to obtain a unified feature vector v.
[0100] v = [Service life; Manufacturer vector; Batch vector; Fault symptom vector; Reason for disassembly vector]
[0101] (3) Perform density clustering on the reclaimed sample vectors in the reclaimed sample dataset to obtain multiple sample clusters;
[0102] Clustering is performed using the DBSCAN algorithm (Density-Based Spatial Clustering of Applications with Noise), grouping the sample vectors with similar features together.
[0103] In power fault data, common scenarios (such as "high humidity-high frequency disconnection") have a large sample size, but long-tail scenarios (such as "lightning strike-rural power grid") have a small sample size. DBSCAN can effectively identify small clusters.
[0104] (4) Take the sample cluster with more than a set threshold as the target sample cluster, and extract the typical values of multiple parameter values corresponding to each feature dimension of multiple split sample vectors from the target sample cluster. The typical values of the feature dimensions of text type are obtained by annotation, and the typical values of the feature dimensions of numerical type are obtained by calculating the average value.
[0105] Clusters with a number of samples within a cluster greater than a threshold (e.g., 50) are retained to filter out noisy clusters. Typical methods for extracting values across multiple feature dimensions are as follows:
[0106] Table 2. Examples of Typical Value Extraction
[0107] Feature Dimension Type Typical value calculation method Example Numerical type (service life) Mean of samples within a cluster 32.7 years (high humidity cluster) Text-based (Fault Symptoms) Typical descriptions annotated by experts (non-average) "Water ingress due to aging of the sealing ring" (a high-frequency description extracted from 100 samples) Text-based (Reason for withdrawal) Same as above "High humidity environment causes seal failure" Text format (manufacturer / batch) The most frequent tags (such as "Waison Group") appeared in 65% of cases. "XX Group" (High Humidity Cluster)
[0108] (5) Construct scene labels and feature vector templates for typical split-off scenarios based on the typical values corresponding to the feature dimensions.
[0109] Finally, typical values are combined into semantic labels (e.g., "high humidity - high frequency dismantling"), and baseline values for feature vectors are defined for each scene label, for example:
[0110] vtemplate=[Average service life; Manufacturer typical value; Batch typical value; Fault phenomenon typical value; Reason for disassembly typical value]
[0111] Based on the typical retraction scenarios constructed above, the retraction feature vector is matched with a pre-constructed typical retraction scenario information database to obtain the current scenario label, including:
[0112] S231, calculate the cosine similarity between the re-recovery feature vector and multiple feature vector templates in the typical re-recovery scenario information database;
[0113] S232, take the scene label corresponding to the feature vector template with the highest similarity as the current scene label.
[0114] In this application, feature vectors are matched using pre-similarity to obtain corresponding scene labels.
[0115] S240, an input vector is constructed based on the operating environment sampling data corresponding to multiple sampling time points, the measurement error data of multiple sampling time points, the year of use, the manufacturer entity, and the batch entity. The current scene label is then fused into the input vector and input into a pre-constructed root cause localization model to obtain the root cause of the fault. The root cause localization model represents the probability distribution of multiple root cause labels corresponding to the input vector.
[0116] In this application, scene labels are used to perform root cause localization, making the root cause of the fault more accurate. This application utilizes a decision tree model to construct a root cause model that integrates scene labels. Figure 3 This is a flowchart illustrating the construction of a root cause model in one embodiment of this application, as follows: Figure 3 As shown, the specific process includes:
[0117] (1) Obtain the fault sample dataset of the electricity meter, wherein the fault sample dataset includes multiple fault sample data, each fault sample data includes the trend feature extracted from the metering error data at the sampling time point, the typical operating environment feature extracted from the operating environment sampling data corresponding to multiple sampling time points, the manufacturer feature, the batch feature, the root cause label and the scene label.
[0118] Specifically, the measurement error data at the sampling time points (such as the error values of 2023-05-01, 2023-07-01, and 2023-09-01) are extracted from the Metrology Automation System (MDS), and the trend characteristics of change are calculated (such as the error slope: (error_09-01 - error_05-01) / (90 days)).
[0119] The operating environment sampling data (temperature, humidity, load) is obtained from the smart meter sensor, and typical operating environment characteristics (such as annual average humidity and humidity standard deviation) are calculated.
[0120] Extract manufacturer characteristics (such as "XX Group") and batch characteristics (such as "202305A") from the electricity meter archive database;
[0121] Obtain root cause labels (such as "poor sealing") and scenario labels (such as "high humidity-high frequency disassembly") from the disassembly work order system.
[0122] The extraction of the changing trend characteristics and typical operating environment characteristics will be introduced later.
[0123] Root cause labels are manually labeled, such as: "poor sealing", "hardware aging", "X component failure", "communication module failure", etc.
[0124] (2) Construct a feature vector based on the aforementioned trend characteristics, typical operating environment characteristics, manufacturer characteristics, and batch characteristics. And based on root cause labels and scene tags Building training data The training data The mathematical expression is:
[0125]
[0126] In the formula, Indicates the sample number. Indicates the total sample size;
[0127] (3) For the scene tags Scene embedding, feature crossing, and feature fusion are performed to obtain the input feature vector. , wherein the input feature vector The mathematical expression is:
[0128]
[0129] In the formula, Represents the scene embedding vector. Represents the feature cross vector;
[0130] Specifically, for the scene tags Scene embedding, feature crossing, and feature fusion are performed to obtain the input feature vector. ,include:
[0131] (3-1) For the scene labels Perform scene embedding to obtain scene embedding vectors. , where the scene embedding vector The mathematical expression is:
[0132]
[0133] In the formula, Represents the scene embedding matrix. Representing scene tags One-hot encoding;
[0134] (3-2) Embed the scene into a vector With the feature vector Perform feature crossing to obtain the feature crossing vector. , where the feature cross vector The mathematical expression is:
[0135]
[0136] (3-3) Embed the scene into a vector The feature cross vector With the feature vector The input feature vector is obtained by fusion. .
[0137] In this application, scene labels are explicitly encoded as model input, rather than as ordinary features. Specifically, scene embeddings and feature cross-features are used to avoid the curse of dimensionality and preserve scene semantics. This ensures semantic fusion between scenes and features (such as automatic enhancement of humidity feature weights in high humidity scenes).
[0138] (4) Construct a loss function and base it on the input feature vector. The input feature vector Root cause label The loss function is used to train the random forest model to obtain the root cause localization model and a set of model parameters representing the probability distribution of multiple root cause labels corresponding to the input vector. The mathematical expression of the loss function is:
[0139]
[0140] The mathematical expression for the output module of the root cause localization model is:
[0141]
[0142]
[0143] In the formula, Indicates the root cause category number. Indicates the total number of root causes. Indicates the first The weights of root causes (their core function is to map discrete scene labels into low-dimensional, dense, learnable continuous vectors (i.e., "embedding vectors") so that the model can efficiently capture the impact of different scenes on the root cause distribution). Indicates the use of the sample for reference Is the root cause... Indicator functions, Samples predicted by the model The root cause Confidence level, Represents the model parameter set, The root of the input feature vector is because The probability, This indicates the index of the decision tree in the random forest. This represents the total number of decision trees in the random forest. Indicates the first Tree root cause The predicted probability, Indicates the root cause of the target. This represents the maximum probability.
[0144] Model parameter set with integrated scene tags It is achieved by encoding scene labels into learnable embeddings. and construct input by crossing features The result was obtained by optimizing the random forest parameters (tree structure, leaf node probability, scene embedding matrix) under conditional cross-entropy loss. It reflects the probability distribution of multiple root cause labels.
[0145] Model output It is directly used to predict the probability of multiple root causes, and the root cause with the highest probability is selected as the root cause of the failure. .
[0146] High confidence (>0.8) directly outputs the root cause, while low confidence (<0.6) triggers manual review. This ensures the accuracy of root cause identification.
[0147] The innovation of the root cause localization model in this application lies in the integration of scene labels. Traditional models treat scenes as ordinary features or ignore them, resulting in the weakening of strong correlation patterns such as "poor sealing in high humidity environments." After scene label integration, the model explicitly models the environment, making it a conditional variable. Scene labels are automatically constructed based on historical fault data (not manually preset), strictly distinguishing between input (observable features + scene) and output (root cause labels), ensuring that the model generalizes in real-world environments.
[0148] The current scene label is fused into the input vector and then input into a pre-built root cause localization model to obtain the root cause of the fault, including:
[0149] S241, the current scene label fusion value is added to the input vector to obtain the fused input vector. ;
[0150] S242, the fused input vector Substituting these values into the root cause localization model yields the root cause of the fault.
[0151] Based on the above model, the input vector will be fused. Substituting these values into the root cause localization model, the output module directly outputs the root cause of the fault with the highest probability.
[0152] In the above process, the methods for constructing the input vector include:
[0153] (1) Calculate the average value of the operating environment sampling data corresponding to multiple sampling time points to obtain the average value of various environmental parameters; and quantize the average value of the various environmental parameters based on a preset quantization relationship table to obtain the operating environment characteristics. The preset quantization relationship table includes multiple value ranges and corresponding quantization values.
[0154] Environmental data fluctuates over time (such as daily humidity variations), and the average value represents typical operating conditions to avoid interference from single-point anomalies.
[0155] (2) Perform linear fitting on the measurement error data at multiple sampling time points to obtain the slope of the linear fitting equation; and use the slope of the linear fitting equation as a feature of the error change trend. ;
[0156] Single-point errors (e.g., +2.8%) cannot distinguish between "slow deterioration" and "rapid failure," and the slope K quantifies the error growth rate; time point selection bias is eliminated through linear fitting.
[0157] (3) Vectorize the year of use, the manufacturer entity, and the batch entity to obtain the year vector. Manufacturer Vector and batch vector ;
[0158] (4) Based on the aforementioned operating environment characteristics Error change trend characteristics Year vector Manufacturer Vector and batch vector Construct the input vector.
[0159] S250, perform a health assessment based on the root cause of the failure to obtain a health score; and perform reuse management based on the root cause of the failure and the health score, wherein the reuse management includes reusability judgment and sorting suggestions.
[0160] Figure 4 This is a diagram illustrating the reuse management process in one embodiment of this application, such as... Figure 4 Finally, after obtaining the root cause of the failure, this application uses the root cause to perform a health score, and uses the health score and the root cause for reuse management. The process of constructing the health score includes:
[0161] S251, determine the recurrence rate of the root cause of the failure under the current scenario label from a pre-constructed multi-scenario recurrence probability table. The multi-scenario recurrence probability table is pre-constructed based on historical reuse data, and includes the failure recurrence rate corresponding to multiple root cause types under multiple scenarios.
[0162] In this application, the recurrence rate directly reflects the probability of the root cause of failure recurrence in a given scenario, serving as an inverse indicator of health (a higher recurrence rate indicates lower health). This application constructs a multi-scenario recurrence probability table to reference the recurrence rates corresponding to the root cause of failure in various scenarios, avoiding a one-size-fits-all approach (e.g., "poor sealing" has a recurrence rate of 12% in a high-humidity scenario but only 2% in a low-temperature scenario), ensuring that the scoring scenario is adaptive. The table below is a multi-scenario recurrence probability table from one embodiment of this application:
[0163] Table 3. Recurrence Probability Table in Multiple Scenarios
[0164] Scene tags Root cause labeling recurrence rate Historical sample size High humidity-high frequency disassembly Poor sealing 0.12 500 Low temperature - long-term light load Hardware aging 0.08 300 Lightning strike - rural power grid Communication module failure 0.05 200 New batch - Manufacturer A Software defects 0.03 150
[0165] S252, based on the recurrence rate Calculate health score ,in, .
[0166] Then, based on the root cause of the failure and the health score, reuse management is performed, including:
[0167] S253, Select the corresponding sorting method based on the root cause of the failure and the pre-built suggestion rule table, wherein the suggestion rule table includes processing and sorting methods corresponding to multiple root causes of failure;
[0168] In this application, based on historical reuse experience, a processing strategy is defined for each root cause label, for example:
[0169] Table 4. Processing Rules Table
[0170] Root cause labeling Reuse strategy Sorting method illustrate Poor sealing No reuse scrapped High recurrence rate (12%) Hardware aging Return to factory for repair Downgraded use Recurrence rate (8%) Software defects Firmware upgrade Reusable Low recurrence rate (3%) Communication module failure Replace module Downgraded use Low recurrence rate (5%) other Manual review Submitted for testing Unclassified faults
[0171] The rule table maps root cause types to executable operations (such as "poor sealing" → scrap), avoiding manual judgment.
[0172] S254, the reusability of the health score and the preset health threshold is determined, wherein when the health score is greater than or equal to the preset health threshold, the meter is determined to be reusable, and when the health score is less than the preset health threshold, the meter is determined to be unreusable.
[0173] Health scoring and reuse management are the ultimate implementation steps of root cause analysis, directly determining operation and maintenance costs and grid security. This solution, through a data-driven recurrence rate table and rule table, enables the reuse of meters to move from "high risk" to "zero risk," providing core support for the full lifecycle management of smart meters.
[0174] This invention discloses a reuse management method for replacing faulty electricity meters. The method involves extracting key entities from the replacement work order of the faulty meter and extracting the usage year, operating environment sampling data corresponding to multiple sampling time points, and metering error data from multiple inspection time points from a historical data recording system. A replacement feature vector is extracted from the basic data, and through feature matching, a scenario label for the faulty meter is obtained. Then, a root cause localization model is used to locate the root cause of the faulty meter. Reuse management is then performed based on the root cause of the fault and a health score. This application, through a root cause localization modeling method that integrates scenario labels, achieves several key technological breakthroughs and significant business value in the task of electricity meter fault diagnosis and health management.
[0175] like Figure 5 As shown, this application also provides a reuse management system for replacing faulty electricity meters, including:
[0176] The acquisition module is used to acquire the work order text of the replacement of the faulty meter, the year of use, the operating environment sampling data corresponding to multiple sampling time points, and the metering error data of multiple sampling time points. The operating environment sampling data includes temperature value, humidity value, and load.
[0177] The entity extraction module is used to extract key entities from the returned work order text, wherein the key entities include the manufacturer entity, batch entity, fault phenomenon entity, and reason for return entity;
[0178] The scene matching module is used to construct a repurchase feature vector based on the key entity and the year of use, and match the repurchase feature vector with a pre-built typical repurchase scene information library to obtain scene tags, wherein the typical repurchase scene information library includes feature vectors of typical repurchase scenes.
[0179] The root cause localization module is used to construct an input vector based on the operating environment sampling data corresponding to multiple sampling time points, the measurement error data of multiple sampling time points, the year of use, the manufacturer entity, and the batch entity, and then input the scene label into the input vector and input it into the pre-constructed root cause localization model to obtain the root cause of the fault. The root cause localization model represents the probability distribution of multiple root cause labels corresponding to the input vector.
[0180] The reuse management module is used to evaluate the health status based on the root cause of the failure and obtain a health status score; and to perform reuse management based on the root cause of the failure and the health status score, wherein the reuse management includes reusability judgment and sorting suggestions.
[0181] This invention discloses a reuse management system for replacing faulty electricity meters. The system extracts key entities from the replacement work orders for faulty meters and extracts the usage year of the replaced meters, operating environment sampling data corresponding to multiple sampling time points, and metering error data from multiple inspection time points from historical data recording systems. It extracts replacement feature vectors from the basic data and obtains the scene label of the replaced faulty meters through feature matching. Then, it uses a root cause localization model to locate the root cause of the replaced faulty meters. Reuse management is based on the root cause of the fault and a health score. This application, through a root cause localization modeling method that integrates scene labels, achieves several key technological breakthroughs and significant business value in the task of electricity meter fault diagnosis and health management.
[0182] Figure 6 A schematic diagram of a computer system suitable for implementing the embodiments of this application is shown. It should be noted that... Figure 6 The computer system of the electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.
[0183] like Figure 6As shown, the computer system includes a Central Processing Unit (CPU) 601, which can perform various appropriate actions and processes, such as executing the methods described in the above embodiments, based on programs stored in Read-Only Memory (ROM) 602 or programs loaded from storage portion 608 into Random Access Memory (RAM) 603. The RAM 603 also stores various programs and data required for system operation. The CPU 601, ROM 602, and RAM 603 are interconnected via a bus 604. An Input / Output (I / O) interface 605 is also connected to the bus 604.
[0184] The following components are connected to I / O interface 605: an input section 606 including a keyboard, mouse, etc.; an output section 607 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 608 including a hard disk, etc.; and a communication section 609 including a network interface card such as a LAN (Local Area Network) card, modem, etc. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to I / O interface 605 as needed. A removable medium 611, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on drive 610 as needed so that computer programs read from it can be installed into storage section 608 as needed.
[0185] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program including a computer program for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 609, and / or installed from removable medium 611. When the computer program is executed by central processing unit (CPU) 601, it performs various functions defined in the system of this application.
[0186] It should be noted that the computer-readable medium shown in the embodiments of this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying a computer-readable computer program. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The computer program contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to wireless, wired, etc., or any suitable combination thereof.
[0187] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. Each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.
[0188] The units described in the embodiments of this application can be implemented in software or hardware, and the described units can also be located in a processor. The names of these units do not necessarily limit the specific unit itself.
[0189] Another aspect of this application provides a computer-readable storage medium storing a computer program thereon, which, when executed by a computer's processor, causes the computer to perform the method as described above. This computer-readable storage medium may be included in the electronic device described in the above embodiments, or it may exist independently and not assembled into the electronic device.
[0190] Another aspect of this application provides a computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the methods provided in the various embodiments described above.
[0191] The above embodiments are merely preferred embodiments provided to fully illustrate this application, and the scope of protection of this application is not limited thereto. Equivalent substitutions or modifications made by those skilled in the art based on this application are all within the scope of protection of this application.
Claims
1. A method for reusing and managing faulty electricity meters, characterized in that, Including the following steps: The system obtains the work order text for replacing faulty meters, the year of use, the operating environment sampling data corresponding to multiple sampling time points, and the metering error data at multiple sampling time points. The operating environment sampling data includes temperature, humidity, and load. Extract the key entities from the returned work order text, wherein the key entities include the manufacturer entity, batch entity, fault phenomenon entity, and reason for return entity; Based on the key entities and the year of use, a repurchase feature vector is constructed, and the repurchase feature vector is matched with a pre-constructed typical repurchase scenario information library to obtain the current scenario label. The typical repurchase scenario information library includes feature vector templates for typical repurchase scenarios. An input vector is constructed based on the operating environment sampling data corresponding to multiple sampling time points, the measurement error data of multiple sampling time points, the year of use, the manufacturer entity, and the batch entity. The current scene label is then fused into the input vector and input into a pre-constructed root cause localization model to obtain the root cause of the fault. The root cause localization model represents the probability distribution of multiple root cause labels corresponding to the input vector. A health assessment is performed based on the root cause of the failure to obtain a health score; and reuse management is performed based on the root cause of the failure and the health score, wherein the reuse management includes reusability judgment and sorting suggestions.
2. The reuse management method for replacing faulty electricity meters according to claim 1, characterized in that, The method for constructing the typical withdrawal scenario information database includes: Obtain the dismantled sample dataset, wherein the dismantled sample dataset includes multiple dismantled sample data, including samples of service life, samples of manufacturers, samples of batches, samples of fault phenomena, and samples of reasons for dismantling; The extracted sample data in the extracted sample dataset is vectorized to obtain the extracted sample vector. Density clustering is performed on the reclaimed sample vectors in the reclaimed sample dataset to obtain multiple sample clusters; Sample clusters with a number of samples within a cluster greater than a set threshold are designated as target sample clusters. Typical values of multiple parameter values corresponding to each feature dimension of multiple decomposed sample vectors are extracted from the target sample clusters. The typical values of text-type feature dimensions are obtained through annotation, while the typical values of numeric-type feature dimensions are obtained by calculating the average value. Based on the typical values corresponding to the feature dimensions, scene labels and feature vector templates for typical split-off scenarios are constructed.
3. The reuse management method for replacing faulty electricity meters according to claim 2, characterized in that, The retraction feature vector is matched with a pre-built database of typical retraction scenarios to obtain scenario labels, including: Calculate the cosine similarity between the re-retrieval feature vector and multiple feature vector templates in the typical re-retrieval scenario information database; The scene label corresponding to the feature vector template with the highest similarity is used as the current scene label.
4. The reuse management method for replacing faulty electricity meters according to claim 1, characterized in that, The method for constructing the root cause localization model includes: Obtain a fault sample dataset of electricity meters, wherein the fault sample dataset includes multiple fault sample data, each fault sample data includes trend features extracted from metering error data at sampling time points, typical operating environment features extracted from operating environment sampling data corresponding to multiple sampling time points, manufacturer features, batch features, root cause labels, and scenario labels. A feature vector is constructed based on the aforementioned trend characteristics, typical operating environment characteristics, manufacturer characteristics, and batch characteristics. And based on root cause labels and scene tags Building training data The training data The mathematical expression is: ； In the formula, Indicates the sample number. Indicates the total sample size; For the scene tags Scene embedding, feature crossing, and feature fusion are performed to obtain the input feature vector. , wherein the input feature vector The mathematical expression is: ； In the formula, Represents the scene embedding vector. Represents the feature cross vector; Construct a loss function based on the input feature vector. The input feature vector Root cause label The loss function is used to train the random forest model to obtain the root cause localization model and a set of model parameters representing the probability distribution of multiple root cause labels corresponding to the input vector. The mathematical expression of the loss function is: ； The mathematical expression for the output module of the root cause localization model is: ； ； In the formula, Indicates the root cause category number. Indicates the total number of root causes. Indicates the first The weight of the root cause, Indicates the use of the sample for reference Is the root cause... Indicator functions, Samples predicted by the model The root cause Confidence level, Represents the model parameter set, The root of the input feature vector is because The probability, This indicates the index of the decision tree in the random forest. This represents the total number of decision trees in the random forest. Indicates the first Tree root cause The predicted probability, Indicates the root cause of the target. Indicates the maximum probability; 5. The reuse management method for replacing faulty electricity meters according to claim 4, characterized in that, For the scene tags Scene embedding, feature crossing, and feature fusion are performed to obtain the input feature vector. ,include: For the scene tags Perform scene embedding to obtain scene embedding vectors. , where the scene embedding vector The mathematical expression is: ； In the formula, Represents the scene embedding matrix. Representing scene tags One-hot encoding; Embed the scene vector With the feature vector Perform feature crossing to obtain the feature crossing vector. , where the feature cross vector The mathematical expression is: ； Embed the scene vector The feature cross vector With the feature vector The input feature vector is obtained by fusion. .
6. A reuse management method for replacing faulty electricity meters according to claim 1 or 4, characterized in that, Methods for constructing input vectors include: The average value of the operating environment sampling data corresponding to multiple sampling time points is calculated to obtain the average value of various environmental parameters; and the average value of the various environmental parameters is quantized based on a preset quantization relationship table to obtain the operating environment characteristics. The preset quantization relationship table includes multiple value ranges and corresponding quantization values. Linear fitting is performed on the measurement error data at multiple sampling time points to obtain the slope of the linear fitting equation; and the slope of the linear fitting equation is used as a feature of the error change trend. ; The usage year, the manufacturer entity, and the batch entity are vectorized to obtain a year vector. Manufacturer Vector and batch vector ; Based on the aforementioned operating environment characteristics Error change trend characteristics Year vector Manufacturer Vector and batch vector Construct the input vector.
7. The reuse management method for replacing faulty electricity meters according to claim 1, characterized in that, The current scene label is fused into the input vector and then input into a pre-built root cause localization model to obtain the root cause of the fault, including: The current scene label fusion value is added to the input vector to obtain the fused input vector. ; The fused input vector Substituting these values into the root cause localization model yields the root cause of the fault.
8. The reuse management method for replacing faulty electricity meters according to claim 1, characterized in that, A health assessment is performed based on the root causes of the failures to obtain a health score, including: The recurrence rate of the root cause of the failure under the current scenario label is determined from a pre-built multi-scenario recurrence probability table. The multi-scenario recurrence probability table is pre-constructed based on historical reuse data, and includes the failure recurrence rate corresponding to multiple root cause types under multiple scenarios. Based on the recurrence rate Calculate health score ,in, .
9. The reuse management method for replacing faulty electricity meters according to claim 1, characterized in that, Reuse management based on the root cause of the failure and the health score includes: Based on the root cause of the failure and the pre-built suggestion rule table, the corresponding sorting method is selected, wherein the suggestion rule table includes processing and sorting methods corresponding to various root causes of failure. The reusability of the health score and the preset health threshold is determined. When the health score is greater than or equal to the preset health threshold, the meter is determined to be reusable. When the health score is less than the preset health threshold, the meter is determined to be unreusable.
10. A reuse management system for replacing faulty electricity meters, characterized in that, include: The acquisition module is used to acquire the work order text of the replacement of the faulty meter, the year of use, the operating environment sampling data corresponding to multiple sampling time points, and the metering error data of multiple sampling time points. The operating environment sampling data includes temperature value, humidity value, and load. The entity extraction module is used to extract key entities from the returned work order text, wherein the key entities include the manufacturer entity, batch entity, fault phenomenon entity, and reason for return entity; The scene matching module is used to construct a repurchase feature vector based on the key entity and the year of use, and match the repurchase feature vector with a pre-built typical repurchase scene information library to obtain scene tags, wherein the typical repurchase scene information library includes feature vectors of typical repurchase scenes. The root cause localization module is used to construct an input vector based on the operating environment sampling data corresponding to multiple sampling time points, the measurement error data of multiple sampling time points, the year of use, the manufacturer entity, and the batch entity, and then input the scene label into the input vector and input it into the pre-constructed root cause localization model to obtain the root cause of the fault. The root cause localization model represents the probability distribution of multiple root cause labels corresponding to the input vector. The reuse management module is used to evaluate the health status based on the root cause of the failure and obtain a health status score; and to perform reuse management based on the root cause of the failure and the health status score, wherein the reuse management includes reusability judgment and sorting suggestions.