A monitoring system for a sorting line of returned smart meters

Abstract

This invention discloses a monitoring system for a sorting line of returned smart meters, comprising a feature fusion sorting module, a digital mapping control module, a chain traceability management module, a status diagnosis module, and an environmental control module. The feature fusion sorting module collects meter data using an industrial camera array, microphone array, structured light scanner, and infrared thermal imager, and performs feature fusion classification. The digital mapping control module uses a distributed control model and optimization algorithm to calculate sorting process control parameters. The chain traceability management module uses blockchain technology to record data and enable information retrieval. The status diagnosis module combines data prediction technology to predict equipment failures. The environmental control module monitors key parameters of the sorting environment. This system improves the automation and intelligence level of the meter sorting line and solves problems such as low sorting efficiency and lagging equipment monitoring in existing systems.

Description

Technical Field

[0001] This invention belongs to the field of smart meter processing technology, specifically relating to a monitoring system for a sorting line for returned smart meters. Background Technology

[0002] With the continuous advancement of smart grid construction, smart meters have been widely used in power systems. In actual operation, due to equipment failures, upgrades, and other reasons, a large number of smart meters need to be returned to service. These returned meters require multiple stages, including sorting, testing, and repair. However, existing systems for handling returned smart meters face the following technical problems:

[0003] 1. Traditional manual sorting methods are inefficient:

[0004] Due to the wide variety of types and models of smart meters, manual sorting is not only time-consuming and labor-intensive, but also prone to classification errors, affecting the accuracy of subsequent processing. Manual recording leads to data distortion, failing to meet the needs of refined management and impacting processing efficiency and quality.

[0005] 2. Automated sorting line systems are complex and difficult to coordinate:

[0006] The processing of returned smart meters involves multiple stages, such as conveying, testing, dismantling, and labeling. These stages require the coordinated control of various types of equipment, including conveyor belts, robots, pressure testing devices, appearance inspection equipment, and laser engraving machines. In existing automated production line systems, the lack of effective integration and coordination among various devices results in complex and error-prone automated systems that make it difficult to guarantee the efficiency and accuracy of the sorting process.

[0007] 3. Lack of a unified monitoring platform, resulting in significant management difficulties:

[0008] Existing systems often fail to provide a centralized and unified monitoring platform, making it difficult to monitor and manage equipment and processes in real time. Because sorting lines involve a wide variety of equipment distributed across a large area, with numerous data collection points, existing systems struggle to detect and handle various anomalies in a timely manner, resulting in delayed fault response and low production efficiency and equipment utilization.

[0009] 4. Difficulty in tracing the entire process:

[0010] The various data generated during the processing of returned smart meters were not effectively correlated and integrated, making it difficult to trace the entire process from when the meter entered the processing stage to when it was finally processed. This not only limits the ability for quality management and continuous improvement, but also increases the difficulty of tracing the source of potential failures.

[0011] 5. Delayed equipment status monitoring and slow maintenance response:

[0012] The existing system lacks real-time monitoring of equipment operating status, resulting in delayed response when equipment failures occur. The lack of effective equipment performance assessment makes it difficult to implement preventative maintenance measures, thereby increasing the frequency of equipment failures and repair costs.

[0013] 6. Lack of security monitoring:

[0014] The process of returning smart meters involves potential hazards such as mechanical movement and high-voltage testing. Existing systems typically lack effective safety monitoring of work areas, personnel access, and abnormal operating conditions, making it impossible to promptly detect and address safety risks, thus affecting operational safety.

[0015] 7. Inadequate environmental monitoring affects treatment quality:

[0016] During the processing of returned smart meters, environmental factors such as temperature, humidity, air pressure, and noise have a significant impact on the processing quality. However, existing systems fail to effectively monitor these environmental factors, resulting in unstable processing outcomes.

[0017] 8. Insufficient remote management capabilities:

[0018] Most current systems can only perform on-site management and lack remote monitoring, diagnostics, and technical support capabilities, which limits management efficiency and technical service levels. In particular, when operating across regions or performing remote equipment diagnostics, timely response and effective intervention are not possible.

[0019] Therefore, there is an urgent need to develop a monitoring system for the sorting line of returned smart meters, which can improve the automation and intelligence level of returned meter processing and solve the above-mentioned technical problems. Summary of the Invention

[0020] This invention provides a monitoring system for a sorting line for returned smart meters, which solves the technical problems of low efficiency of manual sorting, complex line control, lack of traceability of processing, lagging equipment status monitoring, and lack of environmental safety monitoring in the sorting line for returned smart meters.

[0021] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0022] A monitoring system for a sorting line of returned smart meters includes: a feature fusion sorting module, a digital mapping control module, a chain traceability management module, a status diagnosis module, and an environmental control module; wherein:

[0023] The feature fusion sorting module is located at the system input end. It acquires images of the meter's appearance through an industrial camera array, collects acoustic signals of mechanical operation through a microphone array, obtains 3D point cloud data through a structured light scanner, and acquires temperature field data through an infrared thermal imager. It processes the image data using a hierarchical attention network (Swin Transformer), processes the acoustic signals using Mel spectrum analysis, processes the 3D point cloud using a point cloud neural network (PointNet++), processes the temperature field based on a region growing algorithm, and fuses the above features using a weighted attention mechanism. The classification results are then transmitted to the digital mapping control module.

[0024] The digital mapping control module receives classification results and equipment status data, constructs an equipment state transition model using a time Petri net, achieves cross-equipment state consensus based on the Byzantine Fault Tolerance (PBFT) mechanism, calculates sorting process control parameters using a distributed optimization control algorithm based on the Alternating Direction Multiplier (ADMM) method, renders the equipment operating status in real time using WebAssembly technology, and sends the control parameters to the execution equipment.

[0025] The chain traceability management module collects the serial number of the electricity meter through a barcode scanner, obtains the classification identifier and process time data from the digital mapping control module, establishes a data storage structure using sharded blockchain technology, records the electricity meter transfer process using a directed acyclic graph, establishes a time-series index based on a logarithmic structure merge (LSM) tree, and constructs a mapping relationship between electricity meter processing and equipment status through the graph database Neo4j, providing operational data for the status diagnosis module.

[0026] The condition diagnosis module collects equipment operating parameters through speed sensors and vibration sensors, constructs a parameter time series prediction model using a bidirectional gated cyclic network (BiGRU), eliminates parameter distribution differences between devices using a domain adversarial network, extracts fault feature frequency bands through wavelet packet decomposition, and generates equipment maintenance decisions by combining a near-end strategy optimization algorithm, and transmits fault warning information to the digital mapping control module.

[0027] The environmental control module collects temperature and humidity parameters of the sorting environment through a network of temperature and humidity sensors, obtains personnel location data using a lidar scanner, identifies work behaviors using a spatiotemporal convolutional network, predicts motion trajectories using a Kalman filter algorithm, monitors sorting environment parameters based on a hierarchical decomposition computational fluid dynamics model, and transmits the monitoring information of the sorting operation site to the digital mapping control module for linkage processing.

[0028] The beneficial effects of this invention are:

[0029] 1. Improve sorting efficiency and accuracy

[0030] By integrating multiple sensors (industrial camera array, microphone array, structured light scanner, and infrared thermal imager) into a feature fusion sorting module for data acquisition and processing, this invention enables rapid sorting of returned smart meters, improving sorting efficiency and reducing misclassification. This automated approach significantly enhances overall sorting speed and processing accuracy.

[0031] 2. Simplify system control and optimize.

[0032] The digital mapping control module optimizes device state coordination through time-based Petri nets and Byzantine fault-tolerant mechanisms (PBFT), simplifying the control process across devices. Combined with the distributed optimization algorithm of the Alternating Direction Multiplier Method (ADMM), it optimizes control parameters in the sorting process, thereby improving system response speed and reducing inter-device coordination latency.

[0033] 3. Achieve end-to-end data traceability

[0034] The chain-based traceability management module uses sharded blockchain technology and a directed acyclic graph (DAG) to record the meter transfer process. Combined with the graph database Neo4j, it constructs a mapping relationship between meters and equipment status, ensuring full traceability of the returned meters processing. This innovation improves data transparency and provides data support for subsequent quality control and management.

[0035] 4. Improve equipment operation monitoring and fault diagnosis

[0036] The condition diagnosis module collects equipment operating parameters in real time and uses a bidirectional gated cyclic network (BiGRU) to predict equipment status, enabling early identification of potential equipment faults. By combining wavelet packet decomposition technology to extract fault characteristic frequency bands, it optimizes equipment maintenance decisions, helping to reduce equipment failure rates and maintenance costs.

[0037] 5. Strengthen environmental monitoring and safety management

[0038] The environmental management module collects operational environment data through multiple sensors and combines this data with a spatiotemporal convolutional network to identify operational behaviors, monitoring factors such as temperature, humidity, and personnel location in the operational environment in real time. This technology improves environmental safety, ensures safety at the work site, and reduces potential safety risks caused by environmental changes or improper personnel behavior. Attached Figure Description

[0039] Figure 1 A schematic diagram of the system architecture of the present invention is shown;

[0040] Figure 2 A system workflow diagram of the present invention is shown. Detailed Implementation

[0041] Exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0042] Combination Figure 1 This invention provides a monitoring system for a sorting line of returned smart meters, the system of which consists of the following components:

[0043] 1. Feature Fusion Sorting Module

[0044] The feature fusion sorting module, serving as the system's input, uses a multimodal sensor array to collect feature information from returned smart meters. It then extracts and fuses these features using a deep learning model, ultimately achieving accurate meter classification. The implementation process of this module includes the following steps:

[0045] A1). Multimodal data acquisition

[0046] This module simultaneously acquires multimodal data using four sensor arrays. The specific sensor parameters and their data acquisition parameters are set as follows:

[0047] a) Image data acquisition:

[0048] It employs a four-high-resolution industrial camera array, with a resolution of no less than 2048×1536 pixels, arranged in a tetrahedral pattern to obtain omnidirectional images of the electricity meter.

[0049] The images acquired by each camera at time t are:

[0050]

[0051] in, This represents the image captured by the i-th camera at time t. Represents the set of real numbers. Indicates the image height. This indicates the image width, and 3 represents the three RGB channels. Image data is acquired synchronously via hardware to ensure the consistency of image acquisition timing.

[0052] b) Acoustic signal acquisition:

[0053] An acoustic signal data acquired using a circular array of eight omnidirectional microphones at a sampling frequency of 48kHz is represented as follows:

[0054]

[0055] in, This represents the acoustic signal collected by the j-th microphone at time t. This indicates the number of sampling points, typically 2048. The microphone array works synchronously with the image sensor to ensure timing matching between the acoustic signal and image data.

[0056] c) Point cloud data acquisition:

[0057] The three-dimensional point cloud data of the electricity meter was acquired using a combination of structured light scanner and binocular camera, and is represented as follows:

[0058]

[0059] in, This represents the 3D point cloud data of the electricity meter collected at time t. The number of points in the cloud, where each point contains three-dimensional spatial coordinates. k is an index variable, taking values ​​[1, N]. During data acquisition, laser scanning is used in sync with the image to ensure the integrity and accuracy of the 3D point cloud data.

[0060] d) Temperature field data acquisition:

[0061] Temperature field data was acquired using an infrared thermal imager with a resolution of [resolution value missing]. , represented as:

[0062]

[0063] in, This represents the temperature field data collected at time t. This indicates the resolution of the temperature field. The sensor's data acquisition process is synchronized with other sensors to ensure that ambient temperature information is obtained simultaneously within the same timeframe.

[0064] A2). Feature Extraction and Processing

[0065] Different methods are used to extract and process features from the collected multimodal data.

[0066] a) Image data processing:

[0067] For image data, the Swing Transformer is used for processing. The input image patch sequence is... If so, then the formula for calculating attention is:

[0068]

[0069] in, , , These are query, key, and value matrices, respectively. Scaling factor This is the relative position offset matrix. Specific parameter settings for the Swin Transformer include: patch size (16×16), number of Transformer layers (4 layers), and number of hidden units per layer (256), to ensure effective extraction of image features.

[0070] b) Acoustic signal processing:

[0071] For acoustic signals, feature extraction is performed using Mel-spectrum analysis. The formula for calculating the Mel-spectrum is:

[0072]

[0073] in, For linear frequency, The Mel frequency is specified. The parameters for Mel spectrum analysis include the number of Mel filters (40) and the frequency band (0 to 24 kHz) to suit the operating environment of decommissioned meters.

[0074] c) Point cloud data processing:

[0075] For 3D point cloud data, the PointNet++ network is used for processing, employing a hierarchical sampling and grouping strategy. Layer features are used The calculation formula is as follows:

[0076]

[0077] in, For the first Layer features, For the learnable feature extraction function, a multilayer perceptron (MLP) is used here. Point The neighborhood set, x and represent the coordinates of the i-th point and the current point, respectively. PointNet++ parameter settings include neighborhood size (16 points) and number of layers (3 layers) to ensure effective feature extraction from point cloud data.

[0078] A3). Feature fusion decision

[0079] To fuse feature data from different modalities, a weighted attention mechanism is employed. The first... The characteristics of the mode are The formula for calculating its attention weight is:

[0080]

[0081] in, For the first Modal attention weights, , The weight matrix is ​​a learnable matrix. , Respectively The model defines the modalities, the eigenvectors of the k-th modality, and exp as the natural exponential function. This mechanism allows the model to automatically adjust the weights of each modality to ensure that the most informative modality dominates the final decision.

[0082] The fused feature vector is:

[0083]

[0084] in, This is the fused feature vector.

[0085] A4). Classification Decision

[0086] Finally, the Softmax function is used for classification decisions. The fused features are defined as follows: The number of categories is Then the classification probability is:

[0087]

[0088] in, For the first The probability of a class , This is the weight matrix.

[0089] 2. Digital mapping control module

[0090] The digital mapping control module is the core control unit of the system. It receives the classification results from the feature fusion sorting module and interacts with each execution device in real time to ensure the efficient and coordinated operation of the sorting line. This module constructs a virtual mapping using digital twin technology to achieve real-time monitoring and optimized control of the physical equipment. The specific implementation process is as follows:

[0091] B1) Equipment Status Modeling

[0092] A time-based Petri net is used to construct a device state transition model. It is defined as follows:

[0093]

[0094] in:

[0095] This is a set of locations, representing the possible states of the device (such as standby, running, fault, etc.).

[0096] This is a set of transitions, representing events that trigger state changes;

[0097] For flow relationships, describe the state transition path;

[0098] This is the weighting function, representing the priority of the transformation conditions;

[0099] This is the initial identifier, representing the initial state of the system;

[0100] Define the time constraints for state transitions as a time interval function.

[0101] For the first on the assembly line The state transition equation for a device is:

[0102]

[0103] in: express Time device The state vector; Indicates a control action; Indicates the influence of environmental factors; This is the state transition function.

[0104] B2). Cross-device state consensus

[0105] The consensus process leverages the PBFT (Practical Byzantine Fault Tolerance) mechanism to achieve state consistency among devices, ensuring the system continues to operate normally even when some devices fail. The consensus process includes:

[0106] a) Request phase:

[0107]

[0108] in: For equipment Control commands; For timestamps; For equipment identification; For digital signatures.

[0109] b) Response during the preparatory phase:

[0110]

[0111] in: Number the views; For serial number; Request a summary; Sign the master node.

[0112] B3). Distributed Optimization Control

[0113] Distributed optimization control is achieved using the ADMM (Alternating Direction Multiplier Method) algorithm, which decomposes the global control problem into multiple local subproblems.

[0114] a) Objective function:

[0115]

[0116]

[0117] in:

[0118] For the first The local objective function of a device represents the device's local objective function. The optimization objective under specific constraints;

[0119] The overall coordination goal means that the global control system ensures overall coordination while satisfying the local objectives.

[0120] For the first Local variables of a device, representing the device State or behavioral variables;

[0121] This is a global variable, representing a unified control variable for all devices in a distributed system;

[0122] This indicates taking the minimum value of the expression.

[0123] b) Iterative solution process:

[0124] renew:

[0125]

[0126] renew:

[0127]

[0128] renew:

[0129]

[0130] in:

[0131] This represents the value of the i-th variable in the (k+1)-th iteration;

[0132] , Indicates to Find the expression that minimizes the subsequent expression. The value of z;

[0133] , This represents the value of k in the k-th and (k+1)-th iterations;

[0134] Let represent the dual variable of the i-th device in the k-th and k+1-th iterations;

[0135] The step size parameter controls the penalty factor during the update process;

[0136] The iteration number indicates the optimization process.

[0137] For the first The dual variables of each device are used to represent the differences between local and global variables.

[0138] c) Control parameter output:

[0139]

[0140] in:

[0141] For equipment The location parameter indicates the spatial location of the device;

[0142] This is a motion speed parameter, representing the speed at which the equipment moves.

[0143] This is a force parameter, representing the force or torque applied by the equipment;

[0144] These are the operating mode parameters, indicating the device's operating mode.

[0145] B4). Real-time state rendering

[0146] Real-time rendering of device operating status is achieved using WebAssembly technology, with a data update frequency of no less than 60Hz. The rendering frame rate adaptive control equation is as follows:

[0147]

[0148] in:

[0149] The target frame rate represents the frame rate expected by the rendering system.

[0150] The maximum frame rate limit is set to 60fps;

[0151] The minimum frame rate requirement is set to 24fps.

[0152] This represents the current CPU processing power, indicating the adequacy of computing resources.

[0153] This is a rendering load metric that reflects the computational complexity during the rendering process.

[0154] This module, in collaboration with the feature fusion sorting module, generates optimal control strategies in real time and transmits control parameters to the execution equipment. Simultaneously, it provides equipment operating status information to the chain traceability management module, supporting end-to-end data tracking.

[0155] 3. Chain Tracking Management Module

[0156] The chain-tracking management module utilizes multi-source data acquisition, distributed storage, and graph database technology to achieve end-to-end data tracking of returned smart meters in the sorting line. This module ensures that the processing of each returned meter is traceable, queryable, and assessable. The specific implementation process is as follows:

[0157] C1. Collection of meter identification information

[0158] At the entrance of the sorting line, an industrial-grade QR code scanner (resolution ≥ 3072×2048 pixels) is used to collect the meter identification information. The identification information structure for each returned smart meter is as follows:

[0159]

[0160] in:

[0161] MID is a 16-digit meter serial number, using a segmented encoding format "AABB-CCDD-EEFF-GGHH": AA: Manufacturer code (e.g., 01 indicates a Zhejiang manufacturer, 02 indicates a Jiangsu manufacturer), BB: Production year (e.g., 20 indicates 2020), CC-DD: Model code (e.g., 1001 indicates a Class 1 single-phase meter, 2001 indicates a Class 1 three-phase meter), EE-FF: Batch number (e.g., 01-01 indicates the first batch of the year), GG-HH: Serial number (e.g., 00-01 indicates the first meter in the batch).

[0162] Type is a 4-digit meter type code: 1001: single-phase local prepaid smart meter, 1002: single-phase remote prepaid smart meter, 2001: three-phase local prepaid smart meter, 2002: three-phase remote prepaid smart meter;

[0163] Batch is an 8-digit batch code, using the "YYYYMMDD" format to record the production date;

[0164] The system entry timestamp is in milliseconds.

[0165] Status indicates the current state of the meter: 0: To be tested, 1: Testing in progress, 2: Testing completed, 3: Classified, 4: Processing completed.

[0166] c2). Sharded blockchain data storage

[0167] A distributed storage system is constructed using sharded blockchain technology to achieve data immutability and traceability. Block structure definition:

[0168]

[0169] The block header contains key information:

[0170]

[0171] in:

[0172] : A 32-byte preceding block hash value generated using the SHA256 algorithm;

[0173] : UTC timestamp of the generated block (accurate to milliseconds);

[0174] The Merkle root hash value constructed from all transaction data within the block;

[0175] : 4-byte fragment identifier, used to identify the service fragment to which the data belongs;

[0176] Difficulty: The difficulty value of proof of work is dynamically adjusted based on the system's processing capacity.

[0177] The block data body contains:

[0178]

[0179] in:

[0180] Records is a collection of process records;

[0181] States records the status of the relevant equipment;

[0182] Params records environmental parameters.

[0183] c3). Electricity meter circulation record

[0184] A directed acyclic graph (DAG) is used to record the flow of electricity meters between different processes on the sorting line:

[0185]

[0186] in:

[0187] V represents a set of process nodes, comprising 5 key processes: S001 (Appearance Inspection): Inspecting the integrity of the meter casing, the status of the display screen, and the status of the terminal blocks; S002 (Withstand Voltage Test): Performing a 4kV withstand voltage test to ensure insulation performance; S003 (Functional Test): Verifying the metering function, communication function, and fee control function; S004 (Accuracy Verification): Performing error tests at three current points: 0.05Ib, Ib, and Imax; S005 (Classification and Marking): Classifying and laser marking according to the test results.

[0188] The flow edge set E records the transfer relationships between processes:

[0189]

[0190] in, This represents the transfer relationship from process i to process j.

[0191] The weight matrix W records the time information of the process flow:

[0192]

[0193] in This represents the standard turnaround time (in seconds) from process i to process j.

[0194] c4). Time-series data index

[0195] Construct a multi-level time-series index structure based on LSM (Log-Structured Merge-tree):

[0196]

[0197] Configuration parameters for each layer:

[0198] L0 layer: 128MB capacity, resides in memory, and serves as a real-time write buffer;

[0199] L1 layer: 1GB in size, consisting of 10 SSTable files, with a compression ratio of 1:10;

[0200] L2 layer: 10GB capacity, consisting of 100 SSTable files, compression ratio 1:10;

[0201] L3 layer: 100GB capacity, consisting of 1000 SSTable files, compression ratio 1:10.

[0202] c5). Meter processing data association

[0203] Establish a mapping relationship between meter processing and equipment operating status using the Neo4j graph database:

[0204]

[0205] in: v represents the set of meter nodes, which contains meter identification information; This is a collection of device nodes, containing device operating parameters; To process the set of relational edges, record the processing time and parameters; φ is the attribute mapping function, which defines the attribute set of nodes and edges.

[0206] Data from each processing stage is ensured to be tamper-proof through blockchain technology, enabling full traceability. At the same time, based on the correlation analysis capabilities of graph databases, abnormal processes and problematic equipment can be quickly located, improving the operational efficiency and quality control level of the sorting line.

[0207] 4. Status Diagnosis Module

[0208] The condition diagnosis module collects and analyzes equipment operating parameters in real time to monitor the condition and provide early warnings of faults for key equipment in the sorting line. This module integrates data from multiple sensors to build predictive models, providing data support for equipment maintenance decisions.

[0209] D1). Equipment parameter acquisition

[0210] The data collected by the speed sensor is defined as a time series:

[0211]

[0212] in:

[0213] This represents the angular velocity value collected by the i-th speed sensor at time t, in rad / s;

[0214] M represents the total number of speed sensors;

[0215] t is the sampling time point, and the sampling frequency is 1kHz;

[0216] Triaxial acceleration data collected by vibration sensors:

[0217]

[0218]

[0219] in:

[0220] Let represent the triaxial acceleration vector collected by the j-th vibration sensor at time t;

[0221] N is the total number of vibration sensors;

[0222] , , These represent the acceleration values ​​along the X, Y, and Z axes, respectively, in m / s².

[0223] The sampling frequency is 10kHz.

[0224] D2). Time Series Prediction Model

[0225] A parameter prediction model is constructed using a bidirectional gated recurrent unit (BiGRU) network. Its forward propagation process is as follows:

[0226] Reset door calculation:

[0227]

[0228] Update gate calculation:

[0229]

[0230] Candidate hidden state calculation:

[0231]

[0232] Hidden state update:

[0233]

[0234] in:

[0235] The input feature vector at time t contains rotational speed and vibration data;

[0236] , Let be the hidden state vectors at time t and time t-1;

[0237] , , This is the weight matrix;

[0238] , , It is the bias vector;

[0239] σ(·) is the sigmoid activation function;

[0240] ⊙ represents the Hadamard product (element-by-element multiplication).

[0241] D3). Domain Adversarial Adaptation

[0242] To eliminate parameter distribution differences between different devices, a domain adversarial training mechanism is introduced. A feature extractor is defined. Sum Domain Classifier objective function for:

[0243]

[0244] in: Source domain data (with known device parameter distribution)

[0245] For target domain data (distribution of device parameters to be adapted);

[0246] , These represent the expectations for the source domain data and the target domain data, respectively.

[0247] Represents feature extraction mapping;

[0248] The output of the domain discriminator.

[0249] D4). Fault Feature Extraction

[0250] The vibration signal was decomposed using wavelet packet decomposition with 3 decomposition levels, using the db4 wavelet basis. Energy characteristics were calculated.

[0251]

[0252] in: The energy value of the i-th frequency band; These are the wavelet packet coefficients of the i-th frequency band; The number of coefficients in the i-th frequency band.

[0253] Define the fault feature vector:

[0254]

[0255] in:

[0256] Indicates the energy values ​​of different frequency bands;

[0257] denoted as total energy; n represents the number of frequency bands (n=8 in this embodiment).

[0258] D5). Maintenance Decision Generation

[0259] The Proximal Policy Optimization (PPO) algorithm is used to generate equipment maintenance decisions. Value function definition:

[0260]

[0261] in:

[0262] It is a value function, representing the state from the current state. Initial expected return;

[0263] The discount factor, with a value of 0.95, determines the impact of future returns on current decisions.

[0264] This represents the policy ratio at time step t+k.

[0265] Policy function:

[0266]

[0267] in:

[0268] Indicates the current policy (as shown by the parameters) (Control) in state Take action below The probability of;

[0269] Let be the device state vector at time t;

[0270] The maintenance action γ is a discount factor (with a value of 0.95);

[0271] θ represents the policy network parameters;

[0272] This is the output of the policy network.

[0273] Strategy update objectives:

[0274]

[0275] in:

[0276] Let θ be the pruning objective function of PPO, representing the objective function optimized during policy updates. By maximizing this function, the policy parameters θ are adjusted to make policy updates more stable. Ât is the strategy ratio; Ât is the estimated value of the advantage function. In the formula, It is a state-action value function, representing the state... Select action The cumulative returns obtained later It is a state-value function, representing the state... The expected return if the current strategy is continued; This represents the pruning operation in PPO, used to limit the range of changes in the policy ratio and prevent excessive policy updates; ε is the pruning parameter (value 0.2); This represents the expected value.

[0277] D6). Fault warning information

[0278] The generated fault warning information includes the device's unique identifier, warning timestamp, fault type code, fault severity (level 1-5), and recommended maintenance measures.

[0279] Adaptive update of warning thresholds:

[0280]

[0281] in: These represent the old and new thresholds, respectively; α is the smoothing factor (value 0.8); The mean of the most recent N measurements; N is the sliding window size (value 100).

[0282] 5. Environmental Management Module

[0283] The environmental control module is responsible for comprehensive monitoring and management of the sorting line's working environment, ensuring the stability of equipment operation and the safety of operators. This module collects environmental parameters in real time through a multi-layered sensor network and combines deep learning algorithms to achieve intelligent control.

[0284] E1). Environmental parameter acquisition network

[0285] The temperature and humidity sensor network adopts a hierarchical mesh topology, and the collected data is defined as follows:

[0286]

[0287]

[0288] in:

[0289] The total number of sensor nodes (in this embodiment) (Ensure there is at least one monitoring point in a 3m x 3m area).

[0290] For the first The temperature values ​​measured at each node have a measurement range of [−10°C, 50°C] and an accuracy of ±0.3°C.

[0291] The value represents relative humidity, with a measurement range of [0%, 100%] and an accuracy of ±2% RH.

[0292] The value is atmospheric pressure, with a measurement range of [85 kPa, 106 kPa] and an accuracy of ±0.1 kPa.

[0293] The sampling time point is 0.2 Hz.

[0294] The sensor spatial distribution weight matrix is ​​defined as follows:

[0295]

[0296]

[0297] in:

[0298] For nodes and The Euclidean distance between them;

[0299] This is the spatial correlation attenuation coefficient, with a value of 1.5;

[0300] This represents the spatial weight coefficient between nodes.

[0301] E2). Personnel positioning and trajectory tracking

[0302] The parameters of the LiDAR scanner are set as follows: horizontal scanning angle: 360°; vertical scanning angle: 30°; angular resolution: 0.1°; ranging accuracy: ±30mm; scanning frequency: 20Hz.

[0303] Point cloud data preprocessing:

[0304]

[0305] in:

[0306] For a single data point in a point cloud, containing coordinate;

[0307] The centroid of the point cloud in the local area;

[0308] The outlier filtering threshold is set to 0.15 meters.

[0309] This represents the total amount of point cloud data.

[0310] E3). Job Behavior Recognition

[0311] Spatiotemporal Convolutional Network (ST-GCN) is used to process personnel motion data. The graph convolution operation is defined as follows:

[0312]

[0313] in:

[0314] For nodes The neighborhood set; These are the normalization coefficients; For input features; The weight matrix is ​​a learnable matrix; This is a label for spatial relationships.

[0315] The time dimension is modeled as follows:

[0316]

[0317] in:

[0318] For the first The feature matrix of the layer; It is an adjacency matrix; It is a degree matrix; For the first The parameter matrix of the layer; This is the ReLU activation function.

[0319] E4). Motion trajectory prediction

[0320] An improved Kalman filter algorithm is used for trajectory prediction. The state equation is:

[0321]

[0322] The observation equation is:

[0323]

[0324] in:

[0325] for State vector at time step ;

[0326] This is the state transition matrix;

[0327] To control the input matrix;

[0328] For control vectors;

[0329] This is process noise, and it follows a normal distribution. ;

[0330] For observation vectors;

[0331] The observation matrix;

[0332] The observed noise follows a normal distribution. .

[0333] The predicted covariance is updated as follows:

[0334]

[0335] Kalman gain is calculated as follows:

[0336]

[0337] E5). Flow field analysis of environmental parameters

[0338] The computational fluid dynamics model based on hierarchical decomposition mainly includes the following equations:

[0339] a). Continuity equation:

[0340]

[0341] b) Momentum equation:

[0342]

[0343] c) Energy equation:

[0344]

[0345] in:

[0346] The density of air is 1.225 kg / m³.

[0347] It is a velocity vector field;

[0348] For pressure field;

[0349] Dynamic viscosity, 1.789 × 1 Pa·s;

[0350] This is the vector of gravitational acceleration;

[0351] The specific heat capacity at constant pressure is 1005 J / (kg·K);

[0352] The thermal conductivity is 0.0257 W / (m·K);

[0353] It is a viscous dissipation function.

[0354] E6). Environmental Anomaly Warning

[0355] Anomaly detection metrics based on multi-parameter fusion are defined as follows:

[0356]

[0357] in:

[0358] For the first Measured values ​​of several environmental parameters; The historical mean of the parameter; The standard deviation of the parameter; These are the weighting coefficients; This represents the total number of environmental parameters.

[0359] The warning levels are classified as follows:

[0360] Level 1 Warning:

[0361] Level 2 warning:

[0362] Level 3 Warning:

[0363] Level 4 Warning: .

[0364] This module transmits environmental monitoring information to the digital mapping control module in real time, supports environmental adaptive control of the sorting line, and provides environmental parameter support for the status diagnosis module, thereby realizing the correlation analysis between equipment status and environmental conditions.

[0365] Combination Figure 2 The workflow of this invention is as follows:

[0366] S1. Meter Feature Collection and Classification

[0367] The system first collects feature information of decommissioned smart meters using a multimodal sensor array. Four industrial cameras with a resolution of 2048×1536 pixels are arranged in a tetrahedral configuration to achieve omnidirectional image acquisition of the meter's appearance; an array of eight 48kHz sampling frequency microphones arranged in a ring collects 2048 acoustic signals; a structured light scanner generates 3D point cloud data, with each point containing precise spatial coordinate information; an infrared thermal imager collects temperature field data with a resolution of M×M; the collected multi-source data are input into specialized processing models respectively; the image data is processed through a Swing Transformer network, using an attention mechanism to extract hierarchical features; the acoustic signals are converted to the Mel frequency domain through Mel spectrum analysis; the point cloud data is processed by a PointNet++ network, using a hierarchical sampling strategy to extract geometric features; the temperature field is analyzed using a region growing algorithm; finally, a weighted attention mechanism is used to fuse all features into a unified feature vector, and a classification decision is output through a softmax function.

[0368] S2. Equipment Control and Coordinated Scheduling

[0369] The digital mapping control module receives classification results and constructs a device state transition model based on a temporal Petri net. The model includes a location set P (representing device state), a transition set T (representing triggering events), a flow relationship F, and a time interval function I, achieving accurate modeling of device states. A PBFT mechanism is used to achieve state consensus among devices, completing state confirmation through four stages: request, pre-preparation, preparation, and submission. The ADMM algorithm implements distributed optimization control, decomposing the global control problem into multiple local sub-problems and obtaining optimal control parameters through iterative solutions. WebAssembly technology is used to achieve a 60Hz state rendering frequency, ensuring the real-time performance of the monitoring interface.

[0370] S3. End-to-End Data Tracking

[0371] The 16-digit serial number of the electricity meter is collected using an industrial-grade barcode scanner (3072×2048 pixels resolution), which includes the manufacturer code, production year, model code, batch number, and serial number information. Data is stored and processed using sharded blockchain technology, with each block containing a 32-byte preorder hash value, timestamp, Merkle tree root value, and shard identifier. A directed acyclic graph is constructed to record the transfer relationships of five processes (appearance inspection, withstand voltage test, functional test, accuracy verification, and classification and marking). A four-level index structure is established based on an LSM tree to achieve data organization from L0 (128MB) to L3 (100GB). The mapping relationship between meter processing and equipment status is established through the Neo4j graph database to support full-process traceability.

[0372] S4. Equipment Status Monitoring

[0373] Rotation speed sensors (1kHz sampling, ±1rpm accuracy) and triaxial vibration sensors (10kHz sampling, ±50g range) are installed on the core equipment of the production line to collect rotation speed and vibration data, respectively. A BiGRU neural network is used to construct a 24-hour prediction model, employing 128-dimensional hidden layer states for bidirectional feature extraction. Three-layer wavelet packet decomposition is performed using the db4 wavelet basis function to extract energy features in eight frequency bands. A domain adversarial network eliminates individual differences among equipment, achieving unified modeling. A near-end strategy optimization algorithm generates the optimal maintenance strategy within a 4-hour decision cycle, based on equipment health assessment and maintenance costs.

[0374] S5. Environmental Monitoring and Safety Management

[0375] An environmental monitoring network is constructed by deploying 24 sensor nodes within a standard 3m x 3m monitoring unit. Temperature sensors (-10°C to 50°C, ±0.3°C), humidity sensors (0-100%RH, ±2%RH), and pressure sensors (85-106kPa, ±0.1kPa) sample synchronously at a frequency of 0.2Hz. A lidar system (360° horizontal × 30° vertical, 0.1° resolution, 20Hz scanning) tracks the work area in real time. A 12-layer spatiotemporal convolutional network analyzes work behavior, identifying 8 types of standard actions and 5 types of abnormal behaviors. A computational fluid dynamics model (0.1m spatial accuracy, 1s time step) calculates the distribution of environmental parameters, supporting precise control of environmental equipment.

[0376] In summary, the smart meter sorting line monitoring system proposed in this invention achieves comprehensive monitoring and intelligent management of the smart meter processing process through the organic collaboration of five functional modules: feature fusion sorting, digital mapping control, chain tracking management, status diagnosis, and environmental control. The system employs advanced technologies such as multimodal sensing, deep learning, distributed storage, and status prediction to solve technical challenges such as low efficiency in traditional manual sorting, complex line control, difficulty in tracing the processing, lagging equipment status monitoring, and lack of environmental safety monitoring. Practical application shows that the system significantly improves sorting efficiency and accuracy, enables predictive maintenance of equipment status, ensures a safe and stable working environment, and provides reliable technical support for the standardized processing of returned smart meters. This invention has good practicality and promotional value, and is of great significance for improving the management level of power equipment.

[0377] While this document illustrates specific applications of the invention through particular embodiments, these embodiments are for illustrative purposes only and do not imply limitation of the scope of protection of the invention. The scope of protection of the invention is defined by the claims, and any appropriate modifications, equivalent substitutions, or improvements based on the principles of the invention should be considered to fall within the scope of protection of the invention. Therefore, the scope of protection of the invention should be given the broadest interpretation to cover all such modifications, equivalent structures, and functions.

Claims

1. A monitoring system for a sorting line of returned smart meters, characterized in that, include: The feature fusion sorting module includes an industrial camera array, a microphone array, a structured light scanner, and an infrared thermal imager, used to collect multimodal feature information of the electricity meter; The industrial camera array processes image data through a hierarchical attention network, the microphone array processes acoustic signals through Mel spectrum analysis, the structured light scanner processes three-dimensional point clouds through a point cloud neural network, and the infrared thermal imager processes the temperature field through a region growing algorithm. The above features are then fused through a weighted attention mechanism to form a classification result. The digital mapping control module is used to receive the classification results and equipment status data, construct an equipment state transition model through a time Petri net, realize cross-equipment state consensus based on the Byzantine fault tolerance mechanism, and calculate the sorting process control parameters by combining a distributed optimization control algorithm based on the alternating direction multiplier method. The chain-based traceability management module collects the serial number of the electricity meter through a barcode scanner, establishes a data storage structure using sharded blockchain technology, records the electricity meter transfer process using a directed acyclic graph, establishes a time-series index based on a logarithmic merging tree, and constructs a mapping relationship between electricity meter processing and equipment status through a graph database. The condition diagnosis module collects equipment operating parameters through speed and vibration sensors, constructs a parameter time-series prediction model using a bidirectional gated cyclic network, eliminates parameter distribution differences between devices using a domain adversarial network, extracts fault feature frequency bands through wavelet packet decomposition, and generates equipment maintenance decisions by combining a near-end strategy optimization algorithm. The environmental control module collects sorting environmental parameters through a network of temperature and humidity sensors, obtains personnel location data through LiDAR scanning, identifies work behaviors using a spatiotemporal convolutional network, predicts motion trajectories using a Kalman filter algorithm, and monitors sorting environmental parameters based on a hierarchical decomposition computational fluid dynamics model.

2. The monitoring system according to claim 1, characterized in that, The weighted attention mechanism in the feature fusion and sorting module includes a relative position bias matrix and a learnable weight matrix, which is used to calculate the weights of each modality feature.

3. The monitoring system according to claim 1, characterized in that, The Byzantine fault tolerance mechanism in the digital mapping control module includes four processing stages: request, pre-preparation, preparation, and submission.

4. The monitoring system according to claim 1, characterized in that, The directed acyclic graph in the chain traceability management module includes five process nodes: appearance inspection, pressure resistance test, functional test, accuracy verification, and classification and labeling.

5. The monitoring system according to claim 1, characterized in that, The wavelet packet decomposition in the state diagnosis module uses the db4 wavelet basis for three-level decomposition to extract fault features in eight frequency bands.

Images