Cross-regional virtual power plant cooperative scheduling method, device, medium and product

Abstract

The invention discloses a cross-regional virtual power plant cooperative scheduling method and device, a medium and a product, and relates to the field of data processing. The method comprises the following steps: acquiring real-time characteristic data such as space-time positions, output / demand prediction and the like of distributed energy resources and loads, and determining dynamic weights of characteristic dimensions based on a global optimization target and data of a current scheduling period; generating a dynamic resource cluster division instruction containing a member list and a coordination constraint condition according to the dynamic weight and the real-time data, and sending an initial cross-regional coordination scheduling instruction containing a net exchange power target value and the like and a compensation price signal to each dynamic resource cluster local agent; after aggregation response boundary information returned by the local agent is received, an instruction and a signal are updated, a target collaborative scheduling instruction is obtained and finally sent to each dynamic resource cluster for execution, and effective control over cross-regional virtual power plant resources is achieved. According to the method, the problem that the adaptability of the cross-regional virtual power plant collaborative scheduling instruction and the actual resource capacity is insufficient can be relieved.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a cross-regional virtual power plant collaborative scheduling method, equipment, medium and product. Background Art

[0002] A virtual power plant (VPP) is a smart energy system that aggregates and coordinates decentralized resources such as distributed power generation, energy storage, and controllable loads through information and communication technologies and software systems. Its essence is to integrate fragmented energy resources into a "virtual" power plant with flexible adjustment capabilities to participate in the operation of the power system and market transactions, thereby achieving efficient resource utilization and safe and stable operation of the power grid.

[0003] Among them, cross-regional virtual power plant coordinated scheduling is a key means to deal with the contradiction between the high proportion of renewable energy grid connection and the uneven temporal and spatial distribution of energy supply and demand. It breaks the geographical boundary restrictions, integrates the complementarity of wind and solar resources in different climate zones, the differences in load characteristics and cross-regional backup sharing capabilities, optimizes resource allocation on a larger temporal and spatial scale, effectively smoothes the volatility of distributed energy, alleviates local power grid congestion, and improves the overall resilience and economy of the system.

[0004] However, in the coordinated scheduling of cross-regional virtual power plants, related technologies usually adopt a fixed partitioning or typed resource grouping mechanism (such as dividing aggregation units by administrative boundaries or resource types). This static aggregation method is in great conflict with the dynamic and changeable spatiotemporal characteristics, responsiveness and cost structure of distributed resources, resulting in resources being "aggregated but not united" - that is, physically forced aggregated resource groups cannot collaborate efficiently due to lack of target adaptability, resulting in insufficient adaptability between scheduling instructions and actual resource capabilities. Summary of the Invention

[0005] In response to the above-mentioned technical problems and defects, the purpose of the present invention is to provide a cross-regional virtual power plant collaborative scheduling method, equipment, medium and product, which can alleviate the problem of insufficient adaptability between cross-regional virtual power plant collaborative scheduling instructions and actual resource capabilities.

[0006] To achieve the above-mentioned purpose, in a first aspect, the present invention provides a cross-regional virtual power plant collaborative scheduling method, comprising: collecting real-time characteristic data of each distributed energy resource and load, the real-time characteristic data including at least one of spatiotemporal location information, output / demand forecast curve, physical operation parameters, economic characteristic parameters and dispatchable state; determining the dynamic weight of each characteristic dimension based on the global optimization target set in the current scheduling cycle and the real-time characteristic data; generating a dynamic resource cluster partitioning instruction according to the dynamic weight and the real-time characteristic data, the dynamic resource cluster partitioning instruction including a member list of the resource cluster and collaborative constraints; based on the dynamic resource cluster partitioning instruction , sending an initial cross-regional collaborative scheduling instruction and a compensation price signal to the local agent corresponding to each of the dynamic resource clusters, the initial cross-regional collaborative scheduling instruction including the net exchange power target value and the ancillary service demand; receiving the aggregated response boundary information returned by each of the local agents based on the initial cross-regional collaborative scheduling instruction and the compensation price signal, the aggregated response boundary information being used to characterize the feasible domain range of the dynamic resource cluster for the instruction; updating the collaborative scheduling instruction and the compensation price signal according to the aggregated response boundary information to obtain the target collaborative scheduling instruction; sending the target collaborative scheduling instruction to each of the dynamic resource clusters for execution to control the operation of cross-regional virtual power plant resources.

[0007] The present invention adopts the above-mentioned method steps, collects real-time feature data, and determines the dynamic weight of each feature dimension in combination with the global optimization goal, so that the resource division is more in line with the actual working conditions; when generating the dynamic resource cluster division instruction, the resource characteristics and physical constraints are comprehensively considered to ensure that the division result is executable. In the process of dispatch instruction generation and optimization, the initial cross-regional collaborative dispatch instruction and compensation price signal are first sent, and then the collaborative dispatch instruction and compensation price signal are updated according to the aggregated response boundary information returned by the local agent to form a closed-loop feedback mechanism, and the instructions are dynamically adjusted according to the actual feasible domain range of the resource to accurately match the resource capabilities. Finally, the target collaborative dispatch instruction is sent to control the operation of the resource, so as to achieve a high degree of adaptation between the dispatch instruction and the actual capacity of the resource, and improve the effectiveness and reliability of the cross-regional virtual power plant collaborative dispatch.

[0008] Optionally, in some embodiments, the dynamic weight of each feature dimension is determined based on the global optimization goal set in the current scheduling cycle and the real-time feature data, including: determining the key feature dimension identifier according to the global optimization goal of the current scheduling cycle; calculating the influence weight coefficient of each dimension on the achievement of the goal based on the key feature dimension identifier; correcting the influence weight coefficient according to the resource status constraints in the real-time feature data to obtain the dynamic weight; the resource status constraints refer to the physical boundary conditions that limit the feasible domain of resource scheduling, including maximum / minimum power limits, energy storage SOC safety range, equipment start and stop status and network topology connection relationship.

[0009] By adopting the technical solution of the above embodiment, the feature dimension weights are dynamically corrected by combining the current global optimization goals and resource status constraints, ensuring that the feature weights used for resource cluster division can more accurately reflect the actual influence and physical feasibility of resources on the current scheduling goals, thereby improving the effectiveness of subsequent resource aggregation and the rationality of scheduling instructions.

[0010] Optionally, in some embodiments, the calculation of the influence weight coefficient of each dimension on the goal achievement based on the key feature dimension identifier includes: querying a preset target feature mapping rule library to obtain a baseline influence value of each key feature dimension on the optimization target; based on the baseline influence value and the resource dynamic characteristics in the real-time feature data, calculating the actual contribution correction factor of each dimension to the goal achievement; the resource dynamic characteristics refer to the real-time changing operating capability parameters of distributed energy resources, including time-varying attributes such as current output / load curve, response speed, ramp rate and cost volatility; normalizing the actual contribution correction factor to obtain the influence weight coefficient.

[0011] By adopting the technical solution of the above embodiment, by querying the preset rules and correcting the influencing weight coefficient in combination with the dynamic characteristics of the resources, the weight calculation is not only based on static rules, but also can reflect the real-time changing operating capabilities and economy of distributed resources, thereby enhancing the sensitivity of the weight to the current actual state of the resources and laying the foundation for more refined dynamic resource cluster division.

[0012] Optionally, in some embodiments, generating a dynamic resource cluster partitioning instruction based on the dynamic weight and the real-time feature data includes: generating a resource weighted feature vector based on the dynamic weight and the real-time feature data; partitioning the resource weighted feature vector through a constrained clustering algorithm to obtain an initial resource grouping scheme; verifying the physical operability of the initial resource grouping scheme to obtain a resource cluster entity that meets the collaborative constraint conditions; and generating a dynamic resource cluster partitioning instruction based on the member list and constraint boundary of each resource cluster entity.

[0013] By adopting the technical solution of the above embodiment, through a constrained clustering algorithm based on weighted feature vectors and verifying physical operability, it is ensured that the formed resource clusters are not only similar in characteristics, but also physically feasible and meet the coordination constraints, making the resource clusters a scheduling entity that can be actually operated, thereby improving the practicality of the partitioning scheme.

[0014] Optionally, in some embodiments, the partitioning process of the resource weighted eigenvectors using a constrained clustering algorithm to obtain an initial resource grouping scheme includes: constructing a resource similarity matrix based on the resource weighted eigenvectors; correcting the connection weights of the resource similarity matrix according to physical constraints of the power grid to obtain a constraint-corrected similarity matrix; and performing spectral clustering eigendecomposition on the constraint-corrected similarity matrix to obtain an initial resource grouping scheme that satisfies the constraints.

[0015] By adopting the technical solution of the above embodiment, by considering and correcting the physical constraints of the power grid when constructing the resource similarity matrix, the clustering process is directly integrated with factors such as the power grid topology and capacity limitations, ensuring that the generated resource grouping scheme can better adapt to the power grid structure and operating conditions, and reducing the scheduling difficulties caused by network limitations.

[0016] Optionally, in some embodiments, the dynamic resource cluster partitioning instruction is based on, and the initial cross-regional collaborative scheduling instruction and compensation price signal are sent to the local agent corresponding to each dynamic resource cluster, including: generating the power adjustment interval boundary of each dynamic resource cluster based on the topological structure in the dynamic resource cluster partitioning instruction; constructing a distributed scheduling optimization model according to the power adjustment interval boundary; calculating the optimal power instruction initial value within the cluster based on the distributed scheduling optimization model to obtain the collaborative scheduling instruction initial value matrix; generating the initial cross-regional collaborative scheduling instruction and compensation price signal based on the collaborative scheduling instruction initial value matrix, the grid congestion marginal cost and the adjustment cost; and sending the initial cross-regional collaborative scheduling instruction and the compensation price signal based on the mapping relationship between the dynamic resource cluster and the local agent.

[0017] By adopting the technical solution of the above embodiment, by generating power regulation intervals based on the dynamic resource cluster topology and building a distributed model, and calculating the initial instructions and compensation prices in combination with grid congestion and regulation costs, a systematic and economically oriented approach is provided to generate preliminary cross-regional collaborative scheduling instructions and incentive signals, thereby initiating a distributed collaborative optimization process.

[0018] Optionally, in some embodiments, updating the collaborative scheduling instruction and the compensation price signal according to the aggregate response boundary information to obtain the target collaborative scheduling instruction includes: determining a power regulation deviation matrix based on the initial cross-regional collaborative scheduling instruction and the aggregate response boundary information; updating the compensation price signal according to the power regulation deviation matrix and real-time blocking data to obtain a compensation price update signal; and correcting the initial cross-regional collaborative scheduling instruction based on the compensation price update signal feedback to obtain a target collaborative scheduling instruction.

[0019] By adopting the technical solution of the above embodiment, the compensation price is iteratively updated and the instruction is corrected based on the deviation between the initial instruction and the aggregate response boundary and the real-time congestion data, forming a feedback loop, so that the final scheduling instruction can continuously approach the actual response capability of the resource cluster and effectively deal with real-time power grid congestion, which significantly improves the adaptability and effectiveness of scheduling.

[0020] In a second aspect, an embodiment of the present invention provides an electronic device, comprising: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to enable the electronic device to execute the method described in the first aspect or the second aspect, and any possible implementation of the first aspect or the second aspect.

[0021] In a third aspect, the present invention provides a computer-readable storage medium comprising instructions, which, when executed on the electronic device, enables the electronic device to execute the method described in the first aspect or the second aspect, and any possible implementation of the first aspect or the second aspect.

[0022] In a fourth aspect, the present invention provides a computer program product comprising instructions, which, when the computer program product is run on the electronic device, enables the electronic device to execute the method described in the first aspect or the second aspect, and any possible implementation of the first aspect or the second aspect.

[0023] It is understood that the electronic device provided in the second aspect, the storage medium provided in the third aspect, and the computer program product provided in the fourth aspect are all used to execute the method provided by the present invention. Therefore, the beneficial effects that can be achieved can be referred to the beneficial effects of the corresponding methods and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 This is an architecture diagram of a cross-regional virtual power plant collaborative scheduling system according to an embodiment of the present invention; Figure 2 This is a flow chart of a cross-regional virtual power plant collaborative scheduling method according to an embodiment of the present invention; Figure 3 It is a schematic diagram of the architecture of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0025] The terms used in the following embodiments of the present invention are for the purpose of describing specific embodiments only and are not intended to limit the present invention. As used in the specification of the present invention, the singular expressions "a," "an," "above," "the," and "this" are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term "and / or" as used in the present invention refers to any and all possible combinations of one or more of the listed items.

[0026] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying relative importance or implicitly indicating the quantity of the technical features indicated. Thus, a feature designated "first" or "second" may explicitly or implicitly include one or more of such features. In the description of the embodiments of the present invention, unless otherwise specified, "plurality" means two or more.

[0027] It should also be noted that, unless otherwise clearly specified and limited, in the embodiments of the present invention, terms such as "setting" and "connection" should be understood in a broad sense. For example, "connection" can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, or it can be the internal connection of two components; it can be a wired communication connection or a wireless communication connection. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to the specific circumstances. The embodiments of the present invention are described in detail below.

[0028] To address the issues in related technologies regarding cross-regional virtual power plant collaborative scheduling, where the use of fixed partitioning or typed resource grouping mechanisms leads to a mismatch between resource aggregation and dynamic characteristics, and insufficient adaptability between scheduling instructions and actual resource capabilities, an embodiment of the present invention provides a technical solution for cross-regional virtual power plant collaborative scheduling. By introducing a dynamic weight determination mechanism based on real-time feature data and global optimization objectives, and generating dynamic resource cluster partitioning instructions based on this, it enables the flexible and intelligent construction of virtual power plant aggregation units based on the current grid operating status, scheduling objectives, and the real-time characteristics of distributed resources (including spatiotemporal location, output forecasts, operating parameters, economic characteristics, and dispatchability status). This dynamic aggregation approach ensures that members within the same resource cluster have stronger goal consistency and collaborative potential within the current scheduling cycle, thereby overcoming the contradiction in static aggregation, where resources are physically aggregated but lack efficient collaboration. Furthermore, by sending initial scheduling instructions and compensation price signals to the dynamic resource cluster and receiving the aggregate response boundary information fed back by the local agent, information interaction and instruction iterative optimization are achieved between the upper-level scheduling and the dynamic resource cluster, so that the target collaborative scheduling instructions finally issued can more accurately match the actual aggregate response capability and feasible domain range of the dynamic resource cluster.

[0029] The method of combining dynamic aggregation with feedback-based iterative optimization in this embodiment significantly improves the level of refinement and response efficiency of cross-regional virtual power plant resource scheduling, effectively smoothes the volatility of distributed energy, alleviates local power grid congestion, improves the overall resilience and economy of the system, and fundamentally solves the "aggregation but not integration" problem caused by static aggregation, thereby achieving efficient and coordinated utilization of resources.

[0030] Therefore, the embodiment of the present invention provides a cross-regional virtual power plant collaborative scheduling system, the architecture of which is as follows: Figure 1 As shown: The dispatch server at the top level is the hub of the entire system, responsible for receiving instructions or market signals from the power grid and formulating detailed dispatch plans based on the goal of global optimization. At the same time, the dispatch server also exchanges information with the power grid to report the overall operating status and dispatchability of the system.

[0031] Following closely behind is the virtual power plant, which acts as an aggregator, logically integrating distributed energy resources of different geographical locations or types into a controllable whole. The virtual power plant receives instructions from the dispatch server, breaks them down, optimizes them, and then sends them to the next level of local agents.

[0032] A local agent is a regional management unit responsible for resource scheduling and control within its jurisdiction. Each local agent manages several dynamic resource clusters, which are flexibly assembled based on actual needs and resource characteristics. For example, these clusters can include photovoltaic arrays, wind turbines, energy storage systems, and controllable loads.

[0033] Dynamic resource clusters further refine management instructions and ultimately act on the lowest-level distributed energy resources.

[0034] Distributed energy resources are the physical units in the system that actually generate, store or consume energy. They make operational adjustments based on instructions from local agents, such as changing power generation, charging and discharging status, or starting and stopping loads.

[0035] Through this layered, collaborative approach, the entire system effectively integrates and coordinates a large number of distributed energy resources, improving energy efficiency, enhancing the stability and flexibility of the power grid, and enabling better participation in electricity market transactions. The flow of information and control between each layer ensures the system can quickly respond to changes, achieving optimal resource allocation and intelligent management.

[0036] The following combination Figure 2 To be specific, a cross-regional virtual power plant collaborative scheduling method of this embodiment can be executed by a scheduling server, and the method includes the following steps: Step 201: Collect real-time characteristic data of each distributed energy resource and load.

[0037] Among them, distributed energy resources include but are not limited to distributed power generation units such as solar power generation equipment, wind power generation equipment, and gas turbines.

[0038] The real-time characteristic data includes at least one of spatiotemporal location information, output / demand forecast curve, physical operation parameters, economic characteristic parameters and dispatchable status.

[0039] Spatiotemporal location information: refers to the specific location of distributed energy resources or loads in geographic space and their time-related attributes.

[0040] Output / demand forecast curve: refers to the predicted trend of changes in the power generation of distributed energy or the change trend of load electricity demand over a period of time in the future.

[0041] Physical operating parameters: refers to the physical quantities that describe the current working status of distributed energy equipment or loads, such as voltage, current, frequency, temperature, battery state of charge (SOC), etc.

[0042] Economic characteristic parameters: refers to the cost or benefit information related to the participation of distributed energy or loads in power system operation and market transactions, such as power generation cost, regulation cost, compensation price, market quotation, etc.

[0043] Dispatchable status: refers to whether and to what extent the distributed energy resources or loads can accept and execute dispatch instructions at the current moment, such as whether the equipment is online, whether it is under maintenance, and the adjustable power range.

[0044] This step involves data collection, which requires establishing a real-time data collection unit covering all distributed energy resources (such as photovoltaics, wind power, and energy storage) and controllable loads aggregated by the cross-regional virtual power plant. This real-time data collection unit can connect to the local control units or smart terminals of each distributed resource through various communication technologies (such as SCADA, IoT communications, and dedicated networks). The types of data collected include, but are not limited to: GPS coordinates or other positioning information (spatiotemporal location information) of the device; output or demand curves (output / demand forecast curves) for a future period (e.g., the next 24 hours, with a granularity of 15 minutes or one hour) generated by predictive algorithms based on historical data, weather forecasts, user behavior patterns, etc.; real-time operating data (physical operating parameters) such as voltage, current, active / reactive power, frequency, temperature, energy storage system SOC, and battery health status; economic data (economic characteristic parameters) such as the resource's generation cost, start-up and shutdown costs, maintenance costs, regulation costs, quotes for participating in ancillary services, and market transaction prices; and information reflecting the device's current availability and degree of availability (dispatchability), including its current operating mode (on-grid / off-grid), maintenance plan, fault status, upper and lower limits of adjustable power, and ramp rate limits. The collected data requires preprocessing, including data cleaning, formatting, and missing value filling, before being stored in a real-time database or transmitted to a dispatch server to provide accurate and timely input for subsequent dispatch decisions.

[0045] Step 202 : determining the dynamic weight of each feature dimension based on the global optimization target set in the current scheduling period and the real-time feature data.

[0046] Specifically, it is necessary to first define the global optimization objective for the current dispatch cycle (for example, the next 15 minutes or hour). This objective can be a single one, such as minimizing system operating costs, maximizing renewable energy consumption, or alleviating congestion on a specific transmission line; or it can be a combination of multiple objectives, such as balancing economic efficiency, reliability, and environmental benefits.

[0047] Then, based on this global optimization objective, identify the most relevant key feature dimensions. For example, if the goal is to alleviate congestion, then the resource's geographic location (spatiotemporal location information) and adjustable power range (dispatchability) become very important; if the goal is to minimize operating costs, then the resource's economic characteristics (such as generation cost and regulation cost) and output / demand forecast curves become more critical.

[0048] Next, based on these key feature dimensions and in combination with a pre-set rule base or machine learning model, the weight coefficient of each dimension's impact on achieving the current optimization goal is calculated. This calculation process may need to consider the mutual influence between different dimensions and the dynamic characteristics of the resource itself.

[0049] Finally, based on the resource state constraints reflected in the real-time feature data (such as the physical limits of the equipment and the safe operating range), the initially calculated impact weight coefficients are revised to ensure that the weights reflect the true importance of each dimension under the current actual constraints. This ultimately results in the dynamic weights of each feature dimension used for subsequent dynamic resource clustering. This process is dynamic, meaning that the weights of the feature dimensions will be adjusted accordingly during different scheduling cycles or when the grid state changes.

[0050] Step 203: Generate a dynamic resource cluster partitioning instruction based on the dynamic weight and the real-time feature data. The dynamic resource cluster partitioning instruction includes a member list of a resource cluster and coordination constraints.

[0051] A resource cluster refers to a logical collection formed by dynamically aggregating distributed energy resources and loads within the current dispatch cycle using a clustering algorithm based on real-time characteristic data and dynamic weights determined based on global optimization objectives. The physical significance of a resource cluster lies in the fact that the resources within this collection currently possess similar characteristics (e.g., similar responsiveness, economic characteristics, geographic location, or potential contribution to the current dispatch objective) and physically meet certain coordination constraints (such as grid topology and capacity limitations). Therefore, they can participate in dispatch and power market transactions as a whole, achieving stronger regulation capabilities and better overall performance than individual resources. A resource cluster is not a fixed physical entity, but rather a virtual aggregation unit that changes dynamically based on real-time conditions. This aims to overcome the mismatch between traditional static aggregation methods and the dynamic characteristics of resources, achieving "aggregation and integration."

[0052] During the implementation of this step, the real-time characteristic data of each distributed resource (including spatiotemporal location, prediction curve, physical parameters, economic characteristics, dispatchable status, etc.) is first weighted according to dynamic weights to form a weighted characteristic vector. This vector comprehensively reflects the comprehensive characteristics of the resource under the current scheduling target.

[0053] The weighted feature vectors are then clustered using a clustering algorithm (e.g., K-Means, hierarchical clustering, DBSCAN, etc.). During the clustering process, some basic coordination constraints need to be considered. For example, you can set limits such as the maximum number of cluster members and the minimum cluster capacity, or introduce a geographic penalty in the distance calculation to make it easier for geographically close resources to be grouped into the same cluster.

[0054] The clustering algorithm divides resources into clusters. For each cluster, a resource cluster partitioning instruction is generated. This instruction contains the cluster's membership list (i.e., which distributed resources belong to this cluster) and some basic coordination constraints, such as the upper and lower bounds on the total power that the cluster as a whole must meet during the current scheduling period.

[0055] Step 204: Based on the dynamic resource cluster division instruction, an initial cross-region collaborative scheduling instruction and a compensation price signal are sent to the local agent corresponding to each dynamic resource cluster.

[0056] The initial cross-region coordinated scheduling instruction includes a net exchange power target and ancillary service requirements. The net exchange power target refers to the target value for net power exchange between the dynamic resource cluster and the external power grid (or other regions / resource clusters) during the current scheduling cycle (i.e., total power generation minus total power consumption or the target power export / import). The ancillary service requirements refer to the specific types and quantities of ancillary services that the dynamic resource cluster is required to provide to the power grid during the current scheduling cycle, such as frequency regulation, voltage regulation, or reserve capacity.

[0057] This step converts the dynamic resource clustering instructions into initial dispatch instructions and compensation price signals, which are then sent to the local agent. First, based on the resource clustering instructions generated in step 203, the resources included in each dynamic resource cluster are determined. Then, based on the global optimization objectives set for the current dispatch cycle (e.g., minimizing total system cost, maximizing renewable energy consumption, etc.), a preliminary dispatch optimization model is constructed. This model optimizes each dynamic resource cluster, taking into account each cluster's initial aggregation capacity (e.g., based on a simple superposition of the predicted output / demand curves of its members) and some basic grid constraints. Solving this optimization model yields the initial net exchange power target for each dynamic resource cluster (i.e., the power that the cluster needs to inject into or absorb from the grid) as well as preliminary ancillary service requirements (dynamic power support services that grid operators require energy storage clusters to provide through rapid charging and discharging, such as frequency regulation, voltage stabilization, and black starts. Its core value lies in compensating for real-time power imbalances in the grid through cross-regional coordinated dispatch).

[0058] At the same time, based on preliminary dispatch results and the real-time operating status of the power grid, preliminary compensation price signals are calculated to guide resource cluster adjustments. Finally, these initial net exchange power targets, ancillary service requirements, and compensation price signals are packaged and sent via the communication network to the local agent corresponding to each dynamic resource cluster.

[0059] Step 205: Receive aggregate response boundary information returned by each local agent based on the initial cross-region collaborative scheduling instruction and the compensation price signal, wherein the aggregate response boundary information is used to represent the feasible domain range of the dynamic resource cluster for the instruction.

[0060] Aggregate response boundary information is the multi-dimensional feasibility commitment of each dynamic resource cluster to the scheduling instruction. Its core is to dynamically describe the physical limits and economic response boundaries of the cluster under the constraints of coordinated scheduling through mathematical models. Specifically, it includes three dimensions: Power feasible domain (the maximum charging and discharging power range that the cluster can provide under the current SOC and device status, such as -50MW to +80MW, with an additional ramp rate limit such as ±20MW / min); Energy capability boundary (based on the SOC safety window, such as the sustainable output duration curve derived from 25%-85%. For example, if discharging at 80MW, a drop from 85% to 25% SOC can only last 0.9 hours); Service value mapping (a cost-response sensitivity matrix generated based on compensation price signals, e.g., a 40% increase in responsiveness when frequency regulation compensation is >15 RMB / MWh) is used to quantify the extent to which economic incentives can elastically expand the feasible domain.

[0061] A local agent is a software entity deployed at the regional level or within a virtual power plant management platform. It receives commands and price signals from the dispatch server, manages the specific distributed resources within its dynamic resource cluster, and provides feedback to the upper layer regarding the cluster's aggregate capacity. It can be considered the local representative or manager of the resource cluster, responsible for coordinating the operation of resources within the cluster in response to upper layer commands.

[0062] In this step, the local agent calculates and reports the aggregate response boundary information for the resource cluster to the upper layer based on the initial cross-region coordinated dispatch instructions and compensation price signals received, combined with the real-time status and characteristics of the resources within the dynamic resource cluster it manages. Upon receiving the instructions (including the net exchange power target and ancillary service requirements) and compensation price signals, the local agent immediately begins evaluating the overall capacity of its resource cluster.

[0063] First, the dispatch server calculates the power feasible region through a local agent. The local agent traverses all member resources within the resource cluster, obtaining their real-time operating status (e.g., whether generators are online, the current state of charge (SOC) of energy storage devices, and the current power consumption of interruptible loads), physical constraints (e.g., each resource's maximum / minimum output or power consumption, and ramp rate limits), and device health. The local agent then aggregates the capabilities and constraints of these individual resources, while also considering coordination constraints within the resource cluster (e.g., the need for coordinated actions by certain resources to achieve specific cluster-wide functions). Using an internal aggregation algorithm or model, the local agent calculates the maximum charge / discharge power (i.e., the maximum power injected into or absorbed from the grid) and minimum charge / discharge power that the dynamic resource cluster as a whole can provide at the current moment, forming a power feasible region (e.g., -50 MW to +80 MW). The local agent also calculates the ramp rate limit (e.g., ±20 MW / min), the maximum rate of power change, for the cluster as a whole. These calculations reflect the instantaneous power regulation capabilities of the resource cluster under its current physical conditions.

[0064] Secondly, the local agent calculates the energy capability boundary. This is primarily applicable to resource clusters that include energy storage resources. The local agent obtains the real-time SOC of the energy storage resources and combines this with a preset SOC safety window (e.g., 25%-85%). Based on the current SOC level and the power feasible domain, the local agent can infer how long the resource cluster can continuously discharge or charge at different power levels. For example, if the resource cluster's total energy storage capacity is sufficient and at a high SOC level, and the power feasible domain allows for 80 MW of discharge, the local agent calculates how long (e.g., 0.9 hours) it can continuously discharge at 80 MW while maintaining the SOC within the safety window. By calculating the sustainable duration corresponding to different discharge / charging power levels, a sustainable output duration curve is constructed, which is the energy capability boundary. It reflects the energy throughput capacity of the resource cluster over a period of time.

[0065] Finally, a local agent calculates a service value mapping. The local agent analyzes the potential impact of the received compensation price signal on the willingness and ability of each member resource within the resource cluster to respond. Resources may have different cost functions or response strategies to price signals. The local agent assesses how the number and ability of resources within the resource cluster willing to participate in dispatch or provide ancillary services will change under different compensation price levels. For example, if the frequency regulation compensation price is low, only some low-cost resources may be willing to provide frequency regulation services. However, if the price rises above 15 yuan / MWh, more high-cost or incentivized resources may join, increasing the resource cluster's frequency response capability by 40%. The local agent quantifies the relationship between this price signal and the elastic expansion range of the resource cluster's feasible domain, forming a cost-response sensitivity matrix or similar mapping relationship. This reflects the impact of economic incentives on the dispatch flexibility of the resource cluster.

[0066] The local agent integrates the calculated power feasible region (including power range and ramp rate), energy capability boundary (sustainable output duration curve), and service value mapping (cost-response sensitivity matrix) into aggregated response boundary information, which is then sent back to the scheduling server via the communication network. This information represents the dynamic resource cluster's multi-dimensional commitment to the feasibility of its current scheduling instructions, providing critical real-time feedback for the scheduling server's subsequent coordinated scheduling optimization and instruction correction.

[0067] Step 206: Update the coordinated scheduling instruction and the compensation price signal according to the aggregated response boundary information to obtain the target coordinated scheduling instruction.

[0068] Specifically, after the scheduling server receives the aggregated response boundary information fed back by all dynamic resource clusters, this information contains the multi-dimensional constraints and response characteristics of each resource cluster in its current actual state, which is more accurate and closer to reality than the prediction capabilities used in the initial scheduling model. This information includes: Power feasible region: The actual maximum / minimum charge and discharge power range that each resource cluster can provide at the current moment, as well as the ramp rate limit.

[0069] Energy capability boundary: This mainly targets clusters with energy storage and describes the duration of sustainable output or charging at different power levels, reflecting the energy throughput capacity of the cluster over a period of time.

[0070] Service value mapping: describes the relationship between economic compensation price and resource cluster responsiveness (e.g., adjustable power range, willingness to participate in ancillary services).

[0071] The scheduling server uses this feedback information to build an updated collaborative scheduling optimization model. This new model is similar to the initial model, but its constraints are no longer based on predicted ideal capabilities. Instead, it directly uses the actual aggregate response boundary information fed back by each resource cluster.

[0072] Specifically: Power Constraint Updates: The optimization model's power output or absorption constraints for each resource cluster will directly use the reported power feasible range. For example, if a cluster reports its power feasible range as [-50MW, +80MW], the optimization model will constrain the cluster's power variable to this range. Reported ramp rate limits will also be considered.

[0073] Energy Constraint Updates: For resource clusters with energy storage, the optimization model incorporates their reported energy capacity boundaries. This may be achieved by introducing additional energy balance equations or constraints to ensure that the resource cluster's energy schedule meets the actual energy constraints reflected by its sustainable output duration curve during the scheduling period. For example, if a period of continuous high-power discharge is required, the model will check whether the cluster's energy capacity boundaries support this continuous discharge.

[0074] Economic incentive considerations: Optimization models leverage the feedback service value mapping. This can be achieved by introducing a term related to the compensation price into the objective function, or by including the service value mapping as an additional constraint or parameter. For example, the model can predict and utilize the additional responsiveness that a resource cluster can provide based on the currently set compensation price. This allows for more effective use of economic leverage to guide resource behavior while meeting grid demand.

[0075] The dispatch server solves this updated optimization model (typically a large-scale mathematical programming problem) to recalculate the optimal dispatch target for each dynamic resource cluster within the current dispatch cycle, known as the target coordinated dispatch instructions, by taking into account the actual aggregate capacity constraints of all resource clusters, the overall operational requirements of the power grid (such as load balancing, power flow constraints, and voltage stability), and global optimization objectives (such as minimizing total operating costs and maximizing renewable energy consumption). These target instructions are based on the actual capabilities of the resource clusters, making them more feasible and robust.

[0076] The optimization model may also output new compensation price signals during its solution. These price signals reflect the current system demand and value for different types of regulation capabilities. For example, if demand for frequency regulation services is high during a certain period, the optimization model may calculate a higher frequency regulation compensation price to incentivize resource clusters with frequency regulation capabilities to provide more services. These updated price signals are also included in the target coordinated scheduling instructions and sent to the local agent to further guide optimization and response within the resource cluster.

[0077] The dispatching server uses the real capacity information fed back by the resource cluster to correct the optimization model and re-solve to obtain more realistic and optimized dispatching instructions and price signals, thereby improving the effectiveness and reliability of cross-regional virtual power plant collaborative dispatching.

[0078] Step 207: Send the target collaborative scheduling instruction to each of the dynamic resource clusters for execution to control the operation of cross-regional virtual power plant resources.

[0079] Specifically, the dispatch server will calculate the target net exchange power target value, ancillary service requirements, and the final compensation price signal for each dynamic resource cluster and send it to the corresponding local agent via the communication network. After receiving the target collaborative dispatch instruction, the local agent will decompose it into detailed control instructions for each specific distributed resource within its resource cluster. For example, if the target instruction requires a resource cluster to inject 50MW of power into the grid during a certain period of time, the local agent will determine which generators will increase their output, which energy storage devices will discharge, or which interruptible loads will reduce their consumption, as well as their specific output / consumption values, based on the real-time status, capabilities, and priority of each resource in the cluster.

[0080] The local agent will send these detailed control instructions to each distributed resource controller or intelligent terminal in the cluster, which will directly control the operation of the equipment (such as adjusting the power output of the generator, controlling the charging and discharging of energy storage, and controlling the start and stop of the load).

[0081] In this way, the global optimization instructions of the dispatch server are coordinated and executed by local agents, ultimately achieving effective control and coordinated operation of cross-regional virtual power plant resources.

[0082] This embodiment builds a dynamic, feedback-based, cross-regional virtual power plant collaborative scheduling framework, improving the matching of scheduling instructions with the actual capabilities of large-scale distributed resources. Furthermore, the introduction of a dynamic resource clustering mechanism effectively addresses the problem of static aggregation leading to "unbalanced" resources, thereby improving the efficiency of cross-regional virtual power plant collaborative scheduling and resource utilization.

[0083] Specifically, this embodiment first collects real-time characteristic data for each distributed energy resource and load. This data comprehensively reflects the resource's spatiotemporal characteristics, operational status, economic efficiency, and dispatchability. Unlike fixed partitioning or categorized grouping, this embodiment dynamically determines the weights of each characteristic dimension based on the current global optimization objective and real-time characteristic data. This allows for flexible adjustments to the emphasis placed on different resource types and locations based on real-time operational requirements and resource availability.

[0084] Then, based on these dynamic weights and real-time data, dynamic resource cluster partitioning instructions are generated, clarifying the members of each resource cluster and the coordination constraints between them. This dynamic partitioning method ensures that the composition of resource clusters closely matches the current scheduling goals and resource capabilities, avoiding the incompatibility problems caused by static grouping.

[0085] Next, initial cross-regional coordinated dispatch instructions, along with compensation price signals, containing the net exchange power target and ancillary service requirements, are sent to the local agents corresponding to these dynamic resource clusters, initiating preliminary scheduling. Based on the actual response capabilities of the resource clusters they manage, the local agents return aggregated response boundary information, representing the feasible range of the resource clusters' response to the instructions. Based on this feedback, the dispatch server iteratively updates the coordinated dispatch instructions and compensation price signals to obtain target coordinated dispatch instructions that better reflect actual conditions.

[0086] Finally, the target collaborative scheduling instructions are sent to each dynamic resource cluster for execution, realizing precise control and collaborative operation of cross-regional virtual power plant resources.

[0087] Through this dynamic, real-time resource cluster division and feedback-based iterative optimization process, this embodiment can fully utilize the flexibility and complementarity of distributed resources, overcome the limitations of static aggregation, and make scheduling instructions highly adaptable to actual resource capabilities, thereby achieving efficient coordinated scheduling of cross-regional distributed energy resources on a larger time and space scale, effectively smoothing out volatility, alleviating grid congestion, improving the overall resilience and economy of the system, and alleviating the problem of insufficient adaptability between scheduling instructions and actual resource capabilities.

[0088] This embodiment also provides a more specific cross-regional virtual power plant coordinated scheduling method, including the following steps: S301, collecting real-time characteristic data of each distributed energy resource and load.

[0089] The description of this step can refer to the above embodiment and will not be repeated here.

[0090] S302: Determine a key feature dimension identifier based on a global optimization goal of the current scheduling cycle.

[0091] Specifically, based on the global optimization goal of the current dispatch cycle, such as achieving economically optimal dispatch of a cross-regional power system, key characteristic dimensions can be screened from the numerous factors related to this goal. These include the spatiotemporal distribution characteristics of distributed energy resources (power generation output in different regions and time periods), load characteristics (power consumption patterns of various load types, peak and valley periods, etc.), and grid transmission constraints (capacity limitations of different transmission lines, power flow distribution characteristics). By analyzing the potential impact of these factors on achieving the global optimization goal, the key dimensions can be identified and marked as the basis for subsequent calculations. This ensures that subsequent weighting calculations and other work are carried out around the factors that have the greatest impact on achieving the goal, making the overall dispatch decision more focused and more aligned with the target requirements.

[0092] S303: Calculate the influence weight coefficient of each dimension on the achievement of the goal based on the key feature dimension identifier.

[0093] Specifically, after identifying the key feature dimensions, appropriate mathematical methods are used, such as regression analysis based on historical data, expert experience assignment combined with hierarchical analysis method, to calculate the weight coefficient of each dimension's impact on goal achievement.

[0094] Taking regression analysis as an example, a large number of corresponding samples of data on different key feature dimensions and the achievement of global optimization goals (such as economic costs, power balance and other indicators) in past scheduling cycles are collected to establish a regression model. Through model fitting, the weight ratio of each key feature dimension in influencing the achievement of the goal, that is, the influence weight coefficient, is calculated to quantify the importance of each dimension to achieving the global optimization goal, providing an initial basis for the subsequent dynamic adjustment of the weight.

[0095] In some embodiments, this step may specifically include: S3031, querying a preset target feature mapping rule library to obtain a baseline impact value of each key feature dimension on the optimization target.

[0096] In this embodiment, the preset target feature mapping rule base is a pre-built database or knowledge base that stores the mapping relationship between different key feature dimensions and optimization objectives. These rules are usually set in advance based on historical operating data, expert experience, or simulation analysis. For example, for the "economic dispatch target", the rule base may specify a baseline impact value of 0.3 for "distributed power output forecast accuracy" and a baseline impact value of 0.25 for "load peak-to-valley difference".

[0097] During implementation, the dispatch server retrieves the benchmark impact value corresponding to the key feature dimension identifier (such as spatiotemporal location, output forecast curve) from the rule base according to the global optimization goal of the current dispatch cycle (such as economy, reliability or low-carbon goal), which serves as the initial quantitative basis for the contribution of each dimension to the goal.

[0098] The construction process of the target feature mapping rule base includes: first collecting a large amount of correlation data between key feature dimensions (such as the spatiotemporal location of distributed energy, output forecast curve, etc.) and optimization goals (such as economy, reliability) under different scheduling scenarios, and then organizing field experts to evaluate and score the degree of influence of each dimension under different goals. At the same time, using simulation models to simulate the actual contribution of feature dimensions to the goals under various working conditions, and finally systematically sorting out and standardizing these data, expert evaluations and simulation results to form mapping rules or mapping relationships between different optimization goals and key feature dimensions, which are stored in the rule base to provide a basis for subsequent queries on benchmark impact values.

[0099] S3032: Calculate the actual contribution correction factor of each dimension to the goal achievement based on the baseline impact value and the dynamic characteristics of the resources in the real-time feature data.

[0100] The dynamic characteristics of resources refer to the real-time changing operating capacity parameters of distributed energy resources, including time-varying properties such as current output / load curve, response speed, ramp rate and cost volatility.

[0101] The dynamic characteristics of resources change over time, directly impacting their actual ability to contribute to the global optimization goal at the current moment. For example, whether the current photovoltaic output curve deviates from the predicted value due to cloud cover, whether the energy storage system's response speed to dispatch instructions meets the standards, whether the wind turbine ramp rate meets grid regulation requirements, and whether electricity price fluctuations lead to cost changes.

[0102] During implementation, the baseline impact value of each key feature dimension is coupled with the corresponding real-time dynamic characteristic data for analysis: for example, if the current output of a wind farm is 20% lower than the predicted value due to a sudden drop in wind speed, the baseline impact value of its "output prediction curve" dimension needs to be multiplied by a correction factor (such as 0.8); if the response speed of a certain energy storage device is 10% slower than the rated value, the correction factor of its "response speed" dimension may be 0.9.

[0103] Through this type of dynamic matching, the actual contribution correction factor of each dimension to the goal achievement is obtained to reflect the adjustment of the impact of real-time working conditions on the benchmark S3033: Normalize the actual contribution correction factor to obtain the impact weight coefficient.

[0104] Since the actual contribution correction factors of each dimension may be in different dimensions or numerical ranges (such as the output curve correction factor is 0.7-1.2, and the response speed correction factor is 0.8-1.1), they need to be converted into comparable weight coefficients through normalization.

[0105] During implementation, linear normalization methods (such as minimum-maximum normalization) or standardization methods (such as Z-score normalization) are typically used to map all correction factors to the interval [0,1], and ensure that the sum of the weight coefficients of each dimension is 1. For example, if the correction factors of three key dimensions in a certain scheduling cycle are 0.9, 1.1, and 0.8, respectively, the weight coefficients become 0.3, 0.367, and 0.267 after normalization, thus forming a dynamic weight system that reflects the characteristics of real-time resources and provides a quantitative basis for subsequent resource cluster division and scheduling instruction generation.

[0106] S304: Modify the impact weight coefficient according to the resource status constraint in the real-time feature data to obtain the dynamic weight.

[0107] The resource status constraint refers to the physical boundary conditions that limit the feasible domain of resource scheduling. It is an insurmountable constraint range determined by the physical characteristics of the equipment itself and the grid structure. Specifically, it may include rigid constraints such as maximum / minimum power limits, energy storage SOC safety range, equipment start / stop status, and the network topology connection relationship formed by the connection of various components in the grid.

[0108] After obtaining resource status constraint information from real-time feature data, these rigid constraints are combined to modify the previously calculated impact weight coefficients. For example, if a distributed power source is shut down due to equipment start / stop constraints, the impact weight coefficient originally calculated based on the spatiotemporal distribution feature dimension of that power source must be adjusted downward to account for its inability to participate in scheduling. If the energy storage SOC is at the edge of the safe range, the weight coefficients of the relevant feature dimensions related to energy storage must also be adjusted to ensure equipment safety.

[0109] In this way, the physical boundary conditions of the resource scheduling feasible domain (maximum / minimum power limits, energy storage SOC safety range, etc.) are incorporated, allowing the weight coefficient to dynamically reflect the impact of each dimension on the target under the current actual resource constraints, making subsequent scheduling decisions based on dynamic weights more in line with actual operating scenarios, and enhancing the scientificity and feasibility of scheduling. S305: Generate a resource weighted feature vector according to the dynamic weight and the real-time feature data.

[0110] Specifically, the dispatch server first extracts multidimensional features of each distributed energy resource (such as spatiotemporal location coordinates, output forecast curve values, and response speed parameters) from real-time feature data. It then multiplies each feature dimension by its corresponding dynamic weight to obtain a weighted feature value. For example, if the "spatial location" dimension of a photovoltaic power station has a weight of 0.25, and its geographic coordinates are (0.6, 0.3) after normalization, the weighted coordinates are (0.15, 0.075). Similarly, if the "output forecast curve" dimension has a weight of 0.3 and the current prediction error rate is 8%, the weighted value is 0.024.

[0111] The weighted eigenvalues ​​of all dimensions are then combined in order to form the resource weighted eigenvector of the resource point, so that each component in the vector reflects the influence weight of the corresponding feature on the global optimization target.

[0112] S306: Divide the resource weighted feature vectors using a constrained clustering algorithm to obtain an initial resource grouping solution.

[0113] Specifically, the dispatch server can use a clustering algorithm that incorporates physical constraints (such as spectral clustering with power caps or constrained K-means) to group resources using weighted feature vectors as input. When calculating similarity between resources (such as Euclidean distance or cosine similarity), the algorithm simultaneously incorporates grid topology constraints (such as transmission line capacity limits) and equipment operation constraints (such as energy storage SOC safety ranges and wind turbine ramp rate limits).

[0114] For example, during the clustering process, if two resource points have similar characteristics but belong to different electrical islands (constrained by the network topology), the algorithm will force them to be divided into different groups; if the total regulated power of the resources in a group exceeds the line transmission capacity, the grouping will be readjusted.

[0115] By iteratively optimizing cluster centers and group boundaries, an initial resource grouping scheme that satisfies the constraints is ultimately generated. This initial resource grouping scheme is a preliminary resource grouping result. While considering the similarity of multidimensional resource characteristics, it also incorporates physical conditions such as grid topology constraints and equipment operating parameter restrictions. Distributed energy resources, energy storage, and loads are clustered into several groups based on functional complementarity and constraint compatibility. Although this is a preliminary grouping, it initially satisfies some basic constraints. Further optimization is required through physical operability verification to form a viable resource cluster entity.

[0116] In some embodiments, this step may specifically include: S3061: Construct a resource similarity matrix based on the resource weighted feature vector.

[0117] The purpose of this step is to quantify the degree of similarity between any two distributed energy resources or loads in the weighted feature space. Specifically, for each pair of resources, the distance or similarity measure between their respective weighted feature vectors is calculated, for example, using Euclidean distance, cosine similarity, or other metrics appropriate to the feature type. The smaller the Euclidean distance or the larger the cosine similarity, the more similar the two resources are in terms of their weighted features. The similarity values ​​between all pairs of resources are constructed into a symmetric matrix, namely the resource similarity matrix, where each element of the resource similarity matrix represents the similarity between the corresponding two resources.

[0118] Specifically, the dispatch server compares the weighted feature vectors of all distributed energy resources pairwise, using Euclidean distance or cosine similarity algorithms to calculate the similarity between resources, forming an N×N matrix (where N is the total number of resources). Each element S(i, j) in the matrix represents the feature similarity between resource i and resource j. A larger value indicates a higher degree of match between the two in terms of spatial and temporal location, output characteristics, and other dimensions. For example, the weighted feature vectors of a photovoltaic power station and an energy storage device have a similarity of 0.8 in the dimension of "temporal and spatial complementarity" and 0.6 in the dimension of "response speed." The overall similarity, after normalization, is 0.75. This value is then entered into the corresponding position in the matrix to construct a complete resource similarity matrix.

[0119] S3062: According to the physical constraints of the power grid, the connection weights of the resource similarity matrix are modified to obtain a constraint-modified similarity matrix.

[0120] Among them, the physical constraints of the power grid refer to rigid restrictions determined by the power grid topology, equipment operating characteristics, etc., including the transmission capacity of transmission lines, distribution of electrical islands, equipment power limits, etc. These conditions will directly affect the feasibility of resource clustering and scheduling.

[0121] The connection weight is a numerical value in the resource similarity matrix that represents the degree of feature matching between different resources, reflecting the possibility and closeness of actual collaborative scheduling of resources.

[0122] Correction methods typically implement this by adjusting the values ​​of corresponding elements in the similarity matrix. For pairs of resources that must be grouped together, their similarity can be significantly increased (or their distance reduced), making them more likely to be grouped together during clustering. For pairs of resources that must be grouped together, their similarity can be significantly reduced (or their distance increased), making them more likely to be grouped together. The resulting matrix is ​​called a constrained-corrected similarity matrix.

[0123] Specifically, the dispatch server can access the grid topology model and real-time operating data to perform constraint corrections on the initial similarity matrix. If the transmission line between resource i and resource j is blocked (transmission capacity falls below a preset threshold) or belongs to different electrical islands (no physical connection), the matrix S(i, j) is multiplied by a penalty factor (e.g., 0.3) or simply set to 0. If the maximum output limit of resource i and the load demand of resource j, after clustering, could cause local grid overload, the connection weight between them is reduced. For example, if a wind farm and industrial load have similar characteristics, but the current transmission capacity of the line between them is only 20% of the rated value, the similarity weight between the two is corrected from 0.6 to 0.6 × 0.2 = 0.12, forming a corrected similarity matrix that reflects the physical constraints of the grid.

[0124] S3063: Perform spectral clustering eigendecomposition on the constraint-corrected similarity matrix to obtain an initial resource grouping solution that satisfies the constraints.

[0125] Among them, spectral clustering is a clustering method based on graph theory, which regards data points as vertices of the graph and the similarity between points as the weight of the edge.

[0126] In this step, the constraint-corrected similarity matrix is ​​considered to be the adjacency matrix of a similarity graph. The dispatch server converts the constraint-corrected similarity matrix into a Laplace matrix (L=DS, where D is a diagonal matrix and the diagonal elements are the sum of the similarities in each row), performs eigendecomposition on the Laplace matrix, and extracts the eigenvectors corresponding to the first k smallest eigenvalues ​​(k is the preset number of clusters). These eigenvectors are combined into a new matrix, where each row represents the k-dimensional feature representation of a resource. The matrix is ​​then clustered using algorithms such as K-means to obtain an initial resource grouping scheme. For example, after extracting the first three eigenvectors, each resource is mapped to a three-dimensional space and then clustered into three groups to ensure that the characteristics of resources in the same group are highly similar while satisfying the grid constraints. Ultimately, an initial resource grouping that meets the topology constraints and equipment operation restrictions is generated.

[0127] In this embodiment, by encoding constraint information in the similarity matrix, spectral clustering can effectively consider these restrictions in the clustering process.

[0128] S307: Verify the physical operability of the initial resource grouping solution to obtain a resource cluster entity that meets the coordination constraint conditions.

[0129] Specifically, the dispatch server evaluates whether the aggregate characteristics of all resources within the group meet the physical constraints of grid operation, such as total power margin, voltage stability, and power flow limits. Secondly, it checks whether the communication, control, and physical connections between resources within the group support collaborative operation. Finally, it verifies whether each resource's individual operational constraints (such as energy storage SOC range and minimum device operating time) can be met under group collaborative control. Only after passing all these physical verifications can the group be confirmed as a resource cluster entity with collaborative capabilities and participate in subsequent scheduling.

[0130] Multi-dimensional physical constraint verification for each resource cluster in the initial grouping scheme can include: ① Power balance verification, calculating the real-time difference between the power output and load demand within the cluster to ensure that the power limit within the cluster and the interconnected lines is not exceeded; ② Equipment status verification, checking whether the energy storage SOC within the cluster is within the safe range (such as 20%-80%) and whether the generator sets are in a state that allows start and stop; ③ Topology connectivity verification, confirming through the power grid topology model whether the resources within the cluster are connected through physical lines to avoid the formation of undispatched islands; ④ Response capability verification, evaluating whether the comprehensive response speed of the resources within the cluster meets the dispatch instruction requirements (such as a second-level response to frequency regulation requirements).

[0131] If any verification item fails, the grouping parameters are adjusted and clustering is performed again until all resource clusters meet the coordination constraints.

[0132] S308: Generate a dynamic resource cluster partitioning instruction based on the member list and constraint boundary of each resource cluster entity.

[0133] Specifically, first, organize the member list by cluster, and clarify the unique identifiers (such as device ID, geographic location code) and parameters (such as rated power and response speed) of the distributed power supply, energy storage, controllable load and other equipment contained in each cluster; second, determine the cluster-level constraint boundaries, including upper and lower limits of net exchange power, maximum ramp rate, energy storage SOC range and other operating parameters; finally, add coordinated scheduling rules, such as complementary scheduling strategies for power supply and energy storage within the cluster (energy storage discharge when photovoltaic output is low) and load regulation priorities (priority protection for important loads).

[0134] The dynamic resource cluster partitioning instruction is encapsulated in a standardized data format (such as JSON or XML), including metadata such as version number and effective time, and is sent to the corresponding local agent for execution through the scheduling communication network.

[0135] For each resource cluster entity, the dynamic resource cluster partitioning instruction needs to clearly specify the unique identifiers (member list) of the specific resource units it contains, so that these resources can be uniformly managed and controlled during scheduling execution. At the same time, the instruction also needs to provide the equivalent operating constraint boundaries presented to the outside world by the resource cluster as an aggregate entity. This may include the aggregated total output / load range, aggregated response speed, aggregated ramp rate, aggregated cost function, etc. This information constitutes the core content of the dynamic resource cluster partitioning instruction, guiding upper-level scheduling decision makers to optimize the resource cluster as a whole.

[0136] S309: Generate a power adjustment interval boundary of each dynamic resource cluster based on the topology structure in the dynamic resource cluster division instruction.

[0137] Among them, the topological structure refers to the network layout composed of various nodes in the power grid (such as power sources, loads, and energy storage equipment nodes) and their physical connection relationships (such as transmission lines, transformers, etc.), which includes physical attribute information such as the electrical connection method between nodes, line transmission capacity, and electrical island distribution. It is used to clarify the spatial distribution of resource clusters and the constraints of power transmission paths.

[0138] The dispatch server parses the grid topology data (such as node connectivity and line parameters) contained in the dynamic resource cluster partitioning instructions and combines it with the physical operating characteristics and constraints of all resources within the resource cluster, such as the maximum and minimum generator outputs, the load's adjustable range, the energy storage capacity and charge / discharge power limits, and each resource's own ramp rate. By overlaying and comprehensively considering these individual resource constraints, the total power regulation range that the resource cluster as a whole can provide at the current moment or over a period of time in the future is determined—the aggregated power regulation range boundary. This boundary represents the flexibility and capabilities of the resource cluster as a dispatching unit.

[0139] For example, a resource cluster contains two wind turbines (each rated power 1.5MW), one energy storage system (charging and discharging power ±2MW), and the maximum transmission capacity of the transmission line within the cluster is 5MW. In this case, the total power regulation range of the cluster is (-2MW, 5MW), where negative values ​​indicate energy storage discharge or load reduction, and positive values ​​indicate power output or energy storage charging. The boundary value is subject to both line capacity and equipment limits.

[0140] S310: Construct a distributed scheduling optimization model according to the power adjustment interval boundary.

[0141] Among them, the distributed scheduling optimization model can include a global objective function (such as minimizing the total system operating cost, maximizing social welfare, etc.) and system-level operating constraints (such as full network power balance, transmission line flow restrictions, etc.).

[0142] The key to the distributed scheduling optimization model in this embodiment lies in incorporating the aggregate power adjustment range boundaries of resource clusters as cluster-level constraints. By employing a distributed optimization framework such as decomposition and coordination methods (e.g., Lagrangian relaxation, ADMM), the complex network-wide scheduling problem is decomposed into multiple relatively independent intra-cluster subproblems and a coordinator problem. This allows each resource cluster to gradually approach the global optimal solution through interaction with the coordinator, while satisfying its own boundary constraints.

[0143] When adopting a distributed optimization framework (such as ADMM and Lagrangian relaxation), the network-wide scheduling problem is decomposed into multiple intra-cluster sub-problems and a coordinator problem. Each intra-cluster sub-problem uses the aggregate power adjustment interval boundary of the resource cluster as the core constraint, defining the feasible power adjustment range of the cluster under the current coordination signal. The coordinator problem is responsible for handling global constraints such as network-wide balance constraints and network power flow constraints. Through iterative interaction, it coordinates the local decisions of each cluster so that they can gradually converge to the globally optimal scheduling solution that meets the constraints of the entire network while satisfying their respective power adjustment interval boundaries.

[0144] S311, based on the distributed scheduling optimization model, calculate the optimal power instruction initial value in the cluster to obtain the collaborative scheduling instruction initial value matrix.

[0145] Among them, the initial value of the optimal power instruction is the optimal initial power adjustment value of each dynamic resource cluster obtained by solving the constraints such as the power adjustment interval boundary and the transmission capacity of the interconnection line, which provides the basic execution parameters for cross-regional coordinated scheduling.

[0146] After constructing the distributed scheduling optimization model, the scheduling server initiates the model's solution process. During the initial iteration, a coordination signal (such as a shadow price or target power) is generated based on preliminary information and sent to the local agents corresponding to each resource cluster. Upon receiving the coordination signal, each local agent solves the subproblem corresponding to its cluster based on the aggregate power adjustment interval boundaries and the specific characteristics of its member resources, calculating the optimal power response or the optimal output / load instructions for its internal resources under the current coordination signal. These calculated initial optimal power instructions (aggregate values ​​or decomposed values ​​for internal resources) for each resource cluster are aggregated to form an initial value matrix for coordinated scheduling instructions, which serves as the basis for subsequent iterations or actual instructions.

[0147] Specifically, the dispatch server invokes an optimization solver (such as Gurobi or CPLEX) to solve the distributed dispatch optimization model and obtain the optimal initial power regulation command values ​​for each resource cluster. For example, if a cluster's current PV output is 3 MW, the load demand is 5 MW, and the energy storage SOC is 50%, the model calculates that the energy storage needs to discharge 1.5 MW, while the tie line purchases 0.5 MW from a neighboring cluster to achieve power balance and minimize cost. This 1.5 MW discharge command serves as the initial power regulation value for that cluster. The initial values ​​for all clusters are organized in a matrix to form a coordinated dispatch command initial value matrix, where the matrix elements correspond to the active power regulation amount (in MW) for each cluster.

[0148] S312: Generate initial cross-regional collaborative dispatch instructions and compensation price signals based on the collaborative dispatch instruction initial value matrix, grid congestion marginal cost, and regulation cost.

[0149] Among them, the marginal cost of grid congestion refers to the additional economic cost incurred for each additional unit of power transmission when the transmission of electricity is blocked in a specific area or time period due to capacity limitations of transmission lines; the regulation cost refers to the operating cost or compensation fee incurred for regulating distributed energy, energy storage or loads (such as charging and discharging, starting and stopping) in order to achieve grid power balance.

[0150] The initial cross-region coordinated scheduling instructions are refined based on the initial value matrix and take into account the current system operating requirements and constraints. They clearly define the target power or regulation tasks for each resource cluster in the next scheduling cycle. The compensation price signal is typically calculated based on the coordination signal sent by the coordinator during the distributed optimization process (such as the dual variable corresponding to the system-level constraints) and the resource's own regulation cost. It is used to incentivize resource clusters to adjust according to the scheduling instructions and compensate them for the services they provide or the costs they incur.

[0151] Specifically, the dispatch server calculates the power flow of each inter-cluster tie line based on the initial value matrix of the coordinated dispatch instructions. This calculation, combined with the grid's real-time congestion marginal cost (LMP), determines the economic compensation for cross-regional dispatch. For example, if the current LMP difference for a tie line is 0.15 yuan / kWh, and a cluster transmits 1MW of power to a neighboring cluster via this tie line for one hour, the congestion compensation is 150 yuan. Furthermore, the cluster's resource regulation costs (such as the energy storage charging and discharging loss cost of 0.08 yuan / kWh) are taken into account to comprehensively generate the cluster's initial cross-regional coordinated dispatch instructions (such as "discharge 1MW for one hour") and compensation price signals (such as a unit price of 0.23 yuan / kWh).

[0152] In some embodiments, the initial cross-region collaborative scheduling instruction may also include information such as scheduling period and priority, and be packaged in a standardized format.

[0153] S313: Send the initial cross-region collaborative scheduling instruction and the compensation price signal based on the mapping relationship between the dynamic resource cluster and the local agent.

[0154] The dispatch server sends the initial cross-regional collaborative dispatch instructions and compensation price signals to each local agent through the SCADA or EMS communication network based on the predefined mapping table between dynamic resource clusters and local agents (such as the correspondence between cluster ID and regional controller IP address).

[0155] For example, cluster A corresponds to the North China regional agent, and cluster B corresponds to the East China regional agent. The instructions are transmitted through a dedicated communication protocol (such as IEC 61850). After receiving the instructions, the local agent parses them and converts them into control signals for local devices (such as the power setting value of the energy storage inverter), and performs settlement preparation work based on the compensation price signal.

[0156] After receiving the instructions and signals, the local agent is responsible for decomposing and executing them within the cluster, coordinating and managing its member resources to adjust power according to the instructions, and making corresponding economic settlements or responses based on the compensation price signals. The mapping relationship between dynamic resource clusters and local agents ensures that scheduling instructions are accurately transmitted to the entities responsible for execution.

[0157] S314: Receive aggregate response boundary information returned by each of the local agents based on the initial cross-region collaborative scheduling instruction and the compensation price signal.

[0158] This step can refer to the description of the above embodiment and will not be repeated here.

[0159] S315: Determine a power regulation deviation matrix based on the initial cross-region collaborative scheduling instruction and the aggregated response boundary information.

[0160] This step compares the initial cross-region coordinated scheduling instructions sent to each dynamic resource cluster in the previous round with the aggregate response boundary information received from each cluster. The aggregate response boundary information reflects the actual power adjustment range or optimal response that the resource cluster can provide in its current state. By comparing the target power or adjustment amount in the instruction with the range allowed by the response boundary or the actual response value, the power adjustment deviation of each resource cluster can be calculated. This deviation can be the instruction value exceeding the cluster's capabilities or the difference between the cluster's optimal response after receiving the compensation price signal and the instruction value. These deviation values ​​are arranged by resource cluster to form a power adjustment deviation matrix.

[0161] For example, if a cluster's initial dispatch instruction calls for a 2MW discharge, but the cluster's actual adjustable range is 1.5-2.5MW (the aggregate response boundary), the power regulation deviation is 0.5MW. If a tie line is initially scheduled to transmit 3MW, but its current available capacity is only 2.8MW, the deviation is 0.2MW. The deviation values ​​for all clusters and tie lines are arranged according to a preset index to form a power regulation deviation matrix. A positive matrix element indicates that the instruction exceeds the response capacity, while a negative element indicates that the response capacity is redundant.

[0162] S316 , updating the compensation price signal according to the power regulation deviation matrix and the real-time congestion data to obtain a compensation price update signal.

[0163] In distributed optimization-based frameworks, compensation price signals are typically associated with the grid's marginal costs or dual variables of constraints. When there are significant deviations in the power regulation of a resource cluster, or when congestion occurs in a region of the grid, the corresponding compensation price signal can be adjusted to incentivize the resource cluster to change its behavior. For example, if transmission congestion occurs in a region, the compensation prices of the resource clusters associated with that region may be adjusted to encourage them to make adjustments that help alleviate the congestion. This update process is typically iterative, aiming to guide the local optimal decisions of each cluster toward a global optimal solution through adjustments to the price signal, thereby obtaining an updated compensation price signal.

[0164] Specifically, the dispatch server dynamically adjusts compensation prices based on the power regulation deviation matrix and real-time grid congestion data (such as line LMP differences and congestion duration). If a deviation in a region causes increased line congestion (e.g., a deviation increases a line's load factor from 80% to 90%), the congestion compensation price for that region is increased (e.g., from 0.1 yuan / kWh to 0.15 yuan / kWh). If a cluster's insufficient response capacity leads to a scheduling adjustment and requires the transfer of power regulation tasks to another cluster, the receiving cluster is compensated for the regulation service (e.g., 0.08 yuan / kWh). The updated compensation price signal incorporates the latest economic incentive parameters for each region and time period.

[0165] S317: Modify the initial cross-region collaborative scheduling instruction based on the compensation price update signal feedback to obtain a target collaborative scheduling instruction.

[0166] The scheduling server feeds the compensation price update signal back to the local agent corresponding to each dynamic resource cluster. Upon receiving the updated compensation price signal, the local agent recalculates the optimal power response within its resource cluster or optimizes the scheduling of its internal resources based on this signal. This new optimal response or scheduling result reflects the resource cluster's decision after considering the latest compensation price signal.

[0167] The dispatch server collects this new response information (or information implicit in the price signal itself) and, based on this feedback, modifies the initial cross-region coordinated dispatch instructions. The modified instructions are closer to the global optimal solution for the system's current state, taking into account the actual capabilities of each resource cluster and grid constraints. After several iterations of this revision process (in actual applications, this may be achieved through a few iterations or generated all at once based on forecast information), the final converged instructions become the target coordinated dispatch instructions, which are used to guide the actual operation of each resource cluster.

[0168] Specifically, the dispatch server uses the compensation price update signal as an economic constraint, re-solves the distributed dispatch optimization model, and modifies the dispatch instructions to minimize the total cost (including congestion cost and regulation cost). For example, due to a compensation price increase in one area, the originally planned 2MW power regulation task is partially transferred to an adjacent area (for example, 1.2MW). At the same time, the energy storage charging and discharging strategy is adjusted (for example, adding 0.3MW discharge) to balance the deviation. The modified instructions contain the new power regulation amount, regulation period, and priority of each cluster, forming the final target coordinated dispatch instruction, ensuring economically optimal dispatch while meeting physical constraints.

[0169] S318: Send the target collaborative scheduling instruction to each of the dynamic resource clusters for execution to control the operation of cross-regional virtual power plant resources.

[0170] This step can refer to the description of the above embodiment and will not be repeated here.

[0171] The method of this embodiment, by introducing dynamic resource cluster division and a cross-regional collaborative dispatch method based on distributed optimization, can improve the dispatch efficiency, flexibility, and economy of the power system in complex operating environments. Its core technical effects are reflected in the following aspects: First, it improves the refinement and flexibility of dispatching. Traditional dispatching often focuses on individual units or substations, making it inefficient when dealing with massive distributed resources and flexible loads. This solution dynamically groups resources with similar characteristics or close geographical locations into resource clusters, enabling aggregated management of large-scale heterogeneous resources. This aggregation not only simplifies the decision-making burden at the dispatch center but also more accurately understands the overall regulatory capacity of regional resources, enabling more refined dispatching control and improving the system's ability to cope with fluctuations.

[0172] Secondly, it enhances the collaborative optimization capabilities of cross-regional resources. By building a distributed scheduling optimization model, this solution can effectively coordinate resource clusters of different types and regions to participate in system regulation. This distributed optimization framework allows each resource cluster to perform local optimization under the control of a local agent while simultaneously achieving global coordination through information exchange with the coordinator (such as compensation price signals). This breaks down regional barriers, enabling cross-regional resources to more effectively participate in grid balancing and congestion management, improving the overall operational efficiency and economic efficiency of the system.

[0173] Third, it improves the economic efficiency of dispatch. The scheme considers the marginal cost of grid congestion and regulation costs, and uses compensation price signals to guide resource cluster regulation. The compensation price mechanism reflects resource scarcity and the value of regulation, prioritizing resources with low costs and strong regulation capabilities for dispatch, thereby reducing the system's overall operating costs. Especially when addressing grid congestion, compensation prices can effectively guide resources to regulate in a way that helps alleviate the congestion, avoiding costly wind and solar curtailment or power outages.

[0174] Fourth, it improves the system's robustness and adaptability. Dynamic resource clustering allows flexible adjustment of resource cluster composition based on factors such as system operating status and resource availability, enabling scheduling solutions to better adapt to dynamic changes in the grid structure and the uncertainty of renewable energy output. The distributed scheduling architecture delegates some decision-making power to local agents, reducing the computational and communication burdens on central nodes and improving the system's anti-interference capabilities and operational stability.

[0175] Finally, this technology provides an effective dispatching tool for new power systems. With the integration of a high proportion of renewable energy, distributed power sources, and flexible loads, power systems are characterized by high randomness, volatility, and strong coupling. The dynamic aggregation and distributed collaborative dispatching methods provided by this technical solution can effectively manage and utilize these new resources, providing key technical support for building a new power system dominated by renewable energy.

[0176] In summary, this technical solution achieves effective aggregation management and cross-regional optimization and coordination of large-scale heterogeneous resources through dynamic resource cluster division and distributed collaborative scheduling, thereby improving the refinement, flexibility, economy and robustness of power system scheduling.

[0177] The method provided in the above embodiment is executed by the scheduling server. It can be understood that the scheduling server is an electronic device. The following describes the electronic device in the embodiment of the present invention from the perspective of hardware processing. Figure 3 , which is a schematic diagram of a physical device structure of an electronic device in an embodiment of the present invention.

[0178] It should be noted that Figure 3The structure of the electronic device shown is only an example and should not limit the functions and scope of use of the embodiments of the present invention.

[0179] like Figure 3 As shown, the electronic device includes a central processing unit (CPU) 401, which can perform various appropriate actions and processes, such as the methods described in the above embodiments, based on programs stored in a read-only memory (ROM) 402 or programs loaded from a storage unit 408 into a random access memory (RAM) 403. RAM 403 also stores various programs and data required for system operation. CPU 401, ROM 402, and RAM 403 are interconnected via a bus 404. An input / output (I / O) interface 405 is also connected to bus 404.

[0180] The following components are connected to the input / output (I / O) interface 405: an input section 406 including an audio input device, push button switches, and the like; an output section 407 including a display, an audio output device, indicator lights, and the like; a storage section 408 including a hard disk and the like; and a communication section 409 including a network interface card such as a LAN (Local Area Network) card or a modem. The communication section 409 performs communication processing via a network such as the Internet. A drive 410 is also connected to the input / output (I / O) interface 405 as needed. Removable media 411, such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory, is installed in the drive 410 as needed, so that computer programs read from the media can be installed in the storage section 408 as needed.

[0181] In particular, according to an embodiment of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, an embodiment of the present invention includes a computer program product comprising a computer program carried on a computer-readable medium, the computer program including a computer program for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via the communication section 409 and / or installed from removable media 411. When executed by the central processing unit (CPU) 401, the computer program performs the various functions defined in the present invention.

[0182] It should be noted that specific examples of computer-readable storage media may include, but are not limited to, an electrical connection having one or more conductors, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0183] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present invention. Each box in the flowchart or block diagram can represent a module, program segment, or part of the code, and the above-mentioned module, program segment, or part of the code contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings.

[0184] Specifically, the electronic device of this embodiment includes a processor and a memory, the memory is coupled to one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and one or more processors call the computer instructions to enable the electronic device to execute the method provided by the above embodiment.

[0185] As another aspect, the present invention further provides a computer-readable storage medium, which may be included in the electronic device described in the above embodiments, or may exist independently and not incorporated into the electronic device. The storage medium carries one or more computer programs, and when executed by a processor of the electronic device, the electronic device implements the methods provided in the above embodiments.

[0186] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments can still be modified, or some of the technical features thereof can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present invention.

[0187] As used in the above embodiments, the term “when” may be interpreted to mean “if” or “after” or “in response to determining that” or “in response to detecting that”, depending on the context. Similarly, the phrases “upon determining that” or “if (stated condition or event) is detected” may be interpreted to mean “if determining that” or “in response to determining that” or “upon detecting (stated condition or event)” or “in response to detecting (stated condition or event)”, depending on the context.

[0188] Those skilled in the art will appreciate that all or part of the process steps in the above-described method embodiments can be implemented by a computer program instructing the relevant hardware. The program can be stored in a computer-readable storage medium, and when executed, the program can include the process steps in the above-described method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM or random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A cross-regional virtual power plant collaborative scheduling method, characterized in that: include: Collecting real-time characteristic data of each distributed energy resource and load, wherein the real-time characteristic data includes at least one of temporal and spatial location information, output / demand forecast curve, physical operation parameters, economic characteristic parameters, and dispatchable status; Determining the dynamic weight of each feature dimension based on the global optimization goal set in the current scheduling period and the real-time feature data; generating a dynamic resource cluster partitioning instruction according to the dynamic weight and the real-time feature data, wherein the dynamic resource cluster partitioning instruction includes a member list of the resource cluster and coordination constraints; Based on the dynamic resource cluster division instruction, an initial cross-region collaborative scheduling instruction and a compensation price signal are sent to the local agent corresponding to each dynamic resource cluster, wherein the initial cross-region collaborative scheduling instruction includes a net exchange power target value and ancillary service requirements; receiving aggregate response boundary information returned by each of the local agents based on the initial cross-region collaborative scheduling instruction and the compensation price signal, wherein the aggregate response boundary information is used to represent a feasible domain range of the dynamic resource cluster for the instruction; updating the coordinated scheduling instruction and the compensation price signal according to the aggregated response boundary information to obtain the target coordinated scheduling instruction; The target collaborative scheduling instruction is sent to each of the dynamic resource clusters for execution to control the operation of cross-regional virtual power plant resources.

2. The method according to claim 1, characterized in that The determining of the dynamic weight of each feature dimension based on the global optimization goal set in the current scheduling period and the real-time feature data includes: Determine the key feature dimension identifiers based on the global optimization goals of the current scheduling cycle; Based on the key feature dimension identifiers, calculate the weight coefficient of each dimension's impact on goal achievement; According to the resource status constraints in the real-time feature data, the impact weight coefficient is corrected to obtain the dynamic weight; the resource status constraints refer to the physical boundary conditions that limit the feasible domain of resource scheduling, including maximum / minimum power limits, energy storage SOC safety range, equipment start and stop status and network topology connection relationship.

3. The method according to claim 2, characterized in that The calculation of the influence weight coefficient of each dimension on the goal achievement based on the key feature dimension identifier includes: Querying a preset target feature mapping rule library to obtain a baseline impact value of each key feature dimension on the optimization target; Calculate the actual contribution correction factor of each dimension to the goal achievement based on the baseline impact value and the resource dynamic characteristics in the real-time characteristic data; the resource dynamic characteristics refer to the real-time changing operating capability parameters of the distributed energy resource, including the current output / load curve, response speed, ramp rate, and cost volatility; The actual contribution correction factor is normalized to obtain the influence weight coefficient.

4. The method according to claim 1, wherein The generating a dynamic resource cluster partitioning instruction according to the dynamic weight and the real-time feature data includes: generating a resource weighted feature vector according to the dynamic weight and the real-time feature data; The resource weighted feature vector is divided and processed by a constrained clustering algorithm to obtain an initial resource grouping scheme; Verifying the physical operability of the initial resource grouping scheme to obtain a resource cluster entity that meets the coordination constraint conditions; Based on the member list and constraint boundaries of each resource cluster entity, a dynamic resource cluster partitioning instruction is generated.

5. The method according to claim 4, characterized in that The resource weighted feature vector is divided and processed by the constrained clustering algorithm to obtain an initial resource grouping scheme, including: Constructing a resource similarity matrix based on the resource weighted feature vector; According to the physical constraints of the power grid, the connection weights of the resource similarity matrix are modified to obtain a constraint-modified similarity matrix; Spectral clustering eigendecomposition is performed on the constraint-corrected similarity matrix to obtain an initial resource grouping scheme that meets the constraints.

6. The method according to claim 1, characterized in that The sending of the initial cross-region collaborative scheduling instruction and the compensation price signal to the local agent corresponding to each dynamic resource cluster based on the dynamic resource cluster division instruction includes: generating a power adjustment interval boundary of each of the dynamic resource clusters based on a topological structure in the dynamic resource cluster partitioning instruction; Constructing a distributed scheduling optimization model according to the power regulation interval boundary; Based on the distributed scheduling optimization model, the optimal power instruction initial value in the cluster is calculated to obtain the collaborative scheduling instruction initial value matrix; generating an initial cross-regional collaborative dispatch instruction and a compensation price signal based on the collaborative dispatch instruction initial value matrix, the grid congestion marginal cost, and the regulation cost; Based on the mapping relationship between the dynamic resource cluster and the local agent, the initial cross-region collaborative scheduling instruction and the compensation price signal are sent.

7. The method according to claim 1, characterized in that The updating of the coordinated scheduling instruction and the compensation price signal according to the aggregated response boundary information to obtain the target coordinated scheduling instruction includes: Determining a power regulation deviation matrix based on the initial cross-region collaborative scheduling instruction and the aggregated response boundary information; updating the compensation price signal according to the power regulation deviation matrix and the real-time congestion data to obtain a compensation price update signal; The initial cross-region collaborative scheduling instruction is modified based on the compensation price update signal feedback to obtain a target collaborative scheduling instruction.

8. An electronic device, characterized in that: including one or more processors and memory; The memory is coupled to the one or more processors, and is configured to store computer program codes, where the computer program codes include computer instructions. The one or more processors call the computer instructions to enable the electronic device to execute the method according to any one of claims 1 to 7.

9. A computer-readable storage medium storing computer instructions, characterized in that: When the computer instructions are executed on an electronic device, the electronic device is caused to execute the method according to any one of claims 1 to 7.

10. A computer program product, characterized in that When the computer program product is run on an electronic device, the electronic device is enabled to perform the method according to any one of claims 1 to 7.

Images