Transformer substation digital twinborn management and control method, system and equipment based on common inductance calculation integration and medium

Abstract

The invention discloses a transformer substation digital twinborn management and control method, system, equipment and medium based on common inductance calculation integration, and relates to the technical field of digital twinborn management and control, and the method comprises the steps: collecting the on-site multi-modal sensing data of a transformer substation, and completing the data preprocessing at an edge side; performing time standardization processing and fusion analysis on the multi-modal sensing data, and constructing a unified equipment operation time sequence database; constructing a digital twinborn model based on the preprocessed multi-modal sensing data and an equipment operation time sequence database; the virtual engine is driven to be connected with the digital twin model, so that state mapping and bidirectional interaction are carried out between the digital twin model and the virtual engine; a simulation environment is constructed, a multi-working-condition operation state is simulated, potential faults are identified, and a dynamic decision support mechanism is formed. According to the invention, a full-process closed-loop management and control method for the transformer substation is constructed, a fusion channel among multi-source heterogeneous data is opened, and high-precision and high-real-time state visualization and interaction control are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of digital twin management and control, and in particular to a substation digital twin management and control method, system, device and medium based on integrated sensing and computing. BACKGROUND

[0002] With the continuous expansion of the scale of the power system and the continuous improvement of the intelligent level, the safety, stability and management efficiency of the operation of the substation as the core hub in the power grid are directly related to the reliable operation of the entire power grid. However, most of the current substations still mainly rely on traditional operation and maintenance methods such as manual inspection and regular maintenance, which not only has problems such as response lag and insufficient coverage, but also is difficult to find potential risks of equipment in time, which seriously affects the operation and maintenance quality and job safety. At the same time, the types of power equipment are various, and the protocols of manufacturers are different, which leads to large differences in sensing data structure and high integration difficulty, hindering the realization of unified monitoring and collaborative control.

[0003] In recent years, digital twin technology has developed rapidly in the industrial field, and as an important means to realize virtual-real fusion and real-time interaction, it provides a new path for building a high-precision, interactive and predictable virtual mirror for substations. However, existing digital twin applications are mostly focused on visual display, lack of deep integration of sensing, communication, computing, control and other core links, and have not yet formed an intelligent control mechanism driven by a closed loop. In addition, key capabilities such as edge computing, data fusion and simulation prediction have not yet formed effective collaboration at the system level, and cannot meet the comprehensive needs of the new generation of intelligent substations for "global sensing-dynamic analysis-active decision-making-self-adaptive execution". Therefore, there is an urgent need for a substation intelligent management and control method that integrates sensing and computing and digital twin technology to realize full-link closed-loop control from field data collection to system optimization and scheduling, and to improve the operation intelligence level and risk prevention and control capability of the substation. SUMMARY

[0004] In view of the above problems, the present application is proposed.

[0005] Therefore, the problem to be solved by the present application is that existing digital twin applications are mostly focused on visual display, lack of deep integration of sensing, communication, computing, control and other core links, and have not yet formed an intelligent control mechanism driven by a closed loop.

[0006] To solve the above technical problems, the application provides the following technical scheme: a substation digital twin management and control method based on integrated sensing and computing, which comprises the following steps: constructing a distributed sensing and edge computing network, collecting multi-modal sensing data of a substation site through deployment of distributed embedded terminal node devices, and completing data preprocessing on the edge side; performing time standardization processing and fusion analysis on the multi-modal sensing data to construct a unified equipment operation time series database; constructing a digital twin model based on the preprocessed multi-modal sensing data and the equipment operation time series database, and supporting three-dimensional rendering and real-time interaction in a virtual engine; connecting the virtual engine and the digital twin model to enable state mapping and bidirectional interaction between the digital twin model and the virtual engine; constructing a simulation environment through the digital twin model connected with the virtual engine, simulating multi-working-condition operation states and identifying potential faults to form a dynamic decision support mechanism; and when a potential fault is identified, forming a structured operation adjustment scheme and issuing the adjustment scheme to an operation and maintenance terminal.

[0007] As a preferred scheme of the substation digital twin management and control method based on integrated sensing and computing, the data preprocessing on the edge side comprises the following steps: uniformly processing multi-modal sensing data formats by using an adaptive data acquisition protocol; filtering and denoising and extracting features from the collected multi-modal sensing data based on a sliding window mechanism; and preliminarily identifying and marking abnormal states in the multi-modal sensing data in combination with an abnormality detection mechanism to mark abnormal feature values and early warning indicators.

[0008] The preferred technical scheme has the beneficial effect that preprocessing is performed on the edge terminal to reduce the amount of raw data transmission, reduce the bandwidth pressure of the backbone network, and realize communication optimization of “local filtering and key uploading”.

[0009] As a preferred scheme of the substation digital twin management and control method based on integrated sensing and computing, the construction of the unified equipment operation time series database comprises the following steps: identifying equipment data in the multi-modal sensing data, performing field structure analysis and preliminary classification on equipment data of different sources and different protocols; performing time standardization processing on the field structure analyzed equipment data to make the equipment data have a unified time reference and semantic format; and indexing and organizing the time standardized equipment data according to equipment type dimensions and time dimensions to construct an equipment operation time series database.

[0010] The preferred technical scheme has the beneficial effect that time standardization processing and format unification are performed to realize data fusion from different equipment, protocols and manufacturers, construct a time series database with consistent structure, and provide a basic support for subsequent big data analysis and prediction modeling.

[0011] As a preferred scheme of the substation digital twin management and control method based on the integration of common sense algorithm, the constructing a digital twin model comprises: generating a three-dimensional grid model through point cloud splicing and surface reconstruction on the spatial information contained in the preprocessed multi-modal sensing data; mapping the device image data to the three-dimensional grid model in combination with the device surface texture information to construct the appearance of the three-dimensional grid model; extracting device data from the device runtime sequence database and binding the device data to the corresponding components in the three-dimensional grid model in a structured manner to form a digital twin model containing geometric shape, visual feature and semantic attribute.

[0012] As a preferred scheme of the substation digital twin management and control method based on the integration of common sense algorithm, the state mapping and bidirectional interaction between the digital twin model and the virtual engine comprises: loading the constructed digital twin model in the virtual engine, establishing a mapping relationship between each component in the digital twin model and the running state data in the preprocessed multi-modal sensing data; driving the dynamic update and state presentation of the digital twin model in the virtual engine according to the running state data, and mapping and displaying the physical running changes in the virtual space in real time through visual means; setting an interaction trigger logic in the virtual engine, generating a control instruction when an operator operates the virtual model through a client, and returning the control instruction to an edge device through a communication link to form an instruction closed loop from virtual operation to physical feedback, and constructing a bidirectional interaction mechanism between the physical entity and the digital model.

[0013] The preferred technical scheme has the beneficial effects that: by mapping the real-time data of the physical system to the digital twin model and feeding back the state changes in a visual manner, the operator can intuitively master the device state, and the information transparency and operation efficiency are improved.

[0014] As a preferred scheme of the substation digital twin management and control method based on the integration of common sense algorithm, the simulating multi-working-condition running state and identifying potential faults comprises: based on the data input to the digital twin model, dynamically loading the sensing data and physical modeling parameters in different running scenes to simulate the working conditions; in the simulation process, the abnormal characteristic values are analyzed in real time according to the fault identification mechanism; by comparing the abnormal characteristic values with known abnormal patterns in a historical fault sample library, and combining the early warning indexes to determine whether there is a potential risk in the current running working condition, and outputting the fault type.

[0015] As a preferred scheme of the substation digital twin management and control method based on the integration of common sense algorithm, the forming a structured running adjustment scheme and issuing the adjustment scheme to an operation and maintenance terminal comprises: when it is identified that there is a potential fault, generating an optimization scheme based on machine learning; and pushing the optimization scheme to the operation and maintenance terminal device in the form of a structured task package.

[0016] To solve the above technical problems, the present application provides the following technical solutions: a substation digital twin management and control method based on integrated sensing and computing, comprising: a physical layer, a resource layer, a platform layer and an application layer; the physical layer constructs a distributed sensing and edge computing network, collects multi-modal sensing data on the spot of the substation through the deployment of distributed embedded terminal node devices, and completes data preprocessing on the edge side; the resource layer performs time standardization processing and fusion analysis on the multi-modal sensing data, and constructs a unified equipment operation time series database; a digital twin model is constructed based on the preprocessed multi-modal sensing data and the equipment operation time series database, and three-dimensional rendering and real-time interaction in a virtual engine are supported; the platform layer drives the virtual engine to be connected with the digital twin model, so that state mapping and bidirectional interaction are performed between the digital twin model and the virtual engine; a simulation environment is constructed through the digital twin model connected with the virtual engine, potential faults are identified in the simulation of multi-working condition operation states, and a dynamic decision support mechanism is formed; when potential faults are identified, the application layer forms a structured operation adjustment scheme, and the adjustment scheme is issued to an operation and maintenance terminal.

[0017] A computer device comprises a memory and a processor, the memory stores a computer program, and the processor implements the steps of the above-mentioned substation digital twin management and control method based on integrated sensing and computing when executing the computer program.

[0018] A computer readable storage medium stores a computer program, and the computer program implements the steps of the above-mentioned substation digital twin management and control method based on integrated sensing and computing when executed by a processor.

[0019] The present application has the following advantages: the present application fuses the integrated sensing and computing technology and the digital twin model, constructs a full-process closed-loop management and control method for substations, and has a complete system from distributed sensing, edge preprocessing, real-time state mapping, data fusion database construction, working condition simulation analysis to operation optimization issuance. The method not only breaks through the fusion channel between multi-source heterogeneous data, realizes high-precision and high-real-time state visualization and interactive control, but also significantly improves the equipment abnormal prediction and intelligent decision-making ability, promotes the transformation of substation operation and maintenance from "passive response" to "active prediction" and "self-optimization", and improves the safety of the system as a whole. BRIEF DESCRIPTION OF DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0021] Figure 1 A flow chart of a substation digital twin management and control method based on the integration of common sensing and calculation in embodiment 1.

[0022] Figure 2 A structure diagram of a substation digital twin management and control system based on the integration of common sensing and calculation in embodiment 3. DETAILED DESCRIPTION

[0023] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0024] In the following description, many specific details are set forth in order to provide a thorough understanding of the present application, but the present application can also be implemented in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the concept of the present application, therefore the present application is not limited to the specific embodiments disclosed below.

[0025] Embodiment 1, refer to Figure 1 , the first embodiment of the present application, the embodiment provides a substation digital twin management and control method based on the integration of common sensing and calculation, comprising:

[0026] S1: Construct a distributed sensing and edge computing network, collect multi-modal sensing data of the substation site by deploying distributed embedded terminal node devices, and complete data preprocessing on the edge side.

[0027] S2: Time standardization processing and fusion analysis are performed on the multi-modal sensing data to construct a unified equipment operation time series database.

[0028] S3: Construct a digital twin model based on the preprocessed multi-modal sensing data and the equipment operation time series database, and support three-dimensional rendering and real-time interaction in a virtual engine.

[0029] S4: Drive the virtual engine to connect with the digital twin model, and perform state mapping and bidirectional interaction between the digital twin model and the virtual engine.

[0030] S5: Construct a simulation environment through the digital twin model connected with the virtual engine, simulate multi-working condition operation state and identify potential faults, and form a dynamic decision support mechanism.

[0031] S6: When a potential fault is identified, a structured operation adjustment scheme is formed, and the adjustment scheme is issued to the operation and maintenance terminal.

[0032] It should be noted that at present, most substations still mainly rely on traditional operation and maintenance methods such as manual inspection and regular maintenance, not only exist response lag, insufficient coverage and other problems, but also difficult to find potential risks of equipment in time, which seriously affects the operation and maintenance quality and operation safety; in recent years, digital twin technology has developed rapidly in the industrial field, as an important means to realize virtual-real integration and real-time interaction; however, in the existing digital twin application, the key capabilities such as edge computing, data fusion and simulation prediction have not formed effective cooperation at the system level, which cannot meet the comprehensive needs of the new generation of intelligent substations.

[0033] Therefore, in order to solve the above problems, through the steps of S1-S6, first, by deploying multi-source sensing terminals and edge computing nodes, a data acquisition and preliminary screening mechanism with local intelligence is established, the system aggregates and standardizes the key data on the edge side, and then stores it into the device runtime database, which provides support for subsequent multi-scenario simulation and state evolution modeling; then, the collected results are used to build a digital twin model containing geometric structure, running attribute and behavior logic, and loaded into a virtual engine environment to support real-time mapping and human-computer interaction; on this basis, the platform can dynamically simulate multiple working conditions, evaluate system running trend, identify potential risks, and automatically generate operation optimization suggestions based on the analysis results, and push them to the operation and maintenance terminal in the form of instruction task package, to realize intelligent closed-loop regulation and control.

[0034] Embodiment 2, which is different from the first embodiment, is a substation digital twin management and control method based on integrated sensing algorithm, which further comprises, in step S1, deploying a multi-modal sensor array through an embedded terminal node device to realize synchronous measurement of temperature, current, voltage, partial discharge, vibration and other multi-physical quantities.

[0035] Further, the data preprocessing on the edge side includes the following steps A1-A3:

[0036] A1: Adopting an adaptive data acquisition protocol to unify the multi-modal sensing data format.

[0037] A2: Filtering and denoising and feature extraction are performed on the collected multi-modal sensing data based on a sliding window mechanism.

[0038] A3: Combining an abnormality detection mechanism to preliminarily identify and mark abnormal states in the multi-modal sensing data, and marking abnormal feature values and early warning indicators.

[0039] In the embodiment of the application, in step A3, the abnormality detection mechanism adopts an abnormality identification method based on static threshold, including the following steps A311-A313:

[0040] A311: Based on the experience of device operation and historical statistical data, set the warning threshold range for various key indicators (such as current, voltage, vibration intensity, temperature).

[0041] A312: Real-time comparison of extracted feature parameters, if exceeding the set threshold, it is determined as potential abnormality, triggering the marking action.

[0042] A313: Pack the data segment marked as abnormal and upload it to the platform, and attach the abnormal type identifier, realize the rapid reporting of high-risk information.

[0043] In an optional embodiment, the anomaly detection mechanism can also use an anomaly point detection method based on statistical distribution, including the following steps A321-A323:

[0044] A321: Based on the sliding time window, continuously calculate the local mean and standard deviation of the same type of parameters, and build a statistical distribution model;

[0045] A322: Apply the Z-score (standard deviation offset score) method to calculate the standardization offset of the current sample, and when the offset value exceeds the set Z threshold (such as ±2.5), mark it as abnormal;

[0046] A323: Generate an abnormal log for all marked samples, including timestamp, abnormal type and offset degree, for subsequent fusion analysis.

[0047] In another optional embodiment, the anomaly detection mechanism can also use an edge inference recognition method based on a lightweight neural network model, including the following steps A331-A333:

[0048] A331: Deploy a pre-trained lightweight neural network model in the edge computing node, which has learned the normal behavior pattern of device operation on historical data.

[0049] A332: Quickly infer and judge the input real-time feature vector, if the abnormal probability output by the model exceeds the confidence threshold (such as 0.85), it is identified as an abnormal segment.

[0050] A333: Encode the abnormal prediction result and confidence, and send it to the digital twin platform for dynamic visualization highlighting and risk deduction.

[0051] It should be noted that by preprocessing (such as noise reduction, feature extraction, and anomaly screening) in the edge terminal, the amount of raw data transmission is reduced, the backbone network bandwidth pressure is reduced, and the communication optimization of "local filtering and key uploading" is realized; After preprocessing, only key indicators and feature values are uploaded, so that the overall sensing link has low latency and high response capability, meets the industrial real-time requirements, and effectively supports subsequent state mapping and online diagnosis tasks.

[0052] Further, in step S2, constructing the unified device runtime sequence database includes the following steps B1-B3:

[0053] B1: Identify device data in multi-modal perception data, and perform field structure analysis and preliminary classification on device data of different sources and different protocols.

[0054] B2: Perform time standardization processing on the device data after field structure analysis, so that the device data has a unified time reference and semantic format.

[0055] B3: Index the device data after time standardization according to the device type dimension and the time dimension, construct a device runtime sequence database, and support periodic query, trend analysis, and historical backtracking of the operation data of a single device, multiple devices, and a regional subsystem.

[0056] Specifically, the device data includes device number, data type, sampling time, unit format, and sensor identifier.

[0057] In the embodiment of the application, in step B2, the time standardization processing adopts rule template-based data standardization processing, including the following steps B211-B213:

[0058] B211: Predefine a multi-vendor device data field correspondence table and a unit conversion template in the system, and define the standard field name, unit target value, time accuracy requirement, and data type description of each type of data in the form of a configuration file;

[0059] B212: For each uploaded data, call the corresponding standardization rule template according to its source terminal identifier to automatically complete field name remapping (such as “V_rms”→“effective voltage”), data unit conversion (such as °F to °C, kV to V), and decimal precision unification.

[0060] B213: Convert different time formats to a unified UTC standard format through a timestamp analysis module, and fill in missing sampling interval values to ensure the continuity of the time alignment.

[0061] In an optional embodiment, the time standardization processing can also adopt an ETL (Extract-Transform-Load) processing mechanism based on a data pipeline, including the following steps B221-B223:

[0062] B221: Build a data cleaning channel containing an ETL process, use a visualization tool to design a node graph, and import the original data stream into a unified cleaning framework.

[0063] B222: Configure field mapping, type checking, missing value filling, outlier removal and semantic annotation processing modules in the conversion stage, support users to customize conversion logic and adjust rules in real time according to business scenarios.

[0064] B223: The converted data is written to the cache database via the loading node, and the unified structure field includes device number, timestamp, measurement value, unit, sampling frequency, data source and other fields, which are used for subsequent construction of time series index.

[0065] In another optional implementation, the time standardization processing can also use field semantic recognition and standardization based on machine learning, including the following steps B231-B233:

[0066] B231: Train a set of lightweight field semantic recognition models based on historical collection data, use natural language processing and classifiers (such as support vector machines or decision trees) to identify field meanings, for example, automatically identify "U_avg", "avgVoltage", "average voltage" as "average voltage" field;

[0067] B232: Perform semantic reasoning on field names, unit symbols, and value ranges in real-time upload data, and the system automatically matches to the standard field directory according to the field type and confidence output by the model.

[0068] B233: Write the standardized field structure into the standard data format object, such as unified "{device ID, timestamp, index type, value, unit}", and continuously optimize the accuracy of the recognition model through artificial feedback mechanism to improve long-term compatibility.

[0069] It should be noted that through time standardization processing and format unification, data fusion from different devices, protocols and manufacturers is realized, a consistent time series database is constructed, and a foundation is provided for subsequent big data analysis and prediction modeling; after building a unified index system, single device or full station operation history can be accurately traced, supporting operation trend analysis, abnormal evolution path tracking, maintenance record comparison and other high-order functions, helping to realize "visualization, traceability and explainability" of the operation and maintenance system.

[0070] Further, in step S3, constructing the digital twin model includes the following steps C1-C3:

[0071] C1: Generate a three-dimensional mesh model by point cloud stitching and surface reconstruction from the spatial information contained in the preprocessed multi-modal perception data.

[0072] C2: Map the device image data to the three-dimensional mesh model combined with the device surface texture information to construct the appearance of the three-dimensional mesh model.

[0073] C3: Extracting equipment data from the equipment runtime series database and binding it to the corresponding components in the three-dimensional grid model in a structured manner to form a digital twin model containing geometric shape, visual features and semantic attributes.

[0074] Further, in step S4, the state mapping and bidirectional interaction between the digital twin model and the virtual engine include the following steps D1-D3:

[0075] D1: Load the completed digital twin model in the virtual engine, and establish a mapping relationship between each component in the digital twin model and the operating state data in the preprocessed multi-modal perception data.

[0076] D2: According to the operating state data, drive the dynamic update and state presentation of the digital twin model in the virtual engine, and through visual means, map and display the physical operation changes in the virtual space in real time.

[0077] D3: Set up interaction trigger logic in the virtual engine. When the operation and maintenance personnel operate the virtual model through the client, generate control instructions and return them to the edge device through the communication link, form an instruction closed loop from virtual operation to physical feedback, and build a bidirectional interaction mechanism between the physical entity and the digital model.

[0078] In the embodiment of the application, in step D2, the visual means adopts state mapping based on color coding and icon labeling, including the following steps D211-D213:

[0079] D211: In the virtual engine, preset state visualization rules for key components (such as circuit breakers, transformers, and disconnectors) in the digital twin model, and use color coding to classify the operating state, where green represents normal operation, yellow represents a warning state, red represents a fault state, and gray represents an unactivated or offline state.

[0080] D212: After the system receives real-time operating data, it parses the state field of the corresponding equipment and automatically updates the color state of the associated components in the digital twin model, and generates corresponding state icons above the components.

[0081] D213: When the user hovers or clicks on the icon, an information floating window pops up, showing detailed operating data (such as temperature, current, voltage, harmonics, electric arc, vibration, etc.) of the equipment, as well as the current abnormal type, severity level, recommended treatment measures and the last maintenance record.

[0082] In an optional embodiment, the visual means can also use state feedback based on local animation and structural deformation, including the following steps D221-D223:

[0083] D221: Establish animation mapping rules between each component of the virtual model and its typical physical response, such as presenting an outer shell bulging animation when the transformer overheats, generating an arc discharge animation when the lightning arrester breaks down, and presenting a periodic shaking animation when the cable vibrates abnormally. The animation is controlled by a blueprint logic script and is bound to the corresponding three-dimensional component in real time.

[0084] D221: When receiving specific perception data exceeding a threshold, such as temperature rising above 85°C, abnormal voltage amplitude or frequency of partial discharge, abnormal number of device actions, etc., the system automatically triggers the corresponding animation and intensifies the visual feedback in a local deformation manner, such as the model shell turning red and emitting light, the base slightly displacing, and the arc dynamically flashing, etc.

[0085] D221: During the animation process, the system continuously synchronizes the state to the timeline recorder for forming a state playback track; when the data returns to normal or the user manually intervenes, the animation slowly transitions back to the default state, forming a complete closed loop of "abnormality → evolution → recovery" visualization.

[0086] In another optional implementation, the visualization means can also use state display based on data overlay and voice broadcast, including the following steps D231-D233:

[0087] D231: Enable data visualization overlay mechanism in the virtual engine rendering interface, and display various running parameters such as bus voltage, current, temperature and humidity, power factor, frequency, and vibration signal characteristic values on the digital twin model in real time. The parameters are presented in a card-style label manner, and the user can customize the display fields, transparency, and refresh frequency.

[0088] D232: Set a configurable warning rule library, and each type of device parameter is bound to corresponding abnormality judgment logic (such as temperature exceeding 70°C, voltage drop exceeding 15%, and vibration frequency abnormality appearing three times, etc.). Once triggered, the system calls the text-to-speech module (TTS) to broadcast the current abnormality in natural voice, such as "Warning: temperature of No. 3 transformer abnormally rises, please check the cooling system". The voice volume, speech speed, and broadcast strategy can be customized according to user permissions.

[0089] D233: The operation and maintenance personnel can pause the broadcast, retrieve historical trend curves, or export data reports through voice control or manual operation of the terminal. At the same time, the platform records this broadcast behavior into the operation log and binds it to the device, forming an event tracking mechanism for later analysis.

[0090] It should be noted that by mapping the physical system real-time data into the digital twin model and feeding back the state changes in a visual manner, the operation and maintenance personnel can intuitively understand the equipment state, improve information transparency and operation efficiency; an operation and control closed-loop mechanism is established to support the generation of control instructions from the virtual model end and the return to the physical system, realize the bidirectional interactive control of the digital twin model and the real equipment, improve the intelligent regulation and control capability and the emergency response efficiency, and support remote maintenance and risk isolation operation.

[0091] Further, in step S5, simulating multi-working condition operation states and identifying potential faults includes steps E1-E3:

[0092] E1: Based on the data input to the digital twin model, dynamic load of sensing data and physical modeling parameters in different operation scenarios is performed to simulate working conditions such as overload, temperature rise, voltage drop, harmonic surge, and frequent operation of circuit breakers.

[0093] E2: During the simulation process, real-time inference analysis is performed on abnormal characteristic values according to the fault identification mechanism.

[0094] E3: By comparing the abnormal characteristic values with known abnormal patterns in the historical fault sample library, and combining the early warning index, it is determined whether there is a potential risk in the current operation condition, and the fault type is output.

[0095] In the embodiment, in step E2, the fault identification mechanism adopts a feature judgment method based on an expert rule system, including steps E211-E213:

[0096] E211: In the digital twin simulation platform, a rule library of multiple typical fault scenarios is defined, such as “bus voltage drop > 10% and duration > 3s” for instantaneous voltage anomaly, and “transformer temperature rise rate > 5℃ / min and fan not started” for cooling anomaly.

[0097] E212: During the simulation process, feature extraction is performed on the real-time response data of the equipment, including temperature slope, voltage fluctuation amplitude, current ripple value, start-stop frequency, and other parameters.

[0098] E213: The extracted parameters are input into a preset rule judgment module, and whether they meet a certain known fault pattern is verified piece by piece according to the condition matching mode. If the rule condition is met, the corresponding potential fault type and suggestion level are output.

[0099] In an optional embodiment, the fault identification mechanism can also adopt an unsupervised fault identification method based on a statistical anomaly detection algorithm, including steps E221-E223:

[0100] E221: Train an anomaly detection model, such as an Isolation Forest algorithm, a one-class support vector machine, or a clustering boundary method, using historical operational data from the equipment to obtain the boundary of normal state characteristics.

[0101] E222: During simulation, convert the response data of the equipment into statistical feature vectors (such as mean, standard deviation, skewness, kurtosis, etc.) to represent the current operational state.

[0102] E223: Input into the trained model to determine whether it deviates from the normal distribution boundary. If the deviation exceeds the set threshold, output an abnormal state marker and label it as "atypical operational behavior" as a potential fault clue for risk assessment.

[0103] In another optional implementation, the fault identification mechanism can also use a fault association identification method based on knowledge graph and semantic reasoning, including the following steps E231-E233:

[0104] E231: Construct a knowledge graph for the substation system, with nodes including equipment type, operational parameters, typical faults, alarm history, interconnection relationships, etc., and edges containing dependency relationships and event causal paths.

[0105] E232: Convert the real-time generated equipment response data in simulation into event triples (such as "Transformer A - temperature rise - abnormality") and map them to the graph nodes.

[0106] E233: Invoke a semantic reasoning module based on a rule engine, perform chain logic deduction in combination with the graph context, identify potential multi-factor co-seismic or cross-equipment associated faults, and output diagnostic results including risk causes, propagation paths, and reasoning confidence.

[0107] It should be noted that by constructing a multi-working-condition simulation environment and integrating a fault identification model, the platform can identify potential risk points before actual faults occur, achieving a transition from passive repair to proactive warning. By combining time series data with simulation modeling, extracting state characteristics, and performing comparative analysis, fault identification has a data-driven basis, model-driven accuracy, and knowledge reasoning capability, providing high-confidence input for the generation of subsequent adjustment schemes.

[0108] Further, in step S6, a structured operational adjustment scheme is formed, and the adjustment scheme is issued to the operation and maintenance terminal, including the following steps F1-F2:

[0109] F1: When a potential fault is identified, generate an optimization scheme based on machine learning.

[0110] F2: Push the optimization scheme to the operation and maintenance terminal device in the form of a structured task package.

[0111] Specifically, when the simulation analysis detects performance degradation or potential failure risks, the platform generates optimization schemes based on machine learning and knowledge rule systems, including running strategy adjustments, device inspection suggestions, load rescheduling, etc. The optimization schemes are pushed to the operation and maintenance terminal or device control interface in the form of structured task packages, supporting human-machine collaboration or automatic intervention. After executing the optimization scheme, the platform continuously monitors the device running state and related index changes, and judges whether the optimization effect meets the expected target through the effect evaluation module. If the system judges that the optimization effect is significant and the device state tends to be normal without new abnormal trends, the process is temporarily terminated, the platform maintains the current running configuration, and enters the normal monitoring mode; if the system finds that the optimization effect is not ideal, the device still has running risks or new abnormalities are triggered, it returns to step S3, and according to the collected perception data, the twin model is updated and a new round of perception-simulation-evaluation-optimization closed loop is entered, building an intelligent operation and maintenance mechanism of "real-time perception-dynamic decision-continuous optimization".

[0112] The machine learning generally adopts an LSTM (Long Short-Term Memory) model training process as follows:

[0113] 1. Prepare the training data set:

[0114] Extract time series data of various devices (such as transformers, switches, arresters, etc.) from the historical substation operation database.

[0115] Collect signals including voltage, current, active power, temperature, local discharge amplitude, frequency, etc., as well as coping schemes during faults.

[0116] Mark the period when the known fault occurs, and construct "positive samples" and "normal operation" control samples.

[0117] 2. Construct input features and labels:

[0118] Divide the original time series into several segments in a sliding window manner, such as 10 seconds per input sequence.

[0119] Each input sequence is used as the input of LSTM.

[0120] The label is whether the window corresponds to a fault (such as binary classification 0 / 1 or multi-class label).

[0121] 3. Define the network structure:

[0122] Input layer: time series features (such as [60 steps, 6 channels]).

[0123] LSTM layer: 1-2 layers can be stacked, each containing 64-128 neurons.

[0124] Fully connected layer: for classification (Softmax function or Sigmoid function output).

[0125] Loss function: cross-entropy loss.

[0126] Optimizer: adaptive moment estimation optimization algorithm or root mean square propagation optimization algorithm, initial learning rate 0.001.

[0127] 4. Model training:

[0128] Batch training (batch size is generally 32-128).

[0129] Early stopping strategy is used on the validation set to prevent overfitting.

[0130] K-fold cross-validation can be used to improve generalization ability.

[0131] 5. Model deployment and online identification:

[0132] Deploy the trained model to an edge computing node or a simulation platform.

[0133] Real-time input of current device state data.

[0134] The model outputs the fault risk score of the current time window (e.g. probability > 0.8 is judged as potential failure).

[0135] Link with knowledge base or rule system to provide alarm and auxiliary decision.

[0136] Example 3, refer to Figure 2 , which is different from the first two embodiments: a substation digital twin management and control system based on integrated sensing algorithm, including physical layer, resource layer, platform layer and application layer; the physical layer constructs a distributed sensing and edge computing network, collects multi-modal sensing data of the substation site by deploying distributed embedded terminal node devices, and completes data preprocessing on the edge side; the resource layer performs time standardization processing and fusion analysis on the multi-modal sensing data, and constructs a unified device operation time series database; based on the preprocessed multi-modal sensing data and the device operation time series database, a digital twin model is constructed, and three-dimensional rendering and real-time interaction in a virtual engine are supported; the platform layer drives the virtual engine and the digital twin model to connect, so that the digital twin model and the virtual engine perform state mapping and bidirectional interaction; through the digital twin model connected with the virtual engine, a simulation environment is constructed, multi-working condition operation states are simulated and potential failures are identified, forming a dynamic decision support mechanism; when potential failures are identified, the application layer forms a structured operation adjustment scheme and sends the adjustment scheme to the operation and maintenance terminal.

[0137] If the functions are implemented in the form of software function units and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application or the parts of the technical solutions that essentially contribute to the prior art or the parts of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0138] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a list of executable instructions for implementing logic functions, which can be embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus or device, such as a computer-based system, a system including a processor or other system that can fetch the instructions from the instruction execution system, apparatus or device and execute the instructions, or in conjunction with these instructions execution systems, apparatus or devices. For the purpose of this specification, the "computer-readable medium" can be any device that can contain, store, communicate, propagate or transport programs for use by or in connection with an instruction execution system, apparatus or device, or in conjunction with these instruction execution systems, apparatus or devices.

[0139] More specific examples (non-exhaustive list) of the computer-readable medium include the following: an electrical connection having one or more wires (electrical devices), a portable computer diskette (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium can even be paper or other suitable medium on which the program can be printed, because the program can be electronically obtained, for example, by optical scanning of the paper or other medium, followed by editing, interpreting or otherwise processing, if necessary, in other suitable ways, to be electronically obtained and then stored in the computer memory.

[0140] It should be understood that portions of the present application can be implemented in hardware, software, firmware, or combinations thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, implementation can be with a combination of any of the following technologies, which are all well known in the art: discrete logic circuitry having logic gates for implementing logic functions upon an application of data signals, application specific integrated circuits having appropriate combinational logic gates, programmable gate arrays (PGA), field programmable gate arrays (FPGA), and the like.

[0141] It should be noted that the above examples are only used to illustrate the technical solutions of the present application but not limit the present application. Although the present application is described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the present application, and all of them should be covered in the scope of the claims of the present application.

Claims

1. A power substation digital twin management and control method based on integrated common sense algorithm, characterized in that: The application relates to a distributed sensing and edge computing network, and relates to a method for constructing a distributed sensing and edge computing network. The method comprises the following steps: a distributed embedded terminal node device is deployed to collect multi-modal sensing data on a substation site, and data preprocessing is completed on an edge side; multi-modal sensing data is subjected to time standardization processing and fusion analysis to construct a unified equipment operation time sequence database; a digital twin model is constructed based on the preprocessed multi-modal sensing data and the equipment operation time sequence database, and three-dimensional rendering and real-time interaction are supported in a virtual engine; a virtual engine is connected to the digital twin model to realize state mapping and bidirectional interaction between the digital twin model and the virtual engine; a simulation environment is constructed through the digital twin model connected to the virtual engine, potential faults are identified in multi-working-condition simulation, and a dynamic decision support mechanism is formed; 2. The power substation digital twin management and control method based on the integrated common sense algorithm, according to claim 1, characterized in that: when a potential fault is identified, a structured operation adjustment scheme is formed, and the adjustment scheme is issued to an operation and maintenance terminal. The data preprocessing on the edge side comprises the following steps: an adaptive data acquisition protocol is used to unify multi-modal sensing data formats; multi-modal sensing data is filtered, denoised and feature extracted based on a sliding window mechanism; 3. The power substation digital twin management and control method based on the integrated common sense algorithm, according to claim 2, characterized in that: an abnormal state in the multi-modal sensing data is preliminarily identified and marked through an abnormality detection mechanism, and abnormal feature values and early warning indexes are marked. The construction of the unified equipment operation time sequence database comprises the following steps: equipment data in the multi-modal sensing data is identified, field structure analysis and preliminary classification are performed on equipment data from different sources and with different protocols; time standardization processing is performed on the field structure analyzed equipment data, so that the equipment data has a unified time reference and semantic format; 4. The power substation digital twin management and control method based on the integrated common sense algorithm according to claim 3, characterized in that: the time standardized equipment data is indexed and organized according to equipment type dimensions and time dimensions to construct an equipment operation time sequence database. The construction of the digital twin model comprises the following steps: spatial information contained in the preprocessed multi-modal sensing data is used to generate a three-dimensional grid model through point cloud splicing and surface reconstruction; device image data is mapped to the three-dimensional grid model in combination with device surface texture information to construct the appearance of the three-dimensional grid model; 5. The power substation digital twin management and control method based on the integrated common sense algorithm, according to claim 4, characterized in that: equipment data is extracted from the equipment operation time sequence database, and is bound to corresponding components in the three-dimensional grid model in a structured manner to form a digital twin model containing geometric morphology, visual features and semantic attributes. The state mapping and bidirectional interaction between the digital twin model and the virtual engine comprise the following steps: the constructed digital twin model is loaded in the virtual engine, and a mapping relationship is established between each component in the digital twin model and the operation state data in the preprocessed multi-modal sensing data; the dynamic update and state presentation of the digital twin model in the virtual engine are driven according to the operation state data, and physical operation changes are mapped and displayed in a virtual space in real time through a visual means; 6. The power substation digital twin management and control method based on the integrated common sense algorithm, according to claim 5, characterized in that: interaction trigger logic is set in the virtual engine, control instructions are generated when an operation personnel operates the virtual model through a client, and the control instructions are returned to an edge device through a communication link to form an instruction closed loop from virtual operation to physical feedback, thereby constructing a bidirectional interaction mechanism between a physical entity and a digital model. The simulation of multi-working-condition operation states and the identification of potential faults comprise the following steps: Based on the data input to the digital twin model, the sensing data and physical modeling parameters under different operating scenarios are dynamically loaded to simulate the working conditions; During the simulation process, the abnormal characteristic values are analyzed in real time according to the fault identification mechanism; By comparing the abnormal characteristic values with the known abnormal patterns in the historical fault sample library, and combining the early warning indicators, it is determined whether there is a potential risk in the current operating condition, and the fault type is output.

7. The power substation digital twin management and control method based on the integrated common sense algorithm, according to claim 6, characterized in that: The forming of the structured operation adjustment scheme and the issuance of the adjustment scheme to the operation and maintenance terminal include When a potential fault is identified, an optimization scheme is generated based on machine learning; The optimization scheme is pushed to the operation and maintenance terminal device in the form of a structured task package.

8. A power substation digital twin management and control system based on integrated common sense calculation, applying a power substation digital twin management and control method based on integrated common sense calculation according to any one of claims 1-7, characterized in that: It includes a physical layer, a resource layer, a platform layer and an application layer. The physical layer constructs a distributed sensing and edge computing network, collects multi-modal sensing data on the site of the substation by deploying distributed embedded terminal node devices, and completes data preprocessing on the edge side; The resource layer performs time standardization processing and fusion analysis on the multi-modal sensing data, and constructs a unified device operation time series database; Based on the preprocessed multi-modal sensing data and the device operation time series database, a digital twin model is constructed, and three-dimensional rendering and real-time interaction in a virtual engine are supported; The platform layer drives the connection between the virtual engine and the digital twin model, enabling state mapping and bidirectional interaction between the digital twin model and the virtual engine; Through the digital twin model connected with the virtual engine, a simulation environment is constructed to simulate multi-condition operating states and identify potential faults, forming a dynamic decision support mechanism; When a potential fault is identified, the application layer forms a structured operation adjustment scheme and issues the adjustment scheme to the operation and maintenance terminal. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is characterized in that: The processor executes the computer program to realize the steps of the substation digital twin management and control method based on the integrated sensing and computing of any one of claims 1 to 7.

10. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executed by the processor to realize the steps of the substation digital twin management and control method based on the integrated sensing and computing of any one of claims 1 to 7.

Images