Modularized double-layer prefabricated power transformation system based on digital twinning

Abstract

The invention discloses a modular double-layer prefabricated power transformation system based on digital twinning, relates to the technical field of twinning power equipment, and aims to deal with the challenges of new energy access and rapid urban load increase, eliminate data naming conflicts through semantic handshake, solidify and unify a time reference through a time base, reveal the internal relation between data through association modeling, and improve the power transformation efficiency. The method is advantaged in that model and object deviation, version patch synchronization firmware and model, strategy refreshing optimization protection setting value and life evaluation are carried out, deployment efficiency, operation reliability and intelligent level of the transformer substation are improved, a solid foundation is laid for popularization and application of a digital power distribution network, and requirements of a modern power system for efficient and intelligent operation and maintenance are satisfied.

Description

Technical Field

[0001] The present invention relates to the technical field of twin power equipment, and specifically to a modular double-layer prefabricated substation system based on digital twins. Background Art

[0002] Against the backdrop of rapidly growing renewable energy access and urban loads, modular, two-tier prefabricated substations are being widely adopted. Each functional box is pre-commissioned in the factory, then stacked in layers. Once hoisted into place, it can be energized and operated. A digital twin platform maps the status of every primary, secondary, and auxiliary device in real time, enabling remote monitoring and predictive maintenance. Because the boxes are provided by different manufacturers, with varying sensor device models, sampling schedules, and communication semantics, all state variables must be immediately packaged into a unified model at the edge and pushed to the cloud twin within milliseconds once on-site installation is complete. Otherwise, the thermal-mechanical-electrical coupling analysis, protection setting verification, and lifespan assessments that operations and maintenance personnel rely on will be distorted. Existing research shows that prefabricated cabin substations offer a short construction period and a rich set of interfaces, albeit with varying standards. A digital twin framework can provide visualization and analysis support for the entire station lifecycle, with object-based data semantics central to achieving this goal.

[0003] However, the heterogeneous data streams generated when cross-layer cabinets are connected remain a pain point in the industry: There are naming conflicts and time base drift between the high-power data sent by upper-layer primary devices and the operating condition information uploaded by lower-layer secondary and auxiliary devices. Traditional offline resynchronization methods can only align waveforms after the fact. While edge gateways can perform protocol conversion from Modebus to message queue telemetry transmission, they struggle to complete semantic remapping on-site. When this desynchronization continues to accumulate, the twin model gradually deviates from the actual device, ultimately leading to misjudgment of state assessments, delayed fault warnings, and failed self-healing controls. Even introducing Kalman estimation or fault warning algorithms based on sensor drift detection can only alleviate rather than cure these problems, as version updates and manufacturer expansions constantly introduce new differences.

[0004] In current practice, version control concepts should be integrated into asset model management. Using Git-like operations and maintenance methods, each firmware upgrade should be tied to the evolution of the twin object to fundamentally maintain consistency between the virtual and the physical. However, existing prefabricated substations lack a rapid calibration and continuous delivery mechanism for cross-layer modules, making the twin-physical coupling chain most vulnerable during the initial operation of the site. Therefore, how to bridge the cross-layer data gap immediately after installation and ensure stable alignment between the twin and the physical has become a core technical issue that needs to be addressed in modular, two-layer prefabricated substation systems driven by digital twins. Summary of the Invention

[0005] (1) Technical problems solved In response to the shortcomings of the existing technology, the present invention provides a modular two-layer prefabricated substation system based on digital twins, which eliminates data naming conflicts through semantic handshakes, solidifies the time base to unify the time reference, reveals the intrinsic connection between data through association modeling, drift discrimination detects deviations between models and real objects, synchronizes firmware and models with version patches, and optimizes protection settings and life assessments through strategy refreshes, thereby improving the deployment efficiency, operational reliability and intelligence level of substations, meeting the needs of modern power systems for efficient and intelligent operation and maintenance, and solving the technical problems recorded in the background technology.

[0006] (2) Technical solution To achieve the above objectives, the present invention is implemented through the following technical solutions: a modular two-layer prefabricated substation system based on digital twins, including a semantic handshake unit, an edge controller that monitors the first report frame of each cabinet, calls the IEC61850 logical node template to complete the ternary registration of equipment, measurement points, and functions, generates the first version of the semantic list and marks the hierarchical source; The time base curing unit inserts a highly stable timestamp into the edge gateway based on the semantic list, completes the mapping of the Modebus to the message queue telemetry protocol, and outputs a consistent time series stream across layers; Association modeling unit, after the time series stream is written into the semantic buffer, the system constructs a multi-dimensional association matrix hierarchically, mapping the primary power quantity and the secondary operating condition quantity into a dynamic graph in real time; The drift discrimination unit generates a residual vector and simultaneously calculates the residual spectral entropy index and the delay jitter index. After the random forest model outputs the orbit credibility coefficient, the patch signal is triggered at a low threshold. The version patch unit receives a trigger signal for the patch process, submits the current firmware fingerprint and the twin model version to the GitOps pipeline, generates a differential patch after approval, and pushes it back to the edge; The strategy refresh unit refreshes the association matrix as soon as the patch is implemented and broadcasts the status change to the cloud microservice. The engine recalculates the protection constant and life curve accordingly, so that on-site control and twin analysis keep pace with each other.

[0007] Furthermore, the edge controller listens to the first message frame of the newly connected box and receives the data packet, verifies the integrity of the data packet and records the reception time, calls the IEC61850 logical node template to parse the first message frame, extracts the device information, measurement point information and function information, and generates a structured data set containing the device identifier, measurement point and function.

[0008] Furthermore, device registration, measurement point registration and function registration are performed, and the device identifier, measurement point identifier, measurement type, hierarchical source and function description are registered in the database to generate a relational table; Generate the first version of the semantic list based on the relationship table, recording the device identifier, measurement point identifier, measurement type, hierarchical source and functional description.

[0009] Furthermore, the edge gateway extracts the measurement point identifier, measurement type and level source from the first version of the semantic list, and ensures the semantic consistency of the data based on the first version of the semantic list; the edge gateway inserts a high-stability timestamp into each received data packet to record the exact time when the data packet is received.

[0010] Furthermore, the edge gateway converts data packets from the Modbus protocol to the message queue telemetry transmission protocol to generate topics and payloads; the edge gateway sorts data packets according to highly stable timestamps, generates a consistent time series stream across layers, and pushes it to the cloud digital twin platform in real time.

[0011] Furthermore, the edge gateway writes the cross-layer consistent time series stream into the semantic buffer in a highly stable timestamp order, extracts the time series data from the semantic buffer, and distinguishes the primary power measurement data generated by the upper-layer primary equipment from the secondary operation data generated by the lower-layer secondary and auxiliary equipment based on the hierarchical source in the first version of the semantic list; The Pearson correlation coefficient is calculated for each correlation pair of primary power measurement data and secondary operation data to generate a multidimensional correlation matrix. The correlation pairs in the multidimensional correlation matrix and the correlation strength represented by the Pearson correlation coefficient are mapped into a dynamic graph. The dynamic graph is stored and updated in real time using graph database technology.

[0012] Furthermore, the actual correlation strength between the upper-level primary equipment measurement points and the lower-level secondary and auxiliary equipment measurement points is extracted from the dynamic map, and the predicted correlation strength is obtained from the digital twin platform. The difference between the two is calculated to generate a residual vector; Fast Fourier transform is applied to the residual vector to generate frequency domain representation. After calculating the normalized power spectral density, the Shannon entropy formula is used to calculate the residual spectral entropy index.

[0013] Furthermore, the highly stable timestamps of the upper-layer primary equipment measurement points and the lower-layer secondary and auxiliary equipment measurement points are extracted from the time series stream, the cross-layer timestamp differences are calculated, and the improved Allan variance is applied to calculate the delay jitter index; The residual spectral entropy index and delay jitter index are input into the random forest model to output the orbital reliability coefficient. When the orbital reliability coefficient is lower than the preset threshold, the patch signal is triggered.

[0014] Furthermore, the edge controller continuously monitors the patch signal. When the patch signal is 1, the edge controller starts the version patch process, collects the device's firmware fingerprint and the model version identifier obtained from the cloud digital twin platform, packages the firmware fingerprint and model version identifier into a version information package, and submits it to the GitOps pipeline.

[0015] Furthermore, after receiving the version information package, the GitOps pipeline compares the firmware fingerprint with the latest firmware fingerprint in the Git repository and the model version identifier with the latest model version identifier to generate a firmware difference patch and a model difference patch; The GitOps pipeline submits the firmware difference patch and the model difference patch for approval, and integrates them into the final difference patch after approval.

[0016] Furthermore, the GitOps pipeline pushes the final differential patch back to the edge controller; the edge controller applies the firmware differential patch to update the device firmware and applies the model differential patch to update the digital twin model, records the new firmware fingerprint and model version identifier, and notifies the cloud-based digital twin platform to synchronize the update.

[0017] Furthermore, the edge controller receives the differential patch pushed back by the GitOps pipeline, updates the device firmware and digital twin model, generates a new firmware fingerprint and model version identifier, and records them; The edge controller recalculates the association strength in the multidimensional association matrix based on the updated digital twin model to generate a matrix that reflects the device relationship.

[0018] Furthermore, the edge controller packages the new firmware fingerprint, the new model version identifier, and the refreshed multi-dimensional association matrix into a state change package and broadcasts it to the cloud microservice through a secure communication channel; The cloud microservice receives the status change package, updates the firmware fingerprint and model version identifier, adjusts the protection algorithm parameters and life prediction model parameters according to the refreshed multi-dimensional correlation matrix, recalculates the protection constant and life curve, and sends them to the edge controller.

[0019] (3) Beneficial effects The present invention provides a modular two-layer prefabricated power transformation system based on digital twins, which has the following beneficial effects: The semantic handshake monitors the first report frame of the box through the edge controller and calls the IEC61850 logical node template to complete the ternary registration of equipment, measurement points and functions, generate the first version of the semantic list and mark the hierarchical source, effectively eliminating the naming conflict of cross-layer data flow. The time base solidification inserts a high-stable timestamp based on the first version of the semantic list, and completes the mapping of Modebas to the message queue telemetry protocol, outputting a consistent time series stream across layers, eliminating time base drift, and ensuring the data semantic consistency and time accuracy of the system under a complex multi-layer architecture.

[0020] After writing consistent cross-layer time series streams into a semantic buffer, association modeling constructs a multidimensional association matrix, mapping the upper-layer primary power quantities and lower-layer secondary operating conditions into a dynamic graph in real time. This dynamic topology support provides an efficient means for real-time monitoring of system status and anomaly analysis.

[0021] Drift detection generates residual vectors and calculates the residual spectral entropy index and delay jitter index. Using a random forest model, it outputs an alignment confidence coefficient, enabling precise detection and quantification of drift between the digital twin model and the actual device state. After receiving the driving signal of the alignment confidence coefficient, the version patch submits the current firmware fingerprint and twin model version to the GitOps pipeline. After approval, a differential patch is generated and pushed back to the edge, achieving synchronous updates of the device firmware and digital twin model, maintaining virtual-to-physical consistency, ensuring continuous alignment of the digital twin with the actual device state, and improving the system's adaptability.

[0022] Version patches, efficiently managed through the GitOps pipeline, differentially update firmware fingerprints and twin model versions and push them back to edge devices, ensuring the synchronized evolution of assets and models. The integration of version patches with drift detection forms a closed-loop mechanism from drift detection to calibration, providing technical support for the continuous optimization of the modular, two-tier prefabricated substation system.

[0023] After the version patch is implemented, the policy refresh refreshes the multidimensional association matrix, broadcasts the state change to the cloud microservice, and recalculates the protection settings and life curves. This ensures the continuous synchronization of field control and twin analysis, improving safety margins and economic efficiency. The coordination of policy refresh, association modeling, and version patches forms a complete chain from data mapping to policy optimization, ensuring the efficiency and reliability of the system during dynamic operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 This is a structural schematic diagram of the modular double-layer prefabricated substation system based on digital twins in the present invention. DETAILED DESCRIPTION

[0025] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0026] See also Figure 1 The present invention provides a modular double-layer prefabricated power transformation system based on digital twin, comprising: Step 1: In the modular two-layer prefabricated substation system, the edge controller listens to the first message frame of the newly connected cabinet and receives the data packet, verifies the integrity of the data packet and records the reception time. It then uses the IEC61850 logical node template to parse the first message frame, extract device information, measurement point information, and function information, and generate a structured data set containing device identifiers, measurement points, and functions. It then performs device registration, measurement point registration, and function registration, registering the device identifier, measurement point identifier, measurement type, hierarchical source, and function description into the database to generate a relationship table. Generate the first version of the semantic list based on the relationship table, record the device identifier, measurement point identifier, measurement type, hierarchical source and functional description to ensure data semantic consistency, store the first version of the semantic list in the version control database, mark the timestamp and version number, update the version record when the device configuration changes, and retain historical data.

[0027] The step 1 includes the following: Step 101: Device access and first frame monitoring In a modular, double-layer prefabricated substation system, after a new cabinet is hoisted into place and its physical interface is connected to the edge controller, the edge controller activates a preset monitoring mechanism to capture the first message frame sent by the cabinet.

[0028] The first message frame is an initialization data packet proactively sent by a box when it is first connected to the system. It contains the device's unique identifier, type, manufacturer information, and a description of the data point. The edge controller continuously monitors data traffic via a preconfigured communication interface, such as Ethernet or a serial bus, and immediately receives and stores the data packet upon detecting the first message frame. During the reception process, the edge controller verifies the integrity of the data packet to ensure there are no transmission errors and records the reception time to ensure real-time data.

[0029] During use, an automated monitoring mechanism promptly captures the initialization data of newly connected enclosures, enabling automatic device identification and access management. The edge controller continuously scans the status of the communication interface. When it identifies a data packet that meets the first-frame characteristics, it triggers the receiving program, stores the data packet in a temporary buffer, and performs an integrity check. First-frame monitoring enables automatic discovery of new devices, eliminating manual intervention and ensuring a rapid system response to modular enclosure access.

[0030] Step 102: First message frame analysis based on IEC61850 standard After receiving the first message frame, the edge controller calls the logical node template based on the IEC61850 standard to perform structured analysis on the data packet content.

[0031] The IEC61850 standard is a universal communication specification for substation automation, defining the logical nodes and data object models of devices. The parsing process extracts three types of information from the first message frame: device information, including the device's unique identifier, model, and manufacturer name; measurement point information, covering supported measurement points such as voltage, current, and temperature, with each measurement point accompanied by a corresponding data type and unit; and functional information, describing the types of functions supported by the device, such as protection, control, or monitoring. After parsing, the edge controller stores the extracted information according to a predefined structure, generating a structured dataset containing the device identifier, measurement point, and function.

[0032] When in use, the edge controller reads the content of the first message frame field by field based on the data model of the IEC61850 standard, maps each field to the corresponding logical node attribute, and generates a structured list containing device information, measurement point information and function information. This can ensure the interoperability of device data from different manufacturers and ensure the universality and consistency of the analysis results. Through a standardized analysis process, the system can accurately extract key device information and provide a reliable basis for subsequent data management and application.

[0033] Step 103: The edge controller executes a ternary registration process based on the first frame analysis result and registers the extracted information into the system database.

[0034] Ternary registration consists of three sub-processes: first, device registration, which enters the device's unique identifier, model, and manufacturer information into the edge-side device management database to form a device list within the system; second, measurement point registration, which assigns a unique measurement point identifier to each measurement point and records its physical quantity type, such as voltage or current, as well as the data source level, which is divided into upper-level primary devices or lower-level secondary devices; and finally, function registration, which associates the functions supported by the device with the corresponding device identifier and measurement point identifier, and records the applicable scope of the function, such as the triggering conditions of the protection function. After registration is complete, the edge controller generates a relationship table containing the device identifier, measurement point identifier, and function association for subsequent data reference.

[0035] During use, a structured registration process establishes relationships between devices, measurement points, and functions, enabling systematic data management. The edge controller assigns a unique identifier to each device, measurement point, and function, records the attribute information and relationships in a database, and generates a relationship table containing all registered items. This generated relationship table supports rapid query and access of device data, providing efficient data organization for system operation.

[0036] Step 103: Generate semantic list The edge controller generates the first version of the semantic list based on the relationship table completed by the ternary registration, which is used to record the semantic description of the data.

[0037] The semantic list is organized in a table format, with each record containing the following fields: a device identifier, which uniquely identifies the device; a measurement point identifier, which uniquely identifies the measurement point; a measurement type, which describes the measured physical quantity, such as voltage or current; a hierarchical source, which indicates whether the data originates from the primary or secondary layer; and a functional description, which describes the actual function, such as overcurrent protection or condition monitoring. During the generation process, the edge controller uses standardized naming conventions to ensure that there are no conflicts in the naming of different devices and measurement points, thereby ensuring semantic consistency across layers of data. Once completed, the first version of the semantic list is stored as the system's data description baseline.

[0038] When in use, a standardized semantic description is generated based on the registration information to eliminate the data semantic differences between different devices; the edge controller integrates the registration information in the relationship table into a table, recording the device identifier, measurement point identifier, measurement type, hierarchical source and functional description one by one to ensure the integrity and consistency of each record; the generated semantic list provides a standardized reference for system data exchange and analysis, improving the comprehensibility and compatibility of the data.

[0039] After the first version of the semantic manifest is generated, the edge controller stores it in the version control database and marks it as the initial version. To implement version management, each semantic manifest record is accompanied by a timestamp and version number. The identification format is the semantic manifest version number plus the timestamp, for example, "semantic manifest_v1_timestamp_v1." When the device firmware is updated or the configuration changes, the edge controller re-executes the parsing and registration process based on the new first message frame, generates an updated semantic manifest, records it as the new version, and retains a record of historical versions. The version control database maintains the change history of all versions, ensuring that the system can track the evolution of the semantic manifest.

[0040] During use, the semantic manifest's change history is managed through a version control mechanism, ensuring data traceability and consistency. The edge controller assigns a timestamp and version number to each semantic manifest, stores it in the version control database, and updates the version record with each change, preserving historical data. This enhances system maintainability and ensures the accuracy and reliability of the semantic manifest over the long term.

[0041] In the modular double-layer prefabricated substation system, the specific processing technology logic of the semantic handshake realizes the automatic identification, data parsing and unified semantic management of the newly connected cabinet through five steps: device access and first message frame monitoring, first message frame parsing based on the IEC61850 standard, ternary registration, semantic list generation and version control. This solves the problem of inconsistent semantic data of cabinets across layers after the installation is completed, ensures the standardization and traceability of system data, and provides efficient and reliable technical support for the operation of the modular substation system.

[0042] Step 2: The edge gateway extracts the measurement point identifier, measurement type, and layer source from the first version of the semantic list generated in step 1, and ensures data semantic consistency based on the first version of the semantic list. The edge gateway inserts a highly stable timestamp for each received data packet to record the exact moment the data packet was received. The edge gateway converts data packets from the Modebas protocol to the message queue telemetry transmission protocol, generates topics and payloads, and achieves data standardization and efficient transmission; the edge gateway sorts data packets according to highly stable timestamps, generates consistent time series streams across layers, and pushes them to the cloud-based digital twin platform in real time, solving the drift problem of cross-layer data flows in the time base, improving the real-time performance and accuracy of the system, and providing a reliable data foundation for subsequent steps.

[0043] The second step includes the following: Step 201: Semantic list reference The edge gateway first extracts key information from the first version of the semantic list generated in step 1.

[0044] The first version of the semantic manifest contains measurement point identifiers, measurement types, and layer sources. This information is used to identify and process heterogeneous data from upper-layer primary devices and lower-layer secondary and auxiliary devices. The edge gateway queries the first version of the semantic manifest to identify the measurement point identifiers in the data packet and associate them with the corresponding measurement type and layer source. The edge gateway loads the first version of the semantic manifest, extracts the measurement point identifiers, measurement types, and layer sources, and stores them in a local data structure for quick query.

[0045] The first version of the semantic list provides standardized data descriptions to ensure that the edge gateway accurately understands the semantics of cross-layer data. As a unified data reference, the first version of the semantic list can eliminate the semantic differences between device data at different levels and avoid data misunderstandings or processing errors. By providing a clear semantic context, it ensures that the edge gateway can correctly parse and utilize data in subsequent processing, thereby enhancing the compatibility and stability of the system.

[0046] Step 202: Insert high-stability timestamp The edge gateway inserts a highly stable timestamp into each received data packet to record the precise moment of receipt of the data packet.

[0047] The highly stable timestamp comes from the edge gateway's built-in clock source, which is synchronized with an external high-precision clock to maintain microsecond-level time accuracy.

[0048] The processing process includes: the edge gateway parses the data packet and extracts the measurement point identifier; queries the first version of the semantic list based on the measurement point identifier to obtain the corresponding measurement type and layer source; then adds a high-stability timestamp as an additional field to the data packet, and the timestamp format is seconds plus microsecond accuracy.

[0049] After receiving a data packet, the edge gateway parses its measurement point identifier, queries the first-edition semantic list, generates a highly stable timestamp, and updates the packet structure. By inserting the highly stable timestamp, the packet is accurately marked and aligned in the time dimension. The highly stable timestamp resolves inconsistencies in cross-layer data flows caused by time base drift, ensuring the time traceability of all data.

[0050] Step 203: Protocol mapping The edge gateway converts data packets from the Modebus protocol to the Message Queue Telemetry Transport Protocol to achieve standardized and efficient data transmission. The conversion process includes mapping register addresses in the Modebus protocol to Message Queue Telemetry Transport Protocol topics, which are composed of a device identifier and a measurement point identifier; and converting data values ​​in the Modebus protocol into a JSON-formatted payload containing the measurement value and a highly stable timestamp.

[0051] The edge gateway parses the Modebas data packet, generates the corresponding topic and payload, encapsulates it into a message queue telemetry transmission message, and publishes it to the preconfigured message queue; through protocol mapping, heterogeneous data is converted into a unified format to facilitate cross-system transmission and processing. The message queue telemetry transmission protocol is lightweight and highly real-time, which can meet the data transmission needs in IoT scenarios.

[0052] Step 204: Time series stream output The edge gateway sorts data packets from all measurement points based on highly stable timestamps, generating a consistent time series stream across multiple layers. This time series stream, indexed by highly stable timestamps, contains measurement data from both upper-layer primary devices and lower-layer secondary and auxiliary devices, ensuring data alignment in the temporal dimension. The edge gateway utilizes streaming technology to push the time series stream to the cloud-based digital twin platform in real time. After receiving data packets, the edge gateway sorts them by highly stable timestamps, constructs a time series stream, and continuously outputs it to the cloud. This time series stream output ensures temporal continuity and cross-layer consistency of the data.

[0053] When used, time series streams provide a complete data view organized in chronological order, facilitating state analysis and modeling on the cloud platform. The generated time series data streams improve data quality and support the cloud digital twin platform to accurately map and monitor device status in real time.

[0054] Step 3. The edge gateway writes the cross-layer consistent time series stream into the semantic buffer in a highly stable timestamp order, extracts the time series data from the semantic buffer, and distinguishes the primary power measurement data generated by the upper-layer primary equipment from the secondary operation data generated by the lower-layer secondary and auxiliary equipment based on the hierarchical source in the first version of the semantic list. The Pearson correlation coefficient is calculated for each association pair of primary power measurement data and secondary operation data to generate a multidimensional association matrix. The association pairs in the multidimensional association matrix and the association strength represented by the Pearson correlation coefficient are mapped into a dynamic graph. The dynamic graph is stored using graph database technology and updated in real time.

[0055] The step three includes the following: Step 301: Write the time series stream into the semantic buffer The edge gateway writes the cross-layer consistent time series stream into the semantic buffer in the order of highly stable timestamps.

[0056] The semantic buffer is a pre-configured high-speed read-write storage area specifically used to temporarily store time-series stream data. The writing process requires that data packets be arranged in ascending order based on timestamps to ensure that the data maintains temporal continuity and order when stored, so that subsequent steps can extract and process the data as needed.

[0057] By writing the time series stream data into the semantic buffer in an orderly manner, the integrity of the time order of the data is ensured during storage and subsequent processing. After receiving the time series stream, the edge gateway first reads the high-stability timestamp of each data packet, then sorts the data packets according to the size of the timestamp, and finally stores the sorted data in the predetermined position of the semantic buffer in sequence.

[0058] As a transit storage area for data processing, the semantic buffer can provide efficient data access and extraction capabilities. At the same time, it ensures the temporal consistency of data through timestamp sorting, so that subsequent association matrix construction and dynamic graph mapping can be carried out based on accurate time sequence, improving the reliability and accuracy of the entire modeling process.

[0059] Step 302: Construct a multi-dimensional correlation matrix by level After extracting the time series stream data from the semantic buffer, the data is divided into upper-level primary device data and lower-level secondary and auxiliary device data according to the hierarchical sources defined in the first version of the semantic manifest; For each upper-level primary equipment measurement point and lower-level secondary and auxiliary equipment measurement point, a correlation pair is established and the correlation strength between the two is calculated. This correlation strength is calculated using the Pearson correlation coefficient. The specific process is to extract the measurement value sequence for each correlation pair, calculate the covariance of the two sequences, and then calculate the standard deviation of each sequence. The covariance is then divided by the product of the two standard deviations to obtain the Pearson correlation coefficient, which is used as a quantitative value of the correlation strength. All correlation pairs and their correlation strengths form a multidimensional correlation matrix.

[0060] By quantifying the strength of the correlation between upper-layer primary equipment measurement points and lower-layer secondary and auxiliary equipment measurement points, we reveal the inherent connections between cross-layer data and provide a structured data foundation for subsequent dynamic graph mapping. Based on the first-edition semantic list, we distinguish between upper-layer primary equipment and lower-layer secondary and auxiliary equipment measurement points. We extract the corresponding measurement value sequences from the semantic buffer. For each association pair, we calculate the mean of the measurement value sequence. Based on the mean, we calculate the covariance and standard deviation. Finally, we divide the covariance by the product of the two standard deviations to obtain the Pearson correlation coefficient, which is then entered into the corresponding position in the multidimensional correlation matrix.

[0061] Step 303: Real-time mapping to dynamic graph Convert association pairs and association strengths in a multidimensional association matrix into a structured representation of a dynamic graph.

[0062] Specifically, upper-level primary equipment measurement points and lower-level secondary and auxiliary equipment measurement points are defined as nodes in a dynamic graph. Each association pair is defined as an edge connecting two nodes, and the edge weight is determined by the corresponding association strength (i.e., the Pearson correlation coefficient). The dynamic graph is stored using graph database technology, which supports real-time updates and efficient queries of nodes, edges, and their weights. The conversion process involves reading the two measurement points and their association strengths for each association pair from a multidimensional association matrix, mapping the measurement points to nodes, and the association strengths to edge weights. These are stored in the graph database, and the node and edge states are adjusted in real time based on updates to the time series stream.

[0063] Graph database technology is used to transform the associations in a multidimensional association matrix into an intuitive topological structure, facilitating the visualization and analysis of complex connections between cross-layer data. Each association pair is extracted from the multidimensional association matrix, and the upper-layer primary equipment measurement points and lower-layer secondary and auxiliary equipment measurement points within the association pair are identified. These measurement points are then recorded as nodes in the graph database. The association pair is then defined as an edge connecting the two nodes, and the corresponding Pearson correlation coefficient is assigned to the edge as the edge weight. Finally, the data structure of these nodes and edges is stored in the graph database, completing the construction of the dynamic graph.

[0064] Writing time series streams into the semantic buffer ensures temporal consistency across layers of data. A hierarchical multidimensional association matrix is ​​constructed to extract data from the semantic buffer, quantifying the strength of associations between upper-layer primary equipment and lower-layer secondary and auxiliary equipment measurement points, forming a structured association representation. Finally, real-time mapping transforms the quantified results of the multidimensional association matrix into a graph structure, enabling dynamic visualization and real-time management of associations.

[0065] Step 4: Extract the actual correlation strength between the upper-level primary equipment measurement points and the lower-level secondary and auxiliary equipment measurement points from the dynamic map, obtain the predicted correlation strength from the digital twin platform, and calculate the difference between the two to generate a residual vector; Fast Fourier transform is applied to the residual vector to generate a frequency domain representation. After calculating the normalized power spectral density, the Shannon entropy formula is used to calculate the residual spectral entropy index. High-stability timestamps of the upper-layer primary equipment measurement points and the lower-layer secondary and auxiliary equipment measurement points are extracted from the time series stream. The cross-layer timestamp difference is calculated, and the improved Allan variance is applied to calculate the delay jitter index. The residual spectral entropy index and the delay jitter index are input into the random forest model, and the track reliability coefficient is output. When the track reliability coefficient falls below the preset threshold, the patch signal is triggered.

[0066] The step 4 includes the following contents: Step 401: Generate residual vector The actual correlation strength between the upper-level primary equipment measurement points and the lower-level secondary and auxiliary equipment measurement points is extracted from the dynamic map generated in step 3. The actual correlation strength is calculated using the Pearson correlation coefficient. The calculation process of the Pearson correlation coefficient is: The covariance of the time series data for two measurement points is calculated and divided by the product of their standard deviations to obtain a value between -1 and 1. Simultaneously, the predicted correlation strength for the same pair of measurement points is obtained from the cloud-based digital twin platform. This predicted correlation strength is generated by the digital twin model based on historical data and physical rules. For each pair of measurement points, the difference between the actual correlation strength and the predicted correlation strength is calculated and defined as the residual. The residuals of all measurement point pairs are arranged in order to form a residual vector.

[0067] By calculating the difference between the actual and predicted correlation strengths, the degree of deviation in the relationship between the digital twin model and the actual device state is quantified, providing a data basis for subsequent drift identification. For each pair of measurement points, the actual and predicted correlation strengths are obtained, and the difference between the two is calculated as the residual. The residuals of all measurement point pairs are then sequentially combined into a residual vector. The residual vector can intuitively reflect the deviation in the relationship between the digital twin model and the actual device state. The generation of the residual vector provides a quantitative description of the degree of drift, facilitating the system's automated detection and processing.

[0068] Step 402: Calculate the residual spectral entropy index Perform fast Fourier transform on the residual vector to convert the residual vector from the time domain to the frequency domain to obtain the frequency domain representation.

[0069] The Fast Fourier Transform (FFT) process decomposes the residual vector into a series of sine and cosine waves, calculates the amplitude and phase of each frequency component, and generates a frequency-domain representation. Based on this frequency-domain representation, the power spectral density (PSD) is calculated by squaring the amplitude of each frequency component to obtain the power value.

[0070] The power spectral density is normalized by dividing each power value by the sum of the power values ​​to obtain the normalized power spectral density components. The entropy of the normalized power spectral density components is calculated using the Shannon entropy formula and defined as the residual spectral entropy index. The Shannon entropy is calculated by taking the base-2 logarithm of each component in the normalized power spectral density, multiplying it by itself, and then summing all the products and taking the negative value to obtain the residual spectral entropy index.

[0071] The residual spectral entropy index is used to evaluate the complexity of the residual vector in the frequency domain, reflecting the dynamic characteristics of the drift between the digital twin model and the actual equipment state. First, a fast Fourier transform is applied to the residual vector to generate a frequency domain representation. Then, the power spectral density of the frequency domain representation is calculated and normalized. Finally, the entropy value is calculated according to the Shannon entropy formula to obtain the residual spectral entropy index.

[0072] When used, the residual spectral entropy index is used as a dimensionless value, which is convenient for comparison and integration with other indicators, thereby improving the comprehensiveness and accuracy of drift discrimination.

[0073] Step 402: Calculate the delay jitter index Extract the highly stable timestamps of the upper-layer primary equipment measurement points and the lower-layer secondary and auxiliary equipment measurement points from the time series stream generated in step 2. The highly stable timestamps are time stamps calibrated using time synchronization technology. Calculate the timestamp differences between the upper-layer primary equipment measurement points and the lower-layer secondary and auxiliary equipment measurement points to generate a cross-layer timestamp difference sequence. Apply the improved Allan variance calculation method to the cross-layer timestamp difference sequence to obtain the delay jitter index. The improved Allan variance calculation process is as follows: select a fixed average time window length, slide the window across the cross-layer timestamp difference sequence, and calculate the integral of the difference value within each window. Then, calculate the difference between the integral values ​​of adjacent windows, sum and average the squares of these differences to obtain the delay jitter index.

[0074] The delay jitter index is used to evaluate the stability of cross-layer data flows on a time basis, detect delay fluctuations in time series flows, extract highly stable timestamps from the upper-layer primary equipment measurement points and the lower-layer secondary and auxiliary equipment measurement points, and calculate the timestamp difference between the two to generate a cross-layer timestamp differential sequence. The improved Allan variance formula is then applied to calculate the variance of the differential sequence in different time windows to obtain the delay jitter index.

[0075] When used, the delay jitter index, as a dimensionless value, quantifies the time base drift of the time series stream and provides an evaluation basis in the time dimension for drift discrimination.

[0076] Step 403: Calculate the orbit reliability coefficient The residual spectral entropy index and delay jitter index are used as input features and fed into the pre-trained random forest model.

[0077] The random forest model consists of multiple decision trees, each of which makes independent predictions based on input features. The model outputs a dimensionless value between 0 and 1 through majority voting or averaging of the prediction results of all decision trees. This value is defined as the alignment credibility coefficient. The alignment credibility coefficient indicates the degree of alignment between the digital twin model and the actual equipment status. A larger value indicates a higher degree of alignment.

[0078] A random forest model is used to comprehensively evaluate the residual spectral entropy index and the delay jitter index to generate a track reliability coefficient, enabling automated determination of the degree of drift. The residual spectral entropy index and the delay jitter index are input into the random forest model, which then integrates the prediction results of multiple decision trees to calculate and output the track reliability coefficient.

[0079] When used, the calculation of the track credibility coefficient provides a comprehensive quantitative indicator of the degree of drift, which facilitates the system to automatically trigger subsequent calibration operations based on the coefficient, ensuring the continuous consistency of the digital twin model and the actual equipment status.

[0080] Step 404: Patch signal triggering A threshold for the orbit reliability coefficient is set in advance, which is determined based on system requirements and historical data analysis.

[0081] When the track reliability coefficient output by the random forest model is lower than the preset threshold, a patch signal is generated and sent to step five to notify the execution of a difference patch operation, which is intended to calibrate the deviation between the digital twin model and the actual equipment status.

[0082] The track reliability coefficient is determined by a preset threshold, automatically triggering the patch process to ensure timely calibration of the digital twin model with the actual device status. The track reliability coefficient output by the random forest model is compared with the preset threshold. If the track reliability coefficient is less than the threshold, a patch signal is generated and sent to step five.

[0083] When used, the triggering mechanism of the patch signal ensures that the digital twin model is continuously aligned with the actual device status, improving the system's adaptability and long-term reliability.

[0084] Step 5. The edge controller continuously monitors the patch signal generated in step 4. When the patch signal is 1, the edge controller starts the version patch process, collects the device's firmware fingerprint and the model version identifier obtained from the cloud digital twin platform, packages the firmware fingerprint and model version identifier into a version information package and submits it to the GitOps pipeline. After receiving the version information package, the GitOps pipeline compares the firmware fingerprint with the latest firmware fingerprint in the Git repository and the model version identifier with the latest model version identifier, generates a firmware difference patch and a model difference patch, and the GitOps pipeline submits the firmware difference patch and the model difference patch for approval. After approval, they are integrated into the final difference patch. The GitOps pipeline pushes the final difference patch back to the edge controller. The edge controller applies the firmware difference patch to update the device firmware and the model difference patch to update the digital twin model, records the new firmware fingerprint and model version identifier, and notifies the cloud digital twin platform to synchronize the updates.

[0085] The step five includes the following: Step 501: Trigger signal reception The edge controller continuously monitors the patch signal generated in step 4. The patch signal is a binary value, where 1 indicates that the version patch process needs to be initiated, and 0 indicates that it does not. The edge controller periodically reads the patch signal value to make decisions: if the patch signal is 1, the subsequent version patch operation is immediately executed; if the patch signal is 0, the edge controller continues monitoring without executing any operation.

[0086] The edge controller automatically triggers the version patch process by monitoring the patch signal in real time, ensuring timely calibration when the digital twin model drifts from the actual device state. The edge controller reads the patch signal value at regular intervals and decides whether to initiate the version patch process based on the reading: if the value is 1, the process is initiated; if the value is 0, monitoring continues.

[0087] When used, the patch signal, as the output of the drift determination in step 4, accurately reflects the synchronization between the digital twin model and the actual device state. By using it as a trigger, the version patch process is initiated only when necessary, avoiding unnecessary resource usage. This automated triggering mechanism improves the system's responsiveness and adaptability, ensuring timely alignment between the digital twin model and the actual device state.

[0088] Step 502: Firmware fingerprint and model version submission The edge controller first collects the device's current firmware fingerprint. The firmware fingerprint is a fixed-length string generated by hashing the firmware file content and is used to uniquely identify the device's firmware status.

[0089] The hashing process involves inputting the complete firmware file data into a hash function, which outputs a unique string of fixed length as the firmware fingerprint. Simultaneously, the edge controller obtains the current digital twin model's version identifier from the cloud-based digital twin platform. The version identifier is a unique number for the digital twin model. The edge controller packages the firmware fingerprint and model version identifier into a version information package, which consists of a key-value pair containing the firmware fingerprint and model version identifier. This version information package is submitted to the GitOps pipeline via a secure communication channel.

[0090] The edge controller submits firmware fingerprints and model version identifiers to provide the GitOps pipeline with accurate information required for version comparison, ensuring the correctness of subsequent patch generation.

[0091] During use, the edge controller performs a hash operation on the firmware file to generate a firmware fingerprint, reads the model version identifier from the cloud-based digital twin platform, combines the two into a version information package, and sends it to the GitOps pipeline via an encrypted communication protocol. Submitting the version information package provides complete version data to the GitOps pipeline, ensuring that patch generation is based on the latest and most accurate information, thereby improving the reliability and accuracy of version synchronization.

[0092] Step 503: GitOps pipeline processing After receiving the version information package, the GitOps pipeline performs a version comparison operation.

[0093] The specific process is as follows: The GitOps pipeline extracts the firmware fingerprint and model version identifier from the version information package, and then compares them with the latest firmware fingerprint and latest model version identifier stored in the Git repository, respectively. The comparison method is string matching: the current firmware fingerprint is compared character by character with the latest firmware fingerprint. If an inconsistency is found, a firmware difference patch is generated; the current model version identifier is compared character by character with the latest model version identifier. If an inconsistency is found, a model difference patch is generated. The firmware difference patch contains the incremental update content from the current firmware version to the latest version, and the model difference patch contains the incremental update content from the current digital twin model version to the latest version.

[0094] The GitOps pipeline identifies differences between the firmware and the digital twin model through version comparison and generates corresponding patches to ensure that the patch content accurately reflects the version change. The GitOps pipeline extracts the firmware fingerprint and model version identifier from the version information package and compares them with the latest version information in the Git repository. If there are any differences, the version control tool is used to generate the corresponding firmware and model difference patches.

[0095] When used, the processing mechanism of the GitOps pipeline improves the efficiency and accuracy of version management, ensures that the patch content is completely consistent with the actual version changes, and avoids potential errors caused by human operations.

[0096] Step 504: Approval and patch generation The GitOps pipeline submits the generated firmware and model difference patches to the approval process. The approval process verifies the security and compliance of the patches based on preconfigured automated rules or manual review. Automated rules include steps such as checking the completeness of the patch content and verifying version compatibility. Manual review involves operations personnel reviewing the patch content to ensure it meets system requirements. After approval, the GitOps pipeline combines the firmware and model difference patches into a final difference patch, which contains the complete updated data for the firmware and digital twin model.

[0097] The approval process verifies the security and compliance of patches, preventing unauthorized or insecure version changes from entering the system. The GitOps pipeline submits firmware and model differential patches to the approval system. The approval system executes predefined automated rules or accepts manual review results. After approval, the GitOps pipeline integrates the two into the final differential patch.

[0098] When used, the approval and patch generation mechanism enhances the controllability of version updates, ensuring that patch content is strictly verified, thereby safeguarding the stability and security of the system.

[0099] Step 505: Patch rollback and application The GitOps pipeline pushes the final differential patch back to the edge controller via a secure communication channel. After receiving the final differential patch, the edge controller first applies the firmware differential patch, upgrading the device firmware to the latest version by updating the firmware file content. It then applies the model differential patch, upgrading the digital twin model to the latest version by updating the model data. After the update is complete, the edge controller re-hashes the updated firmware file to generate a new firmware fingerprint, records the updated model version identifier, and stores both locally. The edge controller notifies the cloud-based digital twin platform via the communication interface, synchronously updating the new firmware fingerprint and model version identifier to ensure consistency between cloud-based data and edge devices.

[0100] Through patch rollback and application, device firmware and digital twin models are updated synchronously, ensuring consistency between the virtual model and the actual device state. The edge controller receives the final differential patch and sequentially performs firmware and model updates. It generates a new firmware fingerprint for the updated firmware, records the new model version identifier, and notifies the cloud-based digital twin platform of the results for synchronization. This patch rollback and application mechanism enables automatic synchronous updates of devices and models, improving the system's adaptability and long-term reliability.

[0101] Step 6: The edge controller receives the differential patch pushed back by the GitOps pipeline, updates the device firmware and digital twin model, generates a new firmware fingerprint and model version identifier, and records them; The edge controller recalculates the correlation strength in the multidimensional correlation matrix based on the updated digital twin model to generate a matrix reflecting the device relationship; the edge controller packages the new firmware fingerprint, the new model version identifier and the refreshed multidimensional correlation matrix into a state change package, and broadcasts it to the cloud microservice through a secure communication channel; the cloud microservice receives the state change package, updates the firmware fingerprint and model version identifier, adjusts the protection algorithm parameters and life prediction model parameters according to the refreshed multidimensional correlation matrix, recalculates the protection constant and life curve, and sends them to the edge controller.

[0102] The step six includes the following contents: Step 601: Patch landing processing After receiving the differential patch pushed back through the GitOps pipeline, the edge controller starts the patch application operation.

[0103] First, the edge controller decompresses the differential patch, separating it into two parts: the firmware differential patch and the model differential patch. Next, the edge controller applies the firmware differential patch to the device firmware, updating the firmware to its latest state by comparing the differences between the current firmware version and the patch content. Subsequently, the edge controller applies the model differential patch to the digital twin model, also updating the digital twin model to its latest state by comparing the differences between the current model version and the patch content. After the update is complete, the edge controller generates a new firmware fingerprint based on the updated firmware and a new model version identifier based on the updated model. Both are recorded in local storage for subsequent version management and verification.

[0104] The edge controller achieves synchronous updates of device firmware and digital twin models by decompressing the differential patch and applying the firmware differential patch and model differential patch respectively, while generating and recording new firmware fingerprints and model version identifiers.

[0105] A differential patch contains both firmware and model updates. By applying both patches separately, you can ensure that the device's operating status and the digital twin model's status remain consistent after the update. Generating and recording a new firmware fingerprint and model version identifier provides a basis for subsequent version tracking and consistency verification.

[0106] When used, it can achieve accurate updates of firmware and models, avoiding state deviations caused by incomplete updates. At the same time, through the recording of fingerprints and version identification, it can provide a traceable update history, improving the reliability and efficiency of maintenance.

[0107] Step 602: Update the correlation matrix The edge controller recalculates the multidimensional correlation matrix based on the updated digital twin model.

[0108] First, the edge controller extracts the correlation strength between the upper-level primary equipment measurement points and the lower-level secondary and auxiliary equipment measurement points from the updated digital twin model. This correlation strength is derived based on the physical relationships and data dependencies between the measurement points defined in the model. The specific value is determined by analyzing the degree of mutual influence between the measurement point data.

[0109] The edge controller then writes the extracted association strengths into the multidimensional association matrix, updating the values ​​of each element in the matrix. After the update is complete, the multidimensional association matrix reflects the latest topological relationships and data association status of the devices after the update, which is used for subsequent state change broadcasts and cloud-based analysis.

[0110] Based on the updated digital twin model, the edge controller extracts the strength of associations between primary device measurement points and secondary and auxiliary device measurement points. This information is then used to update the multidimensional association matrix, generating a matrix reflecting the latest device relationships. This updated multidimensional association matrix accurately describes the updated relationships between devices, providing reliable data support for state change broadcasts and cloud-based computing, enhancing the accuracy of the digital twin model and the credibility of the analysis results.

[0111] Step 603: Broadcast of status change The edge controller integrates the new firmware fingerprint, the new model version identifier, and the refreshed multi-dimensional association matrix into a state change package, which is used to deliver update information to the cloud microservices.

[0112] Specifically, the edge controller packages the new firmware fingerprint, the new model version identifier, and the multidimensional association matrix into a complete data structure according to a predefined format. Once packaged, the edge controller sends the state change packet to the cloud microservice via a secure communication channel, ensuring data integrity and security during transmission. Upon receiving the state change packet, the cloud microservice can update its own data and models based on the contents.

[0113] The edge controller packages the new firmware fingerprint, new model version identifier, and updated multi-dimensional association matrix into a state change packet and broadcasts it to the cloud microservice via a secure communication channel. This state change packet, which integrates the latest device and model status information and is broadcast to the cloud, ensures data synchronization between the cloud microservice and the edge device. Transmission over a secure communication channel ensures reliable and confidential data transmission. This enables efficient information sharing between the edge and cloud, ensuring that cloud microservices can access updated status data in a timely manner, supporting subsequent calculations and analysis, and improving overall system coordination and responsiveness.

[0114] Step 604: Recalculate protection setting value and life curve After receiving the state change package, the cloud microservice performs update and recalculation operations.

[0115] First, the cloud microservice replaces the current firmware fingerprint and model version identifier with the new firmware fingerprint and new model version identifier in the state change package, respectively, to complete version synchronization. Next, the cloud microservice recalculates the protection constants and life curves based on the refreshed multi-dimensional association matrix in the state change package. For the recalculation of the protection constants, the cloud microservice analyzes the correlation strength in the multi-dimensional association matrix and adjusts the parameters of the protection algorithm to adapt it to the updated device relationships. For the recalculation of the life curve, the cloud microservice also updates the parameters of the equipment life prediction model based on the correlation strength in the multi-dimensional association matrix to generate new life prediction results. After the recalculation is complete, the cloud microservice sends the new protection constants and life curves to the edge controller for on-site control and status monitoring.

[0116] The cloud microservice updates the firmware fingerprint and model version identifier based on the status change packet. It also adjusts the protection algorithm parameters and life prediction model parameters based on the updated multidimensional correlation matrix, recalculates the protection settings and life curve, and then sends them to the edge controller. Device updates can affect the accuracy of protection logic and life predictions. By recalculating the protection settings and life curves based on the multidimensional correlation matrix, the calculation results can be guaranteed to reflect the latest device status, improving control and prediction accuracy.

[0117] When in use, the recalculated protection settings and life curves are consistent with the updated equipment status, improving the reliability of the system protection function and the accuracy of life assessment, and providing precise technical support for on-site control and long-term operation and maintenance.

[0118] Device and model updates are achieved through patch delivery, the latest data relationships are generated through association matrix refreshes, edge-to-cloud synchronization is achieved through state change broadcasts, and control and prediction parameters are optimized through recalculation of protection settings and life curves. This effectively resolves the discrepancy between the digital twin model and actual device state caused by firmware updates and configuration changes, ensuring the continuous consistency of field control and twin analysis, and providing high-precision data support and technical assurance for remote monitoring, state assessment, and predictive maintenance of modular, two-tier prefabricated substation systems.

[0119] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0120] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0121] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is only for some logical functions. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0122] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0123] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A modular, two-tier prefabricated substation system based on digital twins, featuring: include, In the semantic handshake unit, the edge controller monitors the first report frame of each cabinet, calls the IEC61850 logical node template to complete the ternary registration of equipment, measurement points, and functions, generates the first version of the semantic list, and marks the hierarchical source; The time base curing unit inserts a highly stable timestamp into the edge gateway based on the semantic list, completes the mapping of the Modebus to the message queue telemetry protocol, and outputs a consistent time series stream across layers; Association modeling unit, after the time series stream is written into the semantic buffer, the system constructs a multi-dimensional association matrix hierarchically, mapping the primary power quantity and the secondary operating condition quantity into a dynamic graph in real time; The drift discrimination unit generates a residual vector and simultaneously calculates the residual spectral entropy index and the delay jitter index. After the random forest model outputs the orbit credibility coefficient, the patch signal is triggered at a low threshold. The version patch unit receives a trigger signal for the patch process, submits the current firmware fingerprint and the twin model version to the GitOps pipeline, generates a differential patch after approval, and pushes it back to the edge; The strategy refresh unit refreshes the association matrix as soon as the patch is implemented and broadcasts the status change to the cloud microservice. The engine recalculates the protection constant and life curve accordingly, so that on-site control and twin analysis keep pace with each other.

2. The modular double-layer prefabricated power transformation system based on digital twin according to claim 1, characterized in that: The edge controller listens to the first message frame of the newly connected box and receives the data packet, verifies the integrity of the data packet and records the reception time, calls the IEC61850 logical node template to parse the first message frame, extracts device information, measurement point information and function information, and generates a structured data set containing device identifiers, measurement points and functions.

3. The modular double-layer prefabricated power transformation system based on digital twin according to claim 2, characterized in that: Perform device registration, measurement point registration, and function registration, registering device identifiers, measurement point identifiers, measurement types, hierarchical sources, and function descriptions into the database and generating relationship tables; Generate the first version of the semantic list based on the relationship table, recording the device identifier, measurement point identifier, measurement type, hierarchical source and functional description.

4. The modular double-layer prefabricated power transformation system based on digital twin according to claim 3, characterized in that: The edge gateway extracts the measurement point identifier, measurement type, and level source from the first version of the semantic list, and ensures data semantic consistency according to the first version of the semantic list; The edge gateway inserts a highly stable timestamp into each received data packet to record the precise moment the data packet is received.

5. The modular double-layer prefabricated power transformation system based on digital twin according to claim 4, characterized in that: The edge gateway converts the data packets from the Modbus protocol to the Message Queuing Telemetry Transport Protocol, generating topics and payloads; The edge gateway sorts the data packets according to the highly stable timestamps, generates a consistent time series stream across layers, and pushes it to the cloud-based digital twin platform in real time.

6. The modular double-layer prefabricated power transformation system based on digital twin according to claim 5, characterized in that: The edge gateway writes the cross-layer consistent time series stream into the semantic buffer in a highly stable timestamp order, extracts the time series data from the semantic buffer, and distinguishes the primary power measurement data generated by the upper-layer primary equipment from the secondary operation data generated by the lower-layer secondary and auxiliary equipment based on the hierarchical source in the first version of the semantic list; The Pearson correlation coefficient is calculated for each correlation pair of primary power measurement data and secondary operation data to generate a multidimensional correlation matrix. The correlation pairs in the multidimensional correlation matrix and the correlation strength represented by the Pearson correlation coefficient are mapped into a dynamic graph. The dynamic graph is stored in a graph database and updated in real time.

7. The modular double-layer prefabricated power transformation system based on digital twin according to claim 6, characterized in that: The actual correlation strength between the upper-level primary equipment measurement points and the lower-level secondary and auxiliary equipment measurement points is extracted from the dynamic map, and the predicted correlation strength is obtained from the digital twin platform. The difference between the two is calculated to generate a residual vector. Fast Fourier transform is applied to the residual vector to generate frequency domain representation. After calculating the normalized power spectral density, the Shannon entropy formula is used to calculate the residual spectral entropy index.

8. The modular double-layer prefabricated power transformation system based on digital twin according to claim 7, characterized in that: Extract high-stability timestamps of upper-layer primary equipment measurement points and lower-layer secondary and auxiliary equipment measurement points from the time series stream, calculate cross-layer timestamp differences, and apply improved Allan variance to calculate the delay jitter index; The residual spectral entropy index and delay jitter index are input into the random forest model to output the orbital reliability coefficient. When the orbital reliability coefficient is lower than the preset threshold, the patch signal is triggered.

9. The modular double-layer prefabricated power transformation system based on digital twin according to claim 8, characterized in that: The edge controller continuously monitors the patch signal. When the patch signal is 1, the edge controller starts the version patch process, collects the device's firmware fingerprint and the model version identifier obtained from the cloud digital twin platform, packages the firmware fingerprint and model version identifier into a version information package, and submits it to the GitOps pipeline.

10. The modular double-layer prefabricated power transformation system based on digital twin according to claim 9, characterized in that: After receiving the version information package, the GitOps pipeline compares the firmware fingerprint with the latest firmware fingerprint in the Git repository and the model version identifier with the latest model version identifier to generate a firmware difference patch and a model difference patch; The GitOps pipeline submits the firmware difference patch and the model difference patch for approval, and integrates them into the final difference patch after approval.

11. The modular double-layer prefabricated power transformation system based on digital twin according to claim 10, characterized in that: The GitOps pipeline pushes the final differential patch back to the edge controller; the edge controller applies the firmware differential patch to update the device firmware and the model differential patch to update the digital twin model, records the new firmware fingerprint and model version identifier, and notifies the cloud-based digital twin platform to synchronize the update.

12. The modular double-layer prefabricated power transformation system based on digital twin according to claim 11, characterized in that: The edge controller receives the differential patch pushed back by the GitOps pipeline, updates the device firmware and digital twin model, generates a new firmware fingerprint and model version identifier, and records them; The edge controller recalculates the association strength in the multidimensional association matrix based on the updated digital twin model to generate a matrix that reflects the device relationship.

13. The modular double-layer prefabricated power transformation system based on digital twin according to claim 12, characterized in that: The edge controller packages the new firmware fingerprint, new model version identifier, and updated multi-dimensional association matrix into a state change package and broadcasts it to the cloud microservice through a secure communication channel. The cloud microservice receives the status change package, updates the firmware fingerprint and model version identifier, adjusts the protection algorithm parameters and life prediction model parameters according to the refreshed multi-dimensional correlation matrix, recalculates the protection constant and life curve, and sends them to the edge controller.

Images