A multi-energy resource aggregation-oriented virtual power plant collaborative scheduling method and system

Abstract

This invention provides a method and system for collaborative scheduling of virtual power plants oriented towards multi-energy resource aggregation, relating to the field of resource scheduling technology. The method includes initializing a schedulable capability map of the virtual power plant, containing power constraints, regulation costs, and carbon emission characteristics of each type of resource; constructing a two-layer collaborative optimization model of the virtual power plant based on the schedulable capability map, determining the scheduling instructions for each resource cluster in the upper layer and the response capabilities of the corresponding clusters in the lower layer; iteratively solving the two-layer collaborative optimization model, allowing the scheduling instructions in the upper layer and the response capabilities in the lower layer to interactively update, thus obtaining a scheduling scheme for the virtual power plant; and dynamically updating the resource weights in the schedulable capability map based on the response performance and reputation of each resource during the execution of the scheduling scheme. Based on this scheme, collaborative scheduling of the virtual power plant's global low-carbon goals and the autonomous response of distributed resources can be achieved, thereby improving the carbon emission reduction efficiency of the multi-energy aggregation system.

Description

Technical Field

[0001] This invention relates to the field of resource scheduling technology, and more specifically, to a method and system for collaborative scheduling of virtual power plants oriented towards the aggregation of multiple energy resources. Background Technology

[0002] A virtual power plant is an intelligent management system that aggregates and coordinates geographically dispersed energy resources such as distributed power sources, energy storage systems, and controllable loads through advanced information communication and control technologies. A virtual power plant does not change the physical connection of each resource, but forms a virtual entity through a software platform that can be uniformly dispatched and participate in the interaction of the electricity market and the power grid, thereby achieving efficient resource utilization and system balance.

[0003] Traditional centralized dispatching methods, in modeling and optimization, abstract distributed photovoltaic, energy storage, and flexible loads within a virtual power plant into idealized, controllable components that unconditionally obey central commands. This ignores the fact that various resources in a multi-energy resource aggregation environment have independent operational goals and interests. For example, energy storage devices need to consider their own lifespan and charging / discharging costs, while adjustable loads must adhere to user comfort and contractual agreements. These autonomous interests are difficult to effectively coordinate and compensate for under mandatory centralized commands. Furthermore, due to data security and privacy concerns, owners of distributed resources are often unwilling to disclose detailed operational status, cost boundaries, and adjustment potential to the central dispatching agency. This not only reduces the enthusiasm for resource response but also leads to significant deviations between dispatching commands and actual dispatchable capacity due to model information distortion, thus affecting the overall reliability and economic efficiency of the virtual power plant. Therefore, how to achieve coordinated dispatching of the virtual power plant's global low-carbon goals and the autonomous response of distributed resources, thereby improving the carbon emission reduction efficiency of multi-energy aggregation systems, has become a challenge for the industry. Summary of the Invention

[0004] This invention provides a method and system for collaborative scheduling of virtual power plants oriented towards multi-energy resource aggregation. It can realize the collaborative scheduling of global low-carbon goals of virtual power plants and autonomous response of distributed resources, thereby improving the carbon emission reduction efficiency of multi-energy aggregation system operation.

[0005] In a first aspect, the present invention provides a virtual power plant collaborative scheduling method for multi-energy resource aggregation, comprising the following steps: The initial virtual power plant includes a map of dispatchable capabilities for each type of resource, including power constraints, regulation costs, and carbon emission characteristics. Based on the schedulable capability map, a two-layer collaborative optimization model for virtual power plants is constructed. The upper layer of the two-layer collaborative optimization model aims to minimize the total carbon emissions of the virtual power plant and generates scheduling instructions for each resource cluster according to the power load demand. The lower layer of the two-layer collaborative optimization model aims to minimize the response deviation to the scheduling instructions of the upper layer and performs distributed collaborative control of the resources within its own cluster to obtain the response capability of the lower layer's own cluster. The two-layer collaborative optimization model is iteratively solved so that the scheduling instructions of the upper layer and the response capabilities of the lower layer are updated interactively, thereby obtaining the scheduling scheme of the virtual power plant. During the execution of the scheduling scheme, the resource weights in the schedulable capability map are dynamically updated based on the response performance and reputation of each resource, which is used for resource scheduling of the virtual power plant in the next scheduling cycle.

[0006] Furthermore, the initialization of the virtual power plant includes a dispatchable capacity map of each type of resource, encompassing power constraints, regulation costs, and carbon emission characteristics. Specifically, this map includes: Acquire historical operating data and physical parameters of distributed photovoltaic, wind turbines, energy storage batteries and interruptible loads in the virtual power plant; Power constraints, adjustment cost-power relationship, and carbon emission intensity coefficient per unit output are constructed for each type of resource. By aggregating and encoding each power constraint, each regulation cost-power relationship, and all carbon emission intensity coefficients, a dispatchable capacity map is obtained.

[0007] Furthermore, the construction of a two-layer collaborative optimization model for the virtual power plant based on the schedulable capability map specifically includes: Using a schedulable capacity map as input, resources are divided into multiple resource clusters based on their adjustment characteristics and geographical location; A higher-level optimization model is constructed. The decision variable of the higher-level optimization model is the total scheduling command of each resource cluster. The objective function is to minimize the total carbon emissions of the virtual power plant. The constraints include load balancing, cluster scheduling capability, and network transmission security. A lower-level optimization model is constructed. After receiving the scheduling instruction, each resource cluster performs distributed optimization with the goal of minimizing the response deviation within the cluster, and feeds back the actual response capability that the corresponding resource cluster can provide to the upper layer. The two-layer collaborative optimization model of the virtual power plant is determined by the upper-layer optimization model and the lower-layer optimization model.

[0008] Furthermore, the upper layer of the two-layer collaborative optimization model aims to minimize the total carbon emissions of the virtual power plant. Based on the power load demand, it generates scheduling instructions for each resource cluster, specifically including: An objective function is established with the carbon emission intensity of each resource cluster as the weight, and then the cluster carbon emission of the virtual power plant is calculated. The power load demand of the virtual power plant is used as an equality constraint, and the power limits provided by each cluster resource in the dispatchable capacity map are superimposed as inequality constraints. Scheduling optimization is performed based on the cluster carbon emissions, the equality constraints, and the inequality constraints. Power scheduling instructions for each resource cluster that meet both optimal carbon emissions and load requirements are output as scheduling instructions for the corresponding resource clusters and sent to the corresponding lower-level cluster controllers.

[0009] Furthermore, in the two-layer collaborative optimization model, the lower layer aims to minimize the response deviation to the scheduling instructions of the upper layer, and performs distributed collaborative control of the resources within its cluster. The specific response capabilities of the lower-layer cluster include: After receiving scheduling instructions from the upper layer of each resource cluster, the cluster master controller decomposes the scheduling instructions into basic allocation values ​​for each resource within each resource cluster. Each resource performs distributed iterative computation based on its local schedulable capabilities to obtain the output resources of each resource cluster, thereby minimizing the deviation between the overall resource cluster and the instructions. After iterative convergence, the total power range and adjustment cost that each resource cluster can actually provide are calculated and fed back to the upper-layer model as the response capability of the lower-level cluster.

[0010] Furthermore, the two-layer collaborative optimization model is iteratively solved, enabling the scheduling instructions of the upper layer to interact and update the response capabilities of the lower layer, thereby obtaining the scheduling scheme of the virtual power plant, which specifically includes: The upper layer of the two-layer collaborative optimization model sends preliminary scheduling instructions to each resource cluster in the lower layer; Each lower-level resource cluster performs distributed optimization based on the corresponding initial scheduling instructions and feeds back the updated cluster response capabilities to the upper layer. The upper layer updates the constraints in the schedulable capability graph based on the feedback information, and re-solves the optimization problem to generate new scheduling instructions; Repeat the above steps until the difference between the upper and lower layer instructions and response capabilities is less than the preset error threshold, and output the virtual power plant scheduling scheme.

[0011] Furthermore, dynamically updating the resource weights in the schedulable capability map based on the response performance and reputation of each resource specifically includes: For each resource, the actual output curve of the resource is collected, and the actual output curve is compared with the scheduling command of the resource to calculate the response accuracy and response speed index of the resource. The resource's reputation score is determined based on its historical response records; The response accuracy, response speed index and reputation score are weighted and fused to obtain the adjustment cost coefficient of the resource in the schedulable capability map, and then the adjustment cost coefficient of each resource in the schedulable capability map is obtained. The resource weights in the schedulable capacity map are updated using various adjustment cost coefficients.

[0012] Secondly, the present invention provides a virtual power plant collaborative scheduling system for multi-energy resource aggregation, used to execute a virtual power plant collaborative scheduling method for multi-energy resource aggregation, comprising a scheduling unit, the scheduling unit comprising: The initialization module is used to initialize the dispatchable capacity map of each type of resource in the virtual power plant, including power constraints, regulation costs, and carbon emission characteristics. The processing module is used to construct a two-layer collaborative optimization model of the virtual power plant based on the schedulable capability map. The upper layer of the two-layer collaborative optimization model aims to minimize the total carbon emissions of the virtual power plant and generates scheduling instructions for each resource cluster according to the power load demand. The lower layer of the two-layer collaborative optimization model aims to minimize the response deviation to the scheduling instructions of the upper layer and performs distributed collaborative control on the resources within its own cluster to obtain the response capability of the lower layer's own cluster. The processing module is also used to iteratively solve the two-layer collaborative optimization model, so that the scheduling instructions of the upper layer and the response capabilities of the lower layer are updated interactively, thereby obtaining the scheduling scheme of the virtual power plant. The execution module is used to dynamically update the resource weights in the schedulable capability map based on the response performance and reputation of each resource during the execution of the scheduling scheme, for resource scheduling of the virtual power plant in the next scheduling cycle.

[0013] Thirdly, the present invention provides a computer device, the computer device including a memory and a processor, the memory storing code, the processor being configured to acquire the code and execute the above-described virtual power plant collaborative scheduling method for multi-energy resource aggregation.

[0014] Fourthly, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described virtual power plant collaborative scheduling method for multi-energy resource aggregation.

[0015] The technical solution provided by this invention has the following beneficial effects: This invention provides a method and system for collaborative scheduling of virtual power plants oriented towards multi-energy resource aggregation. The system initializes a schedulable capability map of the virtual power plant, including power constraints, regulation costs, and carbon emission characteristics of each resource type. Based on this map, a two-layer collaborative optimization model for the virtual power plant is constructed. The upper layer of the model aims to minimize the total carbon emissions of the virtual power plant and generates scheduling instructions for each resource cluster based on power load demand. The lower layer aims to minimize the response deviation to the upper layer's scheduling instructions and performs distributed collaborative control of resources within its cluster to obtain the response capability of the lower layer's cluster. The two-layer collaborative optimization model is iteratively solved, allowing the upper layer's scheduling instructions and the lower layer's response capabilities to interact and update, thereby obtaining a scheduling scheme for the virtual power plant. During the execution of the scheduling scheme, the resource weights in the schedulable capability map are dynamically updated based on the response performance and reputation of each resource for resource scheduling in the next scheduling cycle.

[0016] Therefore, in this invention, during the execution of the scheduling scheme, the resource weights in the schedulable capability map are dynamically updated based on the response performance and reputation of each resource, for resource scheduling of the virtual power plant in the next scheduling cycle. First, by determining the two-layer collaborative optimization model, a dynamic interactive framework combining an upper-layer optimization model and a lower-layer distributed response model is obtained. This allows the global low-carbon scheduling objective of the virtual power plant to be coordinated with the heterogeneity, autonomy, and privacy protection requirements of distributed resources. The upper-layer centralized optimization ensures the minimization of carbon emissions at the system level, while the lower-layer distributed control guarantees efficient autonomous response of resources while adhering to local constraints and interests. The structured collaborative mechanism effectively resolves the command and capability mismatch problem caused by the oversimplification of resource models in traditional centralized scheduling. Through iterative interaction between the upper and lower-layer models, not only can the global optimal solution be mathematically approximated, but the acceptability of scheduling commands in actual execution is also improved. The feasibility and executability lay the foundation for achieving the global carbon emission reduction target. Then, by determining the scheduling scheme, a final power plan set that has been fully iterated and coordinated and updated based on dynamic feedback information can be obtained. This ensures a high degree of consistency between scheduling instructions and the actual response capabilities of each resource cluster, significantly improving the feasibility and execution accuracy of the scheduling scheme. It integrates real-time information on upper-level carbon emission targets and lower-level resource dynamic response potential, so that the scheme not only meets system-level low-carbon constraints, but also fully respects the real-time operating status and adjustment limits of resources. This not only reduces the scheduling deviation and carbon emission runaway risk caused by the infeasibility of instructions, but also continuously learns and adapts to the performance evolution of resources through a dynamic update mechanism, thereby improving the overall carbon emission reduction efficiency and operational economy of the virtual power plant in long-term operation. In summary, based on the above scheme, the coordinated scheduling of the global low-carbon target and the autonomous response of distributed resources in the virtual power plant can be realized, thereby improving the carbon emission reduction efficiency of the multi-energy aggregation system. Attached Figure Description

[0017] Figure 1 This is an exemplary flowchart of a virtual power plant collaborative scheduling method for multi-energy resource aggregation according to some embodiments of the present invention; Figure 2 This is an exemplary flowchart illustrating the determination of a multi-scale digital surface model according to some embodiments of the present invention; Figure 3 This is a schematic diagram of the structure of a scheduling unit according to some embodiments of the present invention; Figure 4 This is a schematic diagram of the structure of a computer device for implementing a virtual power plant collaborative scheduling method for multi-energy resource aggregation, as shown in some embodiments of the present invention. Detailed Implementation

[0018] To better understand the technical solution of the present invention, the technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0019] refer to Figure 1 The diagram is an exemplary flowchart of a virtual power plant collaborative scheduling method for multi-energy resource aggregation according to some embodiments of the present invention. The diagram mainly includes the following steps: Step 101: Initialize the dispatchable capacity map of the virtual power plant, which includes the power constraints, regulation costs, and carbon emission characteristics of each type of resource.

[0020] In some embodiments, initializing the dispatchable capacity map of a virtual power plant, which includes power constraints, regulation costs, and carbon emission characteristics for each type of resource, can be achieved through the following steps: Acquire historical operating data and physical parameters of distributed photovoltaic, wind turbines, energy storage batteries and interruptible loads in the virtual power plant; Power constraints, adjustment cost-power relationship, and carbon emission intensity coefficient per unit output are constructed for each type of resource. By aggregating and encoding each power constraint, each regulation cost-power relationship, and all carbon emission intensity coefficients, a dispatchable capacity map is obtained.

[0021] It should be noted that, in this invention, the schedulable capacity map is a structured digital mapping used to characterize the schedulable characteristics of various types of resources; historical operating data is a time-series dataset used to quantify the operating patterns of resources; physical parameters are a set of attributes describing the inherent technical characteristics of resources; power constraints are a set of boundary conditions used to limit the output and power consumption range of resources; the adjustment cost-power relationship is a functional relationship used to quantify the economic costs generated by resource adjustment behavior; and the carbon emission intensity coefficient is a conversion factor used to measure the carbon dioxide emissions corresponding to a unit of electricity production and consumption of resources.

[0022] In practice, the process begins by continuously collecting and storing historical operational data of various resources at a uniform time granularity through the existing data acquisition and monitoring system or IoT sensor network of the virtual power plant. For photovoltaic and wind turbines, this primarily involves collecting their historical power output data and irradiance and wind speed data recorded by weather stations. For energy storage batteries, this involves collecting their historical charge and discharge power and state of charge data. For interruptible loads, this involves collecting their historical power consumption curves, interruptible periods, and capacity data. Simultaneously, physical parameters of the resources are obtained from equipment nameplates or technical manuals, such as the rated power and conversion efficiency of photovoltaic panels, the cut-in / cut-out / rated wind speed of wind turbines, the rated capacity and charge and discharge efficiency of energy storage batteries, and the contractually agreed maximum interruption power and minimum interruption duration of interruptible loads. All structured data is then stored in the central database of the virtual power plant as the foundation dataset for subsequent modeling and analysis.

[0023] Then, for each type of resource, power constraints are constructed based on the resource's physical parameters and operating rules. For example, the output constraint for photovoltaics is its rated power multiplied by the real-time availability rate, while the output constraint for energy storage batteries must simultaneously meet power limits and state-of-charge limits. A regulation cost-power relationship is established. For distributed power sources, the cost is mainly operation and maintenance costs, which can be modeled as a linear function related to output. For energy storage, the cost mainly considers battery losses, which can be modeled as a piecewise function related to charge-discharge cycle depth. For interruptible loads, the cost is the compensation fee stipulated in the contract, usually a fixed value or a step function related to the amount of interruption. The carbon emission intensity coefficient is determined. For photovoltaic and wind power, the direct carbon emissions during power generation are zero, but considering the carbon emissions over the entire life cycle, an industry-recognized conversion factor is used. For the grid-purchased portion, the real-time or average emission factor published by the local power grid is used. For energy storage charging and discharging, the indirect carbon emissions are converted based on the emission intensity of the charging source. Thus, the power constraints, regulation cost-power relationship, and carbon emission intensity coefficient per unit output of the resource are obtained.

[0024] Finally, a standardized data structure is designed to integrate and format the power constraints, adjustment cost-power relationships, and carbon emission intensity coefficients established for each type of resource for storage. Specifically, a dispatchable capacity data object is created for each independent resource or resource cluster. In this data object, the power constraints are represented by numerical fields such as minimum output, maximum output, and maximum ramp rate; the adjustment cost-power relationship is represented by the cost function type (e.g., linear, piecewise) and the corresponding parameter list; and the carbon emission intensity coefficient is represented by one or more coefficients related to the operating state. For example, for a certain energy storage power station, its dispatchable capacity data object records its upper and lower limits of charging and discharging power, the unit cost coefficient at different charging and discharging depths, and the grid emission factor used during charging. The collection of dispatchable capacity data objects of all resource individuals in the virtual power plant constitutes the dispatchable capacity map of the entire virtual power plant.

[0025] Step 102: Construct a two-layer collaborative optimization model for the virtual power plant based on the schedulable capability map. The upper layer of the two-layer collaborative optimization model aims to minimize the total carbon emissions of the virtual power plant and generates scheduling instructions for each resource cluster according to the power load demand. The lower layer of the two-layer collaborative optimization model aims to minimize the response deviation to the scheduling instructions of the upper layer and performs distributed collaborative control on the resources within its own cluster to obtain the response capability of the lower layer's own cluster.

[0026] In some embodiments, constructing a two-layer collaborative optimization model for a virtual power plant based on the schedulable capacity map can be achieved through the following steps: Using a schedulable capacity map as input, resources are divided into multiple resource clusters based on their adjustment characteristics and geographical location; A higher-level optimization model is constructed. The decision variable of the higher-level optimization model is the total scheduling command of each resource cluster. The objective function is to minimize the total carbon emissions of the virtual power plant. The constraints include load balancing, cluster scheduling capability, and network transmission security. A lower-level optimization model is constructed. After receiving the scheduling instruction, each resource cluster performs distributed optimization with the goal of minimizing the response deviation within the cluster, and feeds back the actual response capability that the corresponding resource cluster can provide to the upper layer. The two-layer collaborative optimization model of the virtual power plant is determined by the upper-layer optimization model and the lower-layer optimization model.

[0027] It should be noted that the two-layer collaborative optimization model in this invention is a hierarchical decomposition mathematical programming framework. Through the interactive iteration of the upper and lower layers, the two-layer collaborative optimization model coordinates and resolves the contradiction between the global optimization objective of the virtual power plant and the autonomous response of decentralized resources. Specifically, this two-layer collaborative optimization model decomposes the complex centralized optimization problem into two related sub-problems. The upper-layer model acts as the global decision-maker, formulating overall scheduling instructions for each resource cluster based on the schedulable capacity map and system constraints, with the goal of minimizing total carbon emissions. The lower-layer model acts as the local executor. After receiving the instructions, each resource cluster, considering its own adjustment costs and operational constraints, autonomously optimizes its internal resource allocation through a distributed algorithm to minimize the deviation between the actual response and the upper-layer instructions, and feeds back the updated schedulable capacity to the upper layer. The upper and lower layers conduct multiple rounds of information interaction and coordination calculations through a preset distributed algorithm until both sides' decisions converge, thereby ensuring the global low-carbon objective while fully respecting the decentralized autonomous characteristics and local interests of resources.

[0028] In some embodiments, the upper layer of the two-layer collaborative optimization model aims to minimize the total carbon emissions of the virtual power plant. The generation of scheduling instructions for each resource cluster based on power load demand can be achieved in the following manner: An objective function is established with the carbon emission intensity of each resource cluster as the weight, and then the cluster carbon emission of the virtual power plant is calculated. The power load demand of the virtual power plant is used as an equality constraint, and the power limits provided by each cluster resource in the dispatchable capacity map are superimposed as inequality constraints. Scheduling optimization is performed based on the cluster carbon emissions, the equality constraints, and the inequality constraints. Power scheduling instructions for each resource cluster that meet both optimal carbon emissions and load requirements are output as scheduling instructions for the corresponding resource clusters and sent to the corresponding lower-level cluster controllers.

[0029] It should be noted that, in this invention, the scheduling instruction is a control instruction used to directly guide each resource cluster to achieve the total output or total power consumption value in the next scheduling period; cluster carbon emissions are a quantitative indicator used to measure the total carbon dioxide emissions generated by the operation of all resource clusters under a specified scheduling scheme; equality constraints are mathematical conditions that the solution to the optimization problem must strictly satisfy, and equality constraints can ensure the real-time balance between power generation and power consumption; inequality constraints are mathematical conditions used to limit the solution to the optimization problem to be within a specified feasible range, and these inequality constraints can ensure that the scheduling instructions of each cluster are within the limits of its physical capabilities.

[0030] In practical implementation, firstly, the carbon emission intensity coefficient per unit output of each resource cluster is extracted from the dispatchable capacity map. This coefficient represents the carbon emissions corresponding to each kilowatt-hour (or equivalent kilowatt-hour of regulation) generated by the cluster. The core of a mathematical optimization model, namely the objective function, is constructed to minimize the total carbon emissions of the virtual power plant. This objective function takes the form of multiplying the planned dispatch power (as a decision variable) of each resource cluster by its corresponding carbon emission intensity coefficient, and then summing the products of all clusters. Through this weighted summation, the objective function directly links dispatch decisions to carbon emission consequences. For example, if a cluster is primarily photovoltaic and its carbon emission intensity is close to zero, increasing its dispatch power will have a small contribution to total carbon emissions. Conversely, if a cluster includes a large amount of energy storage charging that relies heavily on external power purchases, and the external power carbon emission intensity is high, increasing its dispatch power will significantly increase total carbon emissions. By establishing the objective function, the carbon emission impact of resource clusters with different levels of cleanliness in dispatch decisions can be quantified and compared in a unified manner.

[0031] Then, the most basic equality constraint for power system operation is introduced: in each scheduling period, the sum of the scheduling power of all resource clusters must equal the net load demand that the virtual power plant needs to meet in that period (total load minus the predicted output of uncontrollable power sources). This ensures the instantaneous balance of power supply and demand. A series of inequality constraints are introduced to ensure the feasibility of scheduling instructions. These inequality constraints are directly derived from the power constraints predefined for each resource cluster in the dispatchable capacity map. These mainly include: the planned scheduling power of each resource cluster in each period must not be less than its minimum technically available output (or maximum absorbable power) in that period, and must not be greater than its maximum technically available output (or minimum absorbable power) in that period. For example, for a cluster mainly composed of energy storage systems, its scheduling power (positive values ​​represent discharging, negative values ​​represent charging) must be between the maximum allowable discharge power and the maximum allowable charging power under its current state of charge. The load balance equality constraint and the power upper and lower limit inequality constraints of each cluster together constitute the complete constraint set of the optimization model, restricting the search space of scheduling decisions to a feasible region that satisfies both system balance and the physical capabilities of each cluster.

[0032] Finally, the solver in the upper-level optimization model (such as a linear programming solver or a mixed-integer programming solver) receives the established objective function (minimizing total carbon emissions) and a complete set of constraints (including load balance equality constraints and power upper and lower limit inequality constraints for each cluster). Within the mathematically feasible region enclosed by the constraints, it automatically searches for the set of specified decision variable values ​​that minimize the objective function. This optimal solution is the optimal power plan value for each resource cluster that minimizes the carbon emissions of the entire virtual power plant while meeting real-time load demands. After solving, the optimization model outputs the optimal power plan value. All values ​​correspond to the precise power dispatch instructions for each resource cluster. The central dispatch system of the virtual power plant sends each power dispatch instruction accurately to the local controller of the corresponding lower-level resource cluster through the communication network, serving as the upper-level objective for executing local distributed collaborative control. For example, the central dispatch system issues the instruction "Cluster A's average output is 500 kilowatts in the next hour" to the main controller of Cluster A, thereby completing the deployment of upper-level optimization decisions.

[0033] In some embodiments, the lower layer in the two-layer collaborative optimization model aims to minimize the response deviation to the scheduling instructions of the upper layer, and performs distributed collaborative control of the resources within its cluster. The response capability of the lower layer's cluster can be obtained by the following steps: After receiving scheduling instructions from the upper layer of each resource cluster, the cluster master controller decomposes the scheduling instructions into basic allocation values ​​for each resource within each resource cluster. Each resource performs distributed iterative computation based on its local schedulable capabilities to obtain the output resources of each resource cluster, thereby minimizing the deviation between the overall resource cluster and the instructions. After iterative convergence, the total power range and adjustment cost that each resource cluster can actually provide are calculated and fed back to the upper-layer model as the response capability of the lower-level cluster.

[0034] It should be noted that, in this invention, response capability is a comprehensive indicator used to quantify the actual power that can be provided to the upper-level scheduling system in the next scheduling cycle; the basic allocation value is an initial power adjustment target reference value set for each independent resource in the cluster; and self-output resource refers to the final power adjustment amount that each resource autonomously determines and executes.

[0035] In practice, the cluster master controller first receives a power scheduling instruction for the cluster from the upper-level optimization model via the communication network. This instruction is a total power value, specifying the total output or total power consumption target that the resource cluster needs to achieve in the next scheduling period. The cluster master controller pre-allocates this total power target to each member resource within the cluster, forming a preliminary allocation scheme. The specific decomposition strategy can be based on various principles, such as allocating according to the rated capacity reported by each resource, or according to the actual output ratio of each resource in the previous period, or weighted allocation considering its current adjustability margin. Regardless of the specific rules used, the goal of the decomposition process is to generate a clear and executable basic power allocation value for each resource within the cluster. For example, for a cluster that includes energy storage and flexible loads, if it receives an instruction with a total charging power of 100 kilowatts, the main controller may decompose it into charging energy storage A by 60 kilowatts and reducing the power consumption of flexible load B by 40 kilowatts. This basic allocation value is not required to be strictly and precisely executed, but rather serves as the starting point and reference benchmark for subsequent distributed optimization iterations. The cluster main controller sends the calculated basic allocation values ​​of each resource to the corresponding local resource controller.

[0036] Then, all resources within the cluster (or their local controllers) initiate a distributed negotiation and optimization process based on the base allocation value. Each resource is only aware of its own schedulable capabilities (i.e., power constraints, cost information, etc.) and limited information exchanged with neighboring resources (e.g., an estimate of the total deviation). Through a pre-defined distributed optimization algorithm, such as a consensus algorithm or the alternating direction multiplier method, it continuously exchanges intermediate calculation results with other resources in the cluster across multiple computation rounds, and updates its planned output value accordingly. Each update follows two core principles: first, ensuring that its new planned value does not violate local power constraints; and second, ensuring that its new planned value does not violate local power constraints. The system employs two main strategies: first, to address rate and cost constraints; second, to ensure that the sum of the new planned values ​​of all resources in the entire cluster (i.e., the total output of the cluster) approaches the scheduling instructions issued by the upper layer, thereby minimizing the overall response deviation of the cluster; after multiple rounds of local computation and information exchange, the planned output of all resources within the cluster will gradually become consistent and eventually converge to a stable value. After convergence, the final planned output value determined for each resource is its own output resource, eliminating the need for centralized optimization calculations by the cluster master controller, thus reflecting the characteristics of distributed autonomy; each resource will execute according to its own determined output resource, or report it as the final execution plan; Finally, in the summary and information feedback phase of the lower-level distributed response process, after the distributed iterative computation reaches convergence, the cluster master controller (or obtains it naturally through distributed computation) needs to summarize the final state information of the cluster. Based on the power constraints in the self-output resources and local schedulable capabilities of each resource after convergence, the maximum actual increase (upward adjustment capability) and the maximum actual decrease (downward adjustment capability) of the entire cluster in the next time period under the current state (e.g., considering the latest state of charge of energy storage) are calculated. These two extreme values ​​constitute the total power range that the cluster can actually provide. Based on the determined output adjustment of each resource... The integer quantity and its corresponding adjustment cost-power relationship are summed to calculate the total economic cost required to execute the current scheduling command or make a specified power adjustment, thus obtaining the overall adjustment cost of the cluster. The calculated aggregate information, including the feasible power range and comprehensive economic cost, is more accurate and real-time than the initial schedulable capacity map because it reflects the latest feasibility and economy of resources after local optimization and mutual coordination. The cluster master controller feeds back the updated response capacity information to the upper-level optimization model of the virtual power plant through the communication network, so that the upper level can refresh its understanding of the cluster's capabilities in the next optimization cycle, thereby achieving closed-loop optimization.

[0037] Step 103: Iteratively solve the two-layer collaborative optimization model to enable the scheduling instructions of the upper layer and the response capabilities of the lower layer to interact and update, thereby obtaining the scheduling scheme of the virtual power plant.

[0038] In some embodiments, the two-layer collaborative optimization model is iteratively solved, allowing the scheduling instructions of the upper layer to interact and update the response capabilities of the lower layer, thereby obtaining the scheduling scheme of the virtual power plant, as referenced. Figure 2 The figure is a flowchart illustrating the process of determining a multi-scale digital surface model in some embodiments of the present invention. In this embodiment, determining the multi-scale digital surface model can be achieved through the following steps: Step 1031: The upper layer of the two-layer collaborative optimization model sends preliminary scheduling instructions to each resource cluster in the lower layer; Step 1032: Each lower-level resource cluster performs distributed optimization based on the corresponding preliminary scheduling instructions and feeds back the updated cluster response capability to the upper layer. Step 1033: The upper layer updates the constraints in the schedulable capability graph based on the feedback information, and re-solves the optimization problem to generate new scheduling instructions; Step 1034: Repeat the above steps until the difference between the upper and lower layer instructions and response capabilities is less than the preset error threshold, and output the virtual power plant scheduling scheme.

[0039] It should be noted that, in this invention, the scheduling scheme is a set of power plans used to ultimately guide the operation of all resources in the virtual power plant in the next scheduling cycle; the preliminary scheduling instruction is the initial value of the power adjustment target of various resource clusters used to start the upper and lower layer iterative coordination process; and the error threshold is a precision standard used to determine whether the upper and lower layer models have been fully coordinated and whether the iterative process can be terminated.

[0040] In practical implementation, firstly, in the initial stage of the iterative solution of the two-layer collaborative optimization model, the upper-layer optimization model runs in its initial state. That is, based on the current dispatchable capacity map of the virtual power plant, power load forecast data, and grid operation requirements, it performs an independent centralized optimization calculation with the goal of minimizing carbon emissions. It does not consider the detailed coordination process within the lower-layer resource clusters, but only treats each resource cluster as a whole with fixed regulation capacity and cost. The result obtained is a power allocation plan, that is, it sets a preliminary, globally carbon-optimal power target value for each resource cluster. This power target value is the preliminary dispatch instruction. The central dispatch system of the virtual power plant sends these preliminary dispatch instructions to the cluster master controllers of the corresponding lower-layer resource clusters separately and synchronously through the communication network, as the starting point for subsequent distributed collaborative optimization. For example, the upper layer may calculate that "cluster A should provide 200 kilowatts of net output between 2 pm and 3 pm" and send this value as a preliminary dispatch instruction to the controller of cluster A. Secondly, upon receiving the initial scheduling instruction from the upper layer, the master controller of each lower-level resource cluster immediately uses this instruction as the target and initiates its internal distributed collaborative optimization process. Based on their latest operating status and local constraints, each resource within the cluster (e.g., photovoltaic, energy storage, load) collaboratively determines a set of specific output schemes through information exchange and local calculations. These schemes aim to make the total output of the cluster as close as possible to the instruction, satisfy the individual constraints of all members, and take into account internal economic efficiency. After the internal optimization process is completed, the cluster controller needs to evaluate the final result of the scheme, calculate the upper and lower limits of the power regulation that the cluster can reliably provide in the next scheduling period under this scheme, and the estimated total regulation cost of executing this scheme. The set of power feasible domain and cost information is used as the updated cluster response capability. Each cluster controller feeds back this updated response capability information to the upper-level optimization model through the communication network.

[0041] Then, the upper-level model undergoes a process of self-correction and re-decision-making based on feedback from the lower-level resource clusters. After receiving the cluster response capabilities from all lower-level resource clusters, the central controller of the upper-level optimization model replaces or corrects the original data of the corresponding cluster in the schedulable capability map with the response capability. Specifically, it mainly updates the upper and lower power limit constraints of each cluster, tightening them from a relatively wide range based on historical or rated parameters to a more precise and reliable range based on the current actual state and internal coordination results. The upper-level model keeps the objective function of minimizing carbon emissions unchanged, but uses the updated set of constraints (including more precise cluster power constraints) to re-run the optimization calculation. Since the constraints are more realistic, the solution obtained in this calculation (i.e., the power target value of each cluster) will be more practically executable than the initial scheduling instructions. The recalculated power target value sequence is the new scheduling instruction, and the new instruction sequence will be ready to enter the next round of iterative coordination. Finally, the above process of "instruction issuance - distributed optimization - feedback - recalculation" will be repeated to form a closed-loop iteration. After each iteration, a convergence check will be performed: the difference between the latest dispatch instruction issued by the upper layer and the latest feedback response capability center value (e.g., the midpoint of the available power range) will be compared. For each resource cluster, the absolute value of the difference will be calculated. When the difference of all clusters is less than a pre-set error threshold, it is considered that the upper and lower layer models have achieved sufficient coordination, the upper layer instruction and the actual executable capability of the lower layer are basically consistent, and the iteration process is declared to have converged. Once the convergence condition is met, the iteration loop terminates. The new dispatch instruction generated by the upper layer model in the last iteration and confirmed to be executable by the lower layer capability will be used as the final dispatch scheme of the virtual power plant. This dispatch scheme will be locked and output to the dispatch execution system of the virtual power plant as the basis for the operation of each resource cluster in the next dispatch cycle.

[0042] Step 104: During the execution of the scheduling scheme, the resource weights in the schedulable capability map are dynamically updated based on the response performance and reputation of each resource, for use in the resource scheduling of the virtual power plant in the next scheduling cycle.

[0043] In some embodiments, dynamically updating the resource weights in the schedulable capability map based on the response performance and reputation of each resource can be achieved using the following steps: For each resource, the actual output curve of the resource is collected, and the actual output curve is compared with the scheduling command of the resource to calculate the response accuracy and response speed index of the resource. The resource's reputation score is determined based on its historical response records; The response accuracy, response speed index and reputation score are weighted and fused to obtain the adjustment cost coefficient of the resource in the schedulable capability map, and then the adjustment cost coefficient of each resource in the schedulable capability map is obtained. The resource weights in the schedulable capacity map are updated using various adjustment cost coefficients.

[0044] It should be noted that in this invention, resource weight is an adjustment parameter that comprehensively reflects resource adjustment costs, performance, and reliability, and affects the priority of scheduling decisions; response accuracy is a percentage indicator that quantifies the degree of consistency between the actual execution power of the resource and the scheduling instructions; response speed is an indicator that measures how quickly a resource can achieve the required output from receiving the instruction; reputation score is a cumulative historical score that evaluates the long-term reliability, stability, and integrity of the resource's scheduling instructions; and adjustment cost coefficient is a core parameter that characterizes the economic cost and priority of calling up a unit of power of the resource.

[0045] In practice, firstly, for each resource, at the end of each scheduling cycle, the real-time monitoring system of the virtual power plant collects power measurement data of the resource (e.g., energy storage unit, controllable load) within that cycle, forming a time-series actual output curve. Simultaneously, it acquires the detailed power scheduling command curve issued to the resource at the beginning of the scheduling cycle. The two curves are compared point-by-point on the same time scale, for example, one data point per minute or every 5 minutes. When calculating the response accuracy, the percentage of the sum of the absolute values ​​of the deviations between the actual output and the command value at all time points within the scheduling cycle to the sum of the absolute values ​​of the commands, or the percentage of the area difference between the actual curve and the command curve to the area under the total command curve, can be used as the basis for accuracy calculation. When calculating the response speed index, the time taken for the resource's actual output to first enter and stabilize within the allowable error band of the command value after receiving a new command can be examined. The ratio of this time to a standard or expected response time is used to characterize its speed. Through comparative analysis of the two curves, the response accuracy and response speed index values ​​of the resource within the current scheduling cycle are obtained.

[0046] Secondly, in establishing a long-term reliability evaluation of resources, the resource's reputation score is a comprehensive score dynamically calculated based on its historical response records over multiple scheduling cycles (e.g., the most recent 20 or 30 cycles). The specific calculation can follow these principles: the response accuracy and response speed indicators calculated for each historical cycle are converted into a single-cycle base score according to certain rules (e.g., setting a passing standard; points are added for meeting the standard and deducted for failing to meet it); a time-weighted method is used for accumulation, for example, introducing a decay factor so that the performance of recent historical cycles has a greater weight on the total score than that of distant historical cycles, thus reflecting the timeliness of the score. Furthermore, additional reward or penalty coefficients can be set for consecutive compliance or significant deviations to encourage stable and reliable performance; a normalized score (e.g., between 0 and 100) calculated in the above manner is the resource's current reputation score. This reputation score is independent of individual performance and reflects the resource's overall reliability in long-term scheduling cooperation.

[0047] Then, the real-time performance and historical reputation of resources are transformed into key economic parameters influencing future scheduling decisions. Operationally, a weight value is preset for each of the three evaluation indicators (response accuracy in the current period, response speed in the current period, and historical reputation score). The sum of these three weight values ​​is 1, and their specific values ​​reflect the scheduling system's emphasis on different performance dimensions. For example, the accuracy weight can be set to 0.5, the speed weight to 0.2, and the reputation weight to 0.3. Then, each indicator is normalized so that its value falls within the range of 0 to 1 (1 represents the best). Next, the normalized indicator value is multiplied by its corresponding weight and summed to obtain a comprehensive evaluation score between 0 and 1. Finally, this comprehensive evaluation score is correlated with the resource's basic adjustment cost (e.g., the physical cost of equipment operation and maintenance): the higher the comprehensive evaluation score, the better the performance and the higher the reputation, and the final adjustment cost coefficient is reduced by a certain proportion on top of the basic cost to increase the chance of being prioritized in future scheduling; conversely, the adjustment cost coefficient is increased (as a penalty). Through this process, the resource can calculate an updated adjustment cost coefficient based on its latest performance and long-term reputation. The adjustment cost coefficient of each resource in the schedulable capacity map can then be obtained from the above.

[0048] Finally, the updated adjustment cost coefficient for each resource will be directly used to refresh the relevant fields of the corresponding resource entry in the schedulable capacity graph stored in the virtual power plant central database. Specifically, in the data structure of the schedulable capacity graph, each resource object contains an attribute field for recording its adjustment cost coefficient. The generated new coefficient value is written to overwrite the original old coefficient value of the resource. At the same time, in higher-level aggregation models (e.g., resource cluster models), the overall adjustment cost of the cluster may be recalculated based on the updated cost coefficients of its member resources. Through the periodic and automatic update process, it is ensured that the information on resource economics in the schedulable capacity graph is always up-to-date and can truly reflect the recent execution performance and reliability of the resources. The updated schedulable capacity graph will be used for the two-level collaborative optimization calculation in the next scheduling cycle, thereby realizing closed-loop learning and adaptive optimization of scheduling, execution, evaluation, and updating.

[0049] Furthermore, in another aspect of the present invention, in some embodiments, the present invention provides a virtual power plant collaborative scheduling system for multi-energy resource aggregation, the virtual power plant collaborative scheduling system for multi-energy resource aggregation including a scheduling unit, referenced... Figure 3 The figure is a schematic diagram of the structure of a scheduling unit according to some embodiments of the present invention. The scheduling unit includes: an initialization module 201, a processing module 202, and an execution module 203, which are described below: Initialization module 201, in this invention, is mainly used to initialize the dispatchable capability map of each type of resource in the virtual power plant, which includes the power constraints, regulation costs and carbon emission characteristics of each type of resource. Processing module 202, in this invention, is used to construct a two-layer collaborative optimization model of the virtual power plant based on the schedulable capability map. The upper layer of the two-layer collaborative optimization model aims to minimize the total carbon emissions of the virtual power plant and generates scheduling instructions for each resource cluster according to the power load demand. The lower layer of the two-layer collaborative optimization model aims to minimize the response deviation to the scheduling instructions of the upper layer and performs distributed collaborative control on the resources within its own cluster to obtain the response capability of the lower layer's own cluster. It should be noted that the processing module 202 is also used to iteratively solve the two-layer collaborative optimization model, so that the scheduling instructions of the upper layer and the response capabilities of the lower layer are updated interactively, thereby obtaining the scheduling scheme of the virtual power plant. The execution module 203 in this invention is mainly used to dynamically update the resource weights in the schedulable capability map according to the response performance and reputation of each resource during the execution of the scheduling scheme, for resource scheduling of the virtual power plant in the next scheduling cycle.

[0050] The foregoing detailed examples of a virtual power plant collaborative scheduling system and method for multi-energy resource aggregation provided by embodiments of the present invention. It is understood that the corresponding apparatus, in order to achieve the above functions, includes hardware structures and / or software modules corresponding to the execution of each function. Those skilled in the art should readily recognize that, based on the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein, this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0051] In some embodiments, the present invention also provides a computer device, the computer device including a memory and a processor, the memory for storing a computer program, and the processor for calling and running the computer program from the memory, so that the computer device executes the above-described virtual power plant collaborative scheduling method for multi-energy resource aggregation.

[0052] In some embodiments, reference Figure 4 The dashed lines in the figure indicate that the unit or module is optional. This figure is a structural schematic diagram of a computer device for implementing a virtual power plant collaborative scheduling method for multi-energy resource aggregation according to an embodiment of this application. The virtual power plant collaborative scheduling method for multi-energy resource aggregation described in the above embodiments can be achieved through… Figure 4 The computer device shown is used to implement this, and the computer device includes at least one processor 301, a memory 302 and at least one communication unit 305. The computer device may be a terminal device, a server or a chip.

[0053] Processor 301 can be a general-purpose processor or a special-purpose processor. For example, processor 301 can be a central processing unit (CPU), which can be used to control computer devices, execute software programs, and process data from software programs. The computer device may also include a communication unit 305 for inputting (receiving) and outputting (transmitting) signals.

[0054] For example, the computer device may be a chip, and the communication unit 305 may be the input and / or output circuit of the chip, or the communication unit 305 may be the communication interface of the chip, which may be a component of a terminal device, network device or other device.

[0055] For example, the computer device may be a terminal device or a server, and the communication unit 305 may be a transceiver of the terminal device or the server, or the communication unit 305 may be a transceiver circuit of the terminal device or the server.

[0056] The computer device may include one or more memories 302 storing a program 304. The program 304 can be executed by a processor 301 to generate instructions 303, causing the processor 301 to execute the method described in the above method embodiments according to the instructions 303. Optionally, the memory 302 may also store data (such as a target audit model). Optionally, the processor 301 may also read data stored in the memory 302, which may be stored at the same storage address as the program 304, or it may be stored at a different storage address than the program 304.

[0057] The processor 301 and memory 302 can be configured separately or integrated together, for example, integrated on the system on chip (SOC) of the terminal device.

[0058] It should be understood that each step of the above method embodiment can be completed by hardware logic circuits or software instructions in the processor 301. The processor 301 can be a CPU, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, such as discrete gates, transistor logic devices, or discrete hardware components.

[0059] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0060] For example, in some embodiments, the present invention also provides a computer-readable storage medium storing instructions or code that, when executed on a computer, cause the computer to implement the above-described virtual power plant collaborative scheduling method for multi-energy resource aggregation.

[0061] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0062] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for coordinated scheduling of a virtual power plant oriented to multi-energy resource aggregation, characterized in that, Includes the following steps: The initial virtual power plant includes a map of dispatchable capabilities for each type of resource, including power constraints, regulation costs, and carbon emission characteristics. Based on the schedulable capability map, a two-layer collaborative optimization model for virtual power plants is constructed. The upper layer of the two-layer collaborative optimization model aims to minimize the total carbon emissions of the virtual power plant and generates scheduling instructions for each resource cluster according to the power load demand. The lower layer of the two-layer collaborative optimization model aims to minimize the response deviation to the scheduling instructions of the upper layer and performs distributed collaborative control of the resources within its own cluster to obtain the response capability of the lower layer's own cluster. The two-layer collaborative optimization model is iteratively solved so that the scheduling instructions of the upper layer and the response capabilities of the lower layer are updated interactively, thereby obtaining the scheduling scheme of the virtual power plant. During the execution of the scheduling scheme, the resource weights in the schedulable capability map are dynamically updated based on the response performance and reputation of each resource, which is used for resource scheduling of the virtual power plant in the next scheduling cycle.

2. The method of claim 1, wherein, The initialization of the virtual power plant includes a dispatchable capacity map of power constraints, regulation costs, and carbon emission characteristics for each type of resource, specifically including: Acquire historical operating data and physical parameters of distributed photovoltaic, wind turbines, energy storage batteries and interruptible loads in the virtual power plant; Power constraints, adjustment cost-power relationship, and carbon emission intensity coefficient per unit output are constructed for each type of resource. By aggregating and encoding each power constraint, each regulation cost-power relationship, and all carbon emission intensity coefficients, a dispatchable capacity map is obtained.

3. The method of claim 1, wherein, The two-layer collaborative optimization model for virtual power plants based on the aforementioned schedulable capability map specifically includes: Using a schedulable capacity map as input, resources are divided into multiple resource clusters based on their adjustment characteristics and geographical location; A higher-level optimization model is constructed. The decision variable of the higher-level optimization model is the total scheduling command of each resource cluster. The objective function is to minimize the total carbon emissions of the virtual power plant. The constraints include load balancing, cluster scheduling capability, and network transmission security. A lower-level optimization model is constructed. After receiving the scheduling instruction, each resource cluster performs distributed optimization with the goal of minimizing the response deviation within the cluster, and feeds back the actual response capability that the corresponding resource cluster can provide to the upper layer. The two-layer collaborative optimization model of the virtual power plant is determined by the upper-layer optimization model and the lower-layer optimization model.

4. The method of claim 1, wherein, The upper layer of the two-layer collaborative optimization model aims to minimize the total carbon emissions of the virtual power plant. Based on the power load demand, it generates scheduling instructions for each resource cluster, specifically including: An objective function is established with the carbon emission intensity of each resource cluster as the weight, and then the cluster carbon emission of the virtual power plant is calculated. The power load demand of the virtual power plant is used as an equality constraint, and the power limits provided by each cluster resource in the dispatchable capacity map are superimposed as inequality constraints. Scheduling optimization is performed based on the cluster carbon emissions, the equality constraints, and the inequality constraints. Power scheduling instructions for each resource cluster that meet both optimal carbon emissions and load requirements are output as scheduling instructions for the corresponding resource clusters and sent to the corresponding lower-level cluster controllers.

5. The method of claim 1, wherein, In the two-layer collaborative optimization model, the lower layer aims to minimize the response deviation to scheduling instructions from the upper layer. It performs distributed collaborative control of resources within its cluster, resulting in the following specific response capabilities of the lower-layer cluster: After receiving scheduling instructions from the upper layer of each resource cluster, the cluster master controller decomposes the scheduling instructions into basic allocation values ​​for each resource within each resource cluster. Each resource performs distributed iterative computation based on its local schedulable capabilities to obtain the output resources of each resource cluster, thereby minimizing the deviation between the overall resource cluster and the instructions. After iterative convergence, the total power range and adjustment cost that each resource cluster can actually provide are calculated and fed back to the upper-layer model as the response capability of the lower-level cluster.

6. The method of claim 1, wherein, The two-layer collaborative optimization model is iteratively solved, allowing the scheduling instructions of the upper layer to interact and update the response capabilities of the lower layer, thereby obtaining the scheduling scheme of the virtual power plant, which specifically includes: The upper layer of the two-layer collaborative optimization model sends preliminary scheduling instructions to each resource cluster in the lower layer; Each lower-level resource cluster performs distributed optimization based on the corresponding initial scheduling instructions and feeds back the updated cluster response capabilities to the upper layer. The upper layer updates the constraints in the schedulable capability graph based on the feedback information, and re-solves the optimization problem to generate new scheduling instructions; Repeat the above steps until the difference between the upper and lower layer instructions and response capabilities is less than the preset error threshold, and output the virtual power plant scheduling scheme.

7. The method of claim 1, wherein, The dynamic updating of resource weights in the schedulable capability map based on the response performance and reputation of each resource specifically includes: For each resource, the actual output curve of the resource is collected, and the actual output curve is compared with the scheduling command of the resource to calculate the response accuracy and response speed index of the resource. The resource's reputation score is determined based on its historical response records; The response accuracy, response speed index and reputation score are weighted and fused to obtain the adjustment cost coefficient of the resource in the schedulable capability map, and then the adjustment cost coefficient of each resource in the schedulable capability map is obtained. The resource weights in the schedulable capacity map are updated using various adjustment cost coefficients. 8.A virtual power plant collaborative scheduling system for multi-energy resource aggregation, configured to perform the virtual power plant collaborative scheduling method for multi-energy resource aggregation according to any one of claims 1 to 7, comprising a scheduling unit, characterized in that, The scheduling unit includes: The initialization module is used to initialize the dispatchable capacity map of each type of resource in the virtual power plant, including power constraints, regulation costs, and carbon emission characteristics. The processing module is used to construct a two-layer collaborative optimization model of the virtual power plant based on the schedulable capability map. The upper layer of the two-layer collaborative optimization model aims to minimize the total carbon emissions of the virtual power plant and generates scheduling instructions for each resource cluster according to the power load demand. The lower layer of the two-layer collaborative optimization model aims to minimize the response deviation to the scheduling instructions of the upper layer and performs distributed collaborative control on the resources within its own cluster to obtain the response capability of the lower layer's own cluster. The processing module is also used to iteratively solve the two-layer collaborative optimization model, so that the scheduling instructions of the upper layer and the response capabilities of the lower layer are updated interactively, thereby obtaining the scheduling scheme of the virtual power plant. The execution module is used to dynamically update the resource weights in the schedulable capability map based on the response performance and reputation of each resource during the execution of the scheduling scheme, for resource scheduling of the virtual power plant in the next scheduling cycle.

9. A computer device, comprising: The computer device includes a memory and a processor, the memory storing code, and the processor being configured to retrieve the code and execute the virtual power plant collaborative scheduling method for multi-energy resource aggregation as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 9. When the computer program is executed by the processor, it implements the virtual power plant collaborative scheduling method for multi-energy resource aggregation as described in any one of claims 1 to 7.

Images