Intelligent metering box monitoring method, system and device and storage medium

Abstract

The invention relates to an intelligent metering box monitoring method, system and device and a storage medium, and the method comprises the steps: continuously collecting the operation data of an intelligent metering box in an offline state of the intelligent metering box, carrying out the abnormal trend recognition of the operation data through combining with a pre-trained edge side abnormality recognition model, and obtaining a local judgment result; abnormal record data are generated according to the local judgment result, the abnormal record data are cached in a local storage unit, and the abnormal record data comprise fault types, timestamps and index values; generating a communication connection request signal under the condition that the communication connection condition is met, and controlling the communication channel to establish network connection according to the communication connection request signal; and after the communication response is received, collecting the network signal strength, and uploading the abnormal record data to a remote management platform based on the network signal strength. The intelligent metering box has the effect of improving the monitoring reliability of the intelligent metering box.

Description

Technical Field

[0001] This application relates to the technical field of remote monitoring of power metering equipment, and in particular to a method, system, device and storage medium for monitoring intelligent metering boxes. Background Technology

[0002] Currently, traditional power metering equipment monitoring systems mainly rely on manual inspections and periodic remote meter readings, which cannot achieve real-time perception and continuous monitoring of equipment status. In particular, it is difficult to obtain effective abnormal information when the equipment is disconnected from the network, offline, or has a poor signal. As a result, maintenance personnel cannot detect equipment failures in time, which seriously affects the stability of the power supply system and the service experience on the user side.

[0003] Existing solutions have attempted to achieve some remote monitoring functions by integrating sensors and communication modules, but due to limitations in communication link stability and local data processing capabilities, problems such as inability to monitor offline and omission of key abnormal information still exist, making it difficult to meet the current demand for high reliability and low latency power monitoring.

[0004] The existing technical solutions mentioned above have the following drawbacks: the existing remote monitoring systems lack the ability to collect and compensate for abnormal information in offline states, and the data transmission stability is poor in complex network environments, so there is room for improvement. Summary of the Invention

[0005] To improve the monitoring reliability of smart meter boxes, this application provides a smart meter box monitoring method, system, device, and storage medium.

[0006] The above-mentioned objective of this application is achieved through the following technical solution: A method for monitoring a smart meter box, the method comprising: When the smart meter box is offline, the operating data of the smart meter box is continuously collected. The abnormal trend of the operating data is identified by combining the pre-trained edge-side anomaly identification model to obtain the local judgment result. The operating data includes voltage, current and temperature. Anomaly record data is generated based on the local determination result and cached in the local storage unit. The anomaly record data includes fault type, timestamp, and indicator value. When the communication connection conditions are met, a communication connection request signal is generated, and the communication channel is controlled to establish a network connection based on the communication connection request signal. Upon receiving a communication response, the network signal strength is collected, and the abnormal record data is uploaded to the remote management platform based on the network signal strength.

[0007] By adopting the above technical solutions, continuous collection of operational data from the smart metering box while it is offline, combined with edge-side anomaly identification models for anomaly trend identification, enables continuous monitoring of the operational status during communication interruptions. This effectively reduces the risk of missed anomaly detections due to network issues, thereby improving the reliability and timeliness of power system fault monitoring. By generating local judgment results, constructing anomaly record data, and caching it in local storage, critical fault data is properly preserved, preventing data loss even during offline periods, thus ensuring the integrity of anomaly information. By generating a communication connection request signal and controlling the communication channel to establish a network connection when communication connection conditions are met, communication links can be triggered on demand, reducing communication energy consumption and improving connection success rates. By collecting network signal strength and uploading anomaly record data based on signal strength, the upload strategy can be adaptively adjusted according to the current network status, thereby enhancing the stability and efficiency of data transmission.

[0008] In one example, this application can be further configured such that: the intelligent metering box monitoring method also includes: Collect historical operating data samples, which include data sequences of voltage, current and temperature anomalies with multiple known tags; The historical operational data samples are normalized and sliced ​​using a sliding window to extract multi-dimensional input vectors containing time-related features; Based on the multi-dimensional input vector, a one-dimensional convolutional neural network built on the TinyML framework is used to train the model, and the pre-trained edge anomaly recognition model is obtained. The pre-trained edge-side anomaly recognition model is deployed in an intermediate format to the edge processing unit of the smart metering box for real-time inference and anomaly trend recognition.

[0009] By adopting the above technical solutions, and collecting historical voltage, current, and temperature anomaly event data samples containing known labels, and performing normalization and sliding window slicing, a time-correlated training dataset can be constructed, enhancing the model's sensitivity to trend changes. By using a one-dimensional convolutional neural network built on the TinyML framework for model training, the model's prediction accuracy can be guaranteed while maintaining its lightweight nature, making it suitable for edge device deployment scenarios. By deploying the trained model in an intermediate format to the edge processing unit, local real-time inference and anomaly trend identification can be achieved, thereby improving the timeliness and response speed of anomaly detection.

[0010] In one example, this application can be further configured as follows: the step of combining a pre-trained edge-side anomaly detection model to perform anomaly trend identification on the running data and obtain a local judgment result specifically includes: The running data is continuously segmented based on a sliding time window to form an input sequence with time order; The input sequence is input into the pre-trained edge-side anomaly recognition model to calculate confidence scores for various anomaly trends; Based on the confidence scores of various abnormal trends, the corresponding local judgment results are output.

[0011] By adopting the above technical solution, and by continuously segmenting the operating data based on a sliding time window to form a time-sequential input sequence, the time variation characteristics during equipment operation can be extracted, enhancing the model's ability to identify dynamic trends. By inputting this input sequence into a pre-trained anomaly identification model and calculating confidence scores for various anomaly trends, probabilistic identification of multiple types of faults can be achieved, improving classification accuracy. By outputting local judgment results based on the confidence scores, fault judgments can be made quickly, providing a basis for subsequent recording and uploading, thereby improving the overall intelligence level of the monitoring system.

[0012] In one example, this application can be further configured such that caching the abnormal record data in a local storage unit specifically includes: Construct a local storage structure to store the exception record data; Each abnormal record is marked with an abnormal level, and the priority storage order and index management method of each abnormal record in the local storage structure are determined according to the abnormal level, so as to improve the upload response efficiency and fault identification accuracy after the network is restored.

[0013] By adopting the above technical solutions and constructing a local storage structure for storing abnormal record data, it is possible to achieve orderly archiving of multiple abnormal information generated during offline periods, thereby enhancing the systematic nature of data management. By setting an abnormality level label for each abnormal record and determining its priority storage order and index management method accordingly, it is possible to ensure that high-risk events are saved first and are easy to retrieve and upload. This enables rapid transmission and analysis of high-priority events after network recovery, improving the efficiency of abnormal response and the accuracy of remote identification.

[0014] In one example, this application can be further configured as follows: When the communication connection conditions are met, generating a communication connection request signal and controlling the communication channel to establish a network connection based on the communication connection request signal specifically includes: If the main communication channel of the current smart meter box is available and the abnormality level of the abnormal record data reaches the preset communication trigger threshold, then the communication connection request signal is generated, and the main communication channel is prioritized to establish a network connection. If the number of consecutive failures on the main communication channel reaches a preset threshold, the system will switch to the backup communication channel to re-initiate a connection request in order to establish a network connection and upload the abnormal record data.

[0015] By adopting the above technical solution, the system controls whether to establish communication by judging whether the main communication channel is available and whether the anomaly level reaches the preset communication trigger threshold. This avoids invalid communication connection requests and improves the utilization rate of communication resources. By automatically switching to the backup communication channel to reconnect after the main channel fails a certain number of times, the system's self-recovery capability in a variable network environment can be enhanced, thereby improving the communication success rate and the online stability of the monitoring system.

[0016] In one example, this application can be further configured as follows: the uploading of the abnormal record data to the remote management platform based on the network signal strength specifically includes: Based on the network signal strength of the current communication channel, select the corresponding modulation and demodulation parameters, and use an adaptive modulation and demodulation algorithm to dynamically adjust the data transmission rate; When the network signal strength is weak, reduce the data transmission rate and enhance error correction capabilities.

[0017] By adopting the above technical solution, the modulation and demodulation parameters are selected based on the network signal strength of the current communication channel, and the data transmission rate is dynamically adjusted using an adaptive modulation and demodulation algorithm. The rate and transmission parameters can be flexibly switched according to different network conditions, thereby maintaining the data link without interruption in weak signal environments. When the signal strength is weak, the transmission rate is reduced and the error correction capability is enhanced, which can effectively reduce the transmission error rate and improve data reliability, thereby ensuring the integrity and validity of abnormal data received by the remote platform.

[0018] In one example, this application can be further configured such that: the intelligent metering box monitoring method also includes: A CRC checksum is generated in the abnormal record data corresponding to each of the local determination results, and the CRC checksum is appended to the end of the abnormal record data to form a data packet; When the remote management platform receives the data packet, it regenerates the check code based on the abnormal record data carried in the data packet, and compares the regenerated check code with the target CRC check code attached to the data packet. If the check codes do not match, a retransmission request is generated and sent to the smart metering box until a valid check is completed.

[0019] By adopting the above technical solution, a corresponding CRC checksum is generated in each abnormal record and appended to the end of the data to form a data packet. This enables a pre-set integrity verification mechanism before transmission, providing a basis for subsequent verification. By regenerating the checksum on the remote management platform and comparing it with the target CRC code, it is possible to determine whether the data has been tampered with or damaged during transmission. If the verification is inconsistent, a retransmission request is generated until the valid verification is successful. This significantly improves the accuracy and anti-interference capability of transmitted data, thereby reducing the risk of misjudgment of faults caused by data errors.

[0020] The second objective of this invention is achieved through the following technical solution: A smart meter box monitoring system, the smart meter box monitoring system comprising: An anomaly detection module is used to continuously collect the operating data of the smart meter box when it is offline, and to identify the anomaly trend of the operating data by combining it with a pre-trained edge-side anomaly detection model to obtain a local judgment result. The operating data includes voltage, current and temperature. An anomaly recording module is used to generate anomaly recording data based on the local judgment result and cache the anomaly recording data in a local storage unit. The anomaly recording data includes fault type, timestamp, and indicator value. The communication request module is used to generate a communication connection request signal when the communication connection conditions are met, and to control the communication channel to establish a network connection according to the communication connection request signal; The data upload module is used to collect network signal strength after receiving a communication response, and upload the abnormal record data to the remote management platform based on the network signal strength.

[0021] By adopting the above technical solutions, continuous collection of operational data from the smart metering box while it is offline, combined with edge-side anomaly identification models for anomaly trend identification, enables continuous monitoring of the operational status during communication interruptions. This effectively reduces the risk of missed anomaly detections due to network issues, thereby improving the reliability and timeliness of power system fault monitoring. By generating local judgment results, constructing anomaly record data, and caching it in local storage, critical fault data is properly preserved, preventing data loss even during offline periods, thus ensuring the integrity of anomaly information. By generating a communication connection request signal and controlling the communication channel to establish a network connection when communication connection conditions are met, communication links can be triggered on demand, reducing communication energy consumption and improving connection success rates. By collecting network signal strength and uploading anomaly record data based on signal strength, the upload strategy can be adaptively adjusted according to the current network status, thereby enhancing the stability and efficiency of data transmission.

[0022] The above-mentioned objective three of this application is achieved through the following technical solution: A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described intelligent metering box monitoring method.

[0023] The fourth objective of this application is achieved through the following technical solution: A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described intelligent metering box monitoring method.

[0024] In summary, this application includes the following beneficial technical effects: 1. By continuously collecting operational data in the offline state of the smart metering box and combining it with the edge-side anomaly identification model to identify abnormal trends, continuous monitoring of the operating status can be achieved during communication interruptions, effectively reducing the risk of missed anomaly detection due to network problems, thereby improving the reliability and timeliness of power system fault monitoring. 2. By generating local judgment results and constructing abnormal record data and caching it to the local storage unit, it is possible to ensure that critical fault data is properly preserved, and no data loss will occur even during offline periods, thereby ensuring the integrity of abnormal information; by generating a communication connection request signal and controlling the communication channel to establish a network connection when the communication connection conditions are met, it is possible to trigger the communication link on demand, reduce communication energy consumption, and improve the connection success rate; by collecting network signal strength and uploading abnormal record data based on the signal strength, it is possible to adaptively adjust the upload strategy according to the current network status, thereby enhancing the stability and efficiency of data transmission. Attached Figure Description

[0025] Figure 1 This is a flowchart of a smart metering box monitoring method according to one embodiment of this application; Figure 2 This is a flowchart of step S10 in a smart metering box monitoring method according to an embodiment of this application; Figure 3 This is another implementation flowchart of step S10 in a smart meter box monitoring method according to one embodiment of this application; Figure 4 This is a flowchart illustrating the implementation of step S20 in a smart metering box monitoring method according to an embodiment of this application. Figure 5 This is a flowchart illustrating the implementation of step S30 in a smart metering box monitoring method according to an embodiment of this application. Figure 6This is a flowchart illustrating the implementation of step S40 in a smart metering box monitoring method according to an embodiment of this application. Figure 7 This is another implementation flowchart of a smart meter box monitoring method in one embodiment of this application; Figure 8 This is a schematic diagram of a smart metering box monitoring system according to one embodiment of this application; Figure 9 This is a schematic diagram of a device according to one embodiment of this application. Detailed Implementation

[0026] The present application will be further described in detail below with reference to the accompanying drawings.

[0027] In one embodiment, such as Figure 1 As shown, this application discloses a monitoring method for an intelligent metering box, which specifically includes the following steps: S10: When the smart meter box is offline, continuously collect the operating data of the smart meter box, and combine it with the pre-trained edge-side anomaly identification model to identify the abnormal trend of the operating data and obtain the local judgment result. The operating data includes voltage, current and temperature.

[0028] Specifically, during the continuous collection of operational data from the smart metering box, parameters such as voltage, current, and temperature are sampled at a set sampling frequency. The sampled data is then input into an anomaly detection model pre-deployed at the edge for real-time analysis. This anomaly detection model employs a lightweight deep learning network trained on historical fault data to identify nonlinear anomaly patterns in parameter change trends. The model infers whether an anomaly trend exists in the current sampling period and generates a corresponding local judgment result. In actual operation, trigger conditions for anomaly trend detection can be set, such as current deviation exceeding a set threshold or temperature rise rate within several consecutive periods. The system incorporates features such as accelerated anomaly detection to improve the accuracy and efficiency of anomaly identification. For example, if a voltage signal experiences amplitude fluctuations exceeding ±15% within 30 seconds, and the temperature curve shows a significant upward trend within the same time window, the model can identify it as a potential localized overheating anomaly caused by poor contact. Furthermore, the intelligent metering box has a built-in offline monitoring unit, including but not limited to voltage detection circuits, current detection circuits, and temperature sensors, which can continuously collect critical operating data even in the event of a network interruption. It also features a low-power design, utilizing supercapacitors or built-in batteries as backup power to maintain the normal operation of the offline monitoring unit when the main power supply fails or communication is lost.

[0029] S20: Generate abnormal record data based on the local judgment result and cache the abnormal record data in the local storage unit. The abnormal record data includes the fault type, timestamp and indicator value.

[0030] Specifically, when the judgment result indicates the existence of an abnormal trend, the system automatically calls the data structure generation function to combine the current fault type, identification timestamp, and corresponding indicator values ​​to construct structured abnormal record data. This abnormal record data is then written into the abnormal cache area in the local storage unit. This process ensures the data update and storage efficiency of the cache area by setting the cache queue size and the first-in-first-out mechanism. At the same time, each record is attached with a unique local index number for subsequent tracking and location. If multiple abnormal types are identified during the caching process, multiple data items can be recorded in parallel to ensure that all types of fault characteristics can be completely preserved. For example, if a dual indicator offset of a sudden drop in current and an abnormal rise in temperature is detected simultaneously in an abnormal event, the system will generate two records as "current fault" and "overheating risk" respectively, along with the corresponding current amplitude, voltage stability value, and current temperature value of the equipment, forming a local abnormal cache result with dual types coexisting.

[0031] S30: When the communication connection conditions are met, generate a communication connection request signal and control the communication channel to establish a network connection according to the communication connection request signal.

[0032] Specifically, when an abnormal trend is detected in the current operating state through the edge model, a communication connection process is immediately triggered to attempt to establish a data upload link with the remote management platform. This includes determining the current network connection status of the communication module. If it is determined that the current network is disconnected or communication is unreachable, a reconnection process is initiated while waiting for the connection conditions to be met. If an available cellular network or WiFi network is detected and the signal strength is higher than a set threshold, a communication connection request is automatically generated and pushed to the communication module driver logic. Subsequently, the connection initialization process is initiated to complete the establishment of the physical link and network protocol authentication, thereby creating a stable communication channel for the remote reporting of abnormal data. For example, after detecting abnormal voltage fluctuations, if the network is not currently connected and the LTE signal strength reaches -68 dBm with a latency of less than 40ms, the 4G communication module is immediately activated to complete the VPN channel dialing. After successful establishment, the communication status is marked as "connected" for subsequent use in uploading abnormal record data.

[0033] S40: After receiving the communication response, collect the network signal strength and upload the abnormal record data to the remote management platform based on the network signal strength.

[0034] Specifically, after the remote management platform returns a successful connection response, the system collects real-time network signal strength parameters, including but not limited to RSSI value, packet loss rate, and connection latency. Based on the signal strength level, it selects the optimal data upload strategy. Under good signal conditions, batch uploads are performed; under edge signal conditions, segmented uploads and confirmation mechanisms are used to ensure data transmission integrity. Furthermore, an adaptive modulation and demodulation algorithm is introduced during data upload to dynamically adjust the data modulation method and transmission rate parameters to adapt to different channel environments. High-speed modulation schemes are used to improve transmission efficiency when the signal is strong, and low-order modulation schemes are automatically switched to enhance anti-interference and reduce packet loss probability when the signal is weak. Through these mechanisms, locally cached abnormal record data is sequentially uploaded to the data receiving interface specified by the remote management platform, while recording the upload time and confirmation receipt to ensure the reliability and traceability of data synchronization. For example, during an upload process, if the current RSSI value is detected to be -87 dBm, the packet loss rate to be 3%, and the latency to be 85ms, the adaptive algorithm adjusts the modulation method to QPSK and reduces the transmission rate from the original 512 kbps to 128 kbps. The system achieves speeds of kbps, enabling successful uploading of a single fault record even in weak signal environments and receiving a "successful upload" status receipt from the remote platform. This receipt includes the receiving time and verification signature, used to synchronize data flags and clear the corresponding local cache.

[0035] By adopting the above technical solutions, continuous collection of operational data from the smart metering box while it is offline, combined with edge-side anomaly identification models for anomaly trend identification, enables continuous monitoring of the operational status during communication interruptions. This effectively reduces the risk of missed anomaly detections due to network issues, thereby improving the reliability and timeliness of power system fault monitoring. By generating local judgment results, constructing anomaly record data, and caching it in local storage, critical fault data is properly preserved, preventing data loss even during offline periods, thus ensuring the integrity of anomaly information. By generating a communication connection request signal and controlling the communication channel to establish a network connection when communication connection conditions are met, communication links can be triggered on demand, reducing communication energy consumption and improving connection success rates. By collecting network signal strength and uploading anomaly record data based on signal strength, the upload strategy can be adaptively adjusted according to the current network status, thereby enhancing the stability and efficiency of data transmission.

[0036] In one embodiment, such as Figure 2 As shown, this intelligent metering box monitoring method also includes: S101: Collect historical operation data samples, which include voltage, current and temperature anomaly event data sequences from multiple known tags.

[0037] Specifically, by retrieving the operation logs recorded by the smart metering box under different operating conditions from the historical database, voltage, current, and temperature data sequence samples containing labeled abnormal events are selected. These samples are then clustered and archived according to event type and time period. Each data sequence is accompanied by an anomaly type label, occurrence time, and contextual parameters. For example, a sample might be "July 15, 2023, 15:32:48, high temperature anomaly, temperature rises to 75℃, lasts for 60 seconds". The data collection process includes preprocessing actions such as deduplication, anomaly segment indexing, and uniform sampling rate resampling to ensure that the training data has high confidence and representativeness, and is used for feature extraction and label-supervised training in the subsequent modeling stage.

[0038] S102: Normalize and slice the historical operation data samples using a sliding window to extract multi-dimensional input vectors containing time-related features.

[0039] Specifically, each historical data sample undergoes time normalization and amplitude standardization to ensure that different dimensional features (voltage, current, temperature) express their changing trends on the same scale. Then, a sliding window mechanism is used to divide the continuous time series into data segments of equal length. Each window contains a fixed number of time step data points, thereby constructing a sequence of information vectors with time correlation. During the sliding process, a certain overlap rate is maintained to enhance sample density and model generalization ability. For example, by using a sliding window with a length of 30 and a step size of 10 to slice a temperature anomaly event lasting 300 seconds, 27 local sequences can be generated. Each sequence represents the temperature change state before and after the current moment and is bound to the corresponding anomaly label, providing time-sensitive input features for subsequent network modeling.

[0040] S103: Based on multi-dimensional input vectors, a one-dimensional convolutional neural network built on the TinyML framework is used for model training to obtain a pre-trained edge anomaly recognition model.

[0041] Specifically, training and validation datasets are constructed based on preprocessed multi-dimensional input vectors and input into a lightweight one-dimensional convolutional neural network model. The training engine under the TinyML framework is used for edge-friendly model training. The network structure includes several convolutional layers, batch normalization layers, and activation function layers. The output layer uses a Softmax classifier to determine the anomaly category. The weight parameters are optimized through backpropagation and cross-entropy loss function. After each training round, the validation set is evaluated and the F1 score, recall, and precision are recorded. When the training converges or meets the preset threshold conditions, the model parameters and structure definition file are output. For example, in a certain training, the number of rounds is set to 100 and the learning rate is 0.001. Finally, a model with a precision of 95.6% and an F1 score of 0.921 is obtained, which has edge inference capabilities and is adapted to embedded computing resources.

[0042] S104: Deploy the pre-trained edge-side anomaly recognition model in an intermediate format to the edge processing unit of the smart metering box for real-time inference and anomaly trend recognition.

[0043] Specifically, the trained one-dimensional convolutional neural network model is quantized and pruned, then converted into an intermediate format file, such as TensorFlow Lite or ONNX. This file is then loaded into the edge processing unit of the smart metering box via remote deployment or local burning. This processing unit has a built-in lightweight model inference engine that can receive running data input in real time and perform model forward propagation to output abnormal trend prediction results. After deployment, a test inference operation is performed to verify the correctness of the model call and the response latency. For example, in one deployment, the model is packaged in ONNX format and uploaded to the device firmware. The inference engine performs forward calculations based on the CPU, with each round of inference taking about 32ms, which meets the real-time requirements. The judgment result can be fed back to the data acquisition module through the internal signal bus to link subsequent actions.

[0044] In one embodiment, such as Figure 3 As shown, in step S10, the pre-trained edge-side anomaly detection model is used to identify anomaly trends in the running data to obtain a local judgment result, specifically including: S11: The running data is continuously segmented based on the sliding time window to form an input sequence with time order.

[0045] Specifically, by applying a fixed-length sliding time window mechanism to the real-time acquired voltage, current, and temperature data streams, a continuous data segment before and after each moment is extracted to form a time-ordered input segment. As time progresses, the sliding window continuously slides forward and generates new sequence segments. All sequence segments form a set of input sequences according to the generated time order, which are used for subsequent model inference operations. This sliding mechanism can be set with a window length of 60 seconds and a step size of 10 seconds, thereby enabling fine-grained perception and dynamic analysis of the operating status within each minute. For example, in a certain current curve, the fluctuation amplitude and average level within every 60 seconds are used as the unit feature extraction window to continuously detect abrupt events.

[0046] S12: Input the input sequence into the pre-trained edge-side anomaly recognition model and calculate the confidence scores for various anomaly trends.

[0047] Specifically, the constructed time-sequential input sequence is sequentially fed into the anomaly recognition model deployed in the edge processing unit. Through the model's forward propagation process, multi-class prediction results for various potential anomalies, such as high temperature, overvoltage, and leakage, are output. A corresponding confidence score is calculated for each category. This score reflects the model's confidence in whether the current input exhibits this type of anomaly. The confidence score range is usually between 0 and 1, representing a probability-level judgment. For example, in the model's inference output for a certain sequence segment, if the confidence score for high temperature anomaly is 0.92, the leakage score is 0.07, and the overvoltage score is 0.01, then it can be determined that the current sample exhibits a significant high temperature trend.

[0048] S13: Output the corresponding local judgment result based on the confidence score of various abnormal trends.

[0049] Specifically, based on the confidence scores of various abnormal trends output by the model, the anomaly type with the highest score that exceeds the set threshold is selected as the current local judgment result. This judgment result is then bound to the current timestamp to form a set of anomaly identification records. These records will serve as the input basis for subsequent upload judgments and local alarm triggering. If all score values ​​are lower than the judgment threshold, a "normal" judgment is output. For example, when the confidence score for high temperature is 0.93, exceeding the 0.9 threshold, a "high temperature anomaly" judgment is generated and the timestamp of August 27, 2025, at 10:36:45 is recorded. Such local judgment results can be used for linkage processing of upload strategies, protection mechanisms, and operation logs.

[0050] In one embodiment, such as Figure 4 As shown, in step S20, the abnormal record data is cached in the local storage unit, which specifically includes: S21: Construct a local storage structure to store exception record data.

[0051] Specifically, a lightweight local data table structure is initialized at the edge device to build an exception record cache system for categorized storage and retrieval. In this structure, a unique index number is assigned to each exception record and a first-in-first-out queue is maintained according to its generation timestamp order to facilitate subsequent querying and upload management. The constructed local storage structure supports breakpoint resume and data integrity verification mechanisms. For example, a combination of a circular queue and a hash mapping table is used to achieve efficient writing and positioning functions, ensuring that the most recent N key exception information records can be retained even under network interruption or frequent power outages. At the same time, this local storage structure can be configured according to device resource constraints. It can adopt a limited-capacity circular cache structure combined with a high-level priority upload strategy to ensure that no key data is missed, or it can adopt an scalable or dynamically managed structure to save all exception record data when resources allow.

[0052] S22: Set an anomaly level label for each abnormal record data, and determine the priority storage order and index management method of each abnormal record data in the local storage structure according to the anomaly level, so as to improve the upload response efficiency and fault identification accuracy after the network is restored.

[0053] Specifically, a predefined anomaly level label is set for each anomaly record stored locally. The level classification is based on the confidence score of the anomaly trend, the anomaly severity label in the historical pattern, and the stability index of the current monitoring status. The levels are set into three levels: high, medium, and low. Each level corresponds to a different cache priority and index position strategy. High-level anomalies are prioritized to be stored in the main cache and kept active in real time, while low-level anomalies can be delayed or uploaded in batches when there is low load. For example, if an anomaly record is identified as a "high temperature + overcurrent" composite event and the confidence score exceeds 0.95, it is marked as high-level and inserted at the front of the index list. At the same time, an upload flag is attached to prioritize the transmission operation when the network recovers. Under the premise that the cache structure type is configurable, the anomaly level can also be linked with the cache capacity strategy to realize differentiated processing strategies to improve storage utilization and upload response efficiency.

[0054] In one embodiment, such as Figure 5 As shown, in step S30, that is, when the communication connection conditions are met, a communication connection request signal is generated, and the communication channel is controlled to establish a network connection according to the communication connection request signal, specifically including: S31: If the main communication channel of the current smart meter box is available and the abnormality level of the abnormal recorded data reaches the preset communication trigger threshold, a communication connection request signal is generated, and the main communication channel is prioritized to establish a network connection.

[0055] Specifically, when the anomaly level corresponding to the local judgment result output by the edge inference is higher than or equal to the preset communication trigger level threshold, and the main communication channel is currently detected to be in a connected and available state, the action of actively establishing network communication is triggered. Specifically, a communication connection request signal is generated according to the transmission priority corresponding to the anomaly level, and the network access process is driven to initialize the authentication and handshake process of the main communication channel. At the same time, the current load of the communication channel and the network signal quality are judged in combination with the access policy. After confirming that the basic connection conditions are met, the main communication channel is prioritized to establish a data path and enter the data upload preparation state. During this process, the anomaly record data is marked as "pending upload" to prevent duplicate queuing.

[0056] S32: If the number of consecutive failures of the main communication channel reaches a preset threshold, switch to the backup communication channel to re-initiate the connection request in order to establish a network connection and upload the abnormal record data.

[0057] Specifically, if the main communication channel encounters consecutive timeout responses, interruption returns, or network layer handshake failures during the connection initiation process, and the cumulative number of failures reaches a preset threshold, the current channel failure risk level is immediately determined according to the retry policy, and the communication path is automatically switched to the backup communication channel. On this basis, the connection request data packet is reconstructed and the access process of the backup channel is initiated. A different authentication method or frequency band configuration than the main channel is used for independent connection attempts. For example, the main channel is a cellular data link while the backup channel is a Wi-Fi link. During the switching process, the current communication status table and the transmission channel field of the abnormal record are updated synchronously to ensure that the record upload action has a complete backtracking path and reduce the risk of data upload interruption caused by communication failure.

[0058] In one embodiment, such as Figure 6 As shown, in step S40, which involves uploading the abnormal record data to the remote management platform based on the network signal strength, the specific steps include: S41: Select the corresponding modulation and demodulation parameters based on the network signal strength of the current communication channel, and dynamically adjust the data transmission rate using an adaptive modulation and demodulation algorithm.

[0059] Specifically, once the network connection is successfully established, based on the network signal strength data of the current communication channel, such as Received Signal Strength Indicator (RSSI) or Channel Quality Indicator (CQI), the corresponding modulation and demodulation parameter combination is automatically selected, including modulation method, code rate, and coding scheme. The data modulation strategy used in the actual transmission process is dynamically adjusted through a preset adaptive modulation and demodulation algorithm. When the signal strength changes, the bit error rate and throughput of the current parameter combination are evaluated in real time, and the transmission rate and modulation order are adjusted as needed. This maximizes data transmission efficiency while maintaining connection stability. For example, when the signal is strong, the system automatically switches to 16QAM or 64QAM modulation mode to increase throughput, and actively downgrades to QPSK mode to reduce the risk of bit error when the signal weakens.

[0060] S42: When the network signal strength is weak, reduce the data transmission rate and enhance error correction capability.

[0061] Specifically, when the network signal strength is lower than the set weak signal threshold, such as RSSI less than -100dBm or SINR close to the noise lower limit, the adaptive modulation and demodulation algorithm actively reduces the data transmission rate according to the current channel conditions. At the same time, it selects a coding method with stronger error correction capabilities, such as low code rate Turbo coding or LDPC coding, to resist the spread of bit errors caused by channel fading. By introducing redundant bits and a longer error correction interval in the modulation parameters, the error detection and recovery capabilities are improved. When necessary, the packet retransmission mechanism and the intra-packet verification mechanism can also be enabled to ensure that the basic data upload capability is still available when the channel quality deteriorates. For example, the original 20KB data packet is split into multiple 4KB small packets and uploaded in segments with CRC verification to enhance reliability.

[0062] In one embodiment, such as Figure 7 As shown, this intelligent metering box monitoring method also includes: S50: Generate a CRC checksum in the abnormal record data corresponding to each local judgment result, and append the CRC checksum to the end of the abnormal record data to form a data packet.

[0063] Specifically, after completing the local anomaly trend identification and confidence score, a verification preparation operation is performed on each generated local judgment result and its corresponding anomaly record data. First, a cyclic redundancy check (CRC) code is generated based on all the content bits of the anomaly record data. A fixed-length verification sequence is extracted using a polynomial generation algorithm for subsequent consistency comparison. The CRC check code is then inserted as an additional field at the end of the corresponding anomaly record data to form a fully encapsulated data packet. The data packet structure includes the original anomaly data segment and the appended CRC segment. This operation ensures that the data integrity verification capability is fast and effective during the subsequent remote platform reception stage.

[0064] S60: When the remote management platform receives the data packet, it regenerates the check code based on the abnormal record data carried in the data packet, and compares the regenerated check code with the target CRC check code attached to the data packet. If the check codes do not match, a retransmission request is generated and sent to the smart metering box until a valid check is completed.

[0065] Specifically, after the remote management platform completes the data packet reception, it automatically calls the data unpacking and verification module to re-execute the CRC calculation process for the abnormal record data segments in the data packet. It uses the same polynomial algorithm as the smart meter box to generate the receiving side CRC check code and compares the calculation result bit by bit with the target CRC check code attached to the original data packet. If the two match, the data integrity is confirmed and the abnormal record storage process is initiated. If the comparison does not match, it is determined that the data packet has a bit error or information loss during transmission. The corresponding data packet retransmission request instruction is immediately constructed and sent back to the smart meter box. After receiving the retransmission request, the smart meter box retransmits the corresponding data packet until the platform side completes the valid verification.

[0066] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0067] In one embodiment, a smart meter box monitoring system is provided, which corresponds one-to-one with the smart meter box monitoring method described in the above embodiments. For example... Figure 8 As shown, this intelligent metering box monitoring system includes an anomaly identification module, an anomaly recording module, a communication request module, and a data upload module. Detailed descriptions of each functional module are as follows: The anomaly detection module is used to continuously collect the operating data of the smart meter box when it is offline. It combines the pre-trained edge-side anomaly detection model to identify anomaly trends in the operating data and obtain local judgment results. The operating data includes voltage, current and temperature. The anomaly logging module is used to generate anomaly logging data based on local judgment results and cache the anomaly logging data in the local storage unit. The anomaly logging data includes fault type, timestamp and indicator value. The communication request module is used to generate a communication connection request signal when the communication connection conditions are met, and to control the communication channel to establish a network connection based on the communication connection request signal. The data upload module is used to collect network signal strength after receiving a communication response, and upload abnormal record data to the remote management platform based on the network signal strength.

[0068] Optionally, the intelligent metering box monitoring system also includes: The sample acquisition module is used to collect historical operating data samples, which include voltage, current and temperature anomaly event data sequences of multiple known tags; The data processing module is used to normalize and slice historical operational data samples using a sliding window, and extract multi-dimensional input vectors containing time-related features. The training module is used to train a model based on a one-dimensional convolutional neural network built on the TinyML framework, using multi-dimensional input vectors, to obtain a pre-trained edge anomaly recognition model. The model deployment module is used to deploy the pre-trained edge-side anomaly recognition model in an intermediate format to the edge processing unit of the smart metering box for real-time inference and anomaly trend recognition.

[0069] Optionally, the anomaly detection module includes: The sequence construction submodule is used to continuously segment the running data based on a sliding time window to form an input sequence with time order. The model recognition submodule is used to input the input sequence into a pre-trained edge-side anomaly recognition model and calculate the confidence scores of various anomaly trends. The results output submodule is used to output the corresponding local judgment results based on the confidence scores of various abnormal trends.

[0070] Optional, the exception logging module includes: The structure building submodule is used to build the local storage structure for storing exception record data; The storage order confirmation submodule is used to set an anomaly level label for each abnormal record data, and determine the priority storage order and index management method of each abnormal record data in the local storage structure according to the anomaly level, so as to improve the upload response efficiency and fault identification accuracy after the network is restored.

[0071] Optionally, the communication request module includes: The confirmation request submodule is used to generate a communication connection request signal and prioritize the establishment of a network connection on the main communication channel if the current main communication channel of the smart meter box is available and the abnormality level of the abnormal record data reaches the preset communication trigger threshold. The channel switching submodule is used to switch to the backup communication channel to re-initiate the connection request if the number of consecutive failures of the main communication channel reaches a preset threshold, so as to establish a network connection and upload the abnormal record data.

[0072] Optionally, the data upload module includes: The parameter determination submodule is used to select the corresponding modulation and demodulation parameters based on the network signal strength of the current communication channel, and to dynamically adjust the data transmission rate using an adaptive modulation and demodulation algorithm; The adjustment submodule is used to reduce the data transmission rate and enhance error correction capabilities when the network signal strength is weak.

[0073] Optionally, the intelligent metering box monitoring system also includes: The check code generation module is used to generate a CRC check code in the abnormal record data corresponding to each local judgment result, and append the CRC check code to the end of the abnormal record data to form a data packet. The checksum comparison module is used to regenerate a checksum based on the abnormal record data carried in the data packet after the remote management platform receives the data packet, and compare the regenerated checksum with the target CRC checksum attached to the data packet. If the checksums do not match, a retransmission request is generated and sent to the smart metering box until a valid checksum is completed.

[0074] For specific limitations regarding the intelligent meter box monitoring system, please refer to the limitations of the intelligent meter box monitoring method described above, which will not be repeated here. Each module in the aforementioned intelligent meter box monitoring system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0075] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 9 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a smart metering box monitoring method.

[0076] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps: When the smart meter box is offline, the operating data of the smart meter box is continuously collected. Combined with the pre-trained edge-side anomaly recognition model, the operating data is used to identify abnormal trends and obtain local judgment results. The operating data includes voltage, current and temperature. Anomaly log data is generated based on the local judgment results and cached in the local storage unit. The anomaly log data includes the fault type, timestamp, and indicator value. When the communication connection conditions are met, a communication connection request signal is generated, and the communication channel is controlled to establish a network connection based on the communication connection request signal; Upon receiving a communication response, the system collects network signal strength and uploads abnormal record data to the remote management platform based on the network signal strength.

[0077] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor: When the smart meter box is offline, the operating data of the smart meter box is continuously collected. Combined with the pre-trained edge-side anomaly recognition model, the operating data is used to identify abnormal trends and obtain local judgment results. The operating data includes voltage, current and temperature. Anomaly log data is generated based on the local judgment results and cached in the local storage unit. The anomaly log data includes the fault type, timestamp, and indicator value. When the communication connection conditions are met, a communication connection request signal is generated, and the communication channel is controlled to establish a network connection based on the communication connection request signal; Upon receiving a communication response, the system collects network signal strength and uploads abnormal record data to the remote management platform based on the network signal strength.

[0078] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0079] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0080] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A monitoring method for an intelligent metering box, characterized in that, The intelligent metering box monitoring method includes: When the smart meter box is offline, the operating data of the smart meter box is continuously collected. The abnormal trend of the operating data is identified by combining the pre-trained edge-side anomaly identification model to obtain the local judgment result. The operating data includes voltage, current and temperature. Anomaly record data is generated based on the local determination result and cached in the local storage unit. The anomaly record data includes fault type, timestamp, and indicator value. When the communication connection conditions are met, a communication connection request signal is generated, and the communication channel is controlled to establish a network connection based on the communication connection request signal. Upon receiving a communication response, the network signal strength is collected, and the abnormal record data is uploaded to the remote management platform based on the network signal strength.

2. The intelligent metering box monitoring method according to claim 1, characterized in that, The intelligent metering box monitoring method further includes: Collect historical operating data samples, which include data sequences of voltage, current and temperature anomalies with multiple known tags; The historical operational data samples are normalized and sliced ​​using a sliding window to extract multi-dimensional input vectors containing time-related features; Based on the multi-dimensional input vector, a one-dimensional convolutional neural network built on the TinyML framework is used to train the model, and the pre-trained edge anomaly recognition model is obtained. The pre-trained edge-side anomaly recognition model is deployed in an intermediate format to the edge processing unit of the smart metering box for real-time inference and anomaly trend recognition.

3. The intelligent metering box monitoring method according to claim 1, characterized in that, The step of combining a pre-trained edge-side anomaly detection model to identify anomaly trends in the running data and obtaining a local judgment result specifically includes: The running data is continuously segmented based on a sliding time window to form an input sequence with time order; The input sequence is input into the pre-trained edge-side anomaly recognition model to calculate confidence scores for various anomaly trends; Based on the confidence scores of various abnormal trends, the corresponding local judgment results are output.

4. The intelligent metering box monitoring method according to claim 1, characterized in that, The step of caching the abnormal record data in the local storage unit specifically includes: Construct a local storage structure to store the exception record data; Each abnormal record is marked with an abnormal level, and the priority storage order and index management method of each abnormal record in the local storage structure are determined according to the abnormal level, so as to improve the upload response efficiency and fault identification accuracy after the network is restored.

5. The intelligent metering box monitoring method according to claim 1, characterized in that, The step of generating a communication connection request signal when the communication connection conditions are met, and controlling the communication channel to establish a network connection based on the communication connection request signal, specifically includes: If the main communication channel of the current smart meter box is available and the abnormality level of the abnormal record data reaches the preset communication trigger threshold, then the communication connection request signal is generated, and the main communication channel is prioritized to establish a network connection. If the number of consecutive failures on the main communication channel reaches a preset threshold, the system will switch to the backup communication channel to re-initiate a connection request in order to establish a network connection and upload the abnormal record data.

6. The intelligent metering box monitoring method according to claim 1, characterized in that, The process of uploading the abnormal record data to the remote management platform based on the network signal strength specifically includes: Based on the network signal strength of the current communication channel, select the corresponding modulation and demodulation parameters, and use an adaptive modulation and demodulation algorithm to dynamically adjust the data transmission rate; When the network signal strength is weak, reduce the data transmission rate and enhance error correction capabilities.

7. The intelligent metering box monitoring method according to claim 1, characterized in that, The intelligent metering box monitoring method further includes: A CRC checksum is generated in the abnormal record data corresponding to each of the local determination results, and the CRC checksum is appended to the end of the abnormal record data to form a data packet; When the remote management platform receives the data packet, it regenerates the check code based on the abnormal record data carried in the data packet, and compares the regenerated check code with the target CRC check code attached to the data packet. If the check codes do not match, a retransmission request is generated and sent to the smart metering box until a valid check is completed.

8. A smart metering box monitoring system, characterized in that, The intelligent metering box monitoring system includes: An anomaly detection module is used to continuously collect the operating data of the smart meter box when it is offline, and to identify the anomaly trend of the operating data by combining it with a pre-trained edge-side anomaly detection model to obtain a local judgment result. The operating data includes voltage, current and temperature. An anomaly recording module is used to generate anomaly recording data based on the local judgment result and cache the anomaly recording data in a local storage unit. The anomaly recording data includes fault type, timestamp, and indicator value. The communication request module is used to generate a communication connection request signal when the communication connection conditions are met, and to control the communication channel to establish a network connection according to the communication connection request signal; The data upload module is used to collect network signal strength after receiving a communication response, and upload the abnormal record data to the remote management platform based on the network signal strength.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the intelligent metering box monitoring method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the intelligent metering box monitoring method as described in any one of claims 1 to 7.

Images