Intelligent electric meter box operation and maintenance management method and system

Abstract

The invention discloses an intelligent electric meter box operation and maintenance management method and system, and relates to the field of electric power operation and maintenance management. The intelligent electric meter box operation and maintenance management method comprises the steps of multi-source data acquisition and preprocessing, equipment health state dynamic evaluation, fault root cause positioning and prediction, maintenance task dynamic scheduling, adaptive maintenance execution and closed-loop optimization, feedback of a maintenance result to a health evaluation model, updating of a dynamic weight coefficient, and monthly generation of an equipment health white book. And continuously optimizing the operation and maintenance strategy. According to the intelligent electric meter box operation and maintenance management method, the fault prediction model is established based on the LSTM neural network, the typical fault mode is matched in combination with the knowledge graph, the fault prediction accuracy is improved, and non-planned power failure is reduced. The improved genetic algorithm improves the utilization rate of maintenance resources and shortens the emergency response time. Predictive maintenance replaces regular maintenance, invalid inspection is reduced, and the spare part inventory turnover rate is increased. Human errors are reduced through AR assistance; and the maintenance operation accuracy is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of electric power operation and maintenance management, and in particular to an operation and maintenance management method and system for an intelligent electric meter box. Background Art

[0002] With the advancement of smart grid construction, electricity meter boxes, as key equipment for power metering and distribution, face the following challenges in operation and maintenance:

[0003] Limitations of the traditional operation and maintenance model: Relying on manual inspections and regular maintenance, there are problems such as delayed response, high missed detection rate, over-maintenance or under-maintenance.

[0004] Insufficient data utilization: Multi-source information such as sensor data, historical records, and environmental parameters is not effectively integrated, resulting in one-sided equipment status assessment and low fault prediction accuracy.

[0005] Extensive maintenance decision-making: Fault root cause location relies on experience, resource scheduling lacks a quantitative model, and major faults can easily lead to expanded power outages.

[0006] Technological gap: Although technologies such as the Internet of Things (IoT), artificial intelligence (AI), and augmented reality (AR) have developed, they have not yet been deeply integrated with meter box operations and maintenance, and lack a closed-loop optimization mechanism.

[0007] Improvement direction of existing technology:

[0008] Edge computing and data cleaning: Data preprocessing can be achieved through edge gateways to reduce cloud transmission pressure, but the problems of sensor data loss compensation and anomaly detection need to be solved.

[0009] Dynamic weight model: Traditional health assessment relies on fixed weights and cannot adapt to sudden environmental changes (such as extreme weather). It is necessary to introduce an adaptive weight adjustment mechanism.

[0010] Predictive maintenance: The LSTM-based fault prediction model needs to be combined with the knowledge graph to solve the problem of integrating "data-driven" and "knowledge-driven".

[0011] Intelligent scheduling algorithm: Genetic algorithms are widely used in operation and maintenance scheduling, but the encoding method and fitness function need to be optimized to avoid falling into local optimality. Summary of the Invention

[0012] In response to the shortcomings of the existing technology, the present invention provides an operation and maintenance management method and system for an intelligent meter box, which solves the problems of the traditional operation and maintenance mode in which multi-source information such as sensor data, historical records, and environmental parameters are not effectively integrated, resulting in one-sided equipment status assessment, low fault prediction accuracy and reliance on experience in locating the root cause of the fault, lack of a quantitative model for resource scheduling, and major faults easily causing the expansion of the power outage range.

[0013] To achieve the above objectives, the present invention is implemented through the following technical solutions: A smart meter box operation and maintenance management method, comprising the following steps:

[0014] Step 1: Multi-source data collection and preprocessing: deploy IoT sensors to collect multi-dimensional data and transmit it to the edge gateway via LoRaWAN for data cleaning and format standardization;

[0015] Step 2: Dynamically assess the health status of equipment by integrating real-time sensor data, historical operation and maintenance records, and environmental data. The equipment health index (EHI) is calculated using a dynamic weight fusion model, and fault warnings are then made.

[0016] Step 3: Fault root cause location and prediction: A fault prediction model is established based on the LSTM neural network. The knowledge graph is combined to match typical failure modes and output the remaining service life.

[0017] Step 4: Dynamic scheduling of maintenance tasks. Based on the fault type, impact scope, and maintenance resource availability, the maintenance task priority is dynamically calculated and the corresponding maintenance process is triggered.

[0018] Step 5: Adaptive maintenance execution, generating AR-assisted maintenance instructions, recording maintenance process data, and ensuring the accuracy and traceability of maintenance operations;

[0019] Step 6: Closed-loop optimization: Feedback maintenance results to the health assessment model, update the dynamic weight coefficient, generate a monthly equipment health white paper, and continuously optimize the operation and maintenance strategy.

[0020] Preferably, the edge gateway performs data cleaning and format standardization processing, including:

[0021] When the sensor data missing rate is greater than 5%, the adjacent node data interpolation compensation is triggered;

[0022] When an abnormal pulse signal is detected, Fourier transform is started to perform frequency domain analysis;

[0023] Convert the cleaned data into a unified JSON format and add a timestamp.

[0024] Preferably, the equipment health index EHI is calculated using a dynamic weight fusion model, including:

[0025] DS evidence theory is used to handle data conflicts. When the degree of conflict between sensor data and historical records is greater than 30%, the sensor weight coefficient is reduced.

[0026] Set three levels of health thresholds: EHI>0.8 for healthy status, 0.5≤EHI≤0.8 for warning status, and EHI<0.5 for fault status;

[0027] When the EHI shows a downward trend for three consecutive detection cycles and the decrease is greater than 15%, a fault warning is forcibly triggered.

[0028] Preferably, a fault prediction model is established based on an LSTM neural network, and the input features include:

[0029] Time series features: sensor data volatility over the past 72 hours;

[0030] Spatial characteristics: regional failure rate of similar equipment;

[0031] Construct a knowledge graph of power equipment, including the four-tuple relationship of fault phenomenon-fault location-fault cause-solution;

[0032] When the prediction model outputs a failure probability P > 70%, knowledge graph matching is initiated, including:

[0033] Prioritize solutions with a historical repair success rate greater than 80%;

[0034] When there are multiple failure modes, the Bayesian posterior probability is calculated to rank the root causes.

[0035] Preferably, the maintenance task priority P is calculated by an improved genetic algorithm based on the fault type, impact range, and maintenance resource availability:

[0036]

[0037] Where: S is the fault severity coefficient, U is the number of affected users, D is the maintenance resource distance, and I is the spare parts inventory coefficient;

[0038] Dynamic adjustment of weight coefficient: During the power-saving period of major events, w1 will automatically increase by 30%;

[0039] When P>preset threshold, the three-level response mechanism is triggered:

[0040] P>0.9: immediately start the emergency repair process;

[0041] 0.7<P≤0.9: Maintenance resources are dispatched within 2 hours;

[0042] P≤0.7: Included in regular maintenance plan.

[0043] Preferably, generating AR-assisted maintenance instructions includes:

[0044] 3D device anatomical model shows the fault location;

[0045] Overlay sensor data streams in real time;

[0046] When the operation steps are wrong, vibration feedback and voice warning will be triggered;

[0047] Record maintenance process data, including:

[0048] Maintenance personnel biometric information;

[0049] Spare parts use electronic tags;

[0050] Video stream summary of the repair process.

[0051] Preferably, the maintenance results are fed back to the health assessment model to update the dynamic weight coefficient:

[0052] When the actual fault type does not match the prediction, the weight of the corresponding feature dimension is reduced;

[0053] Generate device health white papers monthly, including:

[0054] Equipment health index distribution heat map;

[0055] Failure mode evolution trend analysis;

[0056] Maintenance cost-benefit assessment;

[0057] When the model prediction accuracy is less than 85% for three consecutive months, the model retraining process is initiated.

[0058] Preferably, the dynamic weight fusion model adopts a particle swarm optimization algorithm, and when a sudden change in ambient temperature greater than 10°C / h is detected, the weight coefficient of the temperature sensor is temporarily increased to 0.4.

[0059] Preferably, the improved genetic algorithm comprises:

[0060] Chromosome encoding: real number encoding is used;

[0061] Fitness function: add maintenance cost constraint;

[0062] Mutation operation: When the population diversity is less than 0.2, a large mutation operation is initiated.

[0063] An intelligent electricity meter box operation and maintenance management system, the system comprising:

[0064] Multi-source data acquisition and pre-processing module, used to deploy IoT sensors to collect multi-dimensional data, which can then be transmitted to the edge gateway via LoRaWAN for data cleaning and format standardization;

[0065] The equipment health status dynamic assessment module is used to integrate real-time sensor data, historical operation and maintenance records, and environmental data, and then calculate the equipment health index (EHI) through a dynamic weight fusion model to perform fault warning judgment;

[0066] The fault root cause location and prediction module can establish a fault prediction model based on the LSTM neural network, match typical failure modes with the knowledge graph, and output the remaining service life;

[0067] The maintenance task dynamic scheduling module can dynamically calculate the maintenance task priority and trigger the corresponding maintenance process based on the fault type, impact scope, and maintenance resource availability;

[0068] Adaptive maintenance execution module, used to generate AR-assisted maintenance instructions, record maintenance process data, and ensure the accuracy and traceability of maintenance operations;

[0069] The closed-loop optimization module can feed maintenance results back to the health assessment model, update the dynamic weight coefficient, and generate a monthly equipment health white paper to continuously optimize the operation and maintenance strategy;

[0070] Management database, used to store data generated during the operation and maintenance management process.

[0071] The present invention discloses a method and system for operation and maintenance management of a smart meter box, which has the following beneficial effects:

[0072] This smart meter box operation and maintenance management method uses an LSTM neural network to establish a fault prediction model, combined with a knowledge graph to match typical fault patterns. This improves fault prediction accuracy and reduces unplanned power outages. An improved genetic algorithm increases maintenance resource utilization and shortens emergency response time. Predictive maintenance replaces scheduled maintenance, reducing ineffective inspections and improving spare parts inventory turnover. AR-assisted maintenance reduces human error, improving maintenance accuracy and reducing rework rates.

[0073] 2. This smart meter box operation and maintenance management method uses EHI trend warnings to proactively intervene in potential failures, extending the average lifespan of equipment. Model self-evolution ensures long-term prediction accuracy and avoids misjudgments caused by model aging. Monthly device health distribution heat maps and cost-benefit assessment reports are generated, providing data support for power grid planning. Bayesian ranking reduces the time required to locate the root cause of a fault and reduce troubleshooting time. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0075] Figure 1 It is a schematic diagram of the overall process of the present invention;

[0076] Figure 2 Schematic diagram of the implementation process of the present invention;

[0077] Figure 3 This is a schematic diagram of the daily operation and maintenance process of the present invention;

[0078] Figure 4 Schematic diagram of the closed-loop optimization and report generation process of the present invention. DETAILED DESCRIPTION

[0079] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention are clearly and completely described. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0080] The present invention provides a smart meter operation and maintenance management method and system that addresses the problems of traditional operation and maintenance models, which fail to effectively integrate multiple sources of information, including sensor data, historical records, and environmental parameters. This leads to one-sided equipment status assessments, low fault prediction accuracy, reliance on empirical experience to locate the root cause of faults, a lack of quantitative models for resource scheduling, and the tendency for major faults to cause widespread power outages. A fault prediction model based on an LSTM neural network, combined with a knowledge graph to match typical fault patterns, improves fault prediction accuracy and reduces unplanned power outages. An improved genetic algorithm improves maintenance resource utilization and shortens emergency response time. Predictive maintenance replaces scheduled maintenance, reducing ineffective inspections and improving spare parts inventory turnover. AR-assisted maintenance reduces human error, improving maintenance accuracy and reducing rework rates. EHI trend warnings enable early intervention for potential faults, extending the average lifespan of equipment. The model's self-evolutionary capabilities ensure long-term prediction accuracy and avoid misjudgments caused by model aging. Monthly equipment health distribution heat maps and cost-benefit assessment reports are generated, providing data support for power grid planning. Bayesian ranking shortens the time it takes to locate the root cause of faults, reducing troubleshooting time.

[0081] 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.

[0082] The embodiment of the present invention discloses a method and system for operation and maintenance management of a smart electric meter box.

[0083] According to the attached Figure 1-4 As shown, the following steps are included:

[0084] Step 1: Multi-source data collection and preprocessing: deploy IoT sensors to collect multi-dimensional data and transmit it to the edge gateway via LoRaWAN for data cleaning and format standardization;

[0085] Step 2: Dynamically assess the health status of equipment by integrating real-time sensor data, historical operation and maintenance records, and environmental data. The equipment health index (EHI) is calculated using a dynamic weight fusion model, and fault warnings are then made.

[0086] Step 3: Fault root cause location and prediction: A fault prediction model is established based on the LSTM neural network. The knowledge graph is combined to match typical failure modes and output the remaining service life.

[0087] Step 4: Dynamic scheduling of maintenance tasks. Based on the fault type, impact scope, and maintenance resource availability, the maintenance task priority is dynamically calculated and the corresponding maintenance process is triggered.

[0088] Step 5: Adaptive maintenance execution, generating AR-assisted maintenance instructions, recording maintenance process data, and ensuring the accuracy and traceability of maintenance operations;

[0089] Step 6: Closed-loop optimization: Feedback maintenance results to the health assessment model, update the dynamic weight coefficient, generate a monthly equipment health white paper, and continuously optimize the operation and maintenance strategy.

[0090] Specifically disclosed, the edge gateway performs data cleaning and format standardization processing, including:

[0091] When the sensor data missing rate is greater than 5%, the adjacent node data interpolation compensation is triggered;

[0092] When an abnormal pulse signal is detected, Fourier transform is started to perform frequency domain analysis;

[0093] Convert the cleaned data into a unified JSON format and add a timestamp.

[0094] Specifically disclosed is the calculation of the Equipment Health Index (EHI) through a dynamic weight fusion model, including:

[0095] DS evidence theory is used to handle data conflicts. When the degree of conflict between sensor data and historical records is greater than 30%, the sensor weight coefficient is reduced.

[0096] Set three levels of health thresholds: EHI>0.8 for healthy status, 0.5≤EHI≤0.8 for warning status, and EHI<0.5 for fault status;

[0097] When the EHI shows a downward trend for three consecutive detection cycles and the decrease is greater than 15%, a fault warning is forcibly triggered.

[0098] Specifically disclosed, a fault prediction model is established based on an LSTM neural network, and the input features include:

[0099] Time series features: sensor data volatility over the past 72 hours;

[0100] Spatial characteristics: regional failure rate of similar equipment;

[0101] Construct a knowledge graph of power equipment, including the four-tuple relationship of fault phenomenon-fault location-fault cause-solution;

[0102] When the prediction model outputs a failure probability P > 70%, knowledge graph matching is initiated, including:

[0103] Prioritize solutions with a historical repair success rate greater than 80%;

[0104] When there are multiple failure modes, the Bayesian posterior probability is calculated to rank the root causes.

[0105] Specifically disclosed, the maintenance task priority P is calculated by an improved genetic algorithm based on the fault type, impact range, and maintenance resource availability:

[0106]

[0107] Where: S is the fault severity coefficient, U is the number of affected users, D is the maintenance resource distance, and I is the spare parts inventory coefficient;

[0108] Dynamic adjustment of weight coefficient: During the power-saving period of major events, w1 will automatically increase by 30%;

[0109] When P>preset threshold, the three-level response mechanism is triggered:

[0110] P>0.9: immediately start the emergency repair process;

[0111] 0.7<P≤0.9: Maintenance resources are dispatched within 2 hours;

[0112] P≤0.7: Included in regular maintenance plan.

[0113] Specifically disclosed is the generation of AR-assisted maintenance instructions, including:

[0114] 3D device anatomical model shows the fault location;

[0115] Overlay sensor data streams in real time;

[0116] When the operation steps are wrong, vibration feedback and voice warning will be triggered;

[0117] Record maintenance process data, including:

[0118] Maintenance personnel biometric information;

[0119] Spare parts use electronic tags;

[0120] Video stream summary of the repair process.

[0121] Specifically, the maintenance results are fed back to the health assessment model to update the dynamic weight coefficient:

[0122] When the actual fault type does not match the prediction, the weight of the corresponding feature dimension is reduced;

[0123] Generate device health white papers monthly, including:

[0124] Equipment health index distribution heat map;

[0125] Failure mode evolution trend analysis;

[0126] Maintenance cost-benefit assessment;

[0127] When the model prediction accuracy is less than 85% for three consecutive months, the model retraining process is initiated.

[0128] Furthermore, the dynamic weight fusion model adopts a particle swarm optimization algorithm, and when a sudden change in ambient temperature greater than 10°C / h is detected, the weight coefficient of the temperature sensor is temporarily increased to 0.4.

[0129] Specifically disclosed, the improved genetic algorithm comprises:

[0130] Chromosome encoding: real number encoding is used;

[0131] Fitness function: add maintenance cost constraint;

[0132] Mutation operation: When the population diversity is less than 0.2, a large mutation operation is initiated.

[0133] An intelligent electricity meter box operation and maintenance management system, the system comprising:

[0134] Multi-source data acquisition and pre-processing module, used to deploy IoT sensors to collect multi-dimensional data, which can then be transmitted to the edge gateway via LoRaWAN for data cleaning and format standardization;

[0135] The equipment health status dynamic assessment module is used to integrate real-time sensor data, historical operation and maintenance records, and environmental data, and then calculate the equipment health index (EHI) through a dynamic weight fusion model to perform fault warning judgment;

[0136] The fault root cause location and prediction module can establish a fault prediction model based on the LSTM neural network, match typical failure modes with the knowledge graph, and output the remaining service life;

[0137] The maintenance task dynamic scheduling module can dynamically calculate the maintenance task priority and trigger the corresponding maintenance process based on the fault type, impact scope, and maintenance resource availability;

[0138] Adaptive maintenance execution module, used to generate AR-assisted maintenance instructions, record maintenance process data, and ensure the accuracy and traceability of maintenance operations;

[0139] The closed-loop optimization module can feed maintenance results back to the health assessment model, update the dynamic weight coefficient, and generate a monthly equipment health white paper to continuously optimize the operation and maintenance strategy;

[0140] Management database, used to store data generated during the operation and maintenance management process.

[0141] A fault prediction model based on an LSTM neural network, combined with a knowledge graph to match typical fault patterns, improves fault prediction accuracy and reduces unplanned power outages. An improved genetic algorithm improves maintenance resource utilization and shortens emergency response time. Predictive maintenance replaces scheduled maintenance, reducing ineffective inspections and increasing spare parts inventory turnover. AR-assisted maintenance reduces human error, improving maintenance accuracy and reducing rework rates. EHI trend warnings enable early intervention in potential faults, extending the average lifespan of equipment. The model's self-evolutionary capabilities ensure long-term prediction accuracy and avoid misjudgments caused by model aging. Monthly equipment health distribution heat maps and cost-benefit assessment reports are generated to provide data support for power grid planning. Bayesian ranking reduces the time it takes to locate the root cause of a fault, reducing troubleshooting time.

[0142] Usage process:

[0143] 1. Deployment and initialization

[0144] IoT sensors such as temperature, humidity, current, and vibration sensors are installed in the meter box and connected to the edge gateway via LoRaWAN.

[0145] Configure dynamic weight fusion model parameters, such as the ambient temperature mutation threshold and weight adjustment coefficient.

[0146] Train the LSTM fault prediction model and knowledge graph, and import historical fault data and solution libraries.

[0147] 2. Daily operation and maintenance process

[0148] 2.1 Data collection and cleaning:

[0149] The sensor data is cleaned by the edge gateway, converted into JSON format and timestamped.

[0150] 2.2 Health status assessment:

[0151] The dynamic weight fusion model calculates EHI and triggers a three-level health status warning.

[0152] When EHI is less than 0.5 or decreases continuously, a fault work order is automatically generated.

[0153] 2.3 Fault prediction and root cause location:

[0154] The LSTM model outputs the fault probability P. If P > 70%, the knowledge graph matching priority solution is initiated.

[0155] Compute Bayesian posterior probabilities and rank potential failure causes.

[0156] 2.4 Task Scheduling and Execution

[0157] The improved genetic algorithm calculates the maintenance priority P and triggers the three-level response mechanism.

[0158] Maintenance personnel receive three-dimensional anatomical models and real-time data streams through AR devices and follow the instructions.

[0159] Trigger vibration / voice alerts in case of operational errors, and record biometric information and spare parts electronic tags.

[0160] 3. Closed-loop optimization and report generation

[0161] 3.1 Model Update

[0162] When the actual fault does not match the prediction, the feature weights are dynamically adjusted.

[0163] The model accuracy is evaluated monthly. If it is less than 85% for three consecutive months, the retraining process is initiated.

[0164] 3.2. Health White Paper Generation:

[0165] Summarize equipment health index distribution, failure trends, and maintenance cost-effectiveness to generate visual reports.

[0166] Provide decision-making basis for power grid upgrades, spare parts procurement, and personnel training.

[0167] Example 1: Smart meter box management in daily operation and maintenance scenarios

[0168] A regional power grid contains 500 smart meter boxes, which require daily health monitoring and predictive maintenance.

[0169] Implementation steps:

[0170] Multi-source data collection

[0171] Each meter box is deployed with temperature, humidity, current, and vibration sensors, which upload data to the edge gateway every 15 minutes via LoRaWAN.

[0172] The edge gateway performs data cleaning:

[0173] When the data missing rate of the temperature sensor of a certain meter box reaches 6%, data interpolation compensation of other meter boxes in the same area is triggered.

[0174] When abnormal pulse signals are detected, such as sudden changes in current, Fourier transform is initiated to analyze the frequency domain characteristics and identify them as poor contact interference.

[0175] The data is converted into JSON format and timestamp added before being transmitted to the cloud.

[0176] Equipment health status assessment

[0177] The dynamic weight fusion model integrates real-time data and historical records. The historical records are the operation and maintenance logs of the past year:

[0178] When the current data of a certain meter box is inconsistent with the historical average value by 35% for three consecutive hours, the weight of the current sensor is reduced to 0.3.

[0179] The calculated EHI is 0.85, indicating a healthy state. However, due to a sudden change in ambient temperature, such as a 12°C increase in 2 hours, the temperature sensor weight is temporarily increased to 0.4, and the EHI is adjusted to 0.78, indicating a warning state.

[0180] Fault prediction and root cause location

[0181] The LSTM model inputs the temperature fluctuation rate over the past 72 hours, σ = 8% and the regional failure rate of similar equipment 3%, and predicts the failure probability P = 72%.

[0182] The system starts knowledge graph matching, locates the failure mode of "high temperature causing capacitor aging", and recommends the solution of "replacing high-temperature resistant capacitors".

[0183] Maintenance task scheduling

[0184] The improved genetic algorithm calculates priority P = 0.82, S = 0.9, U = 200 households, D = 5 km, I = 0.8, triggering a 2-hour response process.

[0185] The nearest maintenance team is dispatched, and the spare parts warehouse automatically allocates high-temperature resistant capacitors.

[0186] AR-assisted maintenance execution

[0187] Maintenance personnel wear AR devices to display a three-dimensional anatomical model of the meter box, marking the capacitor location and real-time temperature data.

[0188] In case of an operation error, if the power is not turned off, the device will vibrate and give a voice prompt "Please disconnect the power first".

[0189] The maintenance process records biometric information, spare parts electronic tags and a 10-second video summary.

[0190] Closed-loop optimization

[0191] The maintenance results are fed back to the model, and the actual fault is "capacitor aging", which is consistent with the prediction, and the weight coefficient remains unchanged.

[0192] A health white paper is generated every month, showing a heat map of the EHI distribution of the meter boxes in the area. The average is 0.82, and "capacitor aging" accounts for 30% of the failure modes.

[0193] Example 2: Emergency Operation and Maintenance in Extreme Weather

[0194] The hot summer weather, with daytime temperatures exceeding 38°C, caused three electricity meter boxes in a certain area to report repairs consecutively.

[0195] Implementation steps:

[0196] Dynamic weight adjustment

[0197] When the edge gateway detects a sudden change in ambient temperature (>10°C / h), it increases the weight of the temperature sensor to 0.4 and reduces the weight of the current sensor to 0.2.

[0198] The EHI of a certain electric meter box dropped to 0.45 due to high temperature, indicating a fault state, triggering the emergency repair process.

[0199] Failure root cause prediction

[0200] The LSTM model combines the time series features of the regional failure rate of 5% and the temperature fluctuation rate σ = 15% in the past 72 hours to predict P = 88%.

[0201] The knowledge graph matches "high temperature causes carbonization of insulating materials" and recommends the solution of "replacing insulating parts."

[0202] Resource scheduling optimization

[0203] The improved genetic algorithm calculates the priority P = 0.95, S = 1.0, U = 500 households, D = 2 km, I = 0.6, and initiates emergency repair.

[0204] Drones delivered spare parts for insulating sleeves, and the maintenance team arrived at the site within 30 minutes.

[0205] Example 3: Power supply guarantee scenario for major events

[0206] A city is hosting an international conference and needs to ensure the stable operation of 100 electricity meter boxes around the conference venue.

[0207] Implementation steps:

[0208] Dynamic priority adjustment

[0209] The system automatically increases the weight of the fault severity coefficient S by 30% and reduces the weight of the maintenance resource distance D by 20%.

[0210] The EHI of a certain electricity meter box is 0.6, which indicates an early warning state. The calculated priority is P = 0.88, triggering a 2-hour response.

[0211] AR-assisted remote collaboration

[0212] Experts use AR equipment to remotely mark the fault location and guide on-site personnel in their operations.

[0213] The video stream of the maintenance process is transmitted to the command center in real time, supporting multi-role collaboration.

[0214] The basic principles, main features, and advantages of the present invention are shown and described above. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely illustrative of the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for operation and maintenance management of a smart meter box, characterized in that: The following steps are involved: Step 1: Multi-source data collection and preprocessing: deploy IoT sensors to collect multi-dimensional data and transmit it to the edge gateway via LoRaWAN for data cleaning and format standardization; Step 2: Dynamically assess the health status of equipment by integrating real-time sensor data, historical operation and maintenance records, and environmental data. The equipment health index (EHI) is calculated using a dynamic weight fusion model, and fault warnings are then made. Step 3: Fault root cause location and prediction: A fault prediction model is established based on the LSTM neural network. The knowledge graph is combined to match typical failure modes and output the remaining service life. Step 4: Dynamic scheduling of maintenance tasks. Based on the fault type, impact scope, and maintenance resource availability, the maintenance task priority is dynamically calculated and the corresponding maintenance process is triggered. Step 5: Adaptive maintenance execution, generating AR-assisted maintenance instructions, recording maintenance process data, and ensuring the accuracy and traceability of maintenance operations; Step 6: Closed-loop optimization: Feedback maintenance results to the health assessment model, update the dynamic weight coefficient, generate a monthly equipment health white paper, and continuously optimize the operation and maintenance strategy.

2. The method for operation and maintenance management of a smart meter box according to claim 1, characterized in that: The edge gateway performs data cleaning and format standardization processing, including: When the sensor data missing rate is greater than 5%, the adjacent node data interpolation compensation is triggered; When an abnormal pulse signal is detected, Fourier transform is started to perform frequency domain analysis; Convert the cleaned data into a unified JSON format and add a timestamp.

3. The method for operation and maintenance management of a smart meter box according to claim 1, characterized in that: The equipment health index (EHI) is calculated using a dynamic weight fusion model, including: DS evidence theory is used to handle data conflicts. When the degree of conflict between sensor data and historical records is greater than 30%, the sensor weight coefficient is reduced. Set three levels of health thresholds: EHI>0.8 for healthy status, 0.5≤EHI≤0.8 for warning status, and EHI<0.5 for fault status; When the EHI shows a downward trend for three consecutive detection cycles and the decrease is greater than 15%, a fault warning is forcibly triggered.

4. The method for operation and maintenance management of a smart meter box according to claim 1, characterized in that: A fault prediction model is established based on the LSTM neural network. The input features include: Time series features: sensor data volatility over the past 72 hours; Spatial characteristics: regional failure rate of similar equipment; Construct a knowledge graph of power equipment, including the four-tuple relationship of fault phenomenon-fault location-fault cause-solution; When the prediction model outputs a failure probability P > 70%, knowledge graph matching is initiated, including: Prioritize solutions with a historical repair success rate greater than 80%; When there are multiple failure modes, the Bayesian posterior probability is calculated to rank the root causes.

5. The method for operation and maintenance management of a smart meter box according to claim 1, characterized in that: According to the fault type, impact range, and maintenance resource availability, the maintenance task priority P is calculated using an improved genetic algorithm: Where: S is the fault severity coefficient, U is the number of affected users, D is the maintenance resource distance, and I is the spare parts inventory coefficient; Dynamic adjustment of weight coefficient: During the power-saving period of major events, w1 will automatically increase by 30%; When P>preset threshold, the three-level response mechanism is triggered: P>0.9: immediately start the emergency repair process; 0.7<P≤0.9: Maintenance resources are dispatched within 2 hours; P≤0.7: Included in regular maintenance plan.

6. The method for operation and maintenance management of a smart meter box according to claim 1, characterized in that: Generate AR-assisted maintenance instructions, including: 3D device anatomical model shows the fault location; Overlay sensor data streams in real time; When the operation steps are wrong, vibration feedback and voice warning will be triggered; Record maintenance process data, including: Maintenance personnel biometric information; Spare parts use electronic tags; Video stream summary of the repair process.

7. The method for operation and maintenance management of a smart meter box according to claim 1, characterized in that: Feedback the maintenance results to the health assessment model to update the dynamic weight coefficient: When the actual fault type does not match the prediction, the weight of the corresponding feature dimension is reduced; Generate monthly device health white papers, including: Equipment health index distribution heat map; Failure mode evolution trend analysis; Maintenance cost-benefit assessment; When the model prediction accuracy is less than 85% for three consecutive months, the model retraining process is initiated.

8. The method for operation and maintenance management of a smart meter box according to claim 3, characterized in that: The dynamic weight fusion model adopts a particle swarm optimization algorithm, and when a sudden change in ambient temperature greater than 10°C / h is detected, the weight coefficient of the temperature sensor is temporarily increased to 0.

4.

9. The method for operation and maintenance management of a smart meter box according to claim 5, characterized in that: The improved genetic algorithm comprises: Chromosome encoding: real number encoding is used; Fitness function: add maintenance cost constraint; Mutation operation: When the population diversity is less than 0.2, a large mutation operation is initiated.

10. An intelligent electricity meter box operation and maintenance management system, characterized in that: The system includes: Multi-source data acquisition and pre-processing module, used to deploy IoT sensors to collect multi-dimensional data, which can then be transmitted to the edge gateway via LoRaWAN for data cleaning and format standardization; The equipment health status dynamic assessment module is used to integrate real-time sensor data, historical operation and maintenance records, and environmental data, and then calculate the equipment health index (EHI) through a dynamic weight fusion model to perform fault warning judgment; The fault root cause location and prediction module can establish a fault prediction model based on the LSTM neural network, match typical failure modes with the knowledge graph, and output the remaining service life; The maintenance task dynamic scheduling module can dynamically calculate the maintenance task priority and trigger the corresponding maintenance process based on the fault type, impact scope, and maintenance resource availability; Adaptive maintenance execution module, used to generate AR-assisted maintenance instructions, record maintenance process data, and ensure the accuracy and traceability of maintenance operations; The closed-loop optimization module can feed maintenance results back to the health assessment model, update the dynamic weight coefficient, and generate a monthly equipment health white paper to continuously optimize the operation and maintenance strategy; Management database, used to store data generated during the operation and maintenance management process.

Images