Metrology asset lifecycle management system based on data analysis

Abstract

This application provides a data analysis-based full lifecycle management system for metering assets. Addressing the challenge of highly overlapping characteristics between communication faults and physical asset faults in metering asset diagnosis, it constructs a fault decoupling mechanism based on the fusion of spatial cross-validation and temporal micro-evidence. When a target electricity meter's disconnection alarm is triggered, on the one hand, the system backtracks and extracts the operational parameters before the breakpoint and addresses the communication routing topology. Neighboring nodes on the same branch are used as environmental probes, and spatial weighted fitting is performed to obtain the baseline characteristics of communication interference in the transformer area. On the other hand, transient attenuation gradient extraction and multimodal orthogonal fusion are performed on the parameters before the breakpoint to obtain the breakpoint drop micro-feature tensor, characterizing evidence of inherent equipment degradation. Subsequently, the classification threshold is adjusted using the interference baseline characteristics, and the micro-feature tensor and fault twin map are orthogonally decoupled and matched to output fault decoupling labels. Based on this, differentiated dispatching is triggered and the health profile is updated, achieving precise isolation between communication anomalies and physical asset failures.

Description

Technical Field

[0001] This application relates to the field of power metering asset management technology, and more specifically, to a data analysis-based full lifecycle management system for metering assets. Background Technology

[0002] Power grid companies possess a massive scale of metering assets; the value of electricity meters and data acquisition equipment in operation at the provincial level alone can reach tens of billions of yuan, with high annual maintenance costs. For a long time, the operation and maintenance of metering assets has relied heavily on periodic passive rotation and offline management. Provincial, municipal, and county-level units lack effective tools to support investment decisions. Furthermore, continuous abnormal wear and tear on stored and operating equipment leads to asset losses, making it difficult to align asset records with actual assets and hindering precise investment implementation. To curb abnormal wear and tear, detect the risk of loss, and promote a shift from periodic rotation to precise operation and maintenance based on actual conditions, building a data-driven, full lifecycle management system for metering assets—capable of integrating multi-source business data and enabling online status evaluation and proactive fault diagnosis—has become an urgent need to improve management and revitalize asset value.

[0003] Existing metering asset management systems typically rely on heartbeat timeouts or consecutive days of offline operation at the master station as the basis for fault diagnosis during the online fault diagnosis phase, in conjunction with basic power outage assessment rules for distribution areas. However, whether it's a communication module disconnection caused by a weak network in the basement, electromagnetic interference from nearby frequency converters, or a fault in the meter's microcontroller unit or a burnt-out power module, the manifestation transmitted to the master station collapses into a single data acquisition interruption point. Current technology only considers the result, neglecting the transient micro-processes and spatial network context before the disconnection. Mathematically, the characteristics of these two types of faults highly overlap, leading to essentially guesswork-based judgments. This results in blind dispatching of maintenance orders and the incorrect removal and scrapping of numerous perfectly good energy meters that are merely experiencing communication faults. Even if the spatial interference intensity of the transformer area is introduced as an auxiliary adjustment, since it only relies on a single scalar information of the external environment to drive the judgment threshold, when strong interference such as lightning surges, harmonic leakage and meter hardware failure occur simultaneously, it will still misjudge hardware damage as communication interference by relaxing the threshold. Conversely, in the case of local directional shielding, it will refuse to relax the threshold and make a misjudgment. The root cause is the lack of verification capability for the internal physical constraint of the communication-electrical degradation asymmetry of the equipment itself, resulting in insufficient decoupling accuracy under complex working conditions.

[0004] Therefore, an optimized, data-analysis-based, full lifecycle management system for measurement assets is desired. Summary of the Invention

[0005] To address the aforementioned technical problems, this application is proposed. Embodiments of this application provide a data analysis-based full lifecycle management system for metered assets, comprising: The backtracking and topology addressing module is used to perform backtracking and topology addressing on the real-time operation messages of the energy meter when the heartbeat interruption time of the target energy meter exceeds the set threshold, based on the time of disconnection, in order to obtain the parameter timing matrix and communication route topology snapshot before the breakpoint. The interference baseline feature fitting module is used to perform channel state exploration and spatial distance weighted fitting on the same branch neighbor nodes of the target energy meter based on the cascade hierarchical relationship in the communication routing topology snapshot to obtain the communication interference baseline features of the transformer area. The micro-feature tensor fusion module is used to extract transient decay gradients from the pre-breakpoint parameter time-series matrix and orthogonally fuse them with multimodal channels to obtain the breakpoint drop micro-feature tensor. The fault orthogonal decoupling classification module is used to perform fault orthogonal decoupling classification on the breakpoint drop micro-feature tensor and the standard waveform spectrum in the pre-set fault twin spectrum library based on the communication interference benchmark features of the transformer area to obtain fault decoupling labels. The differentiated management and ledger update module is used to trigger differentiated management logic based on the category attributes of fault decoupling tags to generate targeted operation and maintenance dispatch instructions, and synchronously write the traceability event characteristics into the target asset electronic ledger to update the full life cycle health profile.

[0006] Compared with existing technologies, the data analysis-based full lifecycle management system for metering assets provided in this application addresses the challenge of overlapping and difficult-to-distinguish features between communication failures and hardware failures at the main station after a meter loses connection. It constructs a fault decoupling diagnosis mechanism that uses the moment of disconnection as a benchmark and integrates spatial and temporal evidence. When the target meter's heartbeat interruption exceeds the limit, on the one hand, it uses the moment of disconnection as an anchor point to backtrack and extract multi-dimensional operational messages before the breakpoint and extract the communication routing topology. Using neighboring nodes in the same branch as spatial probes of the communication environment, it obtains the benchmark features of communication interference in the transformer area through channel exploration and spatial distance weighted fitting, characterizing the external spatial interference background. On the other hand, it extracts transient attenuation gradients of parameters before the breakpoint and performs orthogonal fusion of multi-modal channels to capture the microscopic drop trajectory of the device in the final moment before disconnection. Subsequently, spatial environmental evidence and temporal microscopic evidence are input into the decoupling and classification process. The judgment threshold is dynamically adjusted based on interference baseline features and compared with the fault twin map. The fault attribution domain is orthogonally decoupled, thereby triggering differentiated operation and maintenance dispatch and updating the full life cycle health profile simultaneously. This transforms the original black-box judgment, which collapsed into a single breakpoint, into a precise qualitative assessment with spatiotemporal mutual verification, effectively reducing the blindness of dispatch and the ineffective loss of good assets. Attached Figure Description

[0007] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0008] Figure 1 This is a system block diagram of a data analysis-based measurement asset lifecycle management system according to an embodiment of this application; Figure 2 This is a schematic diagram of the data flow in the data analysis-based full lifecycle management system for metered assets according to an embodiment of this application; Figure 3 This is a block diagram of the interference benchmark feature fitting module in the data analysis-based full lifecycle management system for metered assets according to an embodiment of this application; Figure 4 This is a block diagram of the micro-feature tensor fusion module in the data analysis-based full lifecycle management system for metered assets according to an embodiment of this application; Figure 5 This is a block diagram of the fault orthogonal decoupling classification module in the data analysis-based metering asset lifecycle management system according to an embodiment of this application; Figure 6 This is a block diagram of a threshold dynamic adjustment unit in a data analysis-based full lifecycle management system for metered assets according to an embodiment of this application. Detailed Implementation

[0009] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0010] As indicated in this application and claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" are not specifically singular and may include plural forms. Generally speaking, the terms "comprising" and "including" only indicate the inclusion of explicitly identified steps and elements, which do not constitute an exclusive list, and the method or apparatus may also include other steps or elements.

[0011] While this application makes various references to certain modules of the systems according to embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The modules described are merely illustrative, and different aspects of the systems and methods may use different modules.

[0012] Flowcharts are used in this application to illustrate the operations performed by the system according to embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from them.

[0013] Current online diagnostic technologies for metering assets primarily rely on heartbeat timeouts or consecutive days of offline data reception at the master station as the basis for judgment. However, whether it's a communication module failure or a meter hardware malfunction, the manifestation transmitted to the master station collapses into a single data acquisition breakpoint. Existing technologies only consider the disconnection result without examining the transient micro-processes and spatial networking context before the disconnection. This leads to a high degree of overlap between the two types of faults in the feature space, resulting in blind judgment and causing errors in maintenance dispatching. Consequently, many excellent energy meters with only communication faults are mistakenly removed and scrapped. Therefore, the technical solution in this application proposes a data analysis-based full lifecycle management system for metering assets. When the target electricity meter's heartbeat is interrupted, triggering a disconnection alarm, on the one hand, the system backtracks and extracts the operating parameters before the disconnection point based on the time of disconnection, and addresses the communication routing topology. Then, it calls the neighboring nodes in the same branch as communication environment probes, and obtains the base features of communication interference in the transformer area through spatial distance weighted fitting to characterize external spatial evidence. On the other hand, it extracts the transient decay gradient of the parameters before the disconnection point and performs orthogonal fusion of multimodal channels to obtain the disconnection point drop micro-feature tensor to characterize the intrinsic degradation evidence of the equipment itself at the moment of its death. Subsequently, it dynamically adjusts the classification threshold based on the interference base features, and performs orthogonal decoupling matching between the micro-feature tensor and the preset fault twin map to output a high-confidence fault decoupling label. Finally, it triggers differentiated operation and maintenance dispatch and updates the full life cycle health file accordingly, thereby breaking through the black box limitation of single-point information of the main station, realizing the precise isolation between communication anomalies and the failure of the main body, and significantly reducing the invalid dispatch rate and the loss of erroneous dismantling and scrapping of good assets.

[0014] Figure 1 This is a system block diagram of a data analysis-based full lifecycle management system for metered assets according to an embodiment of this application. Figure 2 This is a schematic diagram of the data flow in a data analysis-based full lifecycle management system for metered assets according to an embodiment of this application. Figure 1 and Figure 2As shown, the data analysis-based metering asset lifecycle management system 100 according to an embodiment of this application includes: a backtracking and topology addressing module 110, used to perform backtracking and topology addressing on the real-time operation messages of the electricity meter when the heartbeat interruption duration of the target electricity meter exceeds a set threshold, based on the time of disconnection, to obtain the parameter timing matrix and communication route topology snapshot before the breakpoint; an interference reference feature fitting module 120, used to perform channel state exploration and spatial distance weighted fitting on the neighboring nodes of the same branch of the target electricity meter based on the cascading hierarchical relationship in the communication route topology snapshot to obtain the communication interference reference features of the transformer area; and a micro-feature tensor fusion module. The module 130 is used to extract transient decay gradients and orthogonally fuse multimodal channels from the time-series matrix of parameters before the breakpoint to obtain the breakpoint drop micro-feature tensor; the fault orthogonal decoupling classification module 140 is used to perform fault orthogonal decoupling classification on the breakpoint drop micro-feature tensor and the standard waveform spectrum in the pre-set fault twin spectrum library based on the base features of communication interference in the distribution area to obtain fault decoupling labels; the differentiated management and ledger update module 150 is used to trigger differentiated management logic based on the category attributes of the fault decoupling labels to generate targeted operation and maintenance dispatch instructions, and synchronously write the traceability event features into the target asset electronic ledger to update the full life cycle health file.

[0015] In the aforementioned data analysis-based metering asset lifecycle management system 100, the backtracking and topology addressing module 110 is used to perform backtracking and topology addressing on the real-time operation messages of the electricity meter, based on the moment of disconnection, when the heartbeat interruption duration of the target electricity meter exceeds a set threshold, to obtain the parameter timing matrix and communication route topology snapshot before the breakpoint. It should be noted that since both the electricity meter communication module disconnection and the meter body hardware crash collapse into a single data acquisition breakpoint at the main station, the fault attribution cannot be distinguished solely based on the disconnection result. Therefore, the technical solution of this application first anchors the connection at the moment of disconnection while simultaneously backtracking and preserving the multi-dimensional operation parameters and communication network background before the breakpoint. Through the above processing, original materials of timing and spatial evidence can be provided for subsequent fault decoupling.

[0016] More specifically, in a specific example of this application, the real-time operation message of the energy meter includes the effective value of the meter terminal voltage, the effective value of the meter terminal current, the transient change rate of active power, the signal-to-noise ratio of carrier communication, the packet loss rate of carrier communication, and the retransmission frequency of carrier communication. The first three items characterize the electrical metering status of the meter itself, and the latter three items characterize the transmission quality of the carrier communication link. Together, they constitute the physical basis for determining the fault attribution domain. The arrival time of the jump packet in the real-time operation message center of the energy meter is monitored in real time, the difference between the current system clock and the latest heartbeat time is calculated, and a comparison is made with a preset heartbeat interruption threshold to determine if it exceeds the limit. The determination logic is expressed as follows: ; in, The waiting time difference represents the duration during which no heartbeat packet has been received. This is the current system clock tick. This is the arrival timestamp of the most recent heartbeat packet extracted from the message. When the preset heartbeat interruption threshold is exceeded, a disconnection alarm is triggered. This threshold is set to 2 to 3 times the normal heartbeat cycle according to the carrier communication protocol to filter out false alarms caused by occasional packet loss. After the alarm is triggered, the device primary key address in the message is extracted as the target energy meter identifier, and the system clock at the moment the alarm is triggered is recorded as the disconnection time timestamp.

[0017] Subsequently, based on the target energy meter identifier and the timestamp of the moment of disconnection, the historical data storage location of the target node is located in the high-frequency buffer. Starting from the timestamp of the moment of disconnection, messages are traced back in reverse order in the historical direction for a fixed number of acquisition cycles. The data of 6 parameter dimensions are extracted and concatenated with the parameter dimensions in a two-dimensional manner in forward time order to obtain the parameter time series matrix before the breakpoint. This matrix completely preserves the evolution trajectory of the electrical and communication state of the target energy meter before the complete loss of connection, avoiding the loss of key transient information before the breakpoint due to the interruption of acquisition.

[0018] Meanwhile, based on the target energy meter identifier and the timestamp of the moment of disconnection, the global dynamic routing management table is explored using the target energy meter identifier as the addressing primary key, and the timestamp of the moment of disconnection is used as a temporal constraint condition. The actual link status of the transformer area's communication network at that moment is locked from the continuously dynamically changing global routing table. A graph structure subgraph containing its direct parent node, child nodes, and peer nodes in the same branch box is extracted with the target node as the center to obtain a communication routing topology snapshot. This snapshot solidifies the cascading hierarchy and surrounding network layout of the target node at the moment of disconnection, providing a topology addressing basis for subsequent exploration of the communication environment by calling neighboring nodes.

[0019] In the aforementioned data analysis-based metering asset lifecycle management system 100, the interference benchmark feature fitting module 120 is used to perform channel state exploration and spatial distance weighted fitting on the neighboring nodes of the target energy meter in the same branch based on the cascading hierarchical relationship in the communication routing topology snapshot to obtain the communication interference benchmark features of the transformer area. It should be noted that, given that the parameter time series matrix before the breakpoint only reflects the operating status of the target energy meter itself, it is impossible to independently determine whether the communication link deterioration is due to its own fault or the common environmental interference of the entire transformer area, and the single-point information at the master station has an inherent blind spot. Based on this, the technical solution of this application further analyzes the neighboring nodes in the same branch box or the same phase line as the target energy meter based on the cascading hierarchical relationship in the communication routing topology snapshot, sends a high-priority channel exploration command to the transformer area edge concentrator to sample and obtain the signal strength, signal-to-noise ratio and link packet loss rate of each neighboring node, and performs weighted fitting on the real-time communication status of the neighboring nodes based on the spatial logical distance to obtain the communication interference benchmark features of the transformer area. Through the above processing, the real communication environment around the target node can be quantified into a comparable spatial benchmark, thereby effectively distinguishing whether the communication anomaly of the target energy meter belongs to common interference in the distribution area or a unique fault of a single unit, providing cross-verification evidence in the spatial dimension for subsequent fault decoupling.

[0020] Figure 3 This is a block diagram of the interference benchmark feature fitting module in a data analysis-based full lifecycle management system for metered assets according to an embodiment of this application. Figure 3 As shown, the interference baseline feature fitting module 120 includes: a neighbor node parsing unit 121, used to parse neighbor nodes in the same branch and quantize spatial distances on the communication routing topology snapshot to obtain a set of neighbor nodes in the same branch and a spatial logical distance vector; a channel state probing unit 122, used to send a high-priority channel probing command to the edge concentrator of the substation based on the set of neighbor nodes in the same branch to sample and obtain the signal strength, signal-to-noise ratio and link packet loss rate of each neighbor node to obtain a real-time communication state matrix of the neighbor nodes; and a baseline feature fitting unit 123, used to perform spatial attenuation attention weighting and baseline feature fitting on the spatial logical distance vector and the real-time communication state matrix of the neighbor nodes to obtain the baseline features of substation communication interference.

[0021] In the aforementioned data analysis-based full lifecycle management system for metered assets 100, the neighbor node parsing unit 121 is used to perform same-branch neighbor node parsing and spatial distance quantization on the communication routing topology snapshot to obtain the same-branch neighbor node set and spatial logical distance vector. It should be noted that since the communication routing topology snapshot carries the cascading relationships between nodes in a graph structure, it cannot be directly used for quantitative comparison of the communication environment without parsing. Based on this, the technical solution of this application further performs same-branch neighbor node parsing and spatial distance quantization on the communication routing topology snapshot. Through the above processing, it is possible to identify a group of neighbor nodes that can be used as communication environment probes and quantify their spatial proximity to the target node.

[0022] More specifically, in a concrete example of this application, a graph network traversal is performed on the communication routing topology snapshot to locate the central node corresponding to the target energy meter where the heartbeat interruption occurred. The graph is then expanded outwards along its edge attributes to extract associated nodes within a one-hop physical link level of the target energy meter, including its direct parent node, child nodes, and peer nodes within the same branch box. The primary key addresses of these nodes are encapsulated into a set of neighboring nodes within the same branch. The limitation to a one-hop level is because nodes within the same branch box or phase line share similar power supply and networking paths with the target energy meter, resulting in a highly consistent probability of being affected by sudden electromagnetic interference from the same source, thus accurately reflecting the local communication environment of the target node. The extraction logic for the set of neighboring nodes within the same branch is as follows: ; in, This refers to the set of neighboring nodes extracted from the same branch. For the traversed first One neighboring node, The central node corresponding to the target electricity meter. For the target node to node in the topology graph Shortest link hop count, This is a snapshot of the communication routing topology.

[0023] While resolving neighbor nodes, the routing cost value from the central node to each of the aforementioned neighbor nodes is extracted from the communication routing topology snapshot. This routing cost value characterizes the physical or electromagnetic similarity between the target node and each neighbor node; a smaller routing cost value indicates greater proximity and stronger correlation due to the same interference source. The routing costs corresponding to each neighbor node are arranged in node sequence order to obtain the spatial logical distance vector. The construction logic of the spatial logical distance vector is expressed as follows: ; in, It is a spatial logical distance vector. For the target node and the first The routing cost between neighboring nodes The function for calculating routing costs. This represents the total number of neighboring nodes in the same branch. The resulting set of neighboring nodes in the same branch is used to determine the targets for subsequent channel probing commands, while the spatial logical distance vector provides the basis for calculating the distance weights for subsequent spatial attenuation attention weighting.

[0024] In the aforementioned data analysis-based full lifecycle management system for metered assets 100, the channel state detection unit 122 is used to send high-priority channel detection commands to the edge concentrator of the distribution area based on the set of neighboring nodes in the same branch, in order to sample and obtain the signal strength, signal-to-noise ratio, and link packet loss rate of each neighboring node to obtain the real-time communication state matrix of the neighboring nodes. It should be noted that, since the set of neighboring nodes in the same branch only provides the address index of the detection object, it does not yet contain real-time channel quality data available for quantification and comparison. Based on this, the technical solution of this application further sends high-priority channel detection commands to the edge concentrator of the distribution area based on the set of neighboring nodes in the same branch to sample and obtain the real-time communication state of each neighboring node. Through the above processing, the communication environment around the target node can be collected as structured state data, providing quantitative input for fitting interference benchmark features.

[0025] More specifically, in a concrete example of this application, the primary key of each node in the set of neighboring nodes in the same branch is traversed, and a channel probing command with a high-priority marker is issued to the edge concentrator. This command is scheduled and executed before regular data acquisition tasks to ensure that the current channel quality of neighboring nodes is promptly returned within the limited time window of the target energy meter's disconnection alarm, avoiding the loss of comparison significance due to the queuing of regular polling and the resulting delay in probing data at the moment of disconnection. After receiving the command, the edge concentrator triggers the underlying carrier communication chip to instantly return the received signal strength, signal-to-noise ratio, and link packet loss rate of each neighboring node. The received signal strength reflects the field strength level of the node's transmit and receive link, the signal-to-noise ratio reflects the relative quality of the useful signal to the background noise, and the link packet loss rate reflects the integrity of message transmission. These three factors together characterize the quality of the communication environment at the location of the neighboring node. The three-dimensional channel parameters returned by each neighboring node are row-concatenated according to the node arrangement order of the set of neighboring nodes in the same branch to obtain the real-time communication state matrix of the neighboring nodes, which is constructed as follows: ; in, This is a real-time communication state matrix for neighboring nodes. For the first The received signal strength of each neighboring node, For the first The signal-to-noise ratio of each neighboring node. For the first The packet loss rate of the link to each neighboring node. For the first and second digits of the matrix Row vector, representing the first row vector. Complete three-dimensional channel parameters of each neighboring node The total number of neighboring nodes in the same branch. The obtained real-time communication state matrix of neighboring nodes and the spatial logical distance vector obtained in the previous sequence maintain a strict correspondence on the node sequence. The two together serve as the input for subsequent spatial decay attention weighting and baseline feature fitting.

[0026] In the aforementioned data analysis-based metering asset lifecycle management system 100, the benchmark feature fitting unit 123 is used to perform spatial attenuation attention weighting and benchmark feature fitting on the spatial logical distance vector and the real-time communication state matrix of neighboring nodes to obtain the benchmark features of communication interference in the transformer area. It should be noted that, due to the varying spatial proximity between each neighboring node and the target energy meter, the reliability of their reflection of the target node's actual communication environment differs. If all neighboring nodes are treated with equal weight, noise from distant nodes with weak correlation to the target node will be introduced. Based on this, the technical solution of this application further performs spatial attenuation attention weighting and benchmark feature fitting on the spatial logical distance vector and the real-time communication state matrix of neighboring nodes. Through the above processing, the discrete channel states of multiple surrounding nodes can be fused into a single benchmark characterizing the local environment of the target node, providing spatial interference evidence for subsequent threshold adjustment.

[0027] More specifically, in a concrete example of this application, the spatial logical distance vector is retrieved, and attention weights are mapped to each routing value using a negative exponential decay function. Since neighboring nodes closer to the target node share power supply and network paths with the target node to a greater extent, they are more likely to be affected by burst interference from the same source, and thus contribute more to characterizing the real environment of the target node. Therefore, neighboring nodes with lower routing values ​​are assigned higher weights, and neighboring nodes with higher routing values ​​are assigned lower weights. Subsequently, a normalization operation is used to map each weight to a range where the sum is one. The calculation logic is expressed as follows: ; in, For the first The normalized spatial attention weights assigned to each neighbor node is the spatial attenuation control coefficient, a positive constant used to adjust the severity of the distance penalty. For the target node and the first The routing cost between neighboring nodes For the target node and the first The routing cost between neighboring nodes This represents the total number of neighboring nodes in the same branch.

[0028] After obtaining the normalized spatial attention weights of each neighboring node, these weights are multiplied row-by-row by the three-dimensional channel parameters in the corresponding row of the real-time communication state matrix of the neighboring nodes and then accumulated to obtain a three-dimensional feature row vector that integrates the surrounding environmental state, i.e., the baseline feature of the station area communication interference. Its fitting logic is expressed as follows: ; in, This serves as a baseline characteristic for communication interference in the distribution area. For the first Normalized spatial attention weights of neighboring nodes, The first node in the real-time communication state matrix of neighbor nodes The three-dimensional channel parameter row vectors of each neighboring node This represents the total number of neighboring nodes within the same branch. The obtained baseline characteristics of communication interference in the transformer area accurately characterize the background noise intensity and channel blocking level of the local space where the target energy meter is located, serving as the spatial environment input for dynamic adjustment of the classification threshold in subsequent fault decoupling classification.

[0029] In the aforementioned data analysis-based full lifecycle management system for metered assets 100, the micro-feature tensor fusion module 130 is used to extract transient decay gradients and perform multimodal orthogonal fusion of the pre-breakpoint parameter time series matrix to obtain the breakpoint drop micro-feature tensor. It should be noted that, given that the parameters in the pre-breakpoint parameter time series matrix have different dimensions and are presented only in their original numerical form, it is impossible to directly compare the relative intensity of changes between the electrical and communication domains, nor can it highlight the dynamic evolution characteristics of the target energy meter as it approaches the moment of disconnection. Based on this, the technical solution of this application further performs multi-source heterogeneous parameter quantum matrix segmentation and normalization smoothing on the pre-breakpoint parameter time series matrix to eliminate dimensional differences, performs sliding window difference operations along the time axis to extract the transient decay gradients of each parameter before the moment of disconnection, and performs multimodal orthogonal fusion of the static parameters representing the absolute state and the dynamic gradients representing the evolution rate in the channel dimension to obtain the breakpoint drop micro-feature tensor. Through the above processing, the evolution trajectory of the electrical and communication state of the equipment in the final moment before disconnection can be characterized as a high-dimensional feature expression, thereby effectively preserving the evidence of the equipment's inherent degradation morphology and providing a temporal dimension for subsequent orthogonal decoupling and matching with the fault twin map.

[0030] Figure 4 This is a block diagram of the micro-feature tensor fusion module in a data analysis-based full lifecycle management system for metered assets according to an embodiment of this application. Figure 4As shown, the micro-feature tensor fusion module 130 includes: a normalization and smoothing processing unit 131, used to perform multi-source heterogeneous parameter quantum matrix segmentation and normalization and smoothing processing on the parameter time-series matrix before the breakpoint to obtain a dimensionless smooth parameter matrix; a transient gradient extraction unit 132, used to perform sliding window difference operation and transient decay gradient extraction on the dimensionless smooth parameter matrix to obtain a multidimensional transient decay gradient matrix; and a multimodal orthogonal fusion unit 133, used to perform multimodal orthogonal fusion of static parameters and dynamic gradients on the dimensionless smooth parameter matrix and the multidimensional transient decay gradient matrix to obtain a breakpoint drop micro-feature tensor.

[0031] In the aforementioned data analysis-based full lifecycle management system for metrological assets 100, the normalization and smoothing processing unit 131 is used to perform multi-source heterogeneous parameter quantum matrix segmentation and normalization smoothing processing on the pre-breakpoint parameter time series matrix to obtain a dimensionless smooth parameter matrix. It should be noted that since the pre-breakpoint parameter time series matrix simultaneously contains electrical parameters such as voltage and current, and communication parameters such as signal-to-noise ratio and packet loss rate, the physical dimensions and numerical magnitudes of each parameter are different. If directly involved in subsequent calculations, parameters with larger magnitudes will mask the changes in parameters with smaller magnitudes. Based on this, the technical solution of this application further performs multi-source heterogeneous parameter quantum matrix segmentation and normalization smoothing processing on the pre-breakpoint parameter time series matrix. Through the above processing, the differences in dimensions and magnitudes between different parameters can be eliminated, providing a uniform-scale input for the subsequent extraction of transient decay gradients.

[0032] More specifically, in a concrete example of this application, the parameter time series matrix before the breakpoint is read, and the parameter categories corresponding to its row dimensions are identified. Electrical parameters representing the metering status and communication parameters representing link transmission quality are segmented into corresponding multi-source heterogeneous parameter quantum matrices. For each row of parameters in the matrix, the mean and standard deviation within the backtracking period are extracted along the time series axis. A standard fractional normalization algorithm is used to map the values ​​of each sampling point to eliminate the magnitude differences between different physical dimensions. A minimal positive smoothing constant is introduced into the denominator to prevent computational overflow caused by the standard deviation of a parameter approaching zero under extreme operating conditions such as meter malfunction. The computational logic of the normalized smoothing mapping is expressed as follows: ; in, The dimensionless smooth parameter matrix is ​​the first... Each parameter dimension in The dimensionless element value at time t. The first parameter in the time series matrix before the breakpoint Each parameter dimension in The original sampled values ​​at time 10:00. For the first The arithmetic mean of the parameters over the entire backtracking time window For the first The standard deviation of each parameter over the entire backtracking time window To prevent division by zero errors in the denominator, a minimal positive smoothing constant is used. All elements after the mapping transformation are reorganized according to their original row and column indices to obtain a dimensionless smoothing parameter matrix with uniform dimensions. This ensures that the parameters in the electrical and communication domains are on the same directly comparable scale, serving as input for subsequent sliding window difference operations and transient decay gradient extraction.

[0033] In the aforementioned data analysis-based full lifecycle management system for metered assets 100, the transient gradient extraction unit 132 is used to perform sliding window difference operations and transient decay gradient extraction on the dimensionless smooth parameter matrix to obtain a multidimensional transient decay gradient matrix. It should be noted that, since the dimensionless smooth parameter matrix only depicts the absolute state of each parameter at each moment, it fails to reflect the evolution rate of the sudden changes or precipitous drops in each parameter of the target energy meter as it approaches the moment of disconnection. This evolution rate is a key morphological feature distinguishing communication failures from physical failures. Based on this, the technical solution of this application further performs sliding window difference operations and transient decay gradient extraction on the dimensionless smooth parameter matrix. Through the above processing, the intensity of transient changes of each parameter at the moment of termination can be quantified, providing dynamic evolution dimension features for subsequent multimodal orthogonal fusion.

[0034] More specifically, in a concrete example of this application, a dimensionless smoothed parameter matrix is ​​received, and a sliding data window with a step size of 1 is initialized in its time series dimension, i.e., the column direction. As the sliding window gradually advances from the beginning of the time axis towards the point of disconnection, the first-order backward difference between the current value and the previous value is calculated for each parameter dimension. This first-order difference reflects the instantaneous change amplitude of the parameter between adjacent acquisition times. The calculation logic of the difference is expressed as follows: ; in, The first digit in the multidimensional transient decay gradient matrix Each parameter dimension in The transient decay gradient value at time t. The dimensionless smooth parameter matrix is ​​the first... Each parameter dimension in the current The dimensionless value of time. This represents the dimensionless value of this parameter dimension at the previous acquisition time. When a connection is lost due to pure communication interference, communication parameters such as signal-to-noise ratio, packet loss rate, and retransmission frequency deteriorate sharply just before the connection is lost, while electrical parameters remain relatively stable, and their transient decay gradients show significant differences. However, when a hardware failure causes a complete power outage, communication and electrical parameters collapse synchronously, and their transient decay gradients tend to be consistent. Therefore, this first-order difference can capture the difference in the evolution of the two types of faults. The difference calculation results of each parameter dimension at each time are arranged in a two-dimensional alignment according to time and parameter dimension to obtain a multi-dimensional transient decay gradient matrix, which serves as the dynamic gradient input for subsequent multimodal orthogonal fusion with the dimensionless smoothed parameter matrix.

[0035] In the aforementioned data analysis-based full lifecycle management system for metrological assets 100, the multimodal orthogonal fusion unit 133 is used to perform multimodal orthogonal fusion of static parameters and dynamic gradients on the dimensionless smooth parameter matrix and the multidimensional transient decay gradient matrix to obtain the breakpoint drop micro-feature tensor. It should be noted that since the dimensionless smooth parameter matrix represents the absolute state of the parameter, and the multidimensional transient decay gradient matrix represents the parameter evolution rate, the two characterize the disconnection process from two complementary perspectives—static and dynamic. If used separately, the complete evolutionary picture at the final moment cannot be fully presented. Based on this, the technical solution of this application further performs multimodal orthogonal fusion of static parameters and dynamic gradients on the dimensionless smooth parameter matrix and the multidimensional transient decay gradient matrix. Through the above processing, the absolute state and evolution rate can be aggregated into a unified high-dimensional feature expression, providing a complete discriminative input for subsequent deformation matching with the fault twin map.

[0036] More specifically, in a concrete example of this application, the dimensionless smooth parameter matrix representing the absolute state of static operation and the multidimensional transient decay gradient matrix representing the dynamic evolution rate are respectively expanded in dimension, upgrading them from two-dimensional matrices to single-channel three-dimensional tensor formats. Then, a channel splicing mechanism is used to orthogonally overlap and splice the two expanded tensors along the channel dimension, so that the static parameters and dynamic gradients are carried side-by-side along independent channels while maintaining their respective time-parameter index correspondences, without interfering with each other and complementing each other. The fusion logic is expressed as follows: ; in, For the fused breakpoint drop micro-feature tensor, It is a dimensionless smooth parameter matrix. The transient decay gradient matrix is ​​multidimensional. For tensor concatenation operators, subscript This indicates that the splicing operation involves orthogonal superposition along the channel dimension of the tensor. After splicing and fusion, a three-dimensional data block with a size equal to the number of parameters multiplied by the time length multiplied by the two channels is obtained, namely the breakpoint drop micro-feature tensor. One channel carries the absolute level of each parameter at the moment of loss of connection, while the other channel carries the transient attenuation intensity of each parameter. Together, they characterize the electrical and communication evolution of the target energy meter at the moment of its demise, serving as the standard input feature for nonlinear deformation matching with various standard waveform spectra in the subsequent fault orthogonal decoupling classification process.

[0037] In the aforementioned data analysis-based metering asset lifecycle management system 100, the fault orthogonal decoupling classification module 140 is used to perform fault orthogonal decoupling classification on the breakpoint drop micro-feature tensor and the standard waveform spectrum in the pre-set fault twin spectrum library based on the communication interference benchmark characteristics of the distribution area to obtain fault decoupling labels. It should be noted that, although the breakpoint drop micro-feature tensor carries evidence of the degradation morphology of the target energy meter itself, it is still difficult to determine whether the communication link deterioration is due to its own fault or common environmental interference in the distribution area based solely on the intrinsic characteristics of the equipment. Furthermore, fixed classification criteria cannot adapt to the differences in communication environments of different distribution areas. Based on this, the technical solution of this application further dynamically adjusts the classification activation threshold using the baseline characteristics of communication interference in the distribution area, so that the classification criteria adapt to the level of spatial environmental interference. Furthermore, a dynamic time warping algorithm is used to nonlinearly align the breakpoint drop micro-feature tensor with various standard waveform spectra in a pre-set fault twin graph library on the time axis using nonlinear deformation and Euclidean distance accumulation. The obtained distance is then evaluated and analyzed based on the dynamic activation threshold to complete the qualitative analysis of orthogonal fault decoupling and obtain fault decoupling labels. Through the above processing, spatial environmental evidence and temporal degradation evidence can be fused and mutually verified within the same judgment framework, thereby effectively stripping away the masking effect of external interference on the fault manifestation and achieving precise isolation between communication anomalies and intrinsic failure.

[0038] Figure 5 This is a block diagram of the fault orthogonal decoupling classification module in a data analysis-based metering asset lifecycle management system according to an embodiment of this application. Figure 5 As shown, the fault orthogonal decoupling classification module 140 includes: a threshold dynamic adjustment unit 141, used to perform spatial environmental interference quantification and classification threshold dynamic adjustment on the communication interference benchmark features of the substation area based on a preset benchmark threshold vector to obtain a dynamic activation threshold vector; a deformation alignment matching unit 142, used to perform nonlinear deformation alignment and Euclidean distance accumulation on the time axis of the breakpoint drop micro-feature tensor and various standard waveform spectra in the fault twin graph library to obtain the twin shortest distance vector; and a fault decoupling qualitative unit 143, used to perform multi-dimensional distance evaluation analysis and orthogonal fault decoupling qualitative analysis on the twin shortest distance vector based on the dynamic activation threshold vector to obtain a fault decoupling label.

[0039] In the aforementioned data analysis-based full lifecycle management system for metered assets 100, the threshold dynamic adjustment unit 141 is used to perform spatial environmental interference quantification and classification threshold dynamic adjustment on the communication interference benchmark characteristics of the distribution area based on a preset benchmark threshold vector to obtain a dynamic activation threshold vector. It should be noted that, given the varying electromagnetic environments and communication network conditions of different distribution areas, using fixed judgment thresholds to qualitatively characterize various faults is difficult to adapt to the environmental differences between distribution areas. In areas with strong interference, communication anomalies are easily misjudged as normal fluctuations, while in areas with weak interference, normal fluctuations are easily misjudged as faults. Therefore, the technical solution of this application further uses a preset benchmark threshold vector as a benchmark to quantify the spatial environmental interference of the communication interference benchmark characteristics of the distribution area. The quantified interference intensity is then converted into an adaptive adjustment amount for the communication fault judgment threshold after nonlinear mapping. Based on this, the preset benchmark threshold vector is adjusted element-by-element to obtain the dynamic activation threshold vector. Through the above processing, the classification criteria can be adaptively scaled up and down according to the interference level of the spatial environment in which the target node is located, thereby effectively matching the real communication environment of each station area and providing a judgment threshold that is linked to the environment for subsequent fault decoupling and qualitative analysis.

[0040] Figure 6 This is a block diagram of a threshold dynamic adjustment unit in a data analysis-based full lifecycle management system for metered assets according to an embodiment of this application. Figure 6 As shown, the threshold dynamic adjustment unit 141 includes: a degradation asymmetry ratio extraction subunit 1411, used to extract the communication-electrical degradation asymmetry ratio from the breakpoint drop micro-feature tensor to obtain the communication-electrical degradation asymmetry ratio; an interference consistency coupling quantization subunit 1412, used to perform interference consistency coupling quantization on the base features of communication interference in the distribution area and the communication-electrical degradation asymmetry ratio to obtain the interference consistency coupling factor; and a base threshold dynamic activation subunit 1413, used to dynamically activate a preset base threshold vector based on the base features of communication interference in the distribution area and the interference consistency coupling factor to obtain a dynamic activation threshold vector.

[0041] In the aforementioned data analysis-based full lifecycle management system for metered assets 100, the degradation asymmetry ratio extraction subunit 1411 is used to extract the communication-electrical degradation asymmetry ratio from the breakpoint drop micro-feature tensor to obtain the communication-electrical degradation asymmetry ratio. It should be noted that, given that existing methods only transmit the breakpoint drop micro-feature tensor to the subsequent matching stage during the threshold adjustment phase, the structural difference information between the communication and electrical domains contained within it is not utilized. This leads to a situation where, under conditions of strong interference in the distribution area and concurrent meter hardware failure, adjusting the threshold solely based on the external interference intensity can misjudge meter damage as a communication failure. Therefore, the technical solution of this application further divides the breakpoint drop micro-feature tensor into electrical sub-channel slices and communication sub-channel slices along the feature dimension, calculating the combined degradation intensity of both at the moment of near disconnection and obtaining their ratio. Through the above processing, the intrinsic degradation morphology fingerprint of the target energy meter itself can be mined, providing discriminative evidence from within the device for subsequent causal consistency verification.

[0042] More specifically, in a concrete example of this application, the breakpoint drop micro-feature tensor is segmented along the feature dimension into electrical sub-channel slices and communication sub-channel slices. The electrical sub-channel slices correspond to voltage gradient, current gradient, and power mutation rate gradient, while the communication sub-channel slices correspond to signal-to-noise ratio gradient, packet loss rate gradient, and retransmission frequency gradient. The gradient vectors of both sets of slices at the end of the time dimension, i.e., the moment closest to the loss of connection, are extracted, and their respective L2 norms are calculated to quantify the combined degradation intensity of the communication and electrical domains at the final moment. Then, the degradation intensity of the communication domain is divided by the degradation intensity of the electrical domain to obtain the communication-electrical degradation asymmetry ratio. To prevent computational overflow caused by the electrical domain gradient being zero, a minimal positive smoothing constant is introduced into the denominator. This process is represented as follows: ; in, The output is the ratio of communication electrical degradation asymmetry. The time-end gradient vector is extracted from the communication sub-channel gradient layer of the drop-off micro-feature tensor. It has a dimension of 3 and corresponds to the terminal attenuation rate of signal-to-noise ratio, packet loss rate, and retransmission frequency, respectively. The time-terminal gradient vector, with dimension 3, is extracted from the electrical subchannel gradient layer of the breakpoint drop micro-feature tensor, and corresponds to the terminal decay rate of voltage, current, and power abrupt change rate, respectively. For L2 norm operations, To prevent extremely small positive smoothing constants with a denominator of zero.

[0043] It should be noted that when the ratio is much greater than 1, it indicates that the degradation intensity of the communication domain is much greater than that of the electrical domain, which is highly consistent with the physical scenario of a disconnection caused by pure communication interference while the hardware itself is intact; when the ratio is close to 1, it indicates that the communication domain and the electrical domain collapse almost simultaneously, which is consistent with the physical scenario of a hardware failure caused by an MCU crash or power supply burnout leading to a complete power outage. This ratio provides evidence of the degradation pattern from within the target device for subsequent causal consistency verification. For example, in a certain distribution area, the carrier channel quality deteriorates due to the startup of an adjacent variable frequency air conditioning cluster. At the same time, the internal power module of the target energy meter experiences overvoltage breakdown, resulting in a hardware-level crash. At this time, the spatial interference intensity reported by neighboring nodes is very high. If the threshold for judging communication faults is relaxed based solely on this, it will be misjudged as communication interference. However, by extracting the breakpoint drop micro-feature tensor of the target energy meter, it can be seen that its signal-to-noise ratio, packet loss rate, and terminal attenuation rate of retransmission frequency collapse almost synchronously with the terminal attenuation rate of voltage, current, and power mutation rate. The resulting communication electrical degradation asymmetry ratio is close to 1. Based on this, it can be identified that the loss of connection is actually due to the hardware fault rather than communication interference, providing an internal basis for avoiding misjudgment.

[0044] In the aforementioned data analysis-based full lifecycle management system for metered assets 100, the interference consistency coupling quantification subunit 1412 is used to perform interference consistency coupling quantification on the communication interference baseline characteristics and the communication electrical degradation asymmetry ratio of the distribution area to obtain the interference consistency coupling factor. It should be noted that, given that simply obtaining the communication electrical degradation asymmetry ratio of the target device is insufficient to directly correct the threshold, it is also necessary to establish a causal consistency quantification assessment between it and external spatial interference; otherwise, under contradictory conditions such as concurrent high interference and hardware failure or concurrent low interference and local shielding, the cause of the disconnection cannot be determined. Based on this, the technical solution of this application further calculates the L2 norm of the distribution area communication interference baseline characteristics to obtain the spatial interference scalar intensity, thereby generating a theoretically expected asymmetry ratio, and attenuating the deviation between it and the actual asymmetry ratio. Through the above processing, the causal consistency between the external interference intensity and the internal degradation mode can be transformed into a calculable confidence quantification value, providing a physical consistency criterion for subsequent threshold gating adjustment.

[0045] More specifically, in a concrete example of this application, the L2 norm of the baseline characteristics of communication interference in the distribution area is first calculated to obtain the scalar intensity of spatial interference. Then, this intensity is mapped to the zero-to-one interval using a Sigmoid activation function and multiplied by a saturation scaling factor to generate the theoretically expected asymmetry ratio that should exist under the current interference level if the fault is indeed caused by communication interference. This expected value constitutes the theoretical anchor point for causal consistency verification. Next, the deviation between the actual communication electrical degradation asymmetry ratio and this expected value is calculated. A Gaussian kernel function is used to attenuate this deviation, so that the smaller the deviation, the closer the actual degradation mode matches the theoretical expectation, and the closer the output is to 1; the larger the deviation, the closer the output is to 0. Finally, the interference consistency coupling factor is obtained. This process is represented as follows: ; in, For the scalar intensity of spatial interference in the transformer area, This is the reference feature vector for communication interference in the distribution area. This refers to L2 norm operations.

[0046] ; in, This represents the theoretically expected asymmetric ratio under the current interference intensity. This is the maximum saturation scaling factor for the desired asymmetry ratio, which can be set based on historical calibration data. The Sigmoid activation function smoothly maps the interference intensity to the interval between zero and one.

[0047] ; in, The output interference consistency coupling factor takes values ​​in the left-open, right-closed interval (0,1]. This represents the communication electrical degradation asymmetry ratio obtained previously. This represents the theoretically expected asymmetric ratio under the current disturbance level. This is the bandwidth parameter of the Gaussian kernel function, which controls the strictness of the consistency tolerance.

[0048] It should be noted that the choice of Gaussian kernel function gives the consistency check a soft decision characteristic. It does not completely reject threshold relaxation once a deviation exists, but rather smoothly decays the relaxation range according to the magnitude of the deviation. This has the necessary engineering robustness when facing weak measurement noise and short-term transient interference in real transformer areas. The physical connotation of the interference consistency coupling factor can be understood as a confidence quantification of the causal relationship between the existence of strong interference in the external space and the fact that the degradation mode of the target equipment itself exhibits communication-dominated asymmetric degradation. Taking the concurrent operation of a variable frequency air conditioning cluster starting up, causing channel degradation in the distribution area, and the target energy meter's power module experiencing overvoltage breakdown as an example, the spatial interference scalar intensity is very high at this time. Based on this, the theoretical expected asymmetry ratio is large, which means that if the loss of connection is indeed caused by communication interference, the equipment should exhibit asymmetric degradation in the communication domain far exceeding that in the electrical domain. However, the actual communication-electrical degradation asymmetry ratio of the target energy meter is close to 1, which deviates greatly from the theoretical expectation. After Gaussian kernel attenuation mapping, the interference consistency coupling factor approaches 0, indicating that the strong external interference and the internal degradation mode do not constitute a causal consistency. Based on this, the erroneous relaxation of the threshold can be suppressed in subsequent gating, providing a causal basis for the correct identification of hardware faults.

[0049] In the aforementioned data analysis-based full lifecycle management system for metered assets 100, the benchmark threshold dynamic activation subunit 1413 is used to dynamically activate a preset benchmark threshold vector based on the base characteristics of the transformer area communication interference and the interference consistency coupling factor to obtain a dynamic activation threshold vector. It should be noted that existing methods only use spatial interference intensity as a single factor to drive threshold relaxation. In scenarios with concurrent high interference and hardware failures, this can lead to misjudging physical damage as communication interference, while in scenarios with concurrent low interference and local shielding, it can lead to misjudging communication blockage as hardware damage. Therefore, the technical solution of this application further multiplies the original penalty factor obtained from the base characteristics of the transformer area communication interference through hyperbolic tangent mapping with the interference consistency coupling factor to form a joint gating factor, and then performs a Hadamard product operation with the preset benchmark threshold vector and the fault category mask vector. Through this processing, the threshold adjustment can add a degradation morphology consistency condition as a physical constraint in addition to the spatial interference amplitude condition, thereby effectively blocking the erroneous path of hardware failure being misjudged as communication failure.

[0050] More specifically, in a concrete example of this application, after completing the intrinsic degradation morphology verification and causal consistency quantification, the existing single hyperbolic tangent penalty factor is upgraded to a joint gating penalty factor. The original penalty factor obtained by inputting the L2 norm calculation of the substation communication interference baseline features into the hyperbolic tangent function is then multiplied by the previously output interference consistency coupling factor to form the joint gating factor. Finally, the preset static baseline threshold vector is multiplied by the joint gating factor and the fault category mask vector to output the final dynamic activation threshold vector. This process is represented as follows: ; in, This is the final output dynamic activation threshold vector. The system pre-sets static baseline judgment threshold vectors for each fault category. The Hadamard product operator represents the element-wise multiplication of vectors. The overall adjustment coefficient for threshold relaxation sensitivity. Using a hyperbolic tangent function, the influence of interference intensity is limited to the range of -1 to +1. For the scalar intensity of spatial interference in the transformer area, As a coupling factor for interference consistency, and as a gated multiplier for physical consistency, This is a fault category mask vector that specifies that threshold relaxation should be applied only to communication-related fault entries.

[0051] It should be noted that when the intensity of spatial interference is high, that is... Approaching 1 and having a high consistency factor When the threshold approaches 1, the joint gating factor is fully effective, and the threshold for determining communication-related faults is reasonably relaxed; however, when a contradictory scenario of high interference but low asymmetry occurs, Approaching 0 suppresses the joint gating factor to an extremely low level, and the threshold is hardly relaxed, thus effectively blocking the erroneous path of hardware failure being misjudged as communication failure. As a result, the generation of the dynamic activation threshold vector is simultaneously constrained by both external environmental evidence and internal degradation evidence.

[0052] For example, in situations with low spatial interference but localized communication shielding, the target energy meter is located in a building basement or metal shielding layer and is subject to highly localized directional communication shielding. At this time, the spatial interference scalar intensity reported by neighboring nodes is not high, but the target device itself exhibits communication-dominated asymmetric degradation, that is, the terminal attenuation rate of signal-to-noise ratio, packet loss rate, and retransmission frequency far exceeds the electrical parameters. The obtained communication electrical degradation asymmetric ratio is consistent with the theoretical expectation under this weak interference level, and the interference consistency coupling factor is still maintained at a reasonable level, so that the joint gating factor has a moderate threshold relaxation capability. Combined with the high similarity of the communication fault spectrum in the subsequent dynamic time warping matching, the communication fault can be correctly characterized, avoiding the ineffective dismantling and scrapping of excellent energy meters that are only blocked by communication.

[0053] In the aforementioned data analysis-based full lifecycle management system for metered assets 100, the deformation alignment and matching unit 142 is used to perform nonlinear deformation alignment and Euclidean distance accumulation on the time axis of the breakpoint drop micro-feature tensor and various standard waveform spectra in the fault twin graph library to obtain the twin shortest distance vector. It should be noted that, due to varying degrees of hardware aging among different energy meters, their microscopic response processes before disconnection exhibit time-varying drift. If a rigid alignment method is used to directly compare the breakpoint drop micro-feature tensor and standard waveform spectra, false distance deviations will occur due to time misalignment. Therefore, the technical solution of this application further performs nonlinear deformation alignment and Euclidean distance accumulation on the time axis of the breakpoint drop micro-feature tensor and various standard waveform spectra in the fault twin graph library. Through the above processing, the interference of response time drift between devices on similarity assessment can be eliminated, obtaining a distance metric characterizing the proximity of the target device to various standard faults.

[0054] More specifically, in a specific example of this application, the fault twin graph library is parsed, and standard waveform graph tensors of several pre-set standard typical faults are extracted. These standard typical faults cover external communication network interference faults such as communication module disconnection due to external electromagnetic interference and harmonic crosstalk in the transformer area, as well as hardware damage faults of the core components of the meter body such as core microcontroller unit crashes and power module burnouts. Each standard waveform graph tensor is a fixed benchmark of the microscopic evolution of the corresponding fault before disconnection. For each standard waveform graph tensor, a dynamic time warping algorithm is invoked to nonlinearly stretch and distort the breakpoint drop micro-feature tensor on the time axis to find the optimal alignment path between them, thereby eliminating the microscopic response time drift caused by the aging of different device hardware. Then, the Euclidean distance between the corresponding points along the optimal alignment path is calculated and summed to obtain the warped cumulative distance between the target device and the standard fault of that type. The calculation logic is expressed as follows: ; in, For the breakpoint drop micro-feature tensor and the first The dynamic time-warped cumulative shortest distance between similar standard waveform spectra. To find a set of nonlinear aligned temporal path coordinates in a normalized matrix. For breakpoint drop micro-feature tensors in temporal index The feature vector at that location, For the first Class-standard waveform spectra in time series index The feature vector at that location, This is for Euclidean distance calculation. and These are independent time-series index nodes for the two sequences.

[0055] The regularized cumulative distances calculated for each type of standard fault are concatenated and arranged in order of category index to obtain a one-dimensional twin shortest distance vector, the structure of which is expressed as follows: ; in, The twin shortest distance vector, For the target device and the first The dynamic time-warped cumulative shortest distance between standard fault classes. This represents the total number of pre-defined standard fault categories in the fault twin graph library. The resulting twin shortest distance vector quantitatively characterizes the degree of similarity between the target equipment's final evolution morphology and various standard faults, serving as input for subsequent multi-dimensional distance evaluation analysis and orthogonal fault decoupling qualitative analysis, combined with dynamic activation threshold vectors.

[0056] In the aforementioned data analysis-based full lifecycle management system for metered assets 100, the fault decoupling qualitative unit 143 is used to perform multi-dimensional distance evaluation and analysis and orthogonal fault decoupling qualitative analysis on the twin shortest distance vector based on the dynamic activation threshold vector to obtain fault decoupling labels. It should be noted that since the twin shortest distance vector only reflects the proximity of the target equipment to various standard faults using distance values, if it is not combined with a judgment threshold that is linked to the spatial environment for screening, normal communication fluctuation errors are easily included in the fault candidates in areas with strong interference, making it impossible to separate apparent interference. Based on this, the technical solution of this application further performs multi-dimensional distance evaluation and analysis and orthogonal fault decoupling qualitative analysis on the twin shortest distance vector based on the dynamic activation threshold vector. Through the above processing, effective fault candidates can be screened and uniquely qualitatively analyzed under environmentally adaptive threshold constraints to obtain fault decoupling labels.

[0057] More specifically, in a specific example of this application, the twin shortest distance vector and the dynamic activation threshold vector corresponding to the current station area communication environment are received, and a logical mask comparison of the two is performed in the same dimension. Since the dynamic activation threshold vector has been scaled according to the consistency of spatial interference intensity and degradation mode, each component gives the distance threshold that the corresponding fault category can be accepted in the current environment. Elements in the twin shortest distance vector that are greater than or equal to the corresponding dynamic activation threshold are filtered out. Such elements indicate that the target device is too far away from the standard fault of this type and the similarity is too low, and are judged as an impossible fault category. All elements less than the threshold are retained as valid candidate category distances. The filtering logic is expressed as follows: ; in, To form a valid candidate category index set that meets the threshold interception conditions, For fault category index number, The first of the twin shortest distance vectors The normalized cumulative distance value for each category, For the corresponding number in the dynamic activation threshold vector Threshold values ​​for each category, This represents the total number of pre-defined standard fault categories in the fault twin database.

[0058] The distance component with the global minimum value is parsed from the set of valid candidate categories. Its corresponding fault category primary key is extracted, and superficial interference is removed using a hard classification method to achieve unique qualitative analysis, resulting in a fault decoupling label. The parsing logic is as follows: ; in, This is the fault category identifier corresponding to the final output fault decoupling label. The operation is to find the category index that minimizes the distance. The category index number in the set of valid candidate category indexes. The first of the twin shortest distance vectors The normalized cumulative distance value for each category, This is the set of valid candidate category indices. The obtained fault decoupling labels clearly define whether the fault domain of the target energy meter's disconnection belongs to external communication network interference or hardware damage to the core components of the meter itself, serving as the basis for triggering differentiated management and control logic and generating targeted maintenance dispatch instructions.

[0059] In the aforementioned data analysis-based full lifecycle management system for metered assets 100, the differentiated control and ledger update module 150 is used to trigger differentiated control logic based on the category attributes of fault decoupling tags to generate targeted maintenance dispatch instructions, and synchronously write the traceability event characteristics into the target asset electronic ledger to update the full lifecycle health profile. It should be noted that although fault decoupling tags clearly identify the fault domain of the disconnection, if communication faults and physical faults are not differentiated and handled accordingly, the drawbacks of traditional blind dispatching will persist, and the fault determination results will not be used as a basis for the evolution of asset health. Therefore, the technical solution of this application further triggers differentiated control logic based on the category attributes of fault decoupling tags to generate targeted maintenance dispatch instructions, and synchronously writes the traceability event characteristics into the target asset electronic ledger to update the full lifecycle health profile. Through the above processing, maintenance dispatch and fault attribution can be accurately matched, and the synchronous evolution of the asset health profile can be completed under the premise of transaction consistency.

[0060] More specifically, in a specific example of this application, incoming fault decoupling tags are intercepted and their underlying classification primary keys are parsed. The primary key is compared with a predefined fault attribution domain. If the tag belongs to the external communication network interference domain, the service diversion control code is set to 0. If it belongs to the meter core component hardware damage domain, the service diversion control code is set to 1. At the same time, a multi-dimensional structured traceability event feature vector is extracted based on the target energy meter identifier associated with the tag and the timestamp of the time of disconnection. The service diversion control code and the traceability event feature vector are encapsulated and packaged in a data structure to obtain a control strategy and event union, so as to ensure the transaction atomicity of fault characterization conclusions and traceability evidence in the process of being transmitted downstream.

[0061] Subsequently, the control strategy and event consortium are unpacked, and the business diversion control code is extracted to trigger differentiated business microservices. If the business diversion control code is 0, a work order template for suppressing meter removal and troubleshooting only for the communication channel of the transformer area is matched, thereby preventing good energy meters with only communication faults from being ineffectively removed. If the business diversion control code is 1, a work order template for spare parts allocation and meter replacement construction is matched with the enterprise resource planning system and metering asset management system, so that faulty energy meters that are indeed damaged are promptly taken out of operation. The generated instructions are temporarily stored in the memory queue as temporary dispatch instructions. Simultaneously, the traceability event feature vector in the consortium is unpacked and mapped, and transformed into ledger update-driven features specifically for lifecycle assessment.

[0062] Finally, the system receives the ledger update driver feature, queries the historical health score of the asset batch, calculates the score penalty based on the severity of the current failure, and uses an exponentially weighted decay algorithm to perform weighted decay deduction on the historical score. The calculation logic is as follows: ; in, The latest asset health score, calculated at the current moment, constitutes the core quantitative indicator of the full life-cycle health profile. This refers to the historical health score of the asset in the previous period recorded in the electronic ledger. The historical integral retention factor, also known as the decay memory factor, is a constant close to 1. The penalty weight constant is preset for this fault type. To calculate the specific integral penalty amount using the ledger update-driven feature mapping, As a boundary protection function, it ensures that the asset health score decays to a minimum of zero. The new score status after deduction is persistently written to the device's electronic ledger to generate a full lifecycle health profile. At the moment the ledger update database transaction is successfully committed, the temporarily stored dispatch instructions are authorized and issued, transforming them into final effective targeted maintenance dispatch instructions. This ensures strict synchronization between the electronic ledger records and the outgoing physical work orders, avoiding state misalignments such as the ledger being updated but the work order not being issued, or the work order being issued but not recorded in the ledger.

[0063] In summary, the data analysis-based full lifecycle management system for metrological assets according to the embodiments of this application is explained. It constructs a metrological asset fault decoupling and judgment mechanism based on the fusion of two-dimensional evidence, namely spatial cross-validation and temporal micro-evidence, in order to solve the problem of highly overlapping characteristics and blind judgment of communication faults and physical faults at the main station. When the target electricity meter's heartbeat is interrupted, triggering a disconnection alarm, on the one hand, the system backtracks and extracts the operating parameters before the disconnection point based on the time of disconnection, and addresses the communication routing topology. Then, it calls neighboring nodes in the same branch as communication environment probes, and obtains the base features of communication interference in the transformer area through spatial distance weighted fitting, characterizing external spatial evidence. On the other hand, it extracts transient decay gradients and orthogonally fuses multimodal channels for the parameters before the disconnection point to obtain the micro-feature tensor of the disconnection point drop, characterizing the intrinsic degradation evidence of the equipment itself at the moment of its death. Subsequently, it dynamically adjusts the classification threshold based on the interference base features, and performs nonlinear deformation matching and orthogonal decoupling between the micro-feature tensor and the preset fault twin map, outputting a high-confidence fault decoupling label. Finally, it triggers differentiated operation and maintenance dispatch and updates the full life cycle health file accordingly, thereby breaking through the black box limitation of single-point information of the main station, realizing precise isolation between communication anomalies and the failure of the main body, reducing the invalid dispatch rate and the loss of erroneous dismantling and scrapping of good assets.

Claims

1. A data analysis-based full lifecycle management system for measurement assets, characterized in that, include: The backtracking and topology addressing module is used to perform backtracking and topology addressing on the real-time operation messages of the energy meter when the heartbeat interruption time of the target energy meter exceeds the set threshold, based on the time of disconnection, in order to obtain the parameter timing matrix and communication route topology snapshot before the breakpoint. The interference baseline feature fitting module is used to perform channel state exploration and spatial distance weighted fitting on the same branch neighbor nodes of the target energy meter based on the cascade hierarchical relationship in the communication routing topology snapshot to obtain the communication interference baseline features of the transformer area. The micro-feature tensor fusion module is used to extract transient decay gradients from the pre-breakpoint parameter time-series matrix and orthogonally fuse them with multimodal channels to obtain the breakpoint drop micro-feature tensor. The fault orthogonal decoupling classification module is used to perform fault orthogonal decoupling classification on the breakpoint drop micro-feature tensor and the standard waveform spectrum in the pre-set fault twin spectrum library based on the communication interference benchmark features of the transformer area to obtain fault decoupling labels. The differentiated management and ledger update module is used to trigger differentiated management logic based on the category attributes of fault decoupling tags to generate targeted operation and maintenance dispatch instructions, and synchronously write the traceability event characteristics into the target asset electronic ledger to update the full life cycle health profile.

2. The data analysis-based full lifecycle management system for measurement assets according to claim 1, characterized in that, The real-time operation message of the electricity meter includes the effective value of the meter terminal voltage, the effective value of the meter terminal current, the transient change rate of active power, the signal-to-noise ratio of carrier communication, the packet loss rate of carrier communication, and the retransmission frequency of carrier communication.

3. The data analysis-based full lifecycle management system for measurement assets according to claim 1, characterized in that, The backtracking and topology addressing module includes: The disconnection alarm triggering unit is used to monitor the arrival time of the jump packet of the real-time running message center of the energy meter in real time, calculate the difference between the current system clock and the latest heartbeat time, and compare it with the preset heartbeat interruption threshold to determine the disconnection alarm trigger, and obtain the target energy meter identifier and the disconnection time timestamp. The time-series matrix backtracking unit is used to locate the target energy meter in the high-frequency buffer and backtrack in the historical direction based on the target energy meter identifier and the timestamp of the time of disconnection to obtain the parameter time-series matrix before the breakpoint. The topology snapshot extraction unit is used to extract a graph structure subgraph from the target node's cascading hierarchy relationship and surrounding network layout information at that moment in the global dynamic routing management table, based on the target energy meter identifier and the timestamp of the time of disconnection, to obtain a communication routing topology snapshot.

4. The data analysis-based full lifecycle management system for measurement assets according to claim 1, characterized in that, The interference benchmark feature fitting module includes: The neighbor node resolution unit is used to perform same-branch neighbor node resolution and spatial distance quantization on the communication routing topology snapshot to obtain the same-branch neighbor node set and spatial logical distance vector. The channel state detection unit is used to send high-priority channel detection commands to the edge concentrator of the substation based on the set of neighboring nodes in the same branch, so as to sample and obtain the signal strength, signal-to-noise ratio and link packet loss rate of each neighboring node to obtain the real-time communication state matrix of the neighboring nodes. The baseline feature fitting unit is used to perform spatial attenuation attention weighting and baseline feature fitting on the spatial logical distance vector and the real-time communication state matrix of neighboring nodes to obtain the baseline features of communication interference in the transformer area.

5. The data analysis-based full lifecycle management system for measurement assets according to claim 1, characterized in that, The micro-feature tensor fusion module includes: The normalization and smoothing unit is used to perform multi-source heterogeneous parameter quantum matrix segmentation and normalization and smoothing on the parameter time series matrix before the breakpoint to obtain a dimensionless smooth parameter matrix; The transient gradient extraction unit is used to perform sliding window difference operations and transient decay gradient extraction on the dimensionless smooth parameter matrix to obtain a multidimensional transient decay gradient matrix. The multimodal orthogonal fusion unit is used to perform multimodal orthogonal fusion of static parameters and dynamic gradients on a dimensionless smooth parameter matrix and a multidimensional transient decay gradient matrix to obtain a breakpoint drop micro-feature tensor.

6. The data analysis-based full lifecycle management system for measurement assets according to claim 1, characterized in that, The fault orthogonal decoupling classification module includes: The threshold dynamic adjustment unit is used to dynamically adjust the threshold for spatial environmental interference quantification and classification of the communication interference baseline characteristics of the transformer area based on a preset baseline threshold vector in order to obtain a dynamic activation threshold vector. The deformation alignment matching unit is used to perform nonlinear deformation alignment and Euclidean distance accumulation on the time axis of various standard waveform maps in the breakpoint drop micro-feature tensor and fault twin map library to obtain the twin shortest distance vector. The fault decoupling qualitative unit is used to perform multidimensional distance evaluation and orthogonal fault decoupling qualitative analysis on twin shortest distance vectors based on dynamic activation threshold vectors to obtain fault decoupling labels.

7. The data analysis-based full lifecycle management system for measurement assets according to claim 1, characterized in that, The differentiated management and ledger update module includes: The strategy union construction unit is used to obtain the business diversion control code by parsing the classification primary key of the fault decoupling label and comparing it with the predefined fault attribution domain, and simultaneously extract the structured traceability event feature vector and combine them to obtain the control strategy and event union; The differentiated traffic processing unit is used to trigger differentiated business microservices to perform work order template matching and source traceability feature unpacking and mapping based on the business traffic control code in the control strategy and event union to obtain temporary dispatch instructions and ledger update driving features. The ledger update and dispatch authorization unit is used to calculate the health score decay of the ledger update driving features and persist it to the electronic ledger to generate a full life cycle health score file. After the transaction is successfully confirmed, the temporary dispatch instructions are authorized and issued to generate targeted operation and maintenance dispatch instructions.

8. The data analysis-based full lifecycle management system for measurement assets according to claim 6, characterized in that, The threshold dynamic adjustment unit includes: The degradation asymmetry ratio extraction subunit is used to extract the communication-electrical degradation asymmetry ratio from the breakpoint drop micro-feature tensor to obtain the communication-electrical degradation asymmetry ratio. The interference consistency coupling quantization subunit is used to perform interference consistency coupling quantization on the communication interference reference characteristics and the ratio of communication electrical degradation asymmetry in the transformer area to obtain the interference consistency coupling factor. The reference threshold dynamic activation subunit is used to dynamically activate the preset reference threshold vector based on the reference characteristics of communication interference in the distribution area and the interference consistency coupling factor to obtain the dynamic activation threshold vector.

Images