A Full-Process Simulation and Verification Method for a Digital Twin-Based Scheduling Automation System
By constructing a digital twin-based automated dispatching system, the full-process simulation and verification of high-voltage main grid dispatching automation scenarios were realized. This solved the problems of fragmentation between simulation and execution processes, lagging model updates, and poor coordination of multi-scale simulations in existing technologies, thereby improving the safety, economy, and reliability of power grid dispatching.
Patent Information
- Application Number
- CN202611291947.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-25
- Publication Date
- 2026-11-10
AI Technical Summary
Existing technologies in power system dispatch automation suffer from problems such as the separation of simulation and execution processes, weak model adaptive update capabilities, insufficient synergy in multi-timescale simulations, and a single dimension of evaluation system, making it difficult to adapt to the complex operational characteristics of high-voltage main grid dispatch.
A scheduling automation system based on digital twins is constructed. Through multi-level collaborative simulation strategy and two-way data interaction link, a scheduling strategy to be verified is generated, and multi-scenario simulation is performed on the digital twin. Quantitative deviation index and adaptive filtering algorithm are introduced to correct parameters, realize closed-loop verification of the whole process and incremental updates of equipment changes, and establish a multi-dimensional evaluation system.
It has realized the full-process simulation and verification of the provincial high-voltage main grid dispatch automation scenario, and constructed a three-stage closed-loop verification system of pre-event, in-event, and post-event, which has improved the safety, economy, reliability and low carbon of power grid dispatch and adapted to the multi-level dispatch needs of the new power system.
Smart Images

Figure CN123069894A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of power system dispatch automation technology, and more specifically, to a method for full-process simulation and verification of a dispatch automation system based on digital twins. Background Technology
[0002] With the rapid development of new power systems, the integration of high proportions of renewable energy, and the widespread application of power electronic equipment, the operating characteristics of power grids are becoming increasingly complex, with significantly enhanced randomness, volatility, and uncertainty. This places higher demands on the reliability, security, and adaptability of dispatch automation systems. Digital twin technology, as a key technology connecting the physical world and the information space, has gained widespread attention and application in the fields of power system data monitoring, simulation, and dispatch optimization due to its advantages of high-fidelity mapping, real-time simulation, and dynamic iteration.
[0003] At present, a large number of studies have been conducted at home and abroad on digital twin simulation of power grids, multi-time scale dispatch optimization, and scenario simulation verification. Some patents and journal articles have disclosed relevant technical solutions. However, based on the analysis of existing comparative patents, existing technologies mostly focus on resilient dispatching in distribution networks, microgrids, and disaster scenarios. There is a significant gap in customized full-process simulation verification technology for provincial main grid dispatch automation systems. These technologies cannot adapt to the complex operating characteristics of the main grid, which has large capacity, high voltage, and multiple plant couplings, and are difficult to meet the refined management and control requirements of the main grid dispatching of new power systems.
[0004] Currently, relevant patented technologies have been explored in the field of digital twins for power systems, specifically:
[0005] Existing patent 1, application number CN117150293A, discloses a data control method for digital twins of regional power system sites. It ensures data timeliness and scenario matching requirements by setting scene tags and data signature verification. This technology primarily focuses on verifying the validity of data access, without addressing the full-process simulation verification of scheduling strategies or providing closed-loop optimization logic.
[0006] Existing patent 2, application number CN121485158B, discloses a source-grid-load-storage coordinated optimization operation method based on digital twins. It constructs a digital twin, calculates the system's transient stability margin based on the extended equal-area criterion, and iteratively optimizes and generates a scheduling strategy. This technology focuses on strategy generation, neglecting online correction and post-implementation iterative optimization after strategy execution.
[0007] Existing patent 3 with application number CN117748733A discloses a power grid information control method based on a digital twin model, which includes general processes such as data acquisition, model building, and simulation optimization, but does not solve the refined technical problems such as multi-timescale coordination and dynamic model calibration.
[0008] Existing patent 4 with application number CN121563081A discloses a resource scheduling method with simulation-feedback-optimization logic, but it is applied to the field of rail operation and maintenance and is not applicable to the power grid scheduling scenario.
[0009] Existing patent 5 with application number CN121350961A discloses a power grid intelligent planning simulation platform based on AI and digital twins. It uses multi-source data fusion, graph convolutional neural network topology recognition, and Monte Carlo multi-scenario verification technology to optimize power grid planning schemes. However, this scheme is biased towards power grid planning scenarios, lacks a clear multi-timescale bidirectional interaction mechanism, and does not set quantitative deviation correction indicators.
[0010] Existing patent 6 with application number CN121688972A discloses a method for power distribution network simulation and load control based on IoT data. It constructs extreme fault scenarios and generates control strategies based on reinforcement learning algorithms. This solution is applied to power distribution networks and is optimized for fault scenarios. It does not have a normalized pre-event, in-event, and post-event full-process control architecture.
[0011] Existing patent 7 with application number CN120933922A discloses an SVG power grid dynamic network construction and simulation system that combines digital twins. It has data processing, deviation calibration and model correction functions. However, this solution only achieves single-dimensional deviation calibration, lacks a multi-dimensional evaluation system, and does not adapt to incremental model updates under equipment changes.
[0012] Existing patent 8 with application number CN116911076B discloses a simulation method for the resilience support of multiple microgrids to the distribution network. It constructs a long- and short-cycle coupled simulation model and adopts a rolling optimization strategy to optimize energy storage scheduling. This scheme focuses on the resilience scheduling of microgrids under disaster scenarios. The simulation scale is singular, there is no second-level real-time simulation module, and it is not applicable to high-voltage main grid scheduling scenarios.
[0013] The deficiencies of the aforementioned existing patents 1-8 are reflected in the following aspects:
[0014] Defect 1: The simulation and execution processes are disconnected, lacking closed-loop control logic. Existing patents 5, 6, and 7 all include simulation verification modules, but simulation deduction, on-site execution, and post-optimization are independent of each other, and simulation results cannot provide real-time feedback to correct execution deviations; although some documents have simple feedback mechanisms, there are no quantitative deviation judgment standards, resulting in a strong reliance on manual intervention and low execution accuracy of scheduling strategies.
[0015] Defect 2: Weak adaptive update capability and poor adaptability of the model. Most of the existing digital twin models in the comparative patent documents are based on fixed topology and historical data. Only the existing patent 7 has a simple model correction function. When the power grid experiences abnormal scenarios such as equipment commissioning, decommissioning, and parameter changes, it cannot quickly complete topology updates and parameter iterations. There are no initialization methods such as transfer learning and similar day matching, resulting in high model mismatch.
[0016] Defect 3: Insufficient coordination in multi-timescale simulation and simplistic interaction logic. Existing patent 8 discloses long- and short-cycle coupled simulation, but only achieves one-way data transmission and lacks a real-time, second-level simulation module; the other comparative patent documents only use single-timescale simulation, and day-ahead, intraday, and real-time scheduling data cannot be communicated bidirectionally, resulting in poor consistency of cross-timescale strategies and difficulty in adapting to multi-level power grid scheduling operations.
[0017] Defect 4: The verification and evaluation system has a single dimension and is not well adapted to the new power system. Existing technologies and comparative patent documents mostly focus on the single dimension of safety, stability, and economic cost evaluation. Existing patent 8 focuses on disaster resilience evaluation and does not integrate low-carbon and new energy consumption indicators, which cannot meet the multi-objective optimization and dispatching needs of the power grid, and the decision-making basis is one-sided.
[0018] Defect 5: Limited application scenarios, unable to adapt to high-voltage main grid dispatch. Existing patents 5, 6, 7, and 8 all focus on distribution networks, microgrids, and disaster emergency dispatch scenarios, leaving a gap in technical solutions for provincial 220kV / 500kV main grids and multi-substation collaborative dispatch. The high complexity and high coupling of the main grid's operation characteristics cannot be adapted to existing technologies.
[0019] No effective solutions have yet been proposed to address the problems in the relevant technologies. Summary of the Invention
[0020] In response to the problems in related technologies, this invention proposes a full-process simulation and verification method for a scheduling automation system based on digital twins, in order to overcome the aforementioned technical problems existing in the existing related technologies.
[0021] Therefore, the specific technical solution adopted by the present invention is as follows:
[0022] A method for full-process simulation and verification of a scheduling automation system based on digital twins, the method comprising:
[0023] S1. Construct a digital twin of the scheduling automation system based on multi-source heterogeneous data from the main network;
[0024] S2. Using the digital twin as the simulation operation carrier, execute a multi-level collaborative simulation strategy of day-ahead planning simulation, intraday rolling simulation and real-time control simulation, and configure bidirectional data interaction links and conflict coordination rules between the multi-level collaborative simulation strategies to generate a scheduling strategy to be verified.
[0025] S3. Simulate and extrapolate the scheduling strategy to be verified on the digital twin of the scheduling automation system in multiple scenarios, filter out the scheduling strategies with constraint violations and perform iterative corrections to obtain a set of safe scheduling strategies that have passed the multi-scenario verification.
[0026] S4. The verified safety policy set is sent to the actual scheduling automation system for execution. During the execution of the safety policy set, a quantitative deviation index and an adaptive filtering algorithm are introduced to correct the digital twin parameters. Simultaneously, multi-plant linkage control and deviation compensation operations are executed to obtain the actual running dataset of the entire execution process of the safety policy set.
[0027] S5. Based on the pre-built comprehensive evaluation index system, conduct multi-dimensional evaluation of the actual operation dataset, and feed back the multi-dimensional evaluation results to steps S1 and S2 for the next round of digital twin update and scheduling strategy generation.
[0028] S6. Monitor the abnormal status of power grid equipment in real time. When an abnormality is detected in the power grid equipment, execute the corresponding differentiated incremental update mechanism according to the type of abnormality to dynamically update the digital twin parameters in step S1.
[0029] The beneficial effects of this invention are as follows:
[0030] 1. The present invention proposes a full-process simulation and verification method for a dispatch automation system based on digital twins, focusing on the dispatch automation scenario of provincial high-voltage main grids, constructing a differentiated core technology architecture; establishing a three-stage closed-loop verification system of pre-event, in-event, and post-event, and designing quantitative deviation correction indicators; establishing a three-level bidirectional collaborative simulation mechanism of day-ahead, intraday, and real-time; constructing a four-dimensional evaluation system of safety, economy, reliability, and low carbon; and adding an incremental update mechanism for equipment anomalies, thereby filling the technical gaps in the fields of refined dispatching of the main grid, full life cycle management, and adaptive iterative optimization.
[0031] 2. This invention is applicable to scenarios of pre-simulation, in-process verification, and post-evaluation of dispatch strategies for power grids with high proportion of new energy access and power electronic grids under the background of new power systems. By focusing on the scenario of provincial main grid dispatch automation, it is different from the simulation dispatch schemes of distribution networks and microgrids. It realizes closed-loop management and control of the entire life cycle of dispatch strategies, improves the safety, economy, low carbon emissions, and reliability of power grid dispatch, and solves the problems of existing power grid simulation dispatch technology, such as fragmented simulation execution, lagging model updates, poor multi-scale simulation coordination, and single evaluation system. Attached Figure Description
[0032] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0033] Figure 1This is an overall flowchart of a full-process simulation and verification method for a scheduling automation system based on digital twins according to an embodiment of the present invention;
[0034] Figure 2 This is a schematic diagram of a multi-timescale collaborative simulation mechanism for a full-process simulation and verification method of a scheduling automation system based on digital twins according to an embodiment of the present invention.
[0035] Figure 3 This is a flowchart of a dynamic deviation correction mechanism for a full-process simulation and verification method of a scheduling automation system based on digital twins according to an embodiment of the present invention.
[0036] Figure 4 This is a schematic diagram of the layered architecture of a digital twin in a method for full-process simulation and verification of a scheduling automation system based on digital twins, according to an embodiment of the present invention.
[0037] Figure 5 This is a flowchart of the pre-simulation deduction of scheduling strategy for a full-process simulation and verification method of a scheduling automation system based on digital twin according to an embodiment of the present invention.
[0038] Figure 6 This is a flowchart illustrating the dynamic update process of the equipment anomaly triggering model for a full-process simulation and verification method of a scheduling automation system based on digital twins, according to an embodiment of the present invention. Detailed Implementation
[0039] To further illustrate the various embodiments, the present invention provides accompanying drawings, which are part of the disclosure of the present invention. These drawings are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these drawings, those skilled in the art should be able to understand other possible implementation methods and the advantages of the present invention.
[0040] According to an embodiment of the present invention, a method for full-process simulation and verification of a scheduling automation system based on digital twins is provided.
[0041] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments, such as... Figure 1 As shown in the embodiment of the present invention, a full-process simulation and verification method for a scheduling automation system based on digital twins is provided. The method includes:
[0042] S1. Construct a digital twin of the scheduling automation system based on multi-source heterogeneous data from the main network.
[0043] Among them, such as Figure 4 As shown ( Figure 4This demonstrates a four-layer architecture model, from the bottom physical mirror layer to the top evaluation and decision layer. The bottom layer provides data support to the upper layers, the upper layers issue driving instructions to the lower layers, and the layers interact bidirectionally through standardized interfaces. This constructs a digital twin of the scheduling automation system, including:
[0044] Based on the CIM / SG-CIM model extension, and integrating real-time operation data, historical data, high and low voltage equipment parameters, and complex grid topology of the high-voltage main grid, a digital twin of the dispatch automation system adapted to the provincial main grid is constructed. CIM (Common Information Model) is a power system information model standard developed by the International Electrotechnical Commission (IEC) to describe various devices, objects, and their interrelationships in a power system, enabling information exchange and interoperability between different systems. SG-CIM (Smart Grid Common Information Model) is a smart grid information model standard that is extended and optimized based on the international standard CIM, taking into account the characteristics of the power grid. Unlike existing general-purpose twin models, this invention adds an electromechanical-electromagnetic hybrid simulation engine to adapt to the transient and steady-state hybrid operation characteristics of the main grid. The digital twin includes:
[0045] Physical Mirroring Layer: This layer performs high-fidelity digital mapping of primary and secondary equipment in the 500kV / 220kV power grid, covering the core network structure. The core architecture of this layer is the CIM topology model, which is linked to equipment parameter libraries, primary equipment model libraries, and secondary equipment model libraries. The primary and secondary equipment model libraries input equipment model data into the CIM topology model, while the CIM topology model outputs static topology information, providing the operational status layer with network topology and basic equipment data.
[0046] Operational Status Layer: Real-time synchronization of massive measurement and status data from the SCADA / EMS system, covering over 200,000 measurement points per network. This layer includes three types of data inputs: real-time SCADA measurements, SOE event sequence records, and PMU synchronization phasors, as well as a historical section playback module. Data from SCADA real-time measurements, SOE event sequence records, and PMU synchronization phasors are all integrated into the historical section playback module, supporting the historical backtracking of power grid operation sections. Simultaneously, real-time status data is output to the behavioral simulation layer, providing real-time operational boundaries for simulation calculations.
[0047] Among them, EMS (Energy Management System) is a large-scale computer application system for power system dispatch and control centers. It integrates functions such as data acquisition, state estimation, security analysis, economic dispatch, and automatic generation control to realize real-time monitoring and optimized dispatch of the power grid.
[0048] Behavioral Simulation Layer: Based on the electromechanical-electromagnetic hybrid simulation engine, it simulates the complex dynamic behavioral characteristics of the high-voltage main grid (specifically step six). This layer includes an electromechanical transient simulation engine, a device dynamic model library, an electromagnetic transient simulation engine, and a hybrid simulation engine. The electromechanical transient simulation engine, the device dynamic model library, and the electromagnetic transient simulation engine are all connected to the hybrid simulation engine, providing it with multi-dimensional simulation capabilities and dynamic model support. The hybrid simulation engine receives real-time status data from the operation status layer and strategy verification instructions from the evaluation and decision-making layer. After completing the simulation calculation, it outputs the dynamic behavioral simulation results to the evaluation and decision-making layer.
[0049] Evaluation and Decision Layer: Embedded with a mainnet-specific scheduling rule base, security and stability verification algorithms, and multi-dimensional economic evaluation models, this layer takes the strategy deduction environment as its core, configuring the scheduling rule base and security and stability verification algorithms as underlying support. It also sets up three types of evaluation modules: reliability evaluation model, low-carbon evaluation model, and economic evaluation model. The scheduling rule base and security and stability verification algorithms provide scheduling rules and verification algorithms to support the strategy deduction environment. The three types of evaluation models interact bidirectionally with the strategy deduction environment to complete the multi-dimensional evaluation of the scheduling strategy. The strategy deduction environment sends strategy verification instructions to the behavior simulation layer and receives the dynamic behavior simulation results returned by the behavior simulation layer.
[0050] The specific methods for constructing digital twins include:
[0051] Unlike existing wavelet transform data fusion methods, this invention uses a dedicated power data platform to normalize multi-source heterogeneous data and construct a digital twin of the dispatch automation system, including the following steps:
[0052] Step 1: Multi-source data fusion: Based on the enterprise data platform, access multi-source heterogeneous data from the power grid resource management system, equipment operation and maintenance lean system, meteorological monitoring system, and SCADA / EMS system. Through data cleaning, normalization, and time-series alignment, a unified data view is formed to adapt to the massive measurement data of the high-voltage main grid.
[0053] Step 2, Equipment-level Modeling: Establish high-precision mechanism models for primary equipment such as generators, transformers, transmission lines, circuit breakers, and reactive power compensation devices, as well as secondary equipment such as protection devices, measurement and control devices, and PMUs. The equipment modeling accuracy is higher than that of the general distribution network model.
[0054] Mechanism model construction (for generators, transformers, etc.): A mathematical model is established based on physical equations (e.g., Parker transformation equations, transformer T-type equivalent circuit). Parameter selection criteria: Key parameters (e.g., generator transient reactance Xd`, inertial time constant H, transformer short-circuit voltage percentage Uk%) are directly read from the equipment nameplate, test report, or EMS database.
[0055] Step 3: Site-level Aggregation: Based on the three-level evaluation system of equipment-site-grid, the high-precision mechanism model is aggregated into a site-level model according to electrical topology relationships, reducing the large-scale computational complexity of the main grid while preserving electrical characteristics. Specific implementation process of site-level aggregation:
[0056] 1. Topology analysis: Based on the CIM model, analyze the connection relationships of all equipment in the substation (e.g., XX500kV substation) to form a node-branch association matrix.
[0057] 2. Node merging: Merge multiple topology nodes connected by closed switches / disconnectors within the site into a single computing node.
[0058] 3. Model replacement: Use the equipment-level mechanism model or data-driven model established in step two to replace each corresponding physical device in the site.
[0059] 4. Network simplification: For radial networks or internal interconnects, static network simplification is performed using star-triangle transformation or Gaussian elimination to generate an external equivalent impedance matrix and injected power model, thereby reducing computational complexity.
[0060] Step 4: System-level assembly: According to the actual power grid topology, assemble the models of each station into a complete high-voltage main grid digital twin, and configure boundary conditions such as external power grid equivalent, load model, and new energy output model.
[0061] Step 5: Model calibration and verification: Use historical operating data to calibrate the digital twin, compare the simulation output with the actual operating trajectory, and ensure that the error of the high-voltage main grid model is less than 3% to meet the dispatch accuracy requirements.
[0062] Step Six: Implementation of the Mechatronics-Electromagnetism Hybrid Simulation Engine
[0063] This invention addresses the dynamic behavior characteristics of high-voltage main grids, which exhibit a mixture of steady-state and transient states. It establishes a hybrid electromechanical-electromagnetic transient simulation engine, clearly defining coupling interfaces, partitioning, interaction timing, and computational logic to achieve refined simulation of the multi-scale dynamic behavior of the power grid. The specific implementation method is as follows:
[0064] 1. Power Grid Zoning: The high-voltage main grid is divided into an electromagnetic transient simulation zone and an electromechanical transient simulation zone. The area where new energy power plants, flexible DC transmission, and reactive power compensation power electronic equipment are located is designated as the electromagnetic transient zone. An electromagnetic transient simulation step size (50μs~100μs) is adopted to accurately characterize the fine dynamic characteristics of power electronic devices such as switching transients, instantaneous harmonics, and voltage flicker. The area dominated by synchronous machines, such as conventional thermal power units, transmission backbone, and large-capacity transformers, is designated as the electromechanical transient zone. An electromechanical transient simulation step size (10ms~20ms) is adopted to take into account the calculation efficiency of large-scale power grid simulation.
[0065] 2. Hybrid Coupling Interaction Mechanism: An interface equivalent coupling algorithm is used to complete the data interaction between the two zones. An equivalent impedance interface is built at the boundary of the zones. The electromagnetic zone outputs the instantaneous voltage and current effective values to the electromechanical zone, and the electromechanical zone sends the power frequency power flow and bus voltage steady-state boundary conditions to the electromagnetic zone. A synchronous timing interpolation alignment method is set to complete the data alignment of different simulation step sizes, eliminate cross-zone timing deviations, and ensure synchronous iterative calculation of hybrid simulation.
[0066] The interface equivalence coupling algorithm includes:
[0067] ①Thevenin Equivalent: For electromechanical transient partitions, calculate their Thevenin equivalent potential relative to the boundary nodes of the electromagnetic partitions. and equivalent impedance The calculation formula is: ,in, , These are the fundamental voltage and current phasors of the boundary nodes.
[0068] ② Interactive data: Electromechanical zones send data to electromagnetic zones. and The electromagnetic zone feeds back the instantaneous power injected into the boundary nodes of the electromechanical zone. The fundamental effective value.
[0069] The synchronous timing interpolation alignment method includes: at the beginning of each electromechanical step, the electromechanical partition sends an equivalent parameter to the electromagnetic partition once, serving as a constant external excitation for the electromagnetic partition within that step period. At the end of the electromechanical step, the electromagnetic partition performs Fourier analysis on the calculation results (voltage, current) of its 200 steps, extracts the effective value of the fundamental component, and feeds it back to the electromechanical partition. Between two electromechanical steps, the electromagnetic partition runs independently without waiting.
[0070] 3. Dynamic Behavior Differentiation Simulation: Under steady-state conditions, the electromechanical transient engine simulates slow dynamic behaviors such as main grid power flow distribution, unit ramp-up, and load fluctuations; under transient conditions such as fault disturbances, sudden changes in new energy sources, and switching operations, the electromagnetic transient engine is automatically activated to capture fine dynamic responses such as millisecond-level voltage drops, power oscillations, and transient overshoots.
[0071] 4. Engine parameter configuration: Adapted to provincial 220kV / 500kV main grids, electromechanical transients support up to 2000 nodes for synchronous calculation, and electromagnetic transients are limited to 200 nodes for fine simulation.
[0072] 5. Simulation accuracy correction: Based on the high-frequency measurement data of the main network PMU, the damping coefficient, line impedance, and unit inertial time constant of the hybrid simulation engine are successively corrected to ensure that the transient waveform and steady-state power flow deviation at the time of fault are controlled within 3%, which meets the standards for dispatch simulation engineering applications.
[0073] S2. Using the digital twin as the simulation runtime carrier, execute a multi-level collaborative simulation strategy that includes day-ahead planning simulation, intraday rolling simulation, and real-time control simulation. Configure bidirectional data interaction links and conflict coordination rules between the multi-level collaborative simulation strategies to generate a scheduling strategy to be verified.
[0074] S2 includes:
[0075] S21. Use digital twins as the operating platform for day-ahead planning simulation, intraday rolling simulation, and real-time control simulation;
[0076] S22. Using the first preset time as the resolution, access the predicted and planned boundary data for future time periods, identify potential operational risk events through continuous power flow simulation and safety verification, and generate a risk list to achieve day-ahead plan simulation.
[0077] S22 includes:
[0078] S221. Obtain load forecast data, new energy output forecast data, unit maintenance plan data and tie line exchange plan data within a preset time range, divide the simulation period with the first preset time as the resolution, and generate a day-ahead simulation dataset.
[0079] S222. Using the day-ahead simulation dataset as input, the power flow tracing algorithm and continuous power flow calculation algorithm are used to simulate the medium- and long-term operating trends of the main network in each time period, generating time-series electrical quantity data for each simulation period.
[0080] S222 includes:
[0081] S2221. Extract the load and output forecast dataset for future periods from the current day-ahead simulation dataset, and set the current simulation period as the initial period.
[0082] S2222. Based on the preset scheduling plan of each load node, the power growth mode of each load node is determined by the proportional allocation principle, and the allocation coefficient of each generator node participating in load growth is determined according to the economic scheduling plan and the droop coefficient, forming a power growth vector.
[0083] S2223. Based on the power growth pattern of each load node and the power growth vector of the generator node, construct an extended power flow equation that includes load level scalar parameters, and solve the extended power flow equation to obtain the power flow solution for the current time period.
[0084] S2224. Using the power flow solution as the current solution, the tangent vector is calculated using the tangent prediction method, and the predicted solution for the next load level is predicted. The predicted solution is used as the initial value for iteration and introduced into the local parameterized equation. The Newton-Raphson method is used to iteratively solve the local parameterized equation to obtain the accurate power flow solution for the current period.
[0085] S2225. Determine whether the power flow calculation for all simulation periods has been completed. If not, return to step S2223 to continue the power flow calculation for the next period. If completed, output the timing electrical quantity data corresponding to the accurate power flow solution for each simulation period.
[0086] S223, preset power flow overload threshold, bus voltage over-limit threshold, and static stability margin threshold, and generate a set of day-ahead simulation safety constraint thresholds;
[0087] S224. Based on the day-ahead simulation safety constraint threshold set, perform cross-sectional thermal stability verification and static safety verification on the time-series electrical quantity data for each time period, and identify potential operational risk events in each simulation period, including heavy load cross-sections, voltage bias, power flow exceeding limits and transmission congestion.
[0088] S225. Mark the occurrence time, location, and over-limit value of each potential operational risk event, and summarize them to form a daily operational risk list.
[0089] S23. Using the second preset time as the rolling cycle, the coordinated operation of multiple plant units is replicated through refined power flow calculation. The scheduling strategy is subjected to full-dimensional safety constraint verification and operation timing verification. After eliminating infeasible strategies, the scheduling instruction sequence is output to achieve intraday rolling simulation.
[0090] S23 includes:
[0091] S231. Using the second preset time as the rolling cycle, superimpose real-time meteorological correction data, load fluctuation data and real-time deviation data of new energy output to generate an intraday rolling simulation dataset.
[0092] S232. Using the intraday rolling simulation dataset as boundary conditions, the power flow calculation results of the previous rolling cycle are used as the initial values for iteration, and a preset adjustment variable is introduced to expand and correct the preset Jacobian matrix.
[0093] S233. Iteratively solve the extended and corrected Jacobian matrix to obtain the refined power flow distribution under the current rolling cycle multi-plant cooperative operation state.
[0094] S234. Based on the refined power flow distribution, the current scheduling strategy is sequentially checked for steady-state power flow, short-circuit current and transient stability. The unit ramping constraints, unit minimum start-up and shutdown time constraints, cross-sectional transmission limit constraints and reactive power reserve constraints are checked item by item to generate the constraint check results of the current scheduling strategy.
[0095] S235. Simulate the timing of multi-plant linkage operations involved in the current scheduling strategy to verify the engineering feasibility of the operation sequence, action interval and adjustment range.
[0096] S236. Summarize the constraint verification results and engineering feasibility verification results of the current scheduling strategy, identify and eliminate infeasible scheduling strategies, and output a sequence of scheduling instructions that has been verified for safety.
[0097] S24. With a step size of seconds or sub-seconds, the pulse regulation command issued by AGC / AVC is received to drive the electromechanical-electromagnetic hybrid simulation engine to simulate the dynamic response process of the unit's excitation regulation and speed control system, quantify and evaluate dynamic indicators and generate dynamic response evaluation results to achieve real-time control simulation.
[0098] S25. Establish a bidirectional data exchange interface between day-ahead planning simulation, intraday rolling simulation and real-time control simulation. At the same time, add a simulation result conflict coordination strategy to realize conflict coordination between day-ahead planning simulation, intraday rolling simulation and real-time control simulation, and finally output the scheduling strategy to be verified.
[0099] To facilitate understanding of the above technical solutions of the present invention, the S2 multi-timescale co-simulation of the present invention will be further explained below from the perspective of architecture and principle, as follows:
[0100] like Figure 2 As shown ( Figure 2 This invention demonstrates the time resolution, simulation range, main functions, and bidirectional data interaction relationship among the three levels of simulation: day-ahead, intraday, and real-time. Based on the digital twin constructed in step S1, a three-level collaborative simulation mechanism of day-ahead planning simulation, intraday rolling simulation, and real-time control simulation is established. Unlike the unidirectional transmission logic of long and short cycles in existing technologies, this invention designs a bidirectional closed-loop data flow interaction mechanism and a multi-level result conflict coordination strategy to achieve mutual correction of simulation parameters and mutual coordination of results across the three levels. It clarifies the technical judgment logic and execution process when simulation results are inconsistent. Specific implementations include:
[0101] Current simulation plan:
[0102] Operational trend prediction and potential over-limit risk identification: With a resolution of 15 minutes, load forecasts for the next 24-72 hours, renewable energy output forecasts, unit maintenance plans, and tie-line exchange plan boundary data are accessed; relying on power flow tracing algorithms and continuous power flow calculation methods, the time-series changes in bus voltage, branch power flow, cross-sectional power, and system frequency are simulated time-by-time to complete the medium- and long-term operational trend prediction of the main grid; power flow overload thresholds, voltage upper and lower limits, and static stability margin thresholds are set, and cross-sectional thermal stability verification and N-1 static safety verification are used to traverse the time-series operating conditions to automatically identify potential operational risks such as overloaded cross-sections, voltage bias, power flow over-limits, and transmission congestion, marking the time period, location, and over-limit amplitude of the risk, forming a risk list and pushing it to the dispatch decision-making level.
[0103] The specific implementation steps of the power flow tracking algorithm and continuous power flow calculation include:
[0104] 1. Initialization: Obtain load and output forecast data for the next 24 hours; set the current simulation time. Based on the dispatch plan data (unit combination status, tie line plan) and the current grid topology and equipment parameters, the following data are output: baseline operating section data (including voltage amplitude and phase angle of each node, power flow of each branch, and active power output of each generator), load growth mode parameters (i.e., the growth mode of each load node (constant power factor growth / constant current growth / constant impedance growth) and generator participation factor). These outputs serve as the inputs for step 2.
[0105] 2. Power Flow Tracing (Determining Growth Direction): Using a proportional allocation principle, for each load node, based on its historical data or dispatch plan, determine its active and reactive power growth pattern (e.g., growth based on a constant power factor). For generator nodes, based on their economic dispatch plan and droop coefficient, determine their allocation coefficient for participating in load growth. Forming a power growth vector Allocation coefficient This indicates the share of incremental active power output that generators need to handle when the total system load increases. It defines how the increase in power generation follows the increase in load, thus determining the path of change in the system's operating point. (The last sentence appears to be incomplete and possibly contains errors.) Step 3, continuous power flow, cannot proceed because the increased load cannot be satisfied out of thin air. Throughout the entire continuous power flow calculation process, it serves as a bridge connecting load growth, generation rescheduling, and power flow calculation, directly determining the understanding of... The evolution trajectory is determined in this step. The output of the initial reference section and the power growth vector determines the evolution direction and path of the system operating point, which is the key input of step 3.
[0106] 3. Continuous power flow calculation: Solving for parameters Extended power flow equations ,in, A scalar parameter representing the system load level, where V represents the node voltage amplitude. This represents the phase angle of the node voltage.
[0107] Prediction phase: Using the tangent prediction method, starting from the current solution... Starting from the beginning, calculate the tangent vector. Predict the next solution , Represents the tangent vector component of the voltage magnitude. Represents the voltage phase angle tangent vector component. The tangent vector component represents the load level parameter. The tangent prediction method is a method in continuous power flow calculation that predicts the next solution point along the tangent direction starting from a known solution point.
[0108] Correction step: Using the predicted value as the initial value, add a parameterized equation (e.g., local parameterization), and use the Newton-Raphson method (i.e., the Newton-Raphson iterative algorithm) to solve iteratively and obtain the exact solution.
[0109] To reduce computational complexity, local parameterization is typically used near the limit point, where the load level is selected. Or the voltage at a certain node As a continuity parameter:
[0110] ;
[0111] in, For the selected continuous variable (usually or The most sensitive node voltage ); To predict the step size, a fixed value (such as 0.05) can be used, or it can be adaptively adjusted based on convergence. This equation forces the selected state variables. From the current solution point linearly advance one step This allows us to determine the search hyperplane for the next solution point.
[0112] 4. Iteration by time period: Update the boundary conditions and repeat step 3 until the simulation ends after 24 hours.
[0113] The specific implementation steps for identifying potential operational risks include:
[0114] 1. Define risk thresholds:
[0115] Overload threshold: 90% of the device's long-term allowable current carrying capacity is the alarm threshold, and 100% is the over-limit threshold.
[0116] Voltage over-limit thresholds: The normal range of 500kV bus voltage is [515kV, 525kV] (1.03~1.05pu); the normal range of 220kV bus voltage is [220kV, 242kV] (1.0~1.1pu).
[0117] Static stability margin threshold: load margin At that time, it was determined that there was a risk of voltage instability.
[0118] 2. Cross-sectional thermal stability verification: For each predefined critical transmission section, calculate its total power. ;like If it is >, then it is marked as a heavily loaded section; if > If it is, then it is marked as an over-limit section.
[0119] 3. N-1 Static Safety Verification: Each component (line, transformer, generator) in the power grid is disconnected sequentially, and the power flow calculation is re-performed under the new topology. If any line / transformer exceeds the limit or the bus voltage exceeds the limit after disconnection, the disconnected device is marked as a risk device, and the limit exceedance information is recorded.
[0120] Intraday rolling simulation:
[0121] Refined simulation and strategy feasibility and safety verification: Using a 5-minute rolling cycle, real-time meteorological corrections, load fluctuation disturbances, and real-time deviations in renewable energy output are superimposed. A refined Newton-Raphson method is used to replicate the coordinated start-up and shutdown, output allocation, and reactive power regulation operation processes of multiple plant units. Steady-state power flow verification, short-circuit current verification, and transient stability verification are carried out for the dispatch strategy. The computer group ramp-up constraints, unit minimum start-up and shutdown time constraints, cross-sectional transmission limit constraints, and reactive power reserve constraints are quantified to determine whether there are any constraints in the current dispatch strategy. At the same time, the timing of multi-plant coordinated operation is simulated to verify the engineering feasibility of the operation sequence, action interval, and adjustment range, and unreasonable strategies with operational conflicts, over-limit risks, and adjustment lags are eliminated.
[0122] The specific implementation of the refined Newton-Raphson iterative algorithm (i.e., the Newton-Raphson method) involves introducing adjustment variables (such as the unit's active power output) based on the conventional Newton-Raphson iterative algorithm. Unproductive efforts Transformer turns ratio The corrections to the Jacobian matrix specifically include:
[0123] 1. Initial value setting: The power flow calculation result of the previous moment is used as the initial value for the current moment's iteration.
[0124] 2. Construction of the modified Jacobian matrix:
[0125] ;
[0126] in, To control the amount of the variable correction; To control the residuals of the target (e.g.) Deviation from planned value); This is the node voltage phase angle correction amount; This is the relative correction amount for the node voltage amplitude; These represent the active and reactive power imbalances at the nodes; each submatrix These are sub-blocks of the extended Jacobian matrix, corresponding to the partial derivatives of power with respect to phase angle, voltage, and control variables, as well as the partial derivatives of the control objective with respect to state variables and control variables.
[0127] Determining whether the current scheduling policy violates constraints includes:
[0128] Climbing constraints: ,in, The unit's ramp rate (MW / min) Let t be the active power output of the generator unit. The active power output of the unit at time t-1.
[0129] Reactive power reserve constraint: ,in, This indicates the total system reserve reactive power. This represents the reactive power reserve factor, typically ranging from 5% to 15%, ensuring sufficient margin to cope with uncertain disturbances. This represents the total reactive power requirement of the system.
[0130] Cross-sectional transmission limit constraints: ,in It is usually set at 90%-100% of the rated capacity.
[0131] Minimum start-up and shutdown time constraints for the unit: or If the inequality is not satisfied, the strategy violates the constraint.
[0132] All of the above constraints must be satisfied. Among them, Indicates the lower limit of active power transmitted across the cross section; This indicates the real-time active power transmitted across the cross-section; Indicates the upper limit of active power transmitted across the cross section; Indicates the current continuous operating time of the unit; Indicates the minimum continuous operating time of the unit; Current continuous downtime of the unit; This indicates the minimum continuous downtime of the unit.
[0133] Real-time control simulation (control effect evaluation and dynamic response characteristic quantification):
[0134] Step 1: Using second-level or even sub-second-level steps, pulse regulation commands issued by AGC / AVC are received, and the dynamic response process of the unit's excitation regulation and speed control system is simulated using an electromechanical-electromagnetic hybrid simulation engine. Step 2: Dynamic indicators such as frequency regulation accuracy, voltage control deviation, power tracking lag time, and oscillation convergence time are quantitatively evaluated. Step 3: The system damping ratio, voltage drop depth, and power overshoot under disturbances are calculated to characterize the transient dynamic response characteristics of the main grid. Step 4: The existence of adverse control effects such as over-regulation, response lag, and local oscillations in real-time control commands is determined, providing quantitative basis for command rollback, fine-tuning, and interlocking. Step 1 is the foundation for all subsequent steps. Without the simulation in Step 1, there are no time-series curves to analyze in Step 2, no oscillation waveforms to extract in Step 3, and no numerical values to determine in Step 4. The output quality of Step 1 directly affects the accuracy of all subsequent steps. Steps 2 and 3 work in parallel and are independent of each other. Both stages share the time-series data from Stage 1, but each extracts different time periods and features: Stage 2 focuses on the final steady-state deviation and how quickly it reaches the steady state for the first time; Stage 3 focuses on the peak value and decay rate during oscillation. Both can run simultaneously without interference. Stage 4 only receives the numerical results already extracted from Stages 2 and 3, and does not directly deal with the large number of time-series data points from Stage 1. This design achieves data dimensionality reduction, compressing thousands of time-series points into seven core values, significantly simplifying the judgment logic. The entire chain is unidirectional and serial, with partial parallelism. The downstream cannot start until the upstream is completed (Stages 2 and 3 must wait for Stage 1 to complete), but Stages 2 and 3 can execute simultaneously without affecting each other; Stage 4 must wait for Stages 2 and 3 to complete before a comprehensive judgment can be made.
[0135] Among them, AGC (Automatic Generation Control) is one of the core functions of EMS. It maintains the system frequency at the rated value by adjusting the active power output of the generator set in real time, and controls the power exchange of the inter-regional tie line to operate according to plan. AVC (Automatic Voltage Control) is one of the core functions of EMS. It maintains the system voltage within the qualified range and optimizes the reactive power flow distribution by coordinating and adjusting the generator excitation, transformer tap changers, reactive power compensation devices, etc.
[0136] The three levels of simulations interact bidirectionally through a dedicated data interface. The bidirectional data flow interaction mechanism for the day-to-day-real-time three-level collaborative simulation includes:
[0137] Day-to-day: The unit combinations and tie line plans planned for the day-to-day simulation are used as boundary conditions for the intraday simulation;
[0138] Intraday → Real-time: The load correction and new energy output correction of the intraday simulation serve as the initial state for the real-time simulation;
[0139] Real-time → Intraday / Day-ahead: Real-time control dynamic response deviation feedback corrects intraday and day-ahead simulation model parameters; in the three-level collaborative simulation, the simulations at the three time scales do not share a single model, but rather each runs based on a digital twin sub-model with different focuses. The day-ahead planning simulation relies on a steady-state scheduling model, which is based on the CIM topology and equipment parameters in the physical mirror layer, and overlays load forecast data and renewable energy output forecast data for 96 time periods of the next day to construct a steady-state power flow calculation model for the entire network. This model focuses on the unit output range, fuel cost curves, and tie-line exchange plans, used to evaluate the economy and safety of scheduling plans for each time period of the next day. The intraday rolling simulation relies on a refined rolling model, which, based on the day-ahead model skeleton, reads the latest SCADA section measurement values from the operating state layer as the initial state, replaces the day-ahead forecast data with ultra-short-term forecast data, and overlays the ramp rate constraints of each unit and the transmission limit constraints of the section to form a hybrid model, used to verify the feasibility of plan adjustments in the next few hours. Real-time control simulation relies on a dynamic response model. This model loads dynamic component models such as generator rotor motion equation, speed control system transfer function, and excitation regulator transfer function from the equipment dynamic model library of the behavioral simulation layer. It uses the measured values (power angle, speed, and voltage) provided by the PMU at the current moment as the initial values for dynamic simulation to form a complete closed-loop control simulation model. This model is used to verify whether the response trajectories of the unit's power, frequency, and voltage after the AGC / AVC command is issued meet the dynamic quality requirements.
[0140] The relationship between the three models is as follows: the day-ahead model provides the initial boundary conditions for the intraday model, the intraday model provides the starting point for dynamic simulation for the real-time model, and the real-time model corrects the relevant parameter assumptions in the day-ahead and intraday models by feeding back deviations from the actual execution results, thus forming a simulation verification chain with the model as the carrier, which is refined step by step and corrected in a closed loop.
[0141] Closed-loop iteration: The three-level simulation forms a closed-loop correction chain through bidirectional data flow, which is adapted to the timing scheduling requirements of the high-voltage main grid.
[0142] Meanwhile, a simulation result conflict coordination strategy is added: if there is an inconsistency between the day-ahead, intraday, and real-time simulation results, the coordination logic of prioritizing short-term results over long-term results, correcting predictions based on actual measurements, and strictly judging constraints is adopted; the priority order is: real-time simulation results > intraday rolling simulation results > day-ahead planned simulation results.
[0143] The planned simulation results, intraday rolling simulation results, and real-time control simulation results are all outputs of step S2 (multi-timescale co-simulation), but they play different roles in the overall process:
[0144] The simulation output is a preliminary scheduling plan (including unit combination status, 96-point output curves of each unit, tie line exchange power plan, and reserve capacity arrangement). This result is the baseline strategy for multi-scenario simulation in step S3, but it is not the final strategy to be issued and executed.
[0145] The intraday rolling simulation output is a revised scheduling plan (based on the day-ahead plan, the unit output for the near period is fine-tuned and the risk of exceeding limits is preventively corrected according to the latest ultra-short-term forecasts and actual cross-sections). This result plays a role in revising and refining the day-ahead plan, and is also used as the input for safety verification in step S3.
[0146] The real-time control simulation output is a dynamic effect evaluation of the closed-loop control command (including the power / frequency / voltage response trajectory under AGC / AVC command, the quantified values of various dynamic indicators, and the control effect judgment conclusion). This result is used to verify whether there are dynamic quality problems at the execution level of the intraday revised plan.
[0147] These three elements together constitute the complete output of step S2. The scheduling strategy to be verified is a comprehensive scheduling scheme formed based on the daily plan simulation results and after intraday rolling corrections. In other words, the scheduling strategy to be verified received in step S3 is not a single model output, but a fusion strategy with the daily plan as the framework, intraday corrections as adjustments, and real-time feedback as constraints.
[0148] Specific coordination and implementation methods include:
[0149] 1. Deviation Judgment: Set cross-scale power deviation threshold and voltage deviation threshold. When the deviation of the calculation results of the three-level simulation exceeds the limit at the same time, it is judged as a simulation conflict.
[0150] 2. Data traceability: Tracing back to the four types of conflict causes: load forecasting deviation, new energy forecasting error, model parameter drift, and inconsistent boundary conditions;
[0151] 3. Hierarchical correction: High-precision short-time simulation results are used to correct the boundary parameters of long-period simulations in reverse. For example, the intraday and day-ahead forecast curves are corrected using real-time measured new energy output.
[0152] 4. Strict Constraints: Under conflict conditions, the most stringent safety constraint results are automatically adopted, and simulation schemes with higher risk of exceeding limits are discarded. Simultaneously, the established simulation models at each level are updated to ensure the consistency of the three-level simulation operation status. The conflict coordination logic, judgment formula, and hierarchical correction method are differentiated technical features that are not disclosed in the existing technology.
[0153] S3. Simulate and extrapolate the scheduling strategy to be verified on the digital twin of the scheduling automation system in multiple scenarios, filter out the scheduling strategies with constraint violations and perform iterative corrections to obtain a set of safe scheduling strategies that have passed the multi-scenario verification.
[0154] S3 includes:
[0155] S31. Establish a standardized set of operation scenarios adapted to the high-voltage main grid;
[0156] S32. Write the scheduling strategy to be verified into the digital twin simulation port, and initialize the simulation boundary conditions of each running scenario with the standardized running scenario set as the simulation scope.
[0157] S33. Load normal scenarios and extreme disturbance scenarios from the standardized operation scenario set in order of priority, and perform topology locking, parameter assignment and boundary solidification operations on each loaded simulation scenario.
[0158] S34. In the normal scenario after loading is completed, the Newton-Raphson iterative algorithm is used to simulate and solve the scheduling strategy to be verified; in the extreme disturbance scenario after loading is completed, the electromechanical-electromagnetic hybrid simulation engine is used, and transient simulation is performed on the scheduling strategy to be verified according to the preset fault duration and fixed simulation step size.
[0159] S35. During the simulation, key measurement point data are collected in real time at a uniform sampling frequency, and interpolation fitting technology is used to generate simulation trajectories for each type of scenario.
[0160] The simulation trajectories generated for each scenario using interpolation fitting techniques include:
[0161] The sampled values of each electrical quantity at discrete time points are extracted from the real-time acquisition data of key measurement points to form a discrete simulation dataset.
[0162] Using the time interval between two adjacent discrete time points in the discrete simulation dataset as the interpolation unit, the simulation period is divided into several interpolation intervals that are connected end to end. A cubic polynomial with time as the independent variable is constructed on each interpolation interval.
[0163] By applying function value constraints and derivative continuity constraints to the cubic polynomial, the undetermined coefficients corresponding to each interpolation interval are solved, and the polynomial expression for each interpolation interval is obtained.
[0164] By concatenating the polynomial expressions of the same electrical quantity in each interpolation interval in chronological order, the continuous trajectory curve of the current electrical quantity in the simulation period is obtained. By traversing all electrical quantities in the current scenario, the simulation trajectory of the current scenario is obtained.
[0165] The simulation trajectory generation process is repeated for each type of scenario to obtain the simulation trajectory corresponding to each type of scenario.
[0166] S36. Compare the simulation trajectory of each scenario with the preset safety threshold in each time period, and mark the scheduling strategy that violates the constraint based on the comparison results;
[0167] S37. Use the interior point optimization algorithm to iteratively correct the scheduling strategy with constraint violations, rewrite the corrected scheduling strategy into the digital twin simulation port, return to step S34 to perform simulation verification, until the simulation trajectories of all scenarios meet the preset safety threshold, and output a set of safe scheduling strategies that have passed multi-scenario verification.
[0168] To facilitate understanding of the above technical solutions of the present invention, the pre-simulation derivation of the S3 scheduling strategy of the present invention is further explained below from the perspective of architecture and principle, as follows:
[0169] like Figure 5 As shown ( Figure 5 This invention demonstrates a complete process from strategy input, multi-scenario generation, traversal simulation to risk assessment and strategy adjustment suggestions. It receives scheduling operation plans drafted by the scheduler or automatically generated optimal scheduling strategies from the scheduling strategy generation and execution layer, and performs refined pre-simulation deduction in a digital twin. To eliminate the shortcomings of purely functional descriptions, this invention clarifies the complete technical implementation methods for generating typical operating scenario sets, multi-scenario traversal simulation calculation processes, operating trajectory extraction methods, and risk assessment. Strategy verification is completed based on algorithm constraints, traversal calculations, and trajectory fitting. Specific implementation content includes:
[0170] Typical Operational Scenario Set Structured Generation: This invention integrates historical measured data, meteorological time-series data, and equipment maintenance records. It employs a DCGAN (Deep Convolutional Generative Adversarial Network) to generate a standardized operational scenario set adapted to the high-voltage main grid. Unlike simply manually setting fault scenarios, this standardized operational scenario set possesses both data-driven and physically constrained generation logic. DCGAN is a deep learning model composed of two neural networks: a generator and a discriminator. Through adversarial learning, it generates new samples similar to the real data distribution. The generation process includes:
[0171] ①Dataset construction: Collect time series samples of provincial main grid load, new energy output, bus voltage, meteorological irradiance, and temperature over the past three years, remove outlier values, and construct a training dataset;
[0172] ② Network training: The generator uses a four-layer convolutional structure to extract the spatiotemporal coupling features of load and new energy, and the discriminator completes the scene authenticity judgment and continuously iterates and optimizes the network weights;
[0173] ③ Embedding physical constraints: Power balance constraints, unit ramping constraints, cross-sectional transmission limit constraints, and voltage fluctuation constraints are embedded at the generator output to eliminate physically infeasible scenarios;
[0174] ④ Scenario Classification and Consolidation: Automatically consolidated into two major categories: normal operation scenarios and extreme disturbance scenarios. These include standardized scenarios such as peak load scenarios, off-peak load scenarios, scenarios with significant changes in new energy power generation, N-1 fault scenarios, and extreme weather scenarios, forming a scenario library containing 10 to 15 fixed scenarios that can be used for batch simulation and invocation of scheduling strategies.
[0175] The multi-scenario automatic traversal simulation calculation includes building a multi-scenario traversal simulation architecture with serial and breakpoint-resume calculations to achieve fully automatic scenario polling calculations. The specific process is as follows:
[0176] ① Strategy Injection: Write the scheduling strategies to be verified (unit output plan, reactive power adjustment command, tie line switching power, load control scheme) into the digital twin simulation port in batches and initialize the simulation boundary conditions in a unified manner;
[0177] ②Scene loading: Normal scenes and extreme disturbance scenes are loaded in sequence according to scene priority. A single scene loading completes topology locking, parameter assignment, and boundary fixing.
[0178] ③ Solution calculation: The Newton-Raphson iterative algorithm is used for steady-state conditions, and the convergence accuracy of the iteration is set to 10⁻. 4 The fault disturbance condition uses the electromechanical-electromagnetic hybrid simulation engine, and the fault duration simulation time is set to 5s~20s with a fixed simulation step size.
[0179] ④ Parallel acquisition of measurement points: During the simulation, key measurement point data such as bus voltage, branch power flow, node frequency, unit output, and cross-sectional transmission power of the entire network are collected in real time and synchronously, with the sampling frequency uniformly set to 100Hz.
[0180] The multi-scenario automatic traversal simulation calculation (specific process of Newton-Raphson iterative algorithm) includes:
[0181] 1. Initialization: Provide initial voltage values for each node. and voltage phase angle The phase angle of the voltage at the slack node is usually set. PQ node pu.
[0182] 2. Calculate power deviation: Calculate the injected active power at each node. and no merit The imbalance quantity, among which, This represents the active power imbalance at node i in the kth iteration, in MW or per unit. If it is positive, it means that the injected power is greater than the outflow power, and the node voltage phase angle needs to be increased. This represents the given active power injection at node i, which is positive for generator nodes and negative for load nodes, and is derived from SCADA real-time data or scheduling plans. This represents the reactive power imbalance at node i in the kth iteration, expressed in MVar or per-unit value. This represents the given reactive power injection power at node i.
[0183] 3. Calculate the Jacobian matrix: Based on the current iteration value, calculate the Jacobian matrix. each sub-block .
[0184] 4. Solve the corrected equations: Solve the system of linear equations. This yields the correction values for the voltage amplitude and phase angle at each node.
[0185] 5. Update state variables: , .
[0186] 6. Convergence Criterion: If all nodes... and If the result is found to be converged, output the result; otherwise, return to step 2.
[0187] in, The vector representing the correction of the node voltage phase angle is the unknown quantity to be determined; k represents the k-th iteration; PQ node represents the load node with given active and reactive power, and the voltage amplitude correction vector is also the unknown quantity to be determined. Represents the active power imbalance vector at nodes (except for balanced nodes). Represents the reactive power imbalance vector at nodes P and Q; This represents the current voltage phase angle of node i in the k-th iteration; This represents the voltage phase angle of node i in the (k+1)th iteration; This represents the voltage phase angle correction at node i obtained by solving the modified equation; This represents the current voltage amplitude of node i in the k-th iteration; This represents the voltage magnitude correction at node i obtained by solving the modified equation.
[0188] In addition, multi-scenario operation trajectory extraction and risk assessment correction include:
[0189] ① Operation trajectory construction: With time as the horizontal axis and electrical quantity as the vertical axis, interpolation and fitting are used to generate voltage time sequence trajectory, power flow change trajectory, frequency fluctuation trajectory and unit regulation trajectory for each type of scenario, fully reproducing the dynamic evolution process of scheduling strategy under different operating conditions;
[0190] ② Quantitative risk assessment: Preset high-voltage main grid safety constraint thresholds, including bus voltage 0.95~1.05pu, line power flow ≤90% of rated capacity, system frequency 49.8~50.2Hz, and transient stability margin ≥15%; compare the simulation trajectory with the safety constraint thresholds time by time to determine whether there are any over-limit scenarios. If so, mark the over-limit time, over-limit amplitude, and affected area, and suggest adjusting the scheduling strategy. If not, it means that the simulation has passed.
[0191] ③ Iterative Strategy Correction: For scenarios involving exceeding limits, an interior-point optimization algorithm is used to optimize unit output allocation, reactive power compensation capacity, and cross-sectional transmission limits with the goal of minimizing adjustment costs. The algorithm quantifies the adjustment step size, adjustment sequence, and correction magnitude. After correction, the data is fed back into simulation for verification until all scenarios are free of constraint violations. In pre-simulation, when a scheduling strategy is detected to cause voltage exceeding limits or cross-sectional overload under specific scenarios (e.g., large-scale renewable energy generation, N-1 fault), a correction scheme needs to be automatically generated. The interior-point optimization algorithm is a mathematical programming method introduced to solve this problem. It can optimize unit output allocation with the goal of minimizing adjustment costs while satisfying various grid security constraints. The interior-point algorithm is a tool for solving large-scale nonlinear programming problems. Its core idea is to search for the optimal solution within the feasible region, avoiding wandering on the boundary, thereby achieving a faster convergence speed. The application of interior point optimization algorithm in this invention is to transform the risk of exceeding the limit into mathematical constraints and the correction scheme into the optimal solution. The scheduling correction instruction that minimizes the adjustment cost and satisfies all safety constraints can be obtained through one automatic solution.
[0192] Among them, the interpolation fitting methods for multi-scenario operation trajectory extraction and risk assessment correction include:
[0193] 1. Using cubic spline interpolation, the simulation output is a discrete-time point result. t=0,0.1s,0.2s,... Electrical quantity values;
[0194] 2. Interpolation Fitting: For each electrical quantity y (e.g., the voltage of a certain bus), in each time interval... Above, construct a cubic polynomial ;
[0195] 3. Boundary conditions: Mandatory requirements , And at the connection point First derivative and second derivative Continuous, solve for the coefficients in each interval. This forms a smooth, continuous trajectory curve.
[0196] in, Indicates at time The fitted value of the electrical quantity, for example, representing The voltage amplitude of the bus at that moment; This represents the i-th discrete time point in the simulation output, for example... ; It represents the time at the i-th discrete moment. The electrical quantity fitting value, Similarly; Represents the i-th discrete time. The simulation output value, Similarly; This represents the coefficient of the cubic term, which determines the rate of change of the curve's curvature. The larger the value, the more pronounced the curvature of the curve in that interval. The coefficient represents the quadratic term and determines the acceleration of the curve. A positive value indicates that the curve bends upward (concave), and a negative value indicates that it bends downward (convex). The coefficient of the linear term determines the curve's position. The slope of the tangent at time t, i.e. the rate of change of the electrical quantity at that time; This represents a constant term that determines the curve's position. The intercept at time.
[0197] S4. The verified safety policy set is sent to the actual scheduling automation system for execution. During the execution of the safety policy set, a quantitative deviation index and an adaptive filtering algorithm are introduced to correct the digital twin parameters. Simultaneously, multi-plant linkage control and deviation compensation operations are performed to obtain the actual running dataset of the entire execution process of the safety policy set.
[0198] S4 includes:
[0199] S41. The verified security policy set is sent to the actual scheduling automation system, and measurement data during the execution of the security policy set is collected in real time.
[0200] S42. Compare the actual running trajectory generated based on the measurement data during the execution process with the simulation trajectory of the digital twin point by point, and calculate the deviation between the actual running trajectory and the simulation trajectory.
[0201] S43. When the deviation exceeds the preset threshold, the digital twin parameters are corrected using an adaptive filtering algorithm so that the operating state of the digital twin is consistent with the actual operating state of the power grid.
[0202] S44. For multi-plant coordinated control operations, the plant execution status feedback data of each plant is collected synchronously according to the cross-sectional level. The plant execution status feedback data is compared with the preset operation sequence template to generate the execution lag of each plant.
[0203] S45. When any power station experiences execution lag, the compensation adjustment amount of the adjacent power station is calculated based on the overall power balance of the section where it is located, and the adjacent power stations are linked to issue compensation instructions and update the power station execution status feedback data.
[0204] S46. Summarize the measurement data and real-time updated plant execution status feedback data throughout the entire execution process of the security policy set to form the actual operation dataset of the entire security policy set execution process.
[0205] To facilitate understanding of the above technical solutions of the present invention, the following further explains the real-time online verification of the S4 scheduling strategy of the present invention from the perspective of architecture and principle, as follows:
[0206] The scheduling strategy, verified by the simulation in step S3, is then deployed to the actual scheduling automation system for execution. Figure 3 As shown ( Figure 3 This invention demonstrates the complete process from real-time monitoring, deviation calculation, threshold judgment, model calibration, re-simulation, to correction command generation. It also establishes a quantified deviation correction mechanism. Unlike existing technologies that lack quantified judgment standards, this invention uses a custom deviation index to precisely trigger the correction process.
[0207] Real-time acquisition of measurement data during the execution process, and dynamic comparison with the simulation trajectory of the digital twin;
[0208] When the deviation between the actual trajectory and the simulated trajectory (deviation index) exceeds a preset threshold, an online correction mechanism is automatically triggered without manual intervention. The deviation index is the root mean square value of the normalized deviation of each measurement point. When the root mean square value of a certain quantity is greater than 1, it is generally considered that the quantity has deviated from the unit reference range. Mathematically, this corresponds to the judgment that the overall deviation exceeds the acceptable range.
[0209] ①Based on adaptive unscented Kalman filtering or particle filtering algorithms, the parameters of the digital twin model are corrected online to quickly match the physical power grid operating status;
[0210] ② Resimulate the remaining time period strategy, compare the new simulation trajectory with the original strategy, and determine whether the instruction needs to be adjusted. If so, generate a correction instruction and push it to the scheduling strategy generation and execution layer, so that the dispatcher can intervene through the dispatcher interface or directly issue it to the physical power grid layer. Then update the digital twin parameters and record the correction event. If not, directly update the digital twin parameters and record the correction event.
[0211] The specific implementation method of the dynamic deviation correction mechanism is as follows:
[0212] Define the deviation index:
[0213] ;
[0214] in, Let be the actual value of the i-th key measurement point at time t. These are simulation values. The allowable deviation threshold for this measurement point is defined as D(t), and N is the number of key measurement points. When D(t) > 1, digital twin model correction is triggered. The correction algorithm employs adaptive unscented Kalman filtering to adjust the model parameters online, enabling the simulated trajectory to quickly approximate the actual trajectory.
[0215] For complex operations involving multi-plant coordination, cross-sectional data synchronization technology is employed to verify the execution progress and coordination sequence of each plant in parallel within a virtual environment. The specific implementation details of cross-sectional data synchronization technology (i.e., cross-sectional technology) include:
[0216] 1. Section boundary delineation: Taking the key transmission sections and interconnection sections of the provincial main grid as the synchronization boundaries, the substations, generator units and transmission branches covered by the sections are divided into the same synchronization verification set to clarify the coupling relationship of the sections;
[0217] 2. High-precision time synchronization: Based on the BeiDou / GPS timing mechanism, all network measurement, control commands, and equipment status data are uniformly aligned to millisecond-level timestamps, eliminating timing deviations caused by communication delays and data acquisition delays between different plants and stations;
[0218] 3. Cross-sectional data consistency calibration: The cross-sectional power flow conservation verification algorithm is adopted to verify the power balance relationship of incoming power, outgoing power and network loss power in real time, correct local measurement errors of power plants and stations, and ensure that multi-source data under the same cross-section are consistent.
[0219] The specific implementation process of the cross-sectional power flow conservation verification algorithm is as follows:
[0220] ① Data Acquisition: Obtain the same timestamp from SCADA / PMU Below, all incoming lines constituting the cross-section ( ) and qualifying ( Active power measurement value and .
[0221] Among them, PMU (Phasor Measurement Unit) is a device that uses high-precision time synchronization technology based on Global Positioning System (GPS) or Beidou Satellite Navigation System (BDS) to synchronously measure the amplitude and phase angle of electrical quantities such as voltage and current at various nodes of the power system; SCADA (Supervisory Control And Data Acquisition) is a computer system used for remote monitoring and control of industrial processes. In the power system, the SCADA system is responsible for real-time monitoring, data acquisition, and remote control of the power grid's operating status.
[0222] ② Network loss calculation: Based on the cross-sectional topology and line parameters, the active power loss of the cross-section is estimated according to the real-time current. The formula for calculating network loss is not covered here.
[0223] ③ Calculation of conservation deviation: Calculate the power imbalance. .
[0224] ④ Consistency correction: when Calibration is triggered when the sum of rated power is 0.5%.
[0225] ⑤ Correction method: Use weighted least squares method to adjust the total imbalance. The corrected power values are generated by assigning weights to all incoming and outgoing lines based on the reciprocal of the measurement accuracy of each quantity. .
[0226] 4. Collaborative Execution Progress Verification: For multi-plant coordinated control operations, feedback on switch changes, output adjustments, and reactive power switching at each plant level is collected synchronously at the cross-sectional level. This feedback is compared with a preset operation sequence template to determine the execution lag of each plant. The preset operation sequence template is essentially a structured execution plan table that contains the standard execution sequence benchmark for each operation step in the multi-plant coordinated operation. This template is not a single numerical value, but a judgment rule system consisting of three parts: a time reference point, the relative time offset of each step, and the allowable time deviation range.
[0227] 5. Deviation Linkage Correction: If a certain power station lags behind in its actions or has a deviation in amplitude, the adjacent power stations will be linked to make fine adjustments and compensations based on the overall power balance of the cross section, so as to suppress the power fluctuation of the cross section and ensure the synchronization and execution accuracy of the coordinated operation of multiple power stations in the high-voltage main grid.
[0228] S5. Based on the pre-built comprehensive evaluation index system, conduct multi-dimensional evaluation of the actual operation dataset, and feed back the multi-dimensional evaluation results to steps S1 and S2 for the next round of digital twin update and scheduling strategy generation.
[0229] To facilitate understanding of the above technical solutions of the present invention, the post-evaluation optimization of the S5 scheduling strategy of the present invention will be further explained from the perspective of architecture and principle as follows:
[0230] After the scheduling strategy is executed, actual data from the entire execution process is collected for multi-dimensional evaluation. Unlike the single evaluation dimension of existing technologies, this invention constructs a four-dimensional comprehensive evaluation index system that aligns with the dual-carbon requirements of the new power system. The evaluation index system includes:
[0231] Safety indicators: static safety margin, transient stability margin, voltage stability margin, frequency stability margin;
[0232] Economic indicators: electricity purchase cost, grid loss rate, and renewable energy consumption rate;
[0233] Reliability indicators: power supply availability, load loss risk; Low-carbon indicators: carbon emission intensity, proportion of clean energy.
[0234] Compare the actual operating results with the simulation results, quantify the deviations, and analyze the causes.
[0235] Based on the evaluation results, the parameters of the scheduling strategy generation model are optimized using reinforcement learning or Bayesian optimization algorithms, forming a closed-loop iterative mechanism of simulation-execution-evaluation-optimization.
[0236] S6. Monitor the abnormal status of power grid equipment in real time. When an abnormality is detected in the power grid equipment, execute the corresponding differentiated incremental update mechanism according to the type of abnormality to dynamically update the digital twin parameters in step S1.
[0237] S6 includes:
[0238] S61. Monitor the abnormal status of power grid equipment in real time. When an abnormality is detected in the power grid equipment, identify the type of abnormality event and generate an abnormality message with a unique code.
[0239] S62. Based on the semantic parser, analyze the key fields of equipment anomaly in the anomaly message, and establish the mapping relationship between power grid equipment and digital twin based on the key fields of equipment anomaly.
[0240] S63. Based on the mapping relationship, the location of the abnormal equipment is locked. The adjacency matrix topology splitting algorithm is used to extract the associated bus, branch and circuit breaker nodes, and local topology decoupling is performed on the current connected domain.
[0241] S64. Within the local topology range formed by decoupling, write all parameters of the abnormal equipment into the attribute fields of the corresponding equipment in the digital twin, and update the associated electrical parameters synchronously.
[0242] S65. Compare the topology connectivity matrix before and after the update, verify the consistency of node connection relationships and electrical boundaries, and solidify the topology and parameters after the verification is passed, thus realizing the initial update of the digital twin parameters.
[0243] S66. After the initial update of the digital twin parameters, a differentiated incremental post-processing mechanism is executed according to the type of abnormal event to achieve dynamic updating of the digital twin parameters.
[0244] S66 includes:
[0245] S661. If the change type is parameter equipment change, then use parameter transfer learning technology to transfer the simulation parameters of the equipment before the change to the changed equipment.
[0246] S662. If the change type is new equipment commissioning, extract the feature parameters of the newly commissioned equipment and construct the feature vector of the newly commissioned equipment. At the same time, extract the feature parameters of the same type of equipment already commissioned in the stock equipment library and construct the feature vector of the stock equipment.
[0247] S663. Normalize the feature vectors of newly commissioned equipment and existing equipment respectively, and calculate the correlation coefficient between the newly commissioned equipment and each existing equipment after normalization.
[0248] S664. The correlation coefficients are weighted to obtain the weighted correlation degree between the current stock equipment and the newly commissioned equipment. All stock equipment is sorted from high to low according to the weighted correlation degree, and a preset number of stock equipment is selected as similar equipment.
[0249] S665. Extract historical operating data of similar equipment and use it as the initialization data template for the digital twin of the newly commissioned equipment to achieve the initialization of the digital twin of the newly commissioned equipment.
[0250] To facilitate understanding of the above technical solutions of the present invention, the S6 digital twin dynamic update of the present invention will be further explained below from the perspective of architecture and principle, as follows:
[0251] like Figure 6 As shown ( Figure 6This demonstrates the automatic update process of the digital twin when power grid equipment undergoes changes, such as new commissioning, decommissioning, or parameter changes. This includes the entire process of topology update, parameter synchronization, model initialization / transfer learning, accuracy verification, and final deployment. When power grid equipment undergoes changes (new equipment commissioning, decommissioning of older equipment, or changes in equipment parameters), a differentiated incremental update process is initiated. This differs from the global reconstruction update method of existing technologies, reducing computational power consumption. Specifically, it includes:
[0252] The triggered anomaly events are identified by type and classified into three categories: new equipment, parameter change, and equipment decommissioning. For new equipment anomalies, the CIM topology model is updated first, and the electrical node attributes and topology connection relationships of the new equipment are entered. For parameter change anomalies, the equipment parameter library is updated directly, and the parameters of the target equipment are synchronously corrected. For equipment decommissioning anomalies, the equipment removal operation is performed.
[0253] After the three types of anomalies are pre-processed, they are uniformly integrated into the topology reconstruction and parameter synchronization stage. Through local topology decoupling, batch parameter assignment and consistency verification, the network structure and equipment parameters are synchronized and refreshed. After the topology and parameters are synchronized, it is determined whether the current anomaly scenario needs to be initialized, so as to adapt to the model update requirements of two different scenarios: new equipment commissioning and parameter changes.
[0254] If it is determined that model initialization is required, for newly commissioned equipment without historical operating data, a weighted grey relational similarity day matching algorithm is used to generate the initial behavior model of the equipment using the historical data of existing similar equipment as a template; if it is determined that model initialization is not required, for parameter change-type anomalies, the simulation weights and operating experience of the model before the anomaly are transferred to the updated model through parameter transfer learning technology, while retaining the original simulation convergence characteristics.
[0255] After completing model initialization or parameter inheritance, the simulation output results are compared with the actual operating trajectory based on the actual power grid operation data. The simulation accuracy of the updated digital twin is verified, and the accuracy verification results are judged. If the accuracy meets the standard, the deployment stage is entered; if the accuracy meets the standard, the calibration stage is entered, forming a closed-loop verification mechanism.
[0256] If the accuracy verification fails to meet the standard, a refined calibration process is initiated to iteratively optimize and adjust the model parameters. After calibration, the model accuracy verification is returned to the second verification stage. If the accuracy verification meets the standard, the updated digital twin is officially deployed to replace the original model in the scheduling simulation and verification business operation, and the entire process operation log and notification of this change update are recorded.
[0257] (1) Based on the CIM model synchronization mechanism, the present invention uses the SG-CIM State Grid standard model interaction protocol to realize automatic synchronization of power grid anomaly information, topology reconstruction and parameter refresh, solving the defects of manual input, delayed update and topology disorder in traditional models. The specific implementation process is as follows: ① Anomaly signal capture: Access the equipment anomaly work order of the power grid resource management system, identify three types of anomaly events: equipment commissioning, decommissioning and parameter modification, and generate anomaly messages with unique codes; ② CIM model mapping and parsing: Based on the CIM semantic parser, parse the equipment number, voltage level, connection node, impedance parameter and switch status fields in the anomaly message, and establish physical equipment and ③ Local topology partitioning: Using the adjacency matrix topology partitioning algorithm, the topology connectivity domain where the moving equipment is located is locked, and only the topology nodes of the associated bus, branch, and circuit breaker are decoupled and partitioned without changing the topology of the non-associated network structure; ④ Batch parameter assignment: The equipment nameplate parameters, test parameters, and real-time operating parameters are batch written into the equipment attribute fields of the twin model, and electrical parameters such as line impedance, transformer ratio, unit inertia constant, and reactive power limit are updated synchronously; ⑤ Model consistency verification: The topology connectivity matrix before and after the update is compared to verify the node connection relationship and electrical boundary consistency. The topology connectivity error is controlled within 0.5%. After the verification is passed, the topology and parameters are solidified.
[0258] (2) Transferring simulation experience using transfer learning technology: To avoid the problems of retraining the model after equipment parameter changes, high computing power consumption, and slow convergence speed, parameter transfer learning method is used to transfer the simulation experience of the old model. The specific implementation process is as follows: ① Source domain and target domain division: The mature high-precision model before the change is defined as the source domain model, and the model whose topology remains unchanged and only the parameters change after the change is defined as the target domain model; ② Feature parameter extraction: The convolution feature extraction algorithm is used to extract the damping characteristics, load response characteristics, power flow distribution characteristics, and transient oscillation characteristics in the source domain model. Four types of general simulation experience features are identified; ③ Weight transfer mapping: a feature mapping matrix is constructed to transfer the training weights, convergence thresholds, and simulation iteration step sizes of the source domain model to the target domain model, freezing the general underlying network parameters; ④ Adaptive fine-tuning: using measured steady-state power flow and voltage measurement data after the change as constraints, a small-sample gradient descent algorithm is used to fine-tune the surface weights of the target domain model to eliminate model offset caused by parameter changes; ⑤ Experience solidification and storage: the transferred model simulation experience is stored in the model knowledge base to achieve rapid reuse of similar equipment after changes, improving the model convergence speed by more than 60%.
[0259] The target domain model construction and training includes: the target domain model is defined as follows: under the same topology, only the line impedance parameter changes from Z... old Change to Z new The twin model has a network structure that is completely identical to the source domain. Training is not started from scratch, but rather fine-tuned based on transfer-based weights.
[0260] Among them, convolutional feature extraction algorithms include:
[0261] 1. Input: Transform the voltage amplitude, phase angle, active power, and reactive power time series obtained from the source domain model simulation under typical operating conditions into a two-dimensional image matrix. For example, fold the 24-hour voltage curve into a 24×24 pixel image.
[0262] 2. Feature Extraction: The image from step 1 is input into the first three convolutional layers of a pre-trained VGG-16 network (VGG, Visual GeometryGroup, is a deep convolutional neural network architecture). These three convolutional layers learn general features such as power flow distribution characteristics (edge information) and transient oscillation characteristics (texture information), and output a fixed-dimensional feature vector. .
[0263] The weighted transition mapping includes:
[0264] 1. Construct the mapping matrix Establish the weights of the last fully connected layer in the source domain model. To target domain model weights The mapping relationship. The mapping function is... ; This is the target domain fully connected layer weight matrix, which is the weight parameter of the last fully connected layer of the target domain model. This is the target value that needs to be calculated, representing the simulation experience that the new model should inherit. The target domain feature matrix consists of simulation features extracted from the target domain model (such as power flow distribution features, transient oscillation features, etc.). The weight mapping matrix is the transformation matrix that maps the feature space of the source domain to the feature space of the target domain. Simulation feature matrix extracted for source domain model The inverse matrix.
[0265] 2. Migration Method: Freeze the front of the target domain model The network parameters are adjusted so that they directly inherit the simulation experience from the source domain model. Only the last two layers of the network are randomly initialized and retrained.
[0266] (3) Initialization of the behavioral model of newly commissioned equipment using the similarity date matching method: For newly commissioned electrical equipment without historical operating data, the weighted grey relational degree similarity date matching algorithm is used to initialize the behavioral model. The specific implementation process is as follows:
[0267] ① Feature Factor Screening: Extract electrical features of the new equipment, including five inherent features: equipment capacity, access voltage level, access area, impedance characteristics, and regulation response rate; ② Historical Sample Library Construction: Establish a historical operation sample library of existing equipment of the same voltage level, area, and model, including annual time-series output, load fluctuation, voltage response, and meteorological correlation data; ③ Grey Relationship Calculation: Construct a multi-dimensional feature weighted matrix to calculate the correlation between the new equipment and existing equipment in the sample library; ④ Optimal Similar Day Screening: Select the top 5 similar day time-series data in terms of correlation, overlay meteorological and load condition weights, and synthesize an initial behavior curve; ⑤ Model Initialization Assignment: Import the synthesized behavior curve into the twin model of the new equipment, complete the initialization of excitation regulation, ramp response, and reactive power voltage regulation dynamic parameters, and optimize the model accuracy month by month with subsequent iterations of real operation data until the model error is stable at ≤3%.
[0268] The weighted grey relational degree calculation includes:
[0269] 1. Construct the feature matrix: new device feature vector Where m=5 (five inherent characteristics), the feature vector of the existing equipment library .
[0270] 2. Data normalization: Initialization method is used. Eliminate the influence of dimensions.
[0271] 3. Calculate the correlation coefficient: Calculate the correlation coefficient between the new equipment and the first... The existing equipment in the first Correlation coefficients on each feature:
[0272]
[0273] in, The resolution coefficient is usually set to 0.5; The original value of the kth feature of the new device; The normalized value of the k-th feature of the new device; The original value of the k-th feature of the i-th existing equipment; The normalized value of the k-th feature of the i-th existing device; m represents the feature dimension, the total number of feature parameters selected for each device. In this invention, m=5 (five inherent features: capacity, voltage level, access area, impedance characteristics, and regulation response rate); This represents the correlation coefficient, specifically indicating the relationship between the new device and the first... The existing equipment in the first The similarity on a feature is expressed as follows: the closer the value is to 1, the more similar the feature is; the closer the value is to 0, the greater the difference is.
[0274] 4. Calculate the weighted correlation degree: ,in, The weight is set for the k-th feature. For example, the weight of the access voltage level is set to 0.3, and the weight of the equipment capacity is set to 0.25. The weight setting is determined based on expert experience and the analytic hierarchy process.
[0275] 5. Select similar days: Select relevance. The top 5 existing devices were used as initialization templates based on their historical operating data.
[0276] To address the shortcomings of existing technologies, this invention constructs a logical closed loop of data-driven model building, collaborative simulation, closed-loop verification, and dynamic evolution, wherein:
[0277] Constructing a high-fidelity digital twin (S1): By fusing multi-source data and using a transparent electromechanical-electromagnetic hybrid simulation engine, a twin is constructed, providing a virtual image highly synchronized with the physical power grid for all subsequent steps, and outputting an executable, high-precision digital twin model (the digital twin model and twin model in this invention both correspond to the digital twin).
[0278] Establish a multi-timescale collaborative simulation mechanism (S2): After constructing the digital twin, address the issue of how to simulate in a way that covers the entire scheduling business. Establish a three-level simulation system: day-ahead, intraday, and real-time, and design a two-way data interaction and conflict coordination strategy to break down time scale barriers. Specifically, the model constructed in S1 serves as the carrier for the S2 simulation, and the output of S2 (e.g., day-ahead plan, intraday correction) provides S3 with specific strategies and boundary conditions to be verified.
[0279] Execution of scheduling strategy pre-simulation (S3): The scheduling plan generated in S2 is used to conduct multi-scenario, refined sandbox simulation on the twin of S1. Specifically, the scheduling strategy of S2 is received as input, the simulation engine of S1 is invoked, and through transparent scenario generation, traversal simulation and risk identification algorithms, a set of verified safety policies is output.
[0280] In-process online verification of the scheduling strategy (S4): The strategy verified in S3 is issued for execution. By quantifying the deviation index and using an adaptive filtering algorithm, a real-time closed loop of execution-monitoring-correction is formed to ensure that the virtual image is consistent with the actual actions. Specifically, the safety strategy output by S3 is the basis for the execution of S4, and the real-time measurement data of S4 is used in turn to correct the twin model of S1.
[0281] Post-execution evaluation and optimization of scheduling strategy (S5): After the strategy is executed, the running experience is solidified into model parameters through four-dimensional evaluation and optimization algorithms. Specifically, the actual execution data recorded in S4 is the input of S5. The evaluation results and optimized parameters of S5 will be fed back to S1 and S2 for model updates and strategy generation in the next cycle.
[0282] Achieving dynamic updates of the digital twin (S6): The physical power grid is constantly changing, and the digital twin must evolve synchronously. When equipment changes, an incremental update mechanism is triggered to avoid model mismatch. This step provides a long-term maintenance and update mechanism for S1, ensuring that the twin maintains high fidelity throughout its entire lifecycle.
[0283] The above-described solution of the present invention will be further explained below with reference to specific embodiments.
[0284] This embodiment provides a full-process simulation and verification method for a dispatch automation system based on digital twins, which is applied to a provincial power grid dispatch control center (high-voltage main grid scenario).
[0285] Step 1: Digital Twin Construction: Based on the provincial power grid's CIM model, and integrating multi-source information such as SCADA real-time data (approximately 200,000 monitoring points), PMU synchronous phasor data (approximately 500 monitoring points), equipment ledger data, and meteorological data, a digital twin covering the province's 500kV / 220kV main grid and major power plants and substations is constructed. The digital twin is deployed on the simulation server cluster in the dispatch center, possessing electromechanical transient simulation capabilities of over 2000 nodes and electromagnetic transient simulation capabilities of up to 200 nodes.
[0286] Step 2, Multi-timescale co-simulation:
[0287] Configure three-level simulation parameters:
[0288] The current simulation uses 96 time periods (15-minute intervals) to simulate the operating trend over the next 24-72 hours.
[0289] Intraday simulation: 12 rolling cycles (5-minute intervals) to simulate the refined operation of the next 1-4 hours;
[0290] Real-time simulation: Steps per second, performing advanced simulation of AGC commands for tens of seconds to several minutes;
[0291] A bidirectional data exchange interface is established between the three levels of simulation to enable real-time data-driven back-correction of the medium- and long-term simulation model, ensuring the consistency of scheduling strategies in terms of timing. The tie-line plan and unit combination results from the day-ahead simulation serve as the boundary conditions for the intraday simulation; the prediction and correction results from the intraday simulation serve as the initial state for the real-time simulation; and the control effects from the real-time simulation are fed back to the intraday simulation to adjust the control strategies for subsequent periods.
[0292] Step 3: Pre-dispatch strategy simulation and deduction: Receive the day-ahead power generation plan drafted by the dispatcher and generate 10 typical operating scenarios (including typical daily load scenario, holiday load scenario, large-scale renewable energy generation scenario, small-scale renewable energy generation scenario, thunderstorm weather scenario, high-temperature weather scenario, and N-1 fault scenario). Simulation and deduction of each scenario are performed in a digital twin, revealing a power flow exceeding the limit risk at a certain 500kV section under the large-scale renewable energy generation scenario. The system marks the risk and provides unit output adjustment suggestions; after modification, the risk is eliminated. Compared to the single fault scenario of existing technologies, this embodiment has a wider scenario coverage and is adapted to complex grid operating conditions.
[0293] Step 4: Real-time Online Verification of Scheduling Strategy: The revised day-ahead plan is issued and executed. During actual operation the following day, the system collects measurement data in real time and dynamically compares it with the simulated trajectory. At 10:15, a photovoltaic power station experienced a sudden drop in output due to cloud cover, and the deviation between the actual output curve and the simulated trajectory exceeded the threshold. The system triggered an online correction mechanism, using adaptive unscented Kalman filtering to correct the photovoltaic output model parameters and resimulating the remaining operating conditions. It was found that the output of a certain hydropower unit needed to be increased to maintain power balance. The system pushed the correction suggestion to the dispatcher, who adopted and implemented it, thus avoiding the risk of frequency exceeding limits.
[0294] Step 5: Post-event evaluation and optimization of the scheduling strategy: After the day's operation concludes, the system collects all actual operational data and conducts a comprehensive evaluation from dimensions such as safety (N-1 pass rate, voltage margin, etc.), economic efficiency (electricity purchase cost, grid loss, etc.), and renewable energy absorption rate (wind power / solar power utilization rate). The evaluation results show that the renewable energy absorption rate is higher than the monthly average. The system records the successful experience of this scheduling and optimizes the parameters of the next day's scheduling strategy generation model using reinforcement learning algorithms.
[0295] Step 6: Dynamic Update of the Digital Twin: Two weeks later, a new main transformer is put into operation at a 500kV substation. The equipment change management module triggers the digital twin update process, automatically updating the topology and parameters of the new main transformer based on the CIM model synchronization mechanism. The initial behavior model of the new main transformer is initialized using operating data from similar substations with the same model of main transformer. As operating data accumulates, the model gradually becomes more refined.
[0296] Based on the above technical solution of the present invention, the beneficial effects of the present invention also include:
[0297] 1. Achieve closed-loop verification of the entire high-voltage main grid dispatch strategy: Unlike the fragmented simulation architecture of existing distribution networks, this invention constructs a three-stage closed-loop verification mechanism dedicated to the main grid, solving the problem of the separation between simulation and execution, realizing full lifecycle management of dispatch strategies, and thus improving the security of high-voltage main grid dispatch. 2. Establish a multi-timescale bidirectional collaborative simulation mechanism: Compared with the unidirectional time-series simulation logic of existing technologies, this invention realizes bidirectional data interaction of three levels of simulation: day-ahead, intraday, and real-time, correcting medium- and long-term simulation deviations, ensuring the consistency of dispatch strategies across time scales, and adapting to multi-level dispatch operations. 3. Possess quantification-driven dynamic self-correction capability: Overcoming the shortcomings of existing technologies that lack quantification of deviation judgment and require manual intervention for correction, this invention defines a deviation index and combines it with an adaptive filtering algorithm to achieve fully automatic model correction, ensuring high synchronization between the twin model and the physical power grid, thereby improving simulation accuracy. 4. Construct a four-dimensional comprehensive evaluation system: Breaking through the limitations of the single-dimensional evaluation of existing technologies, this invention adds a low-carbon indicator, aligns with the dual-carbon policy, and takes into account the multiple needs of power grid safety, economy, reliability, and low carbon, providing a comprehensive basis for dispatch decisions in new power systems. 5. Achieve lightweight model dynamic updates: Compared to the global reconstruction update method of existing technologies, incremental updates and transfer learning techniques are adopted to reduce the model iteration cost in scenarios of equipment anomalies and adapt to scenarios with frequent grid topology changes. 6. Enhance renewable energy consumption and grid adaptability: Relying on DCGAN extreme scenario generation and high-precision simulation, the impact of renewable energy fluctuations on the high-voltage main grid is accurately predicted, maximizing the renewable energy consumption space. 7. Differentiated scenario adaptation, filling technological gaps: Addressing the gap in existing technologies regarding the lack of dedicated simulation and scheduling technology for high-voltage main grids, customized adaptation to 220kV / 500kV main grids is provided, solving the challenges of multi-station coordination and large-capacity power flow scheduling.
[0298] According to another embodiment of the present invention, a full-process simulation and verification system for a scheduling automation system based on digital twins is also provided, the system comprising:
[0299] The digital twin building module is used to construct a digital twin of the scheduling automation system based on multi-source heterogeneous data from the main network.
[0300] The multi-timescale collaborative simulation module is used to use the digital twin as a simulation running carrier to execute multi-level collaborative simulation strategies, including day-ahead planning simulation, intraday rolling simulation, and real-time control simulation. It also configures bidirectional data interaction links and conflict coordination rules between multi-level collaborative simulation strategies and generates scheduling strategies to be verified.
[0301] The strategy deduction module is used to perform multi-scenario simulation and deduction of the scheduling strategy to be verified on the digital twin of the scheduling automation system, filter out the scheduling strategies with constraint violations and perform iterative corrections to obtain a set of safe scheduling strategies after passing multi-scenario verification;
[0302] The online verification module is used to send the verified safety policy set to the actual scheduling automation system for execution. During the execution of the safety policy set, a quantitative deviation index and an adaptive filtering algorithm are introduced to correct the digital twin parameters, and multi-plant linkage control and deviation compensation operations are executed simultaneously to obtain the actual running dataset of the entire execution process of the safety policy set.
[0303] The evaluation and optimization module is used to evaluate the actual operating dataset from multiple dimensions based on a pre-built comprehensive evaluation index system. The multi-dimensional evaluation results are fed back to the digital twin construction module and the multi-timescale collaborative simulation module for the next round of digital twin update and scheduling strategy generation.
[0304] The model update module is used to monitor the abnormal status of power grid equipment in real time. When an abnormality is detected in the power grid equipment, the corresponding differentiated incremental update mechanism is executed according to the type of abnormality to dynamically update the digital twin parameters in the digital twin construction module.
[0305] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A method for full-process simulation and verification of a scheduling automation system based on digital twins, characterized in that, The method includes: S1. Construct a digital twin of the scheduling automation system based on multi-source heterogeneous data from the main network; S2. Using the digital twin as the simulation operation carrier, execute a multi-level collaborative simulation strategy of day-ahead planning simulation, intraday rolling simulation and real-time control simulation, and configure bidirectional data interaction links and conflict coordination rules between the multi-level collaborative simulation strategies to generate a scheduling strategy to be verified. S3. Simulate and extrapolate the scheduling strategy to be verified on the digital twin of the scheduling automation system in multiple scenarios, filter out the scheduling strategies with constraint violations and perform iterative corrections to obtain a set of safe scheduling strategies that have passed the multi-scenario verification. S4. The verified safety policy set is sent to the actual scheduling automation system for execution. During the execution of the safety policy set, a quantitative deviation index and an adaptive filtering algorithm are introduced to correct the digital twin parameters. Simultaneously, multi-plant linkage control and deviation compensation operations are executed to obtain the actual running dataset of the entire execution process of the safety policy set. S5. Based on the pre-built comprehensive evaluation index system, conduct multi-dimensional evaluation of the actual operation dataset, and feed back the multi-dimensional evaluation results to steps S1 and S2 for the next round of digital twin update and scheduling strategy generation. S6. Monitor the abnormal status of power grid equipment in real time. When an abnormality is detected in the power grid equipment, execute the corresponding differentiated incremental update mechanism according to the type of abnormality to dynamically update the digital twin parameters in step S1.
2. The method for full-process simulation and verification of a scheduling automation system based on digital twins according to claim 1, characterized in that, S2 includes: S21. Use digital twins as the operating platform for day-ahead planning simulation, intraday rolling simulation, and real-time control simulation; S22. Using the first preset time as the resolution, access the predicted and planned boundary data for future time periods, identify potential operational risk events through continuous power flow simulation and safety verification, and generate a risk list to achieve day-ahead plan simulation. S23. Using the second preset time as the rolling cycle, the coordinated operation of multiple plant units is replicated through refined power flow calculation. The scheduling strategy is subjected to full-dimensional safety constraint verification and operation timing verification. After eliminating infeasible strategies, the scheduling instruction sequence is output to achieve intraday rolling simulation. S24. With a step size of seconds or sub-seconds, the pulse regulation command issued by AGC / AVC is received to drive the electromechanical-electromagnetic hybrid simulation engine to simulate the dynamic response process of the unit's excitation regulation and speed control system, quantify and evaluate dynamic indicators and generate dynamic response evaluation results to achieve real-time control simulation. S25. Establish a bidirectional data exchange interface between day-ahead planning simulation, intraday rolling simulation and real-time control simulation. At the same time, add a simulation result conflict coordination strategy to realize conflict coordination between day-ahead planning simulation, intraday rolling simulation and real-time control simulation, and finally output the scheduling strategy to be verified.
3. The method for full-process simulation and verification of a scheduling automation system based on digital twins according to claim 2, characterized in that, S22 includes: S221. Obtain load forecast data, new energy output forecast data, unit maintenance plan data and tie line exchange plan data within a preset time range, divide the simulation period with the first preset time as the resolution, and generate a day-ahead simulation dataset. S222. Using the day-ahead simulation dataset as input, the power flow tracing algorithm and continuous power flow calculation algorithm are used to simulate the medium- and long-term operating trends of the main network in each time period, and generate time-series electrical quantity data for each simulation period. S223, preset power flow overload threshold, bus voltage over-limit threshold, and static stability margin threshold, and generate a set of day-ahead simulation safety constraint thresholds; S224. Based on the day-ahead simulation safety constraint threshold set, perform cross-sectional thermal stability verification and static safety verification on the time-series electrical quantity data for each time period, and identify potential operational risk events in each simulation period, including heavy load cross-sections, voltage bias, power flow exceeding limits and transmission congestion. S225. Mark the occurrence time, location, and over-limit value of each potential operational risk event, and summarize them to form a daily operational risk list.
4. The method for full-process simulation and verification of a scheduling automation system based on digital twins according to claim 3, characterized in that, S222 includes: S2221. Extract the load and output forecast dataset for future periods from the current day-ahead simulation dataset, and set the current simulation period as the initial period. S2222. Based on the preset scheduling plan of each load node, the power growth mode of each load node is determined by the proportional allocation principle, and the allocation coefficient of each generator node participating in load growth is determined according to the economic scheduling plan and the droop coefficient, forming a power growth vector. S2223. Based on the power growth pattern of each load node and the power growth vector of the generator node, construct an extended power flow equation that includes load level scalar parameters, and solve the extended power flow equation to obtain the power flow solution for the current time period. S2224. Using the power flow solution as the current solution, the tangent vector is calculated using the tangent prediction method, and the predicted solution for the next load level is predicted. The predicted solution is used as the initial value for iteration and introduced into the local parameterized equation. The Newton-Raphson method is used to iteratively solve the local parameterized equation to obtain the accurate power flow solution for the current period. S2225. Determine whether the power flow calculation for all simulation periods has been completed. If not, return to step S2223 to continue the power flow calculation for the next period. If completed, output the timing electrical quantity data corresponding to the accurate power flow solution for each simulation period.
5. The method for full-process simulation and verification of a scheduling automation system based on digital twins according to claim 4, characterized in that, S23 includes: S231. Using the second preset time as the rolling cycle, superimpose real-time meteorological correction data, load fluctuation data and real-time deviation data of new energy output to generate an intraday rolling simulation dataset. S232. Using the intraday rolling simulation dataset as boundary conditions, the power flow calculation results of the previous rolling cycle are used as the initial values for iteration, and a preset adjustment variable is introduced to expand and correct the preset Jacobian matrix. S233. Iteratively solve the extended and corrected Jacobian matrix to obtain the refined power flow distribution under the current rolling cycle multi-plant cooperative operation state. S234. Based on the refined power flow distribution, the current scheduling strategy is sequentially checked for steady-state power flow, short-circuit current and transient stability. The unit ramping constraints, unit minimum start-up and shutdown time constraints, cross-sectional transmission limit constraints and reactive power reserve constraints are checked item by item to generate the constraint check results of the current scheduling strategy. S235. Simulate the timing of multi-plant linkage operations involved in the current scheduling strategy to verify the engineering feasibility of the operation sequence, action interval and adjustment range. S236. Summarize the constraint verification results and engineering feasibility verification results of the current scheduling strategy, identify and eliminate infeasible scheduling strategies, and output a sequence of scheduling instructions that has been verified for safety.
6. The method for full-process simulation and verification of a scheduling automation system based on digital twins according to claim 1, characterized in that, S3 includes: S31. Establish a standardized set of operation scenarios adapted to the high-voltage main grid; S32. Write the scheduling strategy to be verified into the digital twin simulation port, and initialize the simulation boundary conditions of each running scenario with the standardized running scenario set as the simulation scope. S33. Load normal scenarios and extreme disturbance scenarios from the standardized operation scenario set in order of priority, and perform topology locking, parameter assignment and boundary solidification operations on each loaded simulation scenario. S34. In the normal scenario after loading is completed, the Newton-Raphson method is used to simulate and solve the scheduling strategy to be verified; in the extreme disturbance scenario after loading is completed, the electromechanical-electromagnetic hybrid simulation engine is used, and transient simulation is performed on the scheduling strategy to be verified according to the preset fault duration and fixed simulation step size. S35. During the simulation, key measurement point data are collected in real time at a uniform sampling frequency, and interpolation fitting technology is used to generate the simulation trajectory for each type of scenario. S36. Compare the simulation trajectory of each scenario with the preset safety threshold in each time period, and mark the scheduling strategy that violates the constraint based on the comparison results; S37. Use the interior point optimization algorithm to iteratively correct the scheduling strategy with constraint violations, rewrite the corrected scheduling strategy into the digital twin simulation port, return to step S34 to perform simulation verification, until the simulation trajectories of all scenarios meet the preset safety threshold, and output a set of safe scheduling strategies that have passed multi-scenario verification.
7. The method for full-process simulation and verification of a scheduling automation system based on digital twins according to claim 6, characterized in that, The method of generating simulation trajectories for each scenario using interpolation fitting techniques includes: The sampled values of each electrical quantity at discrete time points are extracted from the real-time acquisition data of key measurement points to form a discrete simulation dataset. Using the time interval between two adjacent discrete time points in the discrete simulation dataset as the interpolation unit, the simulation period is divided into several interpolation intervals that are connected end to end. A cubic polynomial with time as the independent variable is constructed on each interpolation interval. By applying function value constraints and derivative continuity constraints to the cubic polynomial, the undetermined coefficients corresponding to each interpolation interval are solved, and the polynomial expression for each interpolation interval is obtained. By concatenating the polynomial expressions of the same electrical quantity in each interpolation interval in chronological order, the continuous trajectory curve of the current electrical quantity in the simulation period is obtained. By traversing all electrical quantities in the current scenario, the simulation trajectory of the current scenario is obtained. The simulation trajectory generation process is repeated for each type of scenario to obtain the simulation trajectory corresponding to each type of scenario.
8. The method for full-process simulation and verification of a scheduling automation system based on digital twins according to claim 1, characterized in that, S4 includes: S41. The verified security policy set is sent to the actual scheduling automation system, and measurement data during the execution of the security policy set is collected in real time. S42. Compare the actual running trajectory generated based on the measurement data during the execution process with the simulation trajectory of the digital twin point by point, and calculate the deviation between the actual running trajectory and the simulation trajectory. S43. When the deviation exceeds the preset threshold, the digital twin parameters are corrected using an adaptive filtering algorithm so that the operating state of the digital twin is consistent with the actual operating state of the power grid. S44. For multi-plant coordinated control operations, the plant execution status feedback data of each plant is collected synchronously according to the cross-sectional level. The plant execution status feedback data is compared with the preset operation sequence template to generate the execution lag of each plant. S45. When any power station experiences execution lag, the compensation adjustment amount of the adjacent power station is calculated based on the overall power balance of the section where it is located, and the adjacent power stations are linked to issue compensation instructions and update the power station execution status feedback data. S46. Summarize the measurement data and real-time updated plant execution status feedback data throughout the entire execution process of the security policy set to form the actual operation dataset of the entire security policy set execution process.
9. The method for full-process simulation and verification of a scheduling automation system based on digital twins according to claim 1, characterized in that, S6 includes: S61. Monitor the abnormal status of power grid equipment in real time. When abnormality of power grid equipment is detected, identify the type of abnormality event and generate an abnormality message with a unique code. S62. Based on the semantic parser, analyze the key fields of equipment anomaly in the anomaly message, and establish the mapping relationship between power grid equipment and digital twin based on the key fields of equipment anomaly. S63. Based on the mapping relationship, the location of the abnormal equipment is locked. The adjacency matrix topology splitting algorithm is used to extract the associated bus, branch and circuit breaker nodes, and local topology decoupling is performed on the current connected domain. S64. Within the local topology range formed by decoupling, write all parameters of the abnormal equipment into the attribute fields of the corresponding equipment in the digital twin, and update the associated electrical parameters synchronously. S65. Compare the topology connectivity matrix before and after the update, verify the consistency of node connection relationships and electrical boundaries, and solidify the topology and parameters after the verification is passed, thus realizing the initial update of the digital twin parameters. S66. After the initial update of the digital twin parameters, a differentiated incremental post-processing mechanism is executed according to the type of abnormal event to achieve dynamic updating of the digital twin parameters.
10. The method for full-process simulation and verification of a scheduling automation system based on digital twins according to claim 9, characterized in that, S66 includes: S661. If the change type is parameter equipment change, then use parameter transfer learning technology to transfer the simulation parameters of the equipment before the change to the changed equipment. S662. If the change type is new equipment commissioning, extract the feature parameters of the newly commissioned equipment and construct the feature vector of the newly commissioned equipment. At the same time, extract the feature parameters of the same type of equipment already commissioned in the stock equipment library and construct the feature vector of the stock equipment. S663. Normalize the feature vectors of newly commissioned equipment and existing equipment respectively, and calculate the correlation coefficient between the newly commissioned equipment and each existing equipment after normalization. S664. The correlation coefficients are weighted to obtain the weighted correlation degree between the current stock equipment and the newly commissioned equipment. All stock equipment is sorted from high to low according to the weighted correlation degree, and a preset number of stock equipment is selected as similar equipment. S665. Extract historical operating data of similar equipment and use it as the initialization data template for the digital twin of the newly commissioned equipment to achieve the initialization of the digital twin of the newly commissioned equipment.
Citation Information
Patent Citations
Simulation method, device and electronic equipment for resilient support of multiple microgrids to distribution network
CN116911076B
Method and system for controlling digital twinborn body data of regional station of power system
CN117150293A
Power grid information control method based on digital twinborn model
CN117748733A
SVG power grid dynamic networking and simulation system combining digital twinning
CN120933922A
Power grid intelligent planning simulation platform based on AI and digital twinning
CN121350961A
