Virtual power plant multi-level architecture model resource collaboration method and system

Abstract

The invention discloses a virtual power plant multi-level architecture model resource collaboration method and system, and the method comprises the steps: carrying out the systematic modeling of distributed resources based on a resource heterogeneity compatibility criterion; a space-time decoupling mechanism of information interaction between hierarchies is adopted, and interaction of a multi-level framework is carried out; the distributed resource physical layer is used for performing unified abstract and standardized access on heterogeneous equipment so as to execute a perception-conversion-execution function; the regional aggregation agent layer executes dynamic aggregation and optimal scheduling of heterogeneous resources in a region through a distributed decision-making mechanism; the global coordination optimization layer executes cross-region and cross-time scale resource coordination optimization; the dynamic communication topology model performs inter-level interaction through an event triggering information updating mechanism and a communication delay compensation algorithm; through systematic modeling of the multi-dimensional fusion framework, the collaborative operation efficiency of different types of distributed resources is remarkably improved, and the system stability during large-scale heterogeneous resource access is ensured.

Description

Technical Field

[0001] This invention belongs to the field of virtual power plant technology, and in particular relates to a resource coordination method and system for a multi-level architecture model of a virtual power plant. Background Technology

[0002] In recent years, with the rapid development of renewable and distributed energy resources, Virtual Power Plants (VPPs), as a key technology for integrating distributed resources, have played an increasingly important role in power systems. By aggregating dispersed distributed generation, energy storage systems, and controllable loads, VPPs participate in electricity market transactions and grid ancillary services, improving energy utilization efficiency and grid operational flexibility. Simultaneously, with advancements in smart grid and Internet of Things (IoT) technologies, the scale and application scenarios of VPPs are continuously expanding, covering multiple sectors including industry, commerce, and residential, becoming a crucial support for future energy systems.

[0003] Currently, research on virtual power plants mainly focuses on resource aggregation and collaborative control. Traditional virtual power plant architectures typically employ a hierarchical and partitioned model, dividing distributed resources into different levels for management and using centralized or distributed optimization algorithms for resource scheduling. In terms of resource modeling, existing methods mostly use unified equivalent models, such as equivalent generator models or equivalent load models, to standardize different types of distributed resources. Meanwhile, information interaction mechanisms primarily rely on periodic polling or event-triggered methods, with a central controller or regional agents responsible for global state updates and command issuance to achieve coordinated resource operation.

[0004] However, existing virtual power plant architectures exhibit significant compatibility issues when dealing with highly heterogeneous distributed resources. Because different types of resources (such as photovoltaics, energy storage, and flexible loads) have varying dynamic characteristics and control requirements, traditional unified equivalent models struggle to accurately describe their behavioral features, leading to low resource coordination efficiency and impacting the operational performance of virtual power plants. Summary of the Invention

[0005] Purpose of the invention: The purpose of this invention is to provide a resource coordination method for a multi-level architecture model of a virtual power plant that can achieve heterogeneous resource compatibility, efficient information interaction, and low-latency communication control; on the other hand, it provides a resource coordination system for a multi-level architecture model of a virtual power plant.

[0006] Technical solution: The resource coordination method for a multi-level architecture model of a virtual power plant as described in this invention includes:

[0007] Based on the resource heterogeneity compatibility criterion, a systematic model of distributed resources is constructed. A resource classification system based on dynamic characteristic spectrum is established in the physical characteristic dimension. The dynamic response time constant of different resources is standardized and characterized by the feature time constant matrix. Where τ i The matrix represents the dynamic response time constant of the i-th type of resource. This matrix serves as the basic criterion for resource clustering in the hierarchical architecture, ensuring that the dynamic characteristics of resources within each autonomous domain are within a compatible range. A dynamic participation factor adjustment model is constructed at the response mechanism level, and a Lyapunov function is introduced to ensure stability constraints. A three-dimensional decision space including technical performance, economic benefits, and carbon emission intensity is established in the value dimension. The optimal compromise solution set is generated by a non-dominated sorting genetic algorithm, and the upper-level optimization objective is transformed into standardized instructions executable by the equipment layer through a value transfer function.

[0008] A spatiotemporal decoupling mechanism for inter-level information interaction is adopted to facilitate multi-level architecture interaction. In the time dimension, a hybrid mechanism of event triggering and periodic sampling is used. The device layer executes local fast control loops, the cluster layer implements model predictive control, and the system layer runs global optimal power flow calculations. Time consistency is ensured through a time-scale alignment algorithm. In the spatial dimension, a dynamic community reorganization strategy is designed to decouple the physical network into electrically tightly coupled autonomous partitions. Each partition constructs a fully connected information interaction subnet, and the partition boundaries adopt a sparse communication coordination state.

[0009] Based on the standardized representation and standardized instructions, the distributed resource physical layer performs unified abstraction and standardized access for heterogeneous devices to execute the "sensing-conversion-execution" function;

[0010] Based on the autonomous partitions and standardized access, the regional aggregation proxy layer performs dynamic aggregation and optimized scheduling of heterogeneous resources within the region through a distributed decision-making mechanism;

[0011] The global coordination and optimization layer performs cross-regional and cross-timescale resource collaborative optimization based on the spatiotemporal decoupling mechanism of information interaction between the layers and the results of dynamic aggregation and optimized scheduling of heterogeneous resources within the region.

[0012] By using a dynamic communication topology model, combined with an event-triggered information update mechanism and a communication delay compensation algorithm, communication guarantees are provided for information exchange between different levels.

[0013] This invention achieves efficient integration and collaborative operation of heterogeneous resources in a virtual power plant through a multi-level collaborative optimization architecture: First, systematic modeling based on dynamic characteristic spectrum and participation factor model ensures the compatibility of various distributed resources in terms of physical characteristics, response mechanisms, and value dimensions, significantly improving the flexibility and stability of resource scheduling; Second, the spatiotemporal decoupling interaction mechanism reduces communication load while ensuring time consistency through hybrid triggering strategies and community partitioning algorithms; Standardized abstraction at the physical layer enables plug-and-play functionality for heterogeneous devices, distributed decision-making at the regional aggregation layer optimizes local resource scheduling efficiency, and the global coordination layer improves overall economy and low-carbon performance through cross-spatiotemporal optimization; Dynamic communication topology model further enhances the system's robustness to communication delays, ultimately forming a fast-responding, stable, economical, efficient, and low-carbon collaborative control system for a virtual power plant.

[0014] Preferably, the unified abstraction and standardized access of heterogeneous devices includes:

[0015] A modular plug-in architecture is adopted to develop device description files based on the IEC 61850 standard, which encapsulate device control interfaces, communication protocols and operating constraint parameters. The interface heterogeneity and access security issues are solved through adaptive protocol conversion middleware, dynamic capability description models and security isolation mechanisms, providing a standardized access foundation for subsequent data interaction and control.

[0016] Based on the standardized access foundation, a multi-rate hybrid measurement system is constructed, a time-scale alignment service is used to ensure the spatiotemporal consistency of data, and a dynamic data quality assessment model is constructed based on the information entropy theory to perform real-time quantitative assessment and screening of the collected data, providing data support for upper-level decision-making.

[0017] Based on the data support, a dual closed-loop instruction adaptation mechanism is designed to convert the standardized adjustment instructions issued by the cluster layer into device-level control parameters, and to implement rapid control in combination with local measurement signals. At the same time, the physical limits of the device are transformed into hard constraints of the control algorithm.

[0018] Through a modular plug-in architecture and the IEC 61850 standardized description document, plug-and-play access for heterogeneous devices is achieved, significantly improving the flexibility and compatibility of resource access. The adaptive protocol conversion middleware and multi-rate hybrid measurement system ensure data consistency under different communication protocols and data acquisition frequencies, while the dynamic data quality assessment model optimizes measurement reliability. The dual closed-loop command adaptation mechanism ensures accurate execution of upper-layer scheduling commands while combining local fast control and physical constraint protection, which improves response speed and ensures safe operation of equipment, thereby achieving efficient, stable and secure resource control capabilities at the device layer.

[0019] Preferably, the dynamic aggregation and optimized scheduling of heterogeneous resources within the execution area includes:

[0020] An improved spectral clustering algorithm is adopted, which comprehensively considers electrical distance, regulation characteristics and communication topology factors, to aggregate heterogeneous resources in the region into multiple dynamic autonomous units, and an online learning mechanism is introduced to enable the clustering results to adaptively track changes in the system's operating status.

[0021] Based on the dynamic autonomous unit, a two-layer robust optimization framework is constructed. The inner layer handles device-level uncertainties, and the outer layer addresses the impact of inter-cluster interactions. Through scenario reduction technology, the complex optimization problem is transformed into a manageable mixed-integer linear programming problem for optimizing the scheduling of resources within the execution area.

[0022] During the optimized scheduling process, an event-triggered mechanism is used to reduce communication overhead, and the global coordination algorithm is only activated when a critical state exceeds a limit.

[0023] By dynamically dividing autonomous units through an improved spectral clustering algorithm, and comprehensively considering electrical characteristics and communication topology, resource aggregation becomes more flexible and adaptable. Meanwhile, the online learning mechanism ensures that the cluster structure is dynamically optimized according to the system state. The two-layer robust optimization framework effectively addresses device-level uncertainties and the impact of inter-cluster interactions. Combined with scenario reduction technology, it significantly improves computational efficiency and enables the solvability of complex scheduling problems. The event triggering mechanism greatly reduces the communication burden, triggering global coordination only when critical states exceed limits. While ensuring optimization accuracy, it improves system response speed and operational economy, ultimately achieving efficient, reliable, and adaptive regional resource collaborative scheduling.

[0024] Preferably, the execution of cross-regional, cross-timescale resource collaborative optimization includes:

[0025] The scene tree-based prediction-correction optimization module adopts a robust model prediction control method, combines Latin hypercube sampling to generate typical scene sets, and combines real-time state estimation to dynamically correct the optimization trajectory, constructing a spatiotemporally decoupled parallel computing architecture to decompose optimization problems at different time scales.

[0026] The cross-regional collaborative scheduling module constructs a generalized Nash equilibrium model and performs inter-provincial power mutual assistance and ancillary service sharing through the alternating direction multiplier method.

[0027] The carbon-energy co-optimization module embeds carbon emission flow calculations into the optimal power flow model, forming a resource scheduling strategy guided by dynamic carbon prices.

[0028] The scenario tree prediction-correction optimization module realizes multi-timescale collaborative decision-making, and combines robust model prediction and dynamic trajectory correction to significantly improve the adaptability and accuracy of scheduling strategies. The cross-regional collaborative scheduling module is based on the generalized Nash equilibrium model and uses distributed optimization algorithms to efficiently coordinate inter-provincial power mutual assistance and ancillary service sharing, achieving global resource optimization while ensuring regional autonomy. The carbon-energy collaborative optimization module deeply integrates carbon emission flows into the scheduling model and guides low-carbon operation through a dynamic carbon price mechanism, enabling virtual power plants to significantly reduce the system's carbon footprint while meeting energy demand, ultimately forming a global collaborative optimization system that takes into account economy, safety and environmental protection.

[0029] Preferably, the event triggering information update mechanism includes: constructing a dual-criteria triggering function based on "state deviation - communication cost", adopting a three-layer cascaded design, deploying a lightweight triggering judgment module at the device layer, realizing event spatiotemporal aggregation at the cluster layer, and constructing a global event priority queue at the system layer; and automatically switching to a strong triggering mode when the system is in an emergency.

[0030] The communication delay compensation algorithm includes: establishing a three-layer time delay differential game model of device-cluster-system; performing delay compensation through a time delay predictor with a sliding time window, a multi-layer state observer, and an adaptive weight adjustment module; and performing communication-control co-optimization using a time delay compensation model predictive control framework.

[0031] Intelligent event judgment is achieved through a dual-criteria trigger function based on "state deviation - communication cost," and communication resource allocation is optimized using a three-layer cascaded architecture. This significantly reduces redundant communication load while ensuring the real-time nature of critical information. Automatic switching of the strong trigger mode ensures the reliability of system response in emergency situations. The communication delay compensation algorithm effectively suppresses the impact of delay in multi-level interactions through a time-delay differential game model and predictive compensation technology. Combined with an adaptive observer and model predictive control framework, a dynamic balance between communication quality and control performance is achieved, significantly improving the robustness and collaborative control accuracy of the virtual power plant in complex network environments. Ultimately, this constructs an efficient and reliable multi-level information interaction system.

[0032] Secondly, the virtual power plant multi-level architecture model resource collaboration system of the present invention includes:

[0033] The distributed resource physical layer is used for unified abstraction and standardized access of heterogeneous devices. It adopts a modular plug-in architecture, a multi-rate hybrid measurement system, and a dual closed-loop command adaptation mechanism.

[0034] The regional aggregation agent layer, as a hub, includes a resource clustering engine, an optimization decision module, and a coordination and control module, which performs dynamic aggregation and optimized scheduling of heterogeneous resources within the region.

[0035] The global coordination and optimization layer, as the highest decision-making center, includes a scenario tree-based prediction-correction optimization module, a cross-regional collaborative scheduling module, and a carbon-energy collaborative optimization module, which perform cross-regional and cross-time scale resource collaborative optimization.

[0036] The dynamic communication topology model includes an event-triggered information update mechanism and a communication delay compensation algorithm, enabling efficient information exchange between execution levels.

[0037] Through a standardized access mechanism in the distributed resource physical layer, plug-and-play and precise control of heterogeneous devices are achieved; the regional aggregation agent layer, relying on intelligent clustering and distributed optimization, significantly improves the collaborative efficiency and responsiveness of local resources; the global coordination and optimization layer adopts multi-timescale prediction and correction and carbon energy coordination strategies to ensure that the system balances economy and low carbon emissions in cross-regional scheduling; the dynamic communication topology model, through intelligent event triggering and latency compensation technology, reduces communication load while ensuring reliable transmission of key information. Ultimately, a virtual power plant collaborative control system that is fast-responding, stable in operation, economical and efficient, and low-carbon and environmentally friendly is constructed, achieving a comprehensive performance improvement in resource aggregation, optimized scheduling, and information interaction.

[0038] Preferably, the distributed resource physical layer includes:

[0039] The standardized access module is configured as follows: It employs a modular plug-in architecture to develop device description files based on the IEC 61850 standard, uniformly encapsulating the device control interfaces, communication protocols, and operational constraint parameters for photovoltaic inverters, energy storage converters, and flexible loads; it supports online hot-switching of Modbus, DNP3, and IEC 104 protocols through adaptive protocol conversion middleware; it defines key device parameters using XML Schema through a dynamic capability description model; and it achieves secure isolation of the production control system through unidirectional data diodes.

[0040] The multi-rate sensing module is configured to: deploy μPMUs on inverter-type devices to achieve cycle-level electrical quantity acquisition; use minute-level frozen data from smart meters for conventional loads; use second-level monitoring data from meteorological sensors for environmental quantities; ensure spatiotemporal consistency of data through time-scale alignment services; and construct a dynamic data quality assessment model based on information entropy theory.

[0041] The dual closed-loop control module is configured as follows: the outer loop uses a dynamic equivalent algorithm to solve the standardized adjustment commands of the cluster layer into equipment-level control parameters; the inner loop implements millisecond-level fast control based on local measurement signals; the equipment overload protection and insulation withstand limit are transformed into hard constraints of the control algorithm, and the operating point is constrained by the Lagrange multiplier method; and the module switches to a supply guarantee mode based on local rules when communication is interrupted.

[0042] The distributed resource physical layer achieves plug-and-play and protocol compatibility for heterogeneous devices such as photovoltaics, energy storage, and loads through standardized access modules, significantly improving system scalability and interoperability. The multi-rate sensing module adopts differentiated acquisition strategies and time-stamp alignment technology to ensure the spatiotemporal consistency of data from various devices, and significantly improves monitoring reliability when combined with dynamic quality assessment. The dual closed-loop control module achieves accurate conversion of upper-layer commands through dynamic equivalent algorithms, and, together with local rapid control and hard constraint protection, ensures both response speed and device safety. The autonomous supply guarantee mode in the event of communication interruption further enhances system robustness, ultimately constructing a safe, reliable, responsive, and highly compatible device-layer control system.

[0043] Preferably, the regional aggregation proxy layer includes:

[0044] The resource clustering engine is configured to: employ an improved spectral clustering algorithm to form dynamic autonomous units by integrating electrical distance, regulation characteristics, and communication topology factors; and integrate an online learning mechanism to enable clustering results to adaptively track changes in system operating status.

[0045] The optimization decision module is configured to: construct a two-layer robust optimization framework, wherein: the inner layer handles device-level uncertainties; the outer layer addresses the impact of inter-cluster interactions; and the complex optimization problem is transformed into a manageable mixed-integer linear programming problem through scenario reduction techniques.

[0046] The coordination and control module is configured to: employ an event-triggered mechanism to activate the global coordination algorithm only when a critical state is detected to exceed its limits; maintain the parallel autonomy of each sub-region during steady-state operation; construct a decentralized information sharing platform based on blockchain distributed ledger technology; and automatically execute the coordination rules of each agent node through smart contracts.

[0047] The regional aggregation agent layer achieves dynamic autonomous unit division through an intelligent resource clustering engine, enabling resource aggregation to have adaptive adjustment capabilities; the optimization decision module adopts a two-layer robust optimization framework to effectively balance device-level uncertainty and the impact of inter-cluster interactions, significantly improving the reliability and computational efficiency of scheduling decisions; the coordination control module reduces communication burden through an event triggering mechanism, combines blockchain technology to ensure secure and reliable information sharing, and smart contracts enable automatic execution of collaborative rules, improving operational autonomy while ensuring system stability, ultimately forming a flexible, computationally efficient, secure, and reliable regional-level collaborative control system.

[0048] Preferably, the global coordination optimization layer includes:

[0049] The prediction-correction optimization module is configured as follows: it adopts a robust model prediction control method based on scene trees; it generates a typical scene set through Latin hypercube sampling; it dynamically corrects the optimized trajectory by combining real-time state estimation; it constructs a spatiotemporally decoupled parallel computing architecture; and it decomposes the optimization problem at three time scales—day-intraday-real-time—into mutually coordinated sub-problems.

[0050] The cross-regional collaborative scheduling module is configured to: construct a generalized Nash equilibrium model that considers network constraints; and realize inter-provincial power mutual assistance and ancillary service sharing through the alternating direction multiplier method.

[0051] The carbon-energy co-optimization module is configured to: embed carbon emission flow calculations into the optimal power flow model; and form a resource scheduling strategy guided by dynamic carbon prices.

[0052] The global coordination and optimization layer achieves multi-timescale collaborative decision-making through the prediction-correction optimization module, significantly improving the adaptability and accuracy of the scheduling strategy; the cross-regional collaborative scheduling module, based on the generalized Nash equilibrium model, efficiently coordinates inter-provincial power mutual assistance and ancillary service sharing, optimizing global resource allocation while ensuring regional autonomy; the carbon-energy collaborative optimization module deeply integrates carbon emission flows into the scheduling model, guiding low-carbon operation through a dynamic carbon price mechanism, enabling virtual power plants to significantly reduce the system's carbon footprint while meeting energy demand, ultimately forming a global collaborative optimization system that balances economy, safety, and environmental protection.

[0053] Preferably, the dynamic communication topology model includes:

[0054] The event trigger information update module is configured as follows: the trigger condition is defined as a dual-criteria function of state deviation and communication cost, that is, the trigger condition for device i at time k is:

[0055]

[0056] Where x i (k) represents the actual state. The state of the last transmission, σ i δ is the relative threshold coefficient. i The absolute threshold constant is used; a three-layer cascaded design is adopted: the device layer deploys a lightweight trigger judgment module; the cluster layer realizes the spatiotemporal aggregation of trigger events; the system layer builds a global event priority queue; and it automatically switches to strong trigger mode in the event of a system emergency.

[0057] The communication delay compensation module is configured to: establish a three-layer delay differential game model of device-cluster-system, and define the communication delay from layer i to j as τ. ij (t), whose dynamic characteristics satisfy:

[0058] τ ij (t)=f(η ij,ω ij ,Λ net )

[0059] Where η ij ω is the path congestion coefficient. ij Represents the link quality factor, Λ net The network topology parameters are characterized; communication delay is predicted by a delay predictor based on a sliding time window; state estimation is performed using a multi-layer state observer; control command weights are corrected by an adaptive weight adjustment module; and a model predictive control framework with delay compensation is introduced.

[0060] The dynamic communication topology model significantly reduces system communication load through an intelligent event triggering mechanism, optimizing network resource utilization while ensuring real-time transmission of critical information. The automatic switching between the multi-layer cascaded architecture and the strong triggering mode ensures the reliability of system response in emergency situations. The delay compensation module effectively suppresses the impact of communication delay through differential game model and predictive control technology, and achieves dynamic optimization of control accuracy by combining adaptive weight adjustment. Ultimately, it constructs a highly efficient, reliable, responsive, and robust multi-level information interaction system, providing a solid communication guarantee for the safe and stable operation of the virtual power plant.

[0061] Thirdly, the present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program capable of being loaded by the processor and executing the virtual power plant multi-level architecture model resource coordination method.

[0062] Fourthly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the resource coordination method of the virtual power plant multi-level architecture model.

[0063] Beneficial Effects: Compared with existing technologies, this invention has the following significant advantages: 1. Through systematic modeling of a multi-dimensional fusion framework, the collaborative operation efficiency of different types of distributed resources is significantly improved, ensuring system stability when large-scale heterogeneous resources are accessed; 2. The innovative spatiotemporal decoupling mechanism effectively reduces system communication load, greatly improves control command transmission speed, and reduces network bandwidth usage; 3. The modular architecture and multi-rate measurement system enable rapid access and accurate monitoring of various energy devices, ensuring the system's basic operational capabilities under abnormal conditions; 4. Advanced clustering algorithms and distributed decision-making mechanisms enhance the autonomy and adaptability of regional resource scheduling, strengthening the system's ability to cope with extreme events; 5. The multi-timescale collaborative optimization framework enables efficient allocation of cross-regional resources, promoting the transformation of system decision-making mode towards proactive prediction; 6. Intelligent event triggering and latency compensation mechanisms ensure control accuracy under complex network conditions, significantly improving system stability during communication anomalies. Attached Figure Description

[0064] Figure 1 This is a schematic diagram of the method flow of the present invention;

[0065] Figure 2 This is a schematic diagram of the system structure of the present invention. Detailed Implementation

[0066] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0067] like Figure 1 As shown, this embodiment of the invention provides a resource coordination method for a multi-level architecture model of a virtual power plant, including the following steps:

[0068] S1. Based on the resource heterogeneity compatibility criterion, systematically model distributed resources;

[0069] In terms of physical characteristics, a resource classification system based on dynamic characteristic spectra is established, and the dynamic response time constants of different resources are standardized and characterized by a feature time constant matrix.

[0070]

[0071] Where τ i This matrix represents the dynamic response time constant of the i-th type of resource. It serves as the basic criterion for resource clustering in the hierarchical architecture, ensuring that the dynamic characteristics of resources within each autonomous domain are within a compatible range.

[0072] At the response mechanism level, a dynamic participation factor adjustment model is constructed, and a Lyapunov function is introduced to ensure stability constraints and achieve adaptive weight allocation.

[0073] In terms of value, a three-dimensional decision space including technical performance, economic benefits and carbon emission intensity is established at the cluster layer. The optimal compromise solution set is generated through a non-dominated sorting genetic algorithm, and a value transfer function is designed to transform the upper-level optimization objective into standardized instructions that can be executed at the device layer.

[0074] S2. Employ a spatiotemporal decoupling mechanism for information interaction between layers to achieve efficient information interaction in a multi-layered architecture;

[0075] The time dimension decoupling adopts a hybrid mechanism of event triggering and periodic sampling. The device layer executes a fast control loop based on local measurement, and only triggers event reporting when the state exceeds the limit. The cluster layer implements model predictive control, and the system layer runs global optimal power flow calculation. The time consistency of cross-level decisions is ensured through a time-scale alignment algorithm.

[0076] The spatial dimension decoupling design dynamic community reorganization strategy divides the physical network into electrically coupled autonomous partitions. Each partition builds a fully connected information interaction subnet, and the partition boundaries achieve state coordination through sparse communication.

[0077] S3, the distributed resource physical layer, performs unified abstraction and standardized access for heterogeneous devices, realizing the "sensing-conversion-execution" function;

[0078] A modular plug-in architecture is adopted to develop device description files based on the IEC 61850 standard, which encapsulate device control interfaces, communication protocols and operating constraint parameters, and solve the interface heterogeneity problem through adaptive protocol conversion middleware, dynamic capability description model and security isolation mechanism.

[0079] A multi-rate hybrid measurement system is constructed, different measurement methods are used for different devices, and the spatiotemporal consistency of data is ensured through time-scale alignment service. At the same time, a dynamic data quality assessment model is constructed based on information entropy theory.

[0080] A dual closed-loop command adaptation mechanism is designed to convert standardized adjustment commands issued by the cluster layer into device-level control parameters and implement rapid control based on local measurement signals. At the same time, the physical limits of the device are transformed into hard constraints of the control algorithm.

[0081] S4. The regional aggregation agent layer realizes the dynamic aggregation and optimized scheduling of heterogeneous resources within the region through a distributed decision-making mechanism.

[0082] The resource clustering engine adopts an improved spectral clustering algorithm, which comprehensively considers factors such as electrical distance, regulation characteristics and communication topology to form dynamic autonomous units, and introduces an online learning mechanism to enable the clustering results to adaptively track changes in the system's operating status.

[0083] The optimization decision module constructs a two-layer robust optimization framework. The inner layer handles device-level uncertainties, while the outer layer addresses the impact of inter-cluster interactions. Through scenario reduction technology, the complex optimization problem is transformed into a manageable mixed-integer linear programming problem.

[0084] The coordination and control module uses an event-triggered mechanism to reduce communication overhead, and only starts the global coordination algorithm when a critical state is detected to exceed the limit.

[0085] S5, the global coordination and optimization layer, enables resource collaborative optimization across regions and time scales;

[0086] The scene tree-based prediction-correction optimization module adopts a robust model predictive control method. It generates a typical scene set through Latin hypercube sampling and dynamically corrects the optimization trajectory by combining real-time state estimation. It constructs a spatiotemporally decoupled parallel computing architecture to decompose optimization problems at different time scales.

[0087] The cross-regional collaborative scheduling module constructs a generalized Nash equilibrium model that considers network constraints, and realizes inter-provincial power mutual assistance and ancillary service sharing through the alternating direction multiplier method;

[0088] The carbon-energy co-optimization module embeds carbon emission flow calculations into the optimal power flow model to form a resource scheduling strategy guided by dynamic carbon prices.

[0089] S6. The dynamic communication topology model achieves efficient information interaction between different levels through an event-triggered information update mechanism and a communication delay compensation algorithm.

[0090] Based on the event-triggered information update mechanism, a dual-criteria trigger function of "state deviation degree - communication cost" is constructed. A three-layer cascaded design is adopted. The device layer deploys a lightweight trigger judgment module, the cluster layer realizes the spatiotemporal aggregation of trigger events, and the system layer constructs a global event priority queue and automatically switches to strong trigger mode when the system is in an emergency state.

[0091] The communication delay compensation algorithm establishes a three-layer time delay differential game model of device-cluster-system. Delay compensation is achieved through a time delay predictor based on a sliding time window, a multi-layer state observer, and an adaptive weight adjustment module. Furthermore, a model predictive control framework for time delay compensation is introduced to achieve communication-control coordination.

[0092] Based on a similar inventive concept, this invention also provides a virtual power plant multi-level architecture model resource coordination system corresponding to the aforementioned virtual power plant multi-level architecture model resource coordination method. This system includes a bottom-layer distributed resource physical layer, a middle-layer regional aggregation proxy layer, and a top-layer global coordination and optimization layer. Each layer interacts with the others through a dynamic communication topology model, with interfaces as follows: Figure 2 As shown.

[0093] In terms of system architecture design principles, the resource heterogeneity compatibility criterion, in terms of physical characteristics, standardizes the second-level fluctuation of photovoltaic power generation, the minute-level ramp-up characteristics of wind power, and the millisecond-level power response capability of energy storage through a characteristic time constant matrix. This matrix serves as the basic criterion for resource clustering in the hierarchical architecture, ensuring that the dynamic characteristics of resources within each autonomous domain are within the compatibility range.

[0094] At the response mechanism level, a dynamic participation factor adjustment model is constructed, and a Lyapunov function is introduced to ensure stability constraints, so that rapid energy storage resources receive a higher weight in the transient process, while cogeneration units with energy sustainability play a leading role in the steady-state regulation stage.

[0095] In terms of value, a three-dimensional decision space including technical performance, economic benefits and carbon emission intensity is established at the cluster layer. The optimal compromise solution set is generated through a non-dominated sorting genetic algorithm, and a value transfer function is designed to transform the upper-level optimization objective into standardized instructions that can be executed at the device layer.

[0096] The spatiotemporal decoupling mechanism for information interaction between layers is implemented in the following time dimension: the device layer executes a microsecond-level fast control loop based on local measurements, triggering event reporting only when the state exceeds the limit; the cluster layer implements millisecond-level model predictive control, achieving multi-step look-ahead optimization through a sliding time window; and the system layer runs minute-level global optimal power flow calculations. A time-stamp alignment algorithm ensures the time consistency of cross-layer decisions, and a hybrid synchronization strategy combining PTPv3 and logical clocks is introduced.

[0097] In the spatial dimension, relying on community detection algorithms from graph theory, the physical network is divided into electrically tightly coupled autonomous partitions. Each partition contains a fully connected information-interaction subnet, and partition boundaries are coordinated through sparse communication. A dynamic community reorganization strategy is introduced to adaptively adjust partition boundaries when network topology changes or power flow patterns are detected.

[0098] The distributed resource physical layer adopts a modular plug-in architecture. For different equipment types such as photovoltaic inverters, energy storage converters, and flexible loads, it develops device description files based on the IEC 61850 standard, uniformly encapsulating device control interfaces, communication protocols, and operational constraint parameters. It supports online hot-switching of industrial protocols such as Modbus, DNP3, and IEC 104 through adaptive protocol conversion middleware; defines key device parameters using XML Schema through a dynamic capability description model; and employs a security isolation mechanism, using unidirectional data diodes to ensure boundary security of the production control system.

[0099] In terms of state awareness, μPMUs are deployed on inverter-type high-speed devices to achieve cycle-level electrical quantity acquisition; smart meters freeze data at the minute level on the conventional load side; and environmental quantity monitoring relies on meteorological sensors for second-level refresh. Time-stamp alignment services ensure spatiotemporal consistency of data, and a dynamic data quality assessment model is constructed based on information entropy theory, automatically switching to state estimation mode when a communication anomaly is detected.

[0100] In the execution control phase, a dual-closed-loop command adaptation mechanism is designed. The outer loop receives standardized adjustment commands from the cluster layer and calculates them into device-level control parameters using a dynamic equivalent algorithm. The inner loop implements millisecond-level rapid control based on local measurement signals. Physical limits such as equipment overload protection and insulation withstand capability are transformed into hard constraints for the control algorithm, and the Lagrange multiplier method ensures that the operating point always remains within the feasible region. When an upper-layer communication interruption is detected, the system automatically switches to a supply guarantee mode based on local rules.

[0101] The resource clustering engine of the regional aggregation agent layer adopts an improved spectral clustering algorithm, which comprehensively considers factors such as electrical distance, adjustment characteristics and communication topology to form dynamic autonomous units, and introduces an online learning mechanism to enable the clustering results to adaptively track changes in the system's operating status.

[0102] The optimization decision module constructs a two-layer robust optimization framework. The inner layer handles device-level uncertainties, while the outer layer addresses the impact of inter-cluster interactions. Through scenario reduction techniques, complex optimization problems are transformed into manageable mixed-integer linear programming problems.

[0103] The coordination and control module employs an event-triggered mechanism to reduce communication overhead, activating the global coordination algorithm only when a critical state is detected to exceed limits, maintaining parallel autonomy of each sub-region during steady-state operation. Simultaneously, it utilizes blockchain-based distributed ledger technology to construct a decentralized information-sharing platform, where each agent node automatically executes coordination rules through smart contracts.

[0104] The scenario tree-based prediction-correction optimization module of the global coordination optimization layer adopts a robust model predictive control method. It generates a typical scenario set through Latin hypercube sampling and dynamically corrects the optimization trajectory by combining real-time state estimation. It constructs a spatiotemporally decoupled parallel computing architecture, decomposing the optimization problem at three time scales—day-intraday-real-time—into mutually coordinated sub-problems.

[0105] The cross-regional collaborative scheduling module constructs a generalized Nash equilibrium model that considers network constraints, and realizes inter-provincial power mutual assistance and ancillary service sharing through the alternating direction multiplier method.

[0106] The carbon-energy co-optimization module embeds carbon emission flow calculations into the optimal power flow model, forming a resource scheduling strategy guided by dynamic carbon prices.

[0107] The dynamic communication topology model facilitates inter-level interaction through an event-triggered information update mechanism and a communication delay compensation algorithm.

[0108] The event trigger information update module is configured as follows: the trigger condition is defined as a dual-criteria function of state deviation and communication cost, that is, the trigger condition for device i at time k is:

[0109]

[0110] Where x i (k) represents the actual state. The state of the last transmission, σ i δ is the relative threshold coefficient. i The absolute threshold constant is used; a three-layer cascaded design is adopted: the device layer deploys a lightweight trigger judgment module; the cluster layer realizes the spatiotemporal aggregation of trigger events; the system layer builds a global event priority queue; and it automatically switches to strong trigger mode in the event of a system emergency.

[0111] The communication delay compensation module is configured to: establish a three-layer delay differential game model of device-cluster-system, and define the communication delay from layer i to j as τ. ij (t), whose dynamic characteristics satisfy:

[0112] τ ij (t)=f(η ij ,ω ij ,Λ net )

[0113] Where η ij ω is the path congestion coefficient. ij Represents the link quality factor, Λ net The network topology parameters are characterized; communication delay is predicted by a delay predictor based on a sliding time window; state estimation is performed using a multi-layer state observer; control command weights are corrected by an adaptive weight adjustment module; and a model predictive control framework with delay compensation is introduced.

[0114] The present invention also discloses an electronic device.

[0115] Specifically, the electronic device can be a desktop computer, laptop computer, handheld computer, or cloud server, etc. This computer device may include, but is not limited to, a processor and memory. The processor and memory can be connected via a bus or other means. The processor can be a Central Processing Unit (CPU). The processor can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs) or other programmable logic devices, graphics processing units (GPUs), embedded neural network processing units (NPUs) or other dedicated deep learning coprocessors, discrete gate or transistor logic devices, discrete hardware components, or combinations of the above types of chips.

[0116] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer-executable programs, and modules. The processor executes various functional applications and data processing by running non-transitory software programs, instructions, and modules stored in memory. Memory may include a program storage area and a data storage area. The program storage area may store the control unit and the application program required for at least one function; the data storage area may store data created by the processor, etc. Furthermore, memory may include high-speed random access memory and non-transitory memory. In some embodiments, memory may optionally include memory remotely located relative to the processor, which can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0117] The present invention also discloses a computer-readable storage medium.

[0118] Specifically, the computer-readable storage medium is used to store a computer program, which, when executed by a processor, implements the methods described in the above method implementation.

[0119] Those skilled in the art will understand that all or part of the processes in the methods described above can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk drive (HDD), or solid-state drive (SSD), etc.; the storage medium can also include combinations of the above types of memory.

Claims

1. A virtual power plant multi-level architecture model resource coordination method, characterized in that, Comprise: Based on the resource heterogeneity compatibility criteria, the distributed resources are systematically modeled, a resource classification system based on dynamic characteristic spectrum is established in the physical characteristic dimension, and the dynamic response time constants of different resources are standardized by a characteristic time constant matrix; a dynamic participation factor adjustment model is constructed at the response mechanism level, and Lyapunov function is introduced to guarantee stability constraints; a three-dimensional decision space including technical performance, economic benefit and carbon emission intensity is established in the value dimension, a non-dominated sorting genetic algorithm is used to generate an optimal compromise solution set, and a value transfer function is used to convert the upper optimization target into standardized instructions executable by the device layer; A space-time decoupling mechanism for inter-level information interaction is used to interact with the multi-level architecture. In the time dimension, a hybrid mechanism of event triggering and periodic sampling is used, the device layer executes a local fast control loop, the cluster layer implements model predictive control, and the system layer runs global optimal power flow calculation, and time consistency is guaranteed by a time alignment algorithm; A dynamic community reorganization strategy is designed for spatial dimension decoupling, which divides the physical network into autonomous partitions with close electrical coupling, builds a fully connected information interaction subnetwork within each partition, and uses sparse communication to coordinate the state of the partition boundary; The physical layer of the distributed resources uniformly abstracts and standardizes access to heterogeneous devices based on the standardized representation and standardized instructions to perform the functions of "perception, conversion and execution"; The regional aggregation agent layer performs dynamic aggregation and optimal scheduling of heterogeneous resources within the region based on the autonomous partition and standardized access through a distributed decision-making mechanism; The global coordination and optimization layer performs resource coordination and optimization across regions and time scales based on the results of the space-time decoupling mechanism for inter-level information interaction and the dynamic aggregation and optimal scheduling of heterogeneous resources within the region; A dynamic communication topology model is used in combination with an event-triggered information update mechanism and a communication delay compensation algorithm to provide communication support for information interaction between levels.

2. The virtual power plant multi-level architecture model resource coordination method according to claim 1, characterized in that, The uniform abstraction and standardized access to heterogeneous devices include: A modular plug-in architecture is used to develop a device description file based on the IEC 61850 standard, encapsulate device control interfaces, communication protocols and operating constraint parameters, and solve interface heterogeneity and access security problems through adaptive protocol conversion middleware, dynamic capability description model and security isolation mechanism to provide a standardized access foundation for subsequent data interaction and control; Based on the standardized access foundation, a multi-rate hybrid measurement system is constructed, a time alignment service is used to ensure data space-time consistency, and a dynamic data quality evaluation model is constructed based on information entropy theory to quantitatively evaluate and filter collected data in real time, providing data support for upper-level decision-making; Based on the data support, a double-loop instruction adaptation mechanism is designed to calculate the standardized regulation instructions issued by the cluster layer into device-level control parameters, and implement fast control combined with local measurement signals, while converting the physical limits of the device into hard constraint conditions for the control algorithm.

3. The virtual power plant multi-level architecture model resource coordination method of claim 1, wherein the dynamic aggregation and optimal scheduling of heterogeneous resources within the region comprises: An improved spectral clustering algorithm is adopted to aggregate heterogeneous resources in a region into multiple dynamic autonomous units, considering electrical distance, regulation characteristics and communication topology, and an online learning mechanism is introduced to adaptively track the changes in system operation state; Based on the dynamic autonomous units, a double-layer robust optimization framework is constructed, the inner layer handles device-level uncertainty, and the outer layer deals with the interaction between clusters, and through scenario reduction technology, the complex optimization problem is transformed into a mixed integer linear programming problem that can be processed, which is used to perform optimization scheduling of regional resources; During the execution of the optimization scheduling, an event-triggered mechanism is used to reduce communication overhead, and only when the key state limit is exceeded, the global coordination algorithm is started.

4. The resource coordination method of the virtual power plant multi-level architecture model according to claim 1, wherein the execution of the resource coordination optimization across regions and time scales comprises: A prediction-correction optimization module based on a scenario tree adopts a robust model predictive control method, generates a typical scenario set in combination with Latin hypercube sampling, and dynamically corrects the optimization trajectory in combination with real-time state estimation, to construct a time-space decoupled parallel computing architecture to decompose optimization problems of different time scales; A cross-regional coordinated scheduling module constructs a generalized Nash equilibrium model to perform inter-provincial power exchange and auxiliary service sharing through an alternating direction multiplier method; A carbon-energy collaborative optimization module embeds carbon emission flow calculation into an optimal power flow model to form a resource scheduling strategy guided by a dynamic carbon price.

5. The virtual power plant multi-level architecture model resource coordination method of claim 1, the event trigger information update mechanism comprising: A "state deviation-communication cost" double-criterion trigger function is constructed, and a three-level cascade design is adopted, a lightweight trigger judgment module is deployed at the device layer, event space-time aggregation is realized at the cluster layer, and a global event priority queue is constructed at the system layer; In the case of system emergency, it is automatically switched to a strong trigger mode; The communication delay compensation algorithm comprises: establishing a device-cluster-system three-layer time delay differential game model; performing delay compensation through a sliding time window delay predictor, a multi-layer state observer and an adaptive weight adjustment module; and executing communication-control collaborative optimization by using a model predictive control framework with delay compensation.

6. A virtual power plant multi-level architecture model resource coordination system, characterized in that, It comprises: A distributed resource physical layer for unified abstraction and standardized access of heterogeneous devices, which adopts a modular plug-in architecture, a multi-rate hybrid measurement system and a double-closed-loop instruction adaptation mechanism; A regional aggregation agent layer, as a hub link, contains a resource clustering engine, an optimization decision module and a coordination control module, and executes dynamic aggregation and optimization scheduling of heterogeneous resources in a region; A global coordination optimization layer, as the highest decision hub, contains a scenario tree-based prediction-correction optimization module, a cross-regional coordinated scheduling module and a carbon-energy collaborative optimization module, and executes resource coordination optimization across regions and time scales; A dynamic communication topology model, including an event-triggered information update mechanism and a communication delay compensation algorithm, performs efficient information interaction between levels.

7. The virtual power plant multi-level architecture model resource coordination system according to claim 7, characterized in that, The distributed resource physical layer comprises: The standardized access module is configured to: develop a device description file based on the IEC 61850 standard by using a modular plug-in architecture, uniformly encapsulate device control interfaces, communication protocols and operating constraint parameters of photovoltaic inverters, energy storage converters and flexible loads, support online hot switching of Modbus, DNP3 and IEC 104 protocols through adaptive protocol conversion middleware, define key device parameters by using a dynamic capability description model in the form of XML Schema, and realize safe isolation of a production control system through a unidirectional data diode; The multi-rate perception module is configured to: deploy a muPMU on an inverter device to collect cycle-level electrical quantity data, use frozen data of a smart meter at a minute level on a conventional load, and use second-level monitoring data of a meteorological sensor on an environmental quantity, ensure data space-time consistency through a time alignment service, and construct a dynamic data quality evaluation model based on an information entropy theory; The double-loop control module is configured to: calculate device-level control parameters from cluster layer standardized regulation instructions through a dynamic equivalent algorithm in an outer loop, implement millisecond-level fast control based on local measurement signals in an inner loop, convert device overload protection and insulation tolerance limits into hard constraint conditions of a control algorithm, and switch to a power supply protection mode based on local rules in a communication interruption. 8.The virtual power plant multi-level architecture model resource coordination system according to claim 7, characterized in that, The regional aggregation agent layer comprises: A resource clustering engine is configured to: form dynamic autonomous units by comprehensively considering electrical distance, regulation characteristics and communication topology factors by using an improved spectral clustering algorithm, and make clustering results adaptively track system operating state changes by integrating an online learning mechanism; An optimization decision module is configured to: construct a double-layer robust optimization framework, in which: an inner layer processes device-level uncertainties; and an outer layer deals with interaction influences between clusters; and convert a complex optimization problem into a processable mixed integer linear programming problem through a scenario reduction technique; A coordinated control module is configured to: adopt an event-triggered mechanism, and start a global coordination algorithm only when a key state limit is detected; maintain parallel autonomy of each sub-region in a steady-state operation; construct a decentralized information sharing platform based on a blockchain distributed ledger technology; and realize automatic execution of collaborative rules of each agent node through a smart contract. 9.The virtual power plant multi-level architecture model resource coordination system according to claim 7, characterized in that, The global coordination optimization layer comprises: A prediction-correction optimization module is configured to: adopt a robust model predictive control method based on a scenario tree; generate a typical scenario set through Latin hypercube sampling; dynamically correct an optimization trajectory in combination with real-time state estimation; construct a time-space decoupled parallel computing architecture; and decompose optimization problems at three time scales of day-ahead, intra-day and real-time into mutually coordinated sub-problems; A cross-regional collaborative scheduling module is configured to: construct a generalized Nash equilibrium model considering network constraints; and realize inter-provincial power exchange and auxiliary service sharing through an alternating direction multiplier method; A carbon-energy collaborative optimization module is configured to: embed carbon emission flow calculation into an optimal power flow model; and form a resource scheduling strategy guided by a dynamic carbon price. 10.The virtual power plant multi-level architecture model resource coordination system according to claim 7, characterized in that, The dynamic communication topology model comprises: An event-triggered information updating module is configured to: define a trigger condition as a double-criterion function of state deviation degree and communication cost, that is, a trigger condition of a device i at a time k is where x i (k) is the actual state, is the last sent state, σ i is the relative threshold coefficient, δ i is the absolute threshold constant; a three-layer cascade design is adopted: a lightweight trigger judgment module is deployed at the device layer; spatiotemporal aggregation of trigger events is realized at the cluster layer; a global event priority queue is constructed at the system layer; automatic switching to a strong trigger mode is realized in a system emergency state; The communication delay compensation module is configured to: establish a delay differential game model of a device-cluster-system three-layer, define a communication delay from a layer i to a layer j as τ ij The dynamic characteristics satisfy: where η ij is the path congestion coefficient, ω ij denotes the link quality factor, Λ net characterizes the network topology parameter; the communication delay is predicted by a sliding time window-based delay predictor; state estimation is performed by a multi-layer state observer; the control instruction weight is corrected by an adaptive weight adjustment module; and a model predictive control framework with delay compensation is introduced.

Images