Park global multi-dimensional real-time digital twinborn collaborative management and control system and method

Abstract

The invention discloses a park global multi-dimensional real-time digital twinborn collaborative management and control system and method, and belongs to the technical field of smart parks, and the system comprises an equipment mapping unit, a data processing unit, a BIM model updating unit, an instruction correction unit, a collaborative management and control unit and an abnormal report generation unit. The equipment mapping unit is used for constructing a space-time coding mapping table and a material library; the data processing unit is used for collecting and filtering IoT data in real time; the BIM model updating unit is used for dynamically updating the BIM model; the instruction correction unit is used for generating and correcting a preliminary cooperation instruction; the collaborative management and control unit is used for obtaining a final control instruction set; and the exception report generation unit is used for generating a fault root cause report as a collaborative management and control result. According to the method, the fault positioning delay is reduced, the red line violation rate is reduced, and the problems of data splitting, response delay and false report and missing report in the prior art are effectively solved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of smart park, and particularly relates to a park global multi-dimensional real-time digital twin collaborative management system and method. BACKGROUND

[0002] At present, with the in-depth promotion of smart park construction, BIM (BIM) and IoT (IoT) technologies are widely used in the planning, construction and operation management of the park. BIM provides accurate three-dimensional digital expression of the physical space of the park, including rich geometric, semantic and attribute information; IoT collects dynamic data such as environment, equipment operation, energy consumption and personnel activities in real time through a sensor network throughout the park. However, the existing technical solutions have significant technical bottlenecks and systematic defects in realizing the global, dynamic and collaborative fine management of the park, mainly in the following aspects:

[0003] 1. Deep data fragmentation and fusion failure.

[0004] Technical separation: The existing platform architecture usually regards BIM static models (stored in CAD / BIM platforms or specific format files) and IoT real-time data streams (stored in time series databases or message queues) as two independent data sources. They lack unified data models, semantic mapping mechanisms and efficient real-time data channels.

[0005] Severe update delay: Dynamic IoT data (such as sensor readings, device status) is difficult to drive the update of BIM models in a timely and automatic manner. Common batch processing or polling methods cause the model state to be seriously out of sync with the physical world, with an update delay of more than 10 minutes, even reaching hours. This makes the situational awareness, analysis and decision-making based on the model (such as emergency evacuation simulation, fault location) lose timeliness and accuracy.

[0006] Data value loss: The semantic information of BIM models (such as device model, material, spatial relationship) cannot effectively enable IoT data analysis (such as understanding which devices or building envelopes are related to the energy consumption anomaly of a certain area). Conversely, the real-time state of IoT cannot be intuitively and dynamically presented in the BIM model (such as real-time temperature cloud map superimposed on the BIM model).

[0007] "Static digital model" rather than "dynamic digital twin": This fragmentation makes the system essentially only a visualization platform or data dashboard, and cannot form a "digital twin" that truly reflects the real-time state of physical entities and supports prediction and optimization.

[0008] 2. Failure of multi-energy flow system collaborative scheduling in sub-regions.

[0009] Subsystem island operation: The park contains complex energy subsystems, such as distributed photovoltaic / wind power (microgrid), electric vehicle charging piles, energy storage systems (batteries, ice storage, etc.), and flexible loads (air conditioning, lighting). Existing management platforms often use a "chimney" architecture, with each subsystem (microgrid control system, charging pile management platform, building automation system) running independently, lacking a unified data interface and collaborative optimization mechanism.

[0010] Lack of global optimization perspective: Each subsystem only focuses on its own goals (such as ensuring charging completion for charging piles and maintaining local balance for microgrids), and cannot coordinate from the perspective of overall energy efficiency, economy (such as responding to time-of-use electricity prices), or stability (such as participating in grid demand response).

[0011] Safety and economy conflict: Lack of coordination may lead to local overload (such as multiple charging piles running at full power simultaneously causing transformer overload) or resource waste (such as photovoltaic power generation being abundant while energy storage is not charged in time).

[0012] 3. Large areas cannot support virtual power plant (virtual power plant) second-level response.

[0013] Virtual power plants require the aggregation of dispersed and heterogeneous distributed energy resources into a controllable whole to participate in power market auxiliary services or main grid dispatching. This requires second-level (even millisecond-level) coordination and control capabilities. Existing technologies cannot achieve fast and accurate collaborative control across systems (such as adjusting charging power, adjusting air conditioning set values, and charging and discharging energy storage) due to data fragmentation, system isolation, and decentralized decision-making, severely restricting the potential of parks to participate in the power market, improve energy income, and enhance flexibility. At the same time, for large parks with multiple land red lines, there are restrictions on cross-regional power transmission within different red line ranges.

[0014] In summary, existing park management platforms have serious deficiencies in the deep integration of BIM and IoT, real-time collaborative control of multi-source heterogeneous systems, and overall virtual power plant dispatching covering multiple red line areas.

[0015] In existing solutions, BIM and IoT data are separated due to storage architecture (BIM is stored in CAD platforms and IoT is stored in time series databases), resulting in: (1) update delay > 10 minutes, low fault location efficiency (average 35 minutes); (2) energy subsystems run independently and cannot respond to second-level virtual power plant dispatching; (3) lack of red line constraint mechanism for cross-plot power transmission, with a violation rate > 23%. SUMMARY

[0016] To solve the above problems, the present application proposes a park global multi-dimensional real-time digital twin collaborative management and control system and method.

[0017] The technical scheme of the present application is: a park global multi-dimensional real-time digital twin collaborative management and control system comprising a device mapping unit, a data processing unit, a BIM model updating unit, an instruction correction unit, a collaborative management and control unit, and an abnormality report generating unit;

[0018] The device mapping unit is used to construct a space-time coding mapping table and a material library.

[0019] The data processing unit is used to collect and filter IoT data in real time according to the cable material library.

[0020] The BIM model updating unit is used to dynamically update the BIM model according to the filtered IoT data.

[0021] The instruction correction unit is used to generate and correct preliminary collaborative instructions using federated learning according to the dynamically updated BIM model.

[0022] The collaborative management and control unit is used to obtain a final control instruction set according to the corrected preliminary collaborative instructions and the park operation and maintenance cost based on spatial accessibility.

[0023] The abnormality report generating unit is used to perform abnormality detection and root cause analysis on the collected IoT data, generate a fault root cause report, and take the final control instruction set and the fault root cause report as the collaborative management and control result.

[0024] Based on the above system, the present application further proposes a park global multi-dimensional real-time digital twin collaborative management and control method, comprising the following steps:

[0025] S1, constructing a space-time coding mapping table and a material library.

[0026] S2, collecting and filtering IoT data in real time according to the cable material library.

[0027] S3, dynamically updating the BIM model according to the filtered IoT data.

[0028] S4, generating and correcting preliminary collaborative instructions using federated learning according to the dynamically updated BIM model.

[0029] S5, obtaining a final control instruction set according to the corrected preliminary collaborative instructions and the park operation and maintenance cost based on spatial accessibility.

[0030] S6, performing abnormality detection and root cause analysis on the collected IoT data, generating a fault root cause report, and taking the final control instruction set and the fault root cause report as the collaborative management and control result.

[0031] Further, S1 comprises the following sub-steps:

[0032] S11, allocate spatial codes for each device in the park, and construct a mapping relationship between the device and the BIM component according to the spatial code;

[0033] S12, calculate the equivalent power transmission distance according to the mapping relationship between the device and the BIM component;

[0034] S13, construct a mapping table between the device and the BIM component according to the equivalent power transmission distance, and generate a cable material library.

[0035] Further, in S12, the equivalent power transmission distance The expression is:

[0036] ;

[0037] ;

[0038] ;

[0039] wherein, represents the number of path points, represents the three-dimensional horizontal coordinates of the th point on the path, represents the three-dimensional horizontal coordinates of the th point on the path, represents the three-dimensional vertical coordinates of the th point on the path, represents the three-dimensional vertical coordinates of the th point on the path, represents the three-dimensional vertical coordinates of the th point on the path, represents the three-dimensional vertical coordinates of the th point on the path, represents the impedance weight, represents the cable thermal resistance influence factor, represents the thermal resistance coefficient of the cable, represents the reference thermal resistance, represents the bending angle influence factor, represents the maximum bending angle of the path segment, represents the space congestion influence factor, represents the number of devices within 5m of the path segment, represents the reference device density, represents the aging sensitivity coefficient, (0.02 / year for copper cable, 0.05 / year for aluminum cable) represents the service life (read from the BIM device attribute library), represents the material reference weight (0.4 for copper cable, 0.8 for aluminum cable).

[0040] Further, S3 comprises the following sub-steps:

[0041] S31, obtaining clean data with space labels from the filtered IoT data;

[0042] S32, deleting invalid components of the BIM model using the spatiotemporal coding mapping table based on the clean data with space labels, calling the inserted components of the BIM model using the spatiotemporal coding mapping table, completing equipment update, and performing lightweight rendering to complete dynamic update of the BIM model.

[0043] Further, S4 comprises the following sub-steps:

[0044] S41, determining a photovoltaic influence factor according to the dynamically updated BIM model;

[0045] S42, taking the photovoltaic influence factor as an input of a plot-level integrated model, and outputting a prediction vector of each plot; the plot-level integrated model adopts a Bagging model;

[0046] S43, splicing the prediction vector of each plot and the BIM space feature of the dynamically updated BIM model to output a preliminary coordination instruction;

[0047] S44, analyzing the preliminary coordination instruction to obtain local real-time data;

[0048] S45, calculating the number of times of ray crossing the red line boundary according to the local real-time data;

[0049] S46, constructing a pre-check red line constraint using federated learning according to the number of times of ray crossing the red line boundary;

[0050] S47, intercepting the red line violation instruction of the park using the pre-check red line constraint to complete correction of the preliminary coordination instruction.

[0051] Bagging is to use several weak machine learning models and aggregate their predictions together to produce the best prediction.

[0052] Further, in S45, the number of times of ray crossing the red line boundary is expressed as:

[0053] ;

[0054] ;

[0055] wherein, represents the number of red line polygon vertices, represents the relationship between the coordinates of the i-th vertex and the j-th vertex of the land-use red line polygon, ​​coordinates of the first vertex of the land redline polygon, coordinates of the first vertex of the land redline polygon, coordinates of the first vertex of the land redline polygon, coordinates of the first vertex of the land redline polygon, three-dimensional horizontal coordinates of the device to be verified, three-dimensional vertical coordinates of the device to be verified, vertical coordinates of the first vertex of the land redline polygon, vertical coordinates of the first vertex of the land redline polygon, vertical coordinates of the first vertex of the land redline polygon, vertical coordinates of the first vertex of the land redline polygon, horizontal coordinates of the first vertex of the land redline polygon, horizontal coordinates of the first vertex of the land redline polygon, horizontal coordinates of the first vertex of the land redline polygon, horizontal coordinates of the first vertex of the land redline polygon, minimum value operation, maximum value operation;

[0056] In S46, the expression of the pre-verification redline constraint is:

[0057] ;

[0058] ;

[0059] wherein, denotes the number of scheduling instructions in the current batch processing, denotes the predicted power value of the th scheduling instruction, denotes the actual schedulable power of the th scheduling instruction, denotes the dynamic penalty coefficient, denotes the set of violation instructions, denotes the device, denotes the three-dimensional spatial coordinate vector of the device, denotes the coordinate of the land redline boundary point closest to the device, denotes the maximum allowed power transmission distance, denotes the initial penalty coefficient, denotes the exponential function, denotes the penalty growth factor, denotes the time, denotes the violation count of the training batch .

[0060] Further, S6 includes the following sub-steps:

[0061] S61, outlier detection is performed on the IoT data collected in real time in the park by using a K-Means clustering algorithm, and abnormal devices and their occurrence time stamps are marked;

[0062] S62, according to the abnormal device list, the device event log of the abnormal device and the BIM space topological relation of the dynamically updated BIM model are queried, and the cause-effect probability is calculated by using a space-time fusion probability function;

[0063] S63, a fault root cause report is generated according to the maximum cause-effect probability, and a fault propagation path is dynamically rendered in the BIM model, and the final control instruction set and the fault root cause report are taken as a collaborative control result;

[0064] In S62, the space-time fusion probability function The expression is:

[0065]

[0066]

[0067] wherein, denotes the cause-effect probability of event causes event , denotes the occurrence time stamp of event , denotes the occurrence time stamp of event , denotes the occurrence time stamp of event , denotes the time decay coefficient, denotes the spatial correlation strength, denotes the spatial normalization factor, denotes the first weight coefficient, denotes the second weight coefficient, denotes the spatial distance between two abnormalities, denotes the reliability of the connecting piece, denotes the topological hop number, denotes the exponential function. The beneficial effects of the present application are: the present application performs BIM-IoT data dynamic fusion, realizes space coding, quantizes BIM attributes such as cable thermal resistance and bending angle into impedance weight, and reduces transmission loss calculation error; the present application also dynamically verifies the land boundary by three-dimensional ray method, and supports large park multi-plot collaborative scheduling. BRIEF DESCRIPTION OF DRAWINGS

[0068]

[0069] Figure 1 It is a structural schematic diagram of a park global multi-dimensional real-time digital twin collaborative control system;

[0070] Figure 2 ​​​​This is a flow chart of the park-wide multi-dimensional real-time digital twin collaborative management and control method. DETAILED DESCRIPTION

[0071] The embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0072] like Figure 1 As shown, the present invention provides a park-wide multi-dimensional real-time digital twin collaborative management and control system, including a device mapping unit, a data processing unit, a BIM model update unit, an instruction correction unit, a collaborative management and control unit, and an abnormality report generation unit;

[0073] The device mapping unit is used to build the spatiotemporal coding mapping table and material library;

[0074] The data processing unit is used to collect and filter IoT data in real time based on the cable material library;

[0075] The BIM model update unit is used to dynamically update the BIM model based on the filtered IoT data;

[0076] The instruction revision unit is used to generate and revise preliminary collaborative instructions using federated learning based on the dynamically updated BIM model;

[0077] The collaborative control unit is used to obtain the final control instruction set based on the revised preliminary collaborative instructions and the operation and maintenance cost of the park based on spatial accessibility;

[0078] The anomaly report generation unit is used to perform anomaly detection and root cause analysis on the collected IoT data, generate a fault root cause report, and use the final control instruction set and the fault root cause report as the collaborative management and control results.

[0079] Based on the above system, such as Figure 2 As shown, the present invention also proposes a park-wide multi-dimensional real-time digital twin collaborative management and control method, comprising the following steps:

[0080] S1, build spatiotemporal coding mapping table and material library;

[0081] S2, collect and filter IoT data in real time based on the cable material library;

[0082] S3. Dynamically update the BIM model based on the filtered IoT data;

[0083] S4. Generate and revise preliminary collaborative instructions using federated learning based on the dynamically updated BIM model;

[0084] S5. Obtain the final control instruction set based on the revised preliminary coordination instructions and the operation and maintenance cost of the park based on spatial accessibility;

[0085] S6, performing anomaly detection and root cause analysis on the collected IoT data to generate a fault root cause report, and taking the final control instruction set and the fault root cause report as the collaborative management result.

[0086] The application realizes real-time synchronization and intelligent collaboration of physical parks and digital twins by constructing three closed loops of data-model, anomaly processing and federal optimization. The core includes: realizing millimeter-level binding of BIM models and IoT devices through a space-time coding engine; adopting a federal learning framework, integrating a dynamic penalty mechanism with exponentially increasing violation times, and realizing collaborative scheduling under cross-plot privacy protection; establishing a cable cluster temperature rise warning model, quantifying historical accumulation and mutual coupling thermal effects, and realizing accurate overheating risk warning; and tracing the fault root cause based on vector clocks and space-time probability functions. The application reduces fault positioning delay and red line violation rate, effectively solving the pain points of data fragmentation, response delay and false positives and false negatives in the prior art. At the same time, the application realizes automatic and accurate tracing of the fault cause-effect chain through a space-time fusion probability function, overcomes the defects of traditional rule engines that rely on manual threshold and poor flexibility, and improves the accuracy of fault root cause analysis.

[0087] In S2, IoT data streams are accessed through MQTT / OPC-UA protocols, Kalman filtering is used for denoising at edge nodes, and spatial tags are attached to output clean data with spatial tags.

[0088] In the embodiment of the application, S1 includes the following sub-steps:

[0089] S11, assigning spatial codes to each device in the park, and constructing a mapping relationship between the devices and BIM components according to the spatial codes;

[0090] S12, calculating the equivalent power transmission distance according to the mapping relationship between the devices and the BIM components;

[0091] S13, constructing a mapping table between the devices and the BIM components according to the equivalent power transmission distance, and generating a cable material library.

[0092] The cable physical properties (outer diameter / conductor size / insulation material) stored in BIM are converted into thermal resistance coefficients , which are the first to unify three types of BIM attributes, cable thermal resistance , bending angle and device density , into spatial impedance coefficients, and use them as spatial impedance weight factors to realize coupling modeling of cable thermal properties and spatial path, high-precision power transmission loss prediction, and prevention of high cable overload.

[0093] In the embodiment of the application, in S12, the expression of the equivalent power transmission distance is:

[0094] ;

[0095] ;

[0096] ;

[0097] wherein, represents the number of waypoints, represents the three-dimensional horizontal coordinate of the point on the path, represents the three-dimensional horizontal coordinate of the point on the path, represents the three-dimensional vertical coordinate of the point on the path, represents the three-dimensional vertical coordinate of the point on the path, represents the three-dimensional vertical coordinate of the point on the path, represents the three-dimensional vertical coordinate of the point on the path, represents the impedance weight, represents the cable thermal resistance impact factor, represents the thermal resistance coefficient of the cable, represents the reference thermal resistance, represents the bending angle impact factor, represents the maximum bending angle of the path segment, represents the space congestion impact factor, represents the number of devices within 5m of the path segment, represents the reference device density, represents the aging sensitivity coefficient, (0.02 / year for copper cable, 0.05 / year for aluminum cable) represents the service life (read from the BIM device attribute library), represents the material reference weight (0.4 for copper cable, 0.8 for aluminum cable).

[0098] bending angle extracted from BIM pipe fittings, adjusted according to the cable aging state, the value of is adaptively adjusted according to the cable material. Dynamic weight adaptation, weight coefficient changes in real time according to BIM attributes: bending angle increases, the impedance weight increases; the device density increases, the impedance weight Increase, ultimately realize the space and electrical combined modeling, first establish the mapping mechanism of "BIM physical properties→space impedance weight", through the three-dimensional innovation of bending angle quantification, device density interference modeling and thermal resistance dynamic coupling, solve the calculation error of power loss caused by ignoring physical characteristics in traditional scheme. The model deeply excavates the semantic information of BIM, and provides accurate spatial constraints for park energy scheduling.

[0099] In the embodiment of the application, S3 comprises the following sub-steps:

[0100] S31, obtaining the clean data with a space label from the filtered IoT data;

[0101] S32, based on the clean data with a space label, deleting the invalid components of the BIM model by using the space-time coding mapping table, calling the inserted components of the BIM model by using the space-time coding mapping table, completing the equipment update, and performing lightweight rendering, and completing the dynamic update of the BIM model.

[0102] In S32, based on the clean data with a space label, the dynamic update and rendering of the BIM model are completed. The core of this step is to adopt an event triggering mechanism and a performance optimization strategy, which ensures the real-time and smoothness of the digital twin update, and solves the technical bottlenecks of high delay and large resource consumption of full model loading in the traditional polling mode.

[0103] In step S3, the event triggering mechanism is specifically: the edge gateway publishes the filtered IoT data to a specific MQTT topic, and the BIM model update service acts as a subscriber and triggers local update only when the data changes. The specific implementation process and optimization strategy are as follows:

[0104] Event-triggered update mechanism (instead of polling)

[0105] Problem: The traditional scheme adopts the method of polling the database at regular intervals to obtain updated data, and the update delay is generally more than 10 minutes, which cannot meet the real-time requirement.

[0106] Solution: The application adopts the message queue telemetry transmission (MQTT) protocol and performs event triggering based on the publish / subscribe mode. After the edge gateway collects the IoT data and filters them through step S2, the clean data with a space label is immediately published as a message to a specific MQTT topic (Topic).

[0107] Effect: The BIM model update service acts as a subscriber and monitors these topics in real time. Once a message arrives (such as a sudden change in the current value of a device), the update service is triggered immediately and only the relevant components that have changed are operated. The test shows that this mechanism reduces the model update delay from more than 10 minutes in the traditional scheme to within 2 seconds, achieving real-time update.

[0108] Lightweight rendering and asynchronous transaction processing

[0109] Problem: BIM models are usually huge in size. Loading and rendering the entire model every time causes client interface lag, CPU usage is extremely high (>80%), and user experience is poor.

[0110] Solution: To ensure system smoothness, the invention uses the following two optimization strategies:

[0111] a. Create a lightweight view template: In the BIM platform (such as Revit), a dedicated view template (ViewTemplate) is created in advance. This template only displays key component categories related to operation and maintenance control such as electrical, heating, and piping, and hides irrelevant components such as architecture and structure. At the same time, the detail level of the view (Detail Level) is set to "coarse" (Coarse) mode to reduce the load of graphic details.

[0112] b. Use asynchronous transaction processing: To avoid blocking the user interface main thread during update operations, a non-blocking asynchronous transaction processing mechanism is used. All model update requests (such as modifying parameters and highlighting devices) are placed in a first-in-first-out (FIFO) task queue. Use the BIM platform's idle event (Idling Event) or background worker thread to take out and process these update requests one by one from the queue.

[0113] Effect: The above strategies greatly reduce the system rendering burden and interface lag. The interface response delay of a single update operation is less than 1 second, ensuring a smooth experience for users during three-dimensional interaction.

[0114] Under the protection of the above mechanism, perform specific update operations:

[0115] Delete invalid components: Use the space-time coding mapping table to locate and delete the BIM components corresponding to the devices that no longer exist through the BIM API (such as the DeleteElement method of Revit API).

[0116] Insert new components: According to the mapping relationship, call the corresponding family type (Family Symbol) from the pre-installed BIM family library (Family Library) and insert new device components at the specified spatial coordinates through methods such as NewFamilyInstance.

[0117] State synchronization and visualization: Write IoT data (such as current and temperature values) into the shared parameters (Shared Parameter) corresponding to the BIM components, and dynamically modify the appearance properties of the components (such as setting the color of overloaded devices to red and flashing), to realize the visualization of the state.

[0118] Through the above technical solutions, the present invention realizes efficient, real-time, stable and dynamic updating of BIM models, providing an accurate digital twin base for subsequent collaborative management and control.

[0119] In this embodiment of the present invention, S4 includes the following sub-steps:

[0120] S41. Determine the photovoltaic impact factor based on the dynamically updated BIM model;

[0121] S42, taking the photovoltaic impact factor as the input of the plot-level integrated model and outputting the prediction vector of each plot; the plot-level integrated model adopts the Bagging model;

[0122] S43, combining the prediction vectors of each block with the BIM spatial features of the dynamically updated BIM model, and outputting preliminary collaboration instructions;

[0123] S44, parsing the preliminary coordination instruction to obtain local real-time data;

[0124] S45. Calculate the number of times the ray crosses the red line boundary based on local real-time data;

[0125] S46. Use federated learning to construct pre-verification redline constraints based on the number of times the ray crosses the redline boundary.

[0126] S47. Use the pre-verification red line constraint to intercept the red line violation instructions of the park and complete the correction of the preliminary coordination instructions.

[0127] Bagging takes several weak machine learning models and aggregates their predictions to produce the best prediction. Bagging ensembles vote / average the results of base models to generate plot-level predictions (e.g., energy storage charging and discharging strategies).

[0128] In the embodiment of the present invention, in S45, the number of times the ray crosses the red line boundary The expression is:

[0129] ;

[0130] ;

[0131] in, Indicates the number of vertices of the red line polygon, Indicates the land use right line polygon The coordinates of the vertex and the The relationship between vertices, Indicates the land use right line polygon The coordinates of the vertices, Indicates the land use right line polygon The coordinates of the vertices, Indicates the three-dimensional horizontal coordinate of the device to be calibrated, Indicates the three-dimensional vertical coordinate of the device to be verified, Indicates the land use right line polygon The vertical coordinates of the vertices, Indicates the land use right line polygon The vertical coordinates of the vertices, Indicates the land use right line polygon The horizontal coordinates of the vertices, Indicates the land use right line polygon The horizontal coordinates of the vertices, represents the minimum operation, Indicates maximum value operation;

[0132] In S46, pre-check red line constraints The expression is:

[0133] ;

[0134] ;

[0135] in, Indicates the number of scheduling instructions processed in the current batch, Indicates the The predicted power value of the scheduling instructions, Indicates the The actual dispatchable power of the scheduling instructions, represents the dynamic penalty coefficient, Represents the set of illegal instructions, Indicates the device, Represents the three-dimensional space coordinate vector of the device, Indicates the coordinates of the land boundary point closest to the equipment. Indicates the maximum allowable transmission distance, represents the initial penalty coefficient, represents the exponential function, represents the penalty growth factor, Indicates time, Represents the training batch Violation count. is a dimensionless system parameter (default 1000) that controls the basic penalty intensity; is a dimensionless system parameter (default 0.3), which is in the range [0.1, 0.5] according to the existing data and adjusts the exponential growth rate.

[0136] The spatial rule micro-transformation pioneered the transformation of land red line constraints into a continuous and differentiable loss term. / This normalization process not only preserves the spatial relationship, but also adapts to machine learning optimization. Compared with the existing 0 / 1 penalty, the method of the present application makes the model gradient update direction more accurate.

[0137] The penalty coefficient increases exponentially with the number of violations. When a certain plot frequently violates the rules, the penalty will be amplified sharply, forcing the model to quickly converge to the safe area, which solves the problem of slow response of static weights. The core value of this function is to convert discrete spatial rules into differentiable mathematical constraints.

[0138] In S46, a pre-check red line constraint is constructed . The constraint consists of a prediction accuracy loss term and a spatial violation penalty term, and the core innovation is the introduction of a dynamic penalty coefficient λ, which can adaptively increase exponentially according to the number of historical violations, thus solving the problem of slow response of traditional static penalty mechanism.

[0139] In the traditional scheme, the check of land red line mostly uses Boolean logic judgment (i.e. '0' or '1'), and once the instruction violates the rules, it is completely discarded. This method has two inherent defects: first, it cannot provide gradient information for machine learning models, and the model cannot learn the continuous concept that 'the closer to the red line, the greater the risk of violation', resulting in blind optimization direction; second, the scheme with fixed penalty weight cannot distinguish between accidental violations and habitual violations, and the constraint force is insufficient for frequently crossing the boundary plots, the model converges slowly, and is easy to oscillate near the violation boundary.

[0140] The working mode and beneficial effects of the dynamic penalty mechanism of the present application are as follows:

[0141] Continuous and differentiable spatial constraint conversion: penalty term , the discrete red line rule is converted into a continuous and differentiable function. The device coordinates are closer to the red line boundary point , the value is larger, which means the risk of violation is higher. This allows the model to perceive the degree of violation through the gradient descent algorithm, rather than just whether it violates the rules, so that it can update the optimization in the correct direction (i.e. away from the red line).

[0142] Adaptive exponential penalty: this is the core of the present application. The dynamic penalty coefficient λ is not a fixed value, its initial value is . It will increase exponentially (controlled by the exp function) with the increase of the total number of historical violations of a certain plot in the training batch . The exponential growth factor is used to adjust the rate of growth.

[0143] Solve the problem of static punishment delay: this design completely solves the problem of static punishment delay mentioned in the background technology.

[0144] For accidental violations: the number of historical violations is small, the value of lambda is small, and the model is moderately punished, which is enough to correct the direction but not overreact.

[0145] For habitual violations: the number of historical violations accumulates, and the value of lambda will be amplified sharply, resulting in a very large punishment to the model. This is equivalent to giving the model a strong signal to force it to quickly converge to the safe compliance solution space, thereby greatly accelerating the training convergence speed of the model and effectively avoiding repeated exploration and oscillation near the compliance boundary.

[0146] Technical effects: This mechanism ensures that the federated learning model not only learns the global energy scheduling rules, but also efficiently and firmly learns and obeys the spatial physical rules (red line) while strictly protecting the privacy of each block data.

[0147] In the embodiment of the application, S5 comprises the following sub-steps:

[0148] S51, according to each parameter of the cable material library, calculate the transformer capacity, cable current, total temperature rise and corrected power as physical constraints;

[0149] S52, according to the physical constraints and the corrected preliminary cooperative instruction, obtain a spatial compliance scheduling instruction;

[0150] S53, calculate the operation and maintenance cost based on spatial accessibility for the park;

[0151] S54, according to the spatial compliance scheduling instruction and the operation and maintenance cost based on spatial accessibility, obtain a final control instruction set.

[0152] In the embodiment of the application, in S51, the transformer capacity The expression is:

[0153] ;

[0154] Wherein, represents the current load of the transformer, represents the change amount of power flowing into the transformer after the instruction is executed, represents the change amount of power flowing out of the transformer after the instruction is executed;

[0155] In S51, the expression of the cable current is:

[0156] ;

[0157] Wherein, represents the active power transmitted, represents the reactive power transmitted, represents the line voltage, represents the power factor;

[0158] In S51, the cable dynamic temperature rise is warned, the spatial thermal field is dynamically transmitted, and the cable temperature rise generates thermal radiation export coupling coefficient , export cable temperature rise increment, and finally obtain the total temperature rise. The expression of the total temperature rise is:

[0159] ;

[0160] ;

[0161] wherein, represents the current effective value of the cable , represents the thermal resistance coefficient of the cable , represents the cumulative effect coefficient of the cable , represents the temperature rise of the cable at moment, represents the current time step, represents the index, represents the heat dissipation rate coefficient of the cable , represents the sampling time interval, represents the total number of surrounding cables, represents the temperature rise of the cable at moment, represents the thermal coupling coefficient of the cable and the cable , represents the effective contact area, represents the spatial distance of the cable and the cable , represents the exponential function, represents the material attenuation factor, represents the insulation layer thickness of the cable ;

[0162] for real-time warning basis; is the main source of heat; from BIM attribute library, determines the self heating efficiency; for quantifying historical temperature rise residue; To control the temperature rise decay rate; From system clock (default 1s), time-varying model parameter; Dimensionless, BIM space topology calculation, innovation in quantifying heat interaction intensity; Determine the coupling range; Calculate the physical contact area, cable outer diameter x parallel length, innovation in quantifying physical contact degree; The closer the distance, the stronger the coupling.

[0163] In S51, the corrected power The expression is:

[0164] ;

[0165] Where, Indicates the original instruction power value, Indicates the maximum allowed power transmission distance, Indicates the actual power transmission distance.

[0166] In S53, the expression of the operation and maintenance cost based on spatial accessibility is:

[0167] ;

[0168] ;

[0169] Where, Indicates the total number of devices, Indicates the time consumed by a single maintenance, Indicates the labor cost per unit time, Indicates the three-dimensional accessibility score, Indicates the first weight coefficient, Indicates the second weight coefficient, Indicates the third weight coefficient, Indicates the path distance to the nearest entrance, data from three-dimensional path planning, Indicates the maximum longitudinal depth of the park, Indicates the number of turns, Indicates the maximum allowed number of turns, Indicates the spatial safety level.

[0170] In the embodiment of the application, S6 includes the following sub-steps:

[0171] S61, using the K-Means clustering algorithm to detect outliers in the real-time collected IoT data of the park, and marking abnormal devices and their occurrence time stamps;

[0172] ​S62, querying the device event log of the abnormal device and the BIM space topology relationship of the dynamically updated BIM model according to the abnormal device list, and calculating the causal probability by using the spatio-temporal fusion probability function;

[0173] S63, generating a fault root cause report according to the maximum causal probability, and dynamically rendering the fault propagation path in the BIM model, taking the final control instruction set and the fault root cause report as the collaborative control result;

[0174] In the S62, the spatio-temporal fusion probability function The expression is:

[0175] ;

[0176] ;

[0177] wherein, represents the causal probability of event causing event , represents the occurrence timestamp of event , represents the occurrence timestamp of event , represents the time decay coefficient, represents the spatial correlation strength, represents the spatial normalization factor, represents the first weight coefficient, represents the second weight coefficient, represents the spatial distance between two abnormalities, represents the connection reliability, represents the topology hop number, represents the exponential function.

[0178] The spatial normalization factor is dimensionless, comes from the maximum device distance, and is used to balance the dimension. The connection reliability comes from the mechanical and electrical property library and is used to quantify the physical connection strength. The spatial distance from device A to device B comes from the device coordinates and is used to calculate that the closer the distance, the stronger the correlation. The topology hop number is dimensionless and comes from the connection graph analysis, which is used to support indirect correlation analysis. The relative weight sum of time and space influence is 1.

[0179] Those skilled in the art will appreciate that the embodiments described herein are intended to aid the reader in understanding the principles of the present application and should not be understood to limit the scope of the present application to such specific embodiments and examples. Those skilled in the art can make various other specific modifications and combinations according to the technical inspiration disclosed in the present application without departing from the essence of the present application, and these modifications and combinations are still within the scope of the present application.

Claims

1. A park-wide multi-dimensional real-time digital twin collaborative management and control system, characterized by: It includes equipment mapping unit, data processing unit, BIM model updating unit, instruction correction unit, collaborative management and control unit and abnormality report generation unit; The device mapping unit is used to construct a spatiotemporal coding mapping table and a material library; The data processing unit is used to collect and filter IoT data in real time based on the cable material library; The BIM model updating unit is used to dynamically update the BIM model according to the filtered IoT data; The instruction correction unit is used to generate and correct preliminary collaborative instructions using federated learning based on the dynamically updated BIM model; The collaborative control unit is used to obtain a final control instruction set based on the revised preliminary collaborative instructions and the operation and maintenance cost of the park based on spatial accessibility; The abnormality report generation unit is used to perform abnormality detection and root cause analysis on the collected IoT data, generate a fault root cause report, and use the final control instruction set and the fault root cause report as the collaborative management and control results.

2. A park-wide multi-dimensional real-time digital twin collaborative management and control method, characterized by: The following steps are involved: S1. Build spatiotemporal coding mapping table and material library; S2, collect and filter IoT data in real time based on the cable material library; S3. Dynamically update the BIM model based on the filtered IoT data; S4. Generate and revise preliminary collaborative instructions using federated learning based on the dynamically updated BIM model; S5. Obtain the final control instruction set based on the revised preliminary coordination instructions and the operation and maintenance cost of the park based on spatial accessibility; S6. Perform anomaly detection and root cause analysis on the collected IoT data, generate a fault root cause report, and use the final control instruction set and fault root cause report as the collaborative management and control results.

3. The park-wide multi-dimensional real-time digital twin collaborative management and control method according to claim 2 is characterized in that: The S1 includes the following sub-steps: S11. Assign space codes to each device in the park and build a mapping relationship between the device and the BIM component based on the space codes; S12. Calculate the equivalent power transmission distance based on the mapping relationship between the equipment and the BIM components; S13. Based on the equivalent transmission distance, a mapping table between the equipment and the BIM construction is constructed to generate a cable material library.

4. The park-wide multi-dimensional real-time digital twin collaborative management and control method according to claim 3 is characterized in that: In S12, the equivalent transmission distance The expression is: ； ； ； in, represents the number of path points, Indicates the path The three-dimensional horizontal coordinates of the points, Indicates the path The three-dimensional horizontal coordinates of the points, Indicates the path The three-dimensional vertical coordinate of the point, Indicates the path The three-dimensional vertical coordinate of the point, Indicates the path The three-dimensional vertical coordinates of the points, Indicates the path The three-dimensional vertical coordinates of the points, represents the impedance weight, Indicates the cable thermal resistance influencing factor, Indicates the thermal resistance coefficient of the cable, represents the base thermal resistance, represents the bending angle influence factor, represents the maximum bending angle of the path segment, represents the spatial crowding impact factor, Indicates the number of devices within 5m of the path segment. represents the baseline device density, represents the aging sensitivity coefficient, Indicates the length of service, Indicates the base weight of the material.

5. The park-wide multi-dimensional real-time digital twin collaborative management and control method according to claim 2 is characterized in that: The S3 includes the following sub-steps: S31, obtaining clean data with spatial labels from the filtered IoT data; S32. Based on the clean data with spatial tags, the spatiotemporal coding mapping table is used to delete the invalid components of the BIM model, and the spatiotemporal coding mapping table is used to call the inserted components of the BIM model to complete the equipment update, and lightweight rendering is performed to complete the dynamic update of the BIM model.

6. The park-wide multi-dimensional real-time digital twin collaborative management and control method according to claim 2 is characterized in that: The S4 includes the following sub-steps: S41. Determine the photovoltaic impact factor based on the dynamically updated BIM model; S42, using the photovoltaic impact factor as input to a plot-level integrated model, and outputting a prediction vector for each plot; the plot-level integrated model adopts a Bagging model; S43, combining the prediction vectors of each block with the BIM spatial features of the dynamically updated BIM model, and outputting preliminary collaboration instructions; S44, parsing the preliminary coordination instruction to obtain local real-time data; S45. Calculate the number of times the ray crosses the red line boundary based on local real-time data; S46. Use federated learning to construct pre-verification redline constraints based on the number of times the ray crosses the redline boundary. S47. Use the pre-verification red line constraint to intercept the red line violation instructions of the park and complete the correction of the preliminary coordination instructions.

7. The park-wide multi-dimensional real-time digital twin collaborative management and control method according to claim 6 is characterized in that: In S45, the number of times the ray crosses the red line boundary The expression is: ； ； in, Indicates the number of vertices of the red line polygon, Indicates the land use right line polygon The coordinates of the vertex and the The relationship between vertices, Indicates the land use right line polygon The coordinates of the vertices, Indicates the land use right line polygon The coordinates of the vertices, Indicates the three-dimensional horizontal coordinate of the device to be calibrated, Indicates the three-dimensional vertical coordinate of the device to be verified, Indicates the land use right line polygon The vertical coordinates of the vertices, Indicates the land use right line polygon The vertical coordinates of the vertices, Indicates the land use right line polygon The horizontal coordinates of the vertices, Indicates the land use right line polygon The horizontal coordinates of the vertices, represents the minimum operation, Indicates maximum value operation; In the S46, the red line constraint is pre-checked The expression is: ； ； in, Indicates the number of scheduling instructions processed in the current batch, Indicates the The predicted power value of the scheduling instructions, Indicates the The actual dispatchable power of the scheduling instructions, represents the dynamic penalty coefficient, Represents the set of illegal instructions, Indicates the device, Represents the three-dimensional space coordinate vector of the device, Indicates the coordinates of the land boundary point closest to the equipment. Indicates the maximum allowable transmission distance, represents the initial penalty coefficient, represents the exponential function, represents the penalty growth factor, Indicates time, Represents the training batch Violation count.

8. The park-wide multi-dimensional real-time digital twin collaborative management and control method according to claim 2 is characterized in that: The S5 comprises the following sub-steps: S51. Calculate transformer capacity, cable current, total temperature rise, and corrected power based on various parameters in the cable material library as physical constraints; S52. Obtaining a spatial compliance scheduling instruction based on the physical constraints and the revised preliminary coordination instruction; S53, calculate the operation and maintenance cost based on spatial accessibility for the park; S54. Obtain the final control instruction set based on the spatial compliance scheduling instructions and the operation and maintenance costs based on spatial accessibility.

9. The park-wide multi-dimensional real-time digital twin collaborative management and control method according to claim 2 is characterized in that: The S6 comprises the following sub-steps: S61. Use the K-Means clustering algorithm to detect outliers in the IoT data collected in real time in the park, marking abnormal devices and their occurrence timestamps; S62. According to the abnormal device list, query the device event log of the abnormal device and the BIM spatial topology relationship of the dynamically updated BIM model, and calculate the causal probability using the spatiotemporal fusion probability function; S63. Generate a fault root cause report based on the maximum causal probability, dynamically render the fault propagation path in the BIM model, and use the final control instruction set and the fault root cause report as the collaborative management and control results; In S62, the spatiotemporal fusion probability function The expression is: ； ； in, Representing an event Leading to the event The causal probability of Representing an event The occurrence timestamp of Representing an event The occurrence timestamp of represents the time decay coefficient, represents the spatial correlation strength, represents the spatial normalization factor, represents the first weight coefficient, represents the second weight coefficient, represents the spatial distance between two anomalies, Indicates the reliability of the connection, Indicates the topological hop count, Represents the exponential function.

Images