A virtual power plant distributed resource scheduling system based on big data analysis

Abstract

The application discloses a kind of virtual power plant distributed resource scheduling systems based on big data analysis, it is related to power system dispatching technical field, including following steps: data acquisition module, for obtaining resource operation data and scheduling execution data;Task segmentation module, for obtaining adjustment task;Double-layer evaluation module, for inputting standard data and adjustment task into double-layer model, generating resource classification result;Resource grouping module, for generating adjustment resource group;Path undertake module, for constructing task undertaking graph, using margin recursion BFS algorithm to form capacity package;Strategy generation module, for generating global scheduling strategy;Closed-loop scheduling module, for outputting distributed resource scheduling result.The application can improve the accuracy of resource capacity identification in virtual power plant distributed resource scheduling process, the reliability of adjustment task undertaking and the dynamic adaptive capacity of scheduling execution.

Description

Technical Field

[0001] This invention relates to the field of power system dispatching technology, and in particular to a virtual power plant distributed resource dispatching system based on big data analysis. Background Technology

[0002] With the expansion of new energy grid connection, the advancement of electricity market transactions, and the continuous increase in demand-side response resources, virtual power plant aggregation and dispatch technology for distributed resources such as photovoltaics, energy storage, charging piles, and building air conditioning has attracted widespread attention. Existing virtual power plant dispatch systems mainly rely on centralized cloud optimization, static resource ledgers, or simple capacity aggregation for resource allocation. However, in practical applications, the following problems are commonly encountered: The data protocols, acquisition cycles, and field formats of various types of distributed resources differ significantly, making it difficult to form a unified data foundation for resource operation data and scheduling execution data. This leads to inaccurate resource status identification and unstable basis for issuing scheduling instructions. There is a discrepancy between the aggregated capacity of virtual power plants and their actual adjustable capacity. Although some resources are connected to the platform, they are affected by response speed, adjustable capacity, duration, and operating boundaries, making it difficult for them to stably undertake regulation tasks. This results in the regulation capacity assessment results deviating from the actual execution capacity. Different resources have significant differences in regulation direction, response time, and duration. The existing method of coarse-grained grouping by resource type or region is difficult to accurately match grid scheduling instructions, which can easily lead to unclear resource acceptance paths, uneven distribution of regulation volume, and low efficiency of multi-resource collaboration. There is a lack of an effective closed loop between scheduling execution results, resource capacity assessment, and scheduling strategies. Resource execution deviations are difficult to be fed back to the subsequent scheduling process in a timely manner, affecting the accuracy, stability, and real-time performance of virtual power plant distributed resource scheduling.

[0003] Therefore, how to provide a distributed resource scheduling system for virtual power plants based on big data analysis is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0004] One objective of this invention is to propose a distributed resource scheduling system for virtual power plants based on big data analytics. This invention standardizes the resource operation data and scheduling execution data of distributed resources, combines them with regulation tasks formed by dividing grid dispatch instructions, establishes a two-layer evaluation process of regulation capacity classification and task matching, generates resource classification results, and forms regulation resource groups based on these results. Furthermore, it constructs a task assignment graph, uses a margin recursive BFS algorithm to filter assignment paths and form capability packages, generates a global scheduling strategy based on these capability packages, and updates the resource classification results, capability packages, and global scheduling strategy based on the execution results of resource scheduling instructions. This addresses the problems of heterogeneous distributed resource data, insufficient matching between regulation capacity and regulation tasks, unclear resource assignment paths, and difficulty in closed-loop correction of scheduling results in existing technologies, achieving accurate assignment, coordinated allocation, and dynamic scheduling of distributed resources in virtual power plants.

[0005] A distributed resource scheduling system for virtual power plants based on big data analysis, according to an embodiment of the present invention, includes the following steps: The data acquisition module is used to deploy edge acquisition nodes to acquire resource operation data and scheduling execution data of distributed resources within the virtual power plant, perform protocol conversion and standardization processing, and obtain standard data. The task segmentation module is used to receive power grid dispatch instructions, segment tasks, and obtain regulation tasks; A two-layer evaluation module is used to input standard data and adjustment tasks into a two-layer model, which includes a resource-level evaluation layer and a task-level evaluation layer. The resource-level evaluation layer classifies the adjustment capabilities of distributed resources and generates capability classification results. The task-level evaluation layer matches the adjustment tasks with the capability classification results and generates resource classification results. The resource grouping module is used to group and aggregate distributed resources based on resource classification results, and generate adjustment resource groups. The path receiving module is used to construct a task receiving graph based on adjustment tasks, adjustment resource groups, edge acquisition nodes and distributed resources. It uses the residual recursive BFS algorithm to traverse the receiving paths in the task receiving graph and selects the receiving paths corresponding to the adjustment tasks to form capability packages. The strategy generation module is used to select the receiving path from the capability package according to the adjustment task, and to allocate the adjustment task to the edge acquisition node and distributed resources corresponding to the receiving path, thereby generating a global scheduling strategy. The closed-loop scheduling module is used to configure instruction packages according to the global scheduling strategy and send the instruction packages to the edge acquisition nodes. The edge acquisition nodes send resource scheduling instructions to the adjustment resource group according to the instruction packages. Based on the execution results of the resource scheduling instructions, the module updates the resource classification results, capability packages and global scheduling strategy, and outputs the distributed resource scheduling results.

[0006] Optionally, the data acquisition module specifically comprises: Deploy edge acquisition nodes and establish an acquisition mapping between distributed resources and edge acquisition nodes; Edge acquisition nodes acquire resource operation data and scheduling execution data of distributed resources according to the acquisition mapping, perform data integrity checks, remove data with missing fields and data with abnormal acquisition time, and generate valid acquisition data; The valid collected data is subjected to protocol identification, and protocol conversion is performed according to the protocol identification results to generate protocol conversion data; Align the acquisition time of the protocol conversion data to generate aligned data under the same acquisition period; The aligned data is then standardized in terms of fields and units to obtain standard data.

[0007] Optionally, the task segmentation module specifically includes: Receive power grid dispatch instructions, verify the instruction format of the power grid dispatch instructions, discard power grid dispatch instructions that lack adjustment direction, adjustment amount, response time limit or duration, and generate valid dispatch instructions; Valid scheduling instructions are parsed to extract the adjustment direction, adjustment amount, response time limit, duration and execution period. Valid scheduling instructions are divided into time periods according to their execution time periods to generate task time periods; The effective scheduling instructions are split into adjustment amounts according to the adjustment amount and task time period, and segmented adjustment amounts corresponding to the task time period are generated. The adjustment direction, segmented adjustment amount, response time limit, duration, and task time period are bound together to obtain adjustment tasks. The adjustment tasks are then sorted according to the response time limit to obtain an adjustment task sequence.

[0008] Optionally, the two-layer evaluation module specifically includes: Standard data is aggregated according to distributed resources, and the running status data and scheduling execution data of distributed resources are extracted. By associating the adjustment direction, segmented adjustment amount, response time limit, duration, and task time period in the adjustment task sequence with distributed resources, resource task association data is obtained. Input the running status data, scheduling execution data, and resource task association data into the resource-level evaluation layer; The resource-level assessment layer classifies the adjustment capabilities of distributed resources according to response speed, adjustment range, stability, and security, and generates capability classification results, which include effective adjustment class, potential adjustment class, and non-adjustable class. Input the capability classification results and adjustment task sequences into the task-level evaluation layer; The task-level evaluation layer matches adjustment tasks with distributed resources corresponding to effective and potential adjustment classes according to adjustment direction, segmented adjustment amount, response time limit, and duration, and eliminates distributed resources corresponding to non-adjustable classes. For the matching distributed resource tags, corresponding adjustment tasks, task time periods, and capability classification results, generate resource classification results.

[0009] Optionally, the resource grouping module specifically includes: Remove distributed resources marked as unadjustable in the resource classification results, and retain distributed resources corresponding to the effective and potential adjustable categories; The reserved distributed resources are marked with adjustment characteristics according to the adjustment direction, segmented adjustment amount, response time limit and duration corresponding to the adjustment task; Based on the adjustment characteristic labels and capability classification results, the retained distributed resources are grouped and aggregated to generate adjustment resource groups; The resource groups are bound to the corresponding adjustment tasks, task time periods, and edge acquisition nodes to obtain the resource grouping results.

[0010] Optionally, the task assignment diagram based on adjustment tasks, adjustment resource groups, edge acquisition nodes, and distributed resources is specifically as follows: Set the adjustment task, adjustment resource group, edge acquisition node, and distributed resource as task node, resource group node, edge node, and resource node, respectively; Directed edges are established sequentially according to the binding relationship between adjustment tasks and adjustment resource groups, the binding relationship between adjustment resource groups and edge acquisition nodes, and the acquisition mapping between edge acquisition nodes and distributed resources. Write the adjustment direction, segmented adjustment amount, response time limit, duration and task time period into the task node, and write the capability classification results and resource classification results into the resource node to obtain the task assignment diagram.

[0011] Optionally, the step of using the residual recursive BFS algorithm to traverse the connection paths in the task connection graph and filter and adjust the connection paths corresponding to the tasks to form a capability package specifically involves: Select adjustment tasks sequentially according to the sorting results in the adjustment task sequence; Establish a queue to be expanded for the selected adjustment tasks, record the segmented adjustment amount corresponding to the adjustment task as the initial adjustment margin, and add the task node corresponding to the adjustment task to the queue to be expanded. The queue to be expanded is expanded layer by layer according to the hierarchical order of task nodes, resource group nodes, edge nodes and resource nodes. When extending to resource group nodes, paths corresponding to adjustment resource groups that are not bound to adjustment tasks are filtered out; When extending to edge nodes, paths corresponding to edge acquisition nodes that are not bound to the adjustment resource group are filtered out. When extending to resource nodes, paths corresponding to distributed resources whose capability classification results are non-adjustable are excluded. Based on standard data and capacity classification results, determine the adjustable capacity and sustainable adjustment duration corresponding to resource nodes; If the duration of sustainable adjustment is less than the duration of the adjustment task, the corresponding path will be filtered out. Subtract the available adjustment amount from the adjustment reserve to obtain the updated adjustment reserve; if the updated adjustment reserve is greater than 0, the formed path is retained and the untraversed path of the same adjustment task is continued to be expanded; if the updated adjustment reserve is less than or equal to 0, the path participating in the reserve recursion is taken as the corresponding receiving path of the adjustment task. The capability package is formed by aggregating the acceptance path, the corresponding distributed resources, the adjustable capacity, the task time period, the response time limit, and the capability classification results.

[0012] Optionally, the strategy generation module specifically includes: According to the sorting results in the adjustment task sequence, select the adjustment tasks in order; extract the adjustment direction, segmented adjustment amount, response time limit, duration and task time period corresponding to the adjustment task; Select and adjust the task time period, response time limit, duration and capability classification results of the task from the capability package to match the receiving path. The segmented adjustment amount is allocated according to the adjustable amount that can be accepted according to the receiving path, and the allocated adjustment amount corresponding to the receiving path is obtained. If the cumulative value of the adjustable amount that can be accepted for the corresponding path is less than the segmented adjustable amount, then continue to select the corresponding path for the same adjustable task from the capability package. If the cumulative value of the adjustable amount that can be accepted for the corresponding path is greater than or equal to the segmented adjustable amount, then stop selecting the corresponding path for the same adjustable task. The adjustment tasks are assigned to the edge acquisition nodes and distributed resources corresponding to the selected receiving paths; By associating adjustment tasks, adjustment directions, receiving paths, edge acquisition nodes, distributed resources, allocated adjustment amounts, response time limits, durations, and task time periods, a global scheduling strategy is generated.

[0013] Optionally, the step of configuring instruction packets according to the global scheduling strategy and sending the instruction packets to the edge acquisition nodes, and the edge acquisition nodes sending resource scheduling instructions to the resource adjustment group according to the instruction packets, specifically involves: The global scheduling strategy is aggregated according to the edge acquisition nodes, and the aggregation results are written into the instruction packet. Write the edge acquisition node, adjustment resource group, distributed resources, adjustment direction, allocated adjustment amount, response time limit, duration, task time period and receiving path into the instruction package; Send the instruction packet to the corresponding edge acquisition node; The edge acquisition node parses the instruction packet to extract the adjustment resource group, distributed resources, adjustment direction, allocated adjustment amount, response time limit, duration and task time period; Edge acquisition nodes issue resource scheduling instructions to the adjustment resource group according to the task time period. The resource scheduling instructions include distributed resources, adjustment direction, allocation adjustment amount, response time limit and duration.

[0014] Optionally, the step of updating the resource classification result, capability package, and global scheduling strategy based on the execution result of the resource scheduling instruction, and outputting the distributed resource scheduling result, specifically involves: Obtain the execution results of resource scheduling instructions. The execution results include distributed resources, adjustment direction, actual adjustment amount, actual response time, actual duration, and task period. The execution results are compared with the distributed resources, adjustment direction, allocation adjustment amount, response time limit, duration and task period in the global scheduling strategy to obtain the execution deviation; The capability classification results in the resource classification results are adjusted according to the execution deviation to obtain the updated resource classification results; Based on the updated resource classification results and execution deviations, the acceptance paths, adjustable quantities, response time limits, and capability classification results in the capability package are adjusted to obtain the updated capability package. Based on the updated capability package, the receiving path, edge acquisition node, distributed resources, and allocation adjustment amount in the global scheduling strategy are adjusted to obtain the updated global scheduling strategy. The execution results, execution deviations, updated resource classification results, updated capability packages, and updated global scheduling policies are correlated to output distributed resource scheduling results.

[0015] The beneficial effects of this invention are: (1) This invention divides the power grid dispatching command into regulation tasks and inputs the standard data and regulation tasks into a two-layer model, so that the response speed, regulation range, stability and security of distributed resources participate in the regulation capability classification. At the same time, the regulation tasks are matched with the capability classification results to generate resource classification results, thereby reducing the deviation between the aggregated capacity of virtual power plants and the real adjustable capacity. (2) By generating adjustment resource groups based on resource classification results, constructing task acceptance graphs based on adjustment tasks, adjustment resource groups, edge acquisition nodes and distributed resources, and using the margin recursive BFS algorithm to filter the acceptance paths corresponding to adjustment tasks to form capability packages, the adjustment tasks can match the acceptable distributed resources according to the segmented adjustment amount, response time limit and duration, thereby improving the accuracy of multi-type distributed resources coordinating to accept adjustment tasks. (3) By selecting the receiving path from the capability package according to the adjustment task and generating a global scheduling strategy, configuring the instruction package according to the global scheduling strategy and issuing resource scheduling instructions, updating the resource classification results, capability package and global scheduling strategy according to the execution results of the resource scheduling instructions, so that the distributed resource scheduling results can be fed back to the resource capability assessment and scheduling allocation process, thereby improving the stability, coordination and dynamic adaptability of the distributed resource scheduling of the virtual power plant. Attached Figure Description

[0016] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a schematic diagram of a distributed resource scheduling system for virtual power plants based on big data analysis proposed in this invention. Figure 2 This is a structural diagram of the two-layer model proposed in this invention; Figure 3 This is a data flow diagram of a virtual power plant distributed resource scheduling system based on big data analysis proposed in this invention. Detailed Implementation

[0017] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0018] refer to Figures 1-3 A distributed resource scheduling system for virtual power plants based on big data analytics, comprising: The data acquisition module is used to deploy edge acquisition nodes to acquire resource operation data and scheduling execution data of distributed resources within the virtual power plant, perform protocol conversion and standardization processing, and obtain standard data. The task segmentation module is used to receive power grid dispatch instructions, segment tasks, and obtain regulation tasks; The two-layer evaluation module is used to input standard data and regulation tasks into the two-layer model, which includes a resource-level evaluation layer and a task-level evaluation layer. The resource-level evaluation layer classifies the regulation capabilities of distributed resources and generates capability classification results. The task-level evaluation layer matches the regulation tasks with the capability classification results and generates resource classification results. The resource grouping module is used to group and aggregate distributed resources based on resource classification results, and generate adjustment resource groups. The path receiving module is used to construct a task receiving graph based on adjustment tasks, adjustment resource groups, edge acquisition nodes and distributed resources. It uses the residual recursive BFS algorithm to traverse the receiving paths in the task receiving graph and selects the receiving paths corresponding to the adjustment tasks to form capability packages. The strategy generation module is used to select the receiving path from the capability package according to the adjustment task, and to allocate the adjustment task to the edge acquisition node and distributed resources corresponding to the receiving path, thereby generating a global scheduling strategy. The closed-loop scheduling module is used to configure instruction packages according to the global scheduling strategy and send the instruction packages to the edge acquisition nodes. The edge acquisition nodes send resource scheduling instructions to the adjustment resource group according to the instruction packages. Based on the execution results of the resource scheduling instructions, the module updates the resource classification results, capability packages and global scheduling strategy, and outputs the distributed resource scheduling results.

[0019] In this embodiment, the data acquisition module specifically includes: Deploy edge acquisition nodes and establish an acquisition mapping between distributed resources and edge acquisition nodes. The acquisition mapping includes the distributed resource number, edge acquisition node number, protocol type, and acquisition cycle. The edge acquisition node obtains the resource operation data and scheduling execution data of the distributed resources according to the acquisition mapping, and writes the distributed resource number, edge acquisition node number and acquisition time into it; Perform data integrity checks on resource operation data and scheduling execution data, remove data with missing fields and data with abnormal collection times, and generate valid collected data; The valid collected data is identified by protocol, and the device address field, time field and data field in the valid collected data are converted into a unified resource identifier field, a unified time field and a unified data field according to the protocol type, generating protocol conversion data; The protocol conversion data is aligned according to the collection period, and the protocol conversion data within the same collection period is grouped into aligned data. The aligned data is standardized in terms of fields and units. Field names are converted to uniform field names, and power data, energy data, and time data are converted to uniform units to obtain standard data.

[0020] In this embodiment, the task segmentation module specifically includes: Receive power grid dispatch instructions, verify the instruction format of the power grid dispatch instructions in terms of adjustment direction, adjustment amount, response time limit, duration and execution period, and remove power grid dispatch instructions with empty fields, adjustment amount less than or equal to zero, response time limit less than or equal to 0, duration less than or equal to 0, or execution period incomplete, and generate valid dispatch instructions. Valid scheduling instructions are parsed to extract the adjustment direction, adjustment amount, response time limit, duration and execution period. The effective scheduling instructions are divided into time periods according to the start and end times of the execution period to generate continuous task time periods; The adjustment amount is divided according to the proportion of the task period in the duration, and the segmented adjustment amount corresponding to the task period is generated. By binding the adjustment direction, segmented adjustment amount, response time limit, duration, and task time period, an adjustment task is obtained; The adjustment tasks are sorted from shortest to longest response time. If the response time is the same, they are sorted from earliest to latest task start time to obtain the adjustment task sequence.

[0021] In this embodiment, the two-layer evaluation module specifically includes: Standard data is aggregated according to distributed resource number, and the running status data and scheduling execution data of distributed resources are extracted. Extract the power status, adjustable capacity, and operating boundary of distributed resources from the operating status data; extract the instruction reception time, start response time, actual adjustment amount, power fluctuation record, and limit overrun record from the scheduling execution data. By associating the adjustment direction, segmented adjustment amount, response time limit, duration, and task time period in the adjustment task sequence with the distributed resource number, resource task association data is obtained. Input the running status data, scheduling execution data, and resource task association data into the resource-level evaluation layer; The resource-level assessment layer subtracts the instruction reception time from the start response time to obtain the response speed; compares the actual adjustment amount with the segmented adjustment amount to obtain the adjustment range; obtains the stability based on the power fluctuation records and duration during the task period; and obtains the safety based on the operating boundary and limit violation records. Distributed resources are classified into three categories based on their adjustment capabilities: response speed, adjustment range, stability, and security. The results of the classification include three categories: effective adjustment, potential adjustment, and non-adjustable. Input the capability classification results and adjustment task sequences into the task-level evaluation layer; The task-level evaluation layer matches adjustment tasks with distributed resources corresponding to effective and potential adjustment classes according to adjustment direction, segmented adjustment amount, response time limit, and duration, and eliminates distributed resources corresponding to non-adjustable classes. For the matching distributed resource tags, corresponding adjustment tasks, task time periods, and capability classification results, generate resource classification results.

[0022] In this embodiment, the two-layer evaluation module aggregates standard data according to distributed resource numbers, ensuring a one-to-one correspondence between operational status data, scheduling execution data, and corresponding distributed resources. This avoids biases in regulation capability judgment caused by the mixing of multiple types of distributed resource data. By extracting power status, adjustable capacity, and operational boundaries from operational status data, and extracting instruction reception time, start response time, actual adjustment amount, power fluctuation records, and limit exceedance records from scheduling execution data, the response speed, adjustment range, stability, and security are all determined by actual operational and execution data, improving the authenticity of the capability classification results. Furthermore, by associating the adjustment direction, segmented adjustment amount, response time limit, duration, and task period in the adjustment task sequence with the distributed resource number, the capability classification results can be matched with specific adjustment tasks, avoiding scheduling based solely on static resource status. Simultaneously, by eliminating distributed resources corresponding to non-adjustable categories and marking the adjustment tasks, task periods, and capability classification results for matched distributed resources, resource classification results that can be directly used for resource grouping, path connection, and global scheduling strategy generation can be formed, improving the accuracy and stability of distributed resource scheduling in virtual power plants.

[0023] In this embodiment, the resource grouping module specifically includes: Extract distributed resources, adjustment tasks, task periods, edge acquisition nodes, capability classification results, response speed, adjustment range, amount of adjustment that can be carried out, and sustainable adjustment duration from the resource classification results; Distributed resources whose capability classification results are classified as non-adjustable are removed, while distributed resources whose capability classification results are classified as effectively adjustable or have the potential to be adjustable are retained. Write an up or down flag to the reserved distributed resources according to the adjustment direction corresponding to the adjustment task; Based on the comparison results between the acceptable adjustment amount and the segmented adjustment amount, write the full acceptance tag or partial acceptance tag to the reserved distributed resources; Based on the comparison between response speed and response time limit, the reserved distributed resources are marked as either meeting the write time limit or failing to meet the time limit. Based on the comparison between the sustainable adjustment duration and the duration, write a sustainable satisfaction flag or a sustainable insufficiency flag to the retained distributed resources; The upward adjustment flag, downward adjustment flag, full acceptance flag, partial acceptance flag, time limit met flag, time limit insufficient flag, continuous met flag, and continuous insufficient flag are used as adjustment characteristic flags; Based on the adjustment task, task time period, edge acquisition node, adjustment characteristic label and capability classification results, the retained distributed resources are grouped and aggregated to generate adjustment resource groups; The resource grouping results are obtained by binding the adjustment resource group with the corresponding adjustment task, task time period, edge acquisition node and distributed resource.

[0024] In this embodiment, a task assignment diagram is constructed based on adjustment tasks, adjustment resource groups, edge acquisition nodes, and distributed resources, specifically as follows: Extract adjustment tasks from the adjustment task sequence and set each adjustment task as a task node; extract adjustment resource groups from the resource grouping results and set each adjustment resource group as a resource group node; extract edge acquisition nodes from the resource grouping results and set each edge acquisition node as an edge node; extract distributed resources from the acquisition mapping and set each distributed resource as a resource node. Write the adjustment direction, segmented adjustment amount, response time limit, duration, and task time period to the corresponding task node; write the adjustment characteristic flag and capability classification result to the corresponding resource group node; write the edge acquisition node number and acquisition cycle to the corresponding edge node; write the capability classification result, resource classification result, acceptable adjustment amount, and sustainable adjustment duration to the corresponding resource node. Based on the binding relationship between adjustment tasks and adjustment resource groups, directed edges are established from task nodes to resource group nodes; based on the binding relationship between adjustment resource groups and edge acquisition nodes, directed edges are established from resource group nodes to edge nodes; based on the acquisition mapping between edge acquisition nodes and distributed resources, directed edges are established from edge nodes to resource nodes. By associating task nodes, resource group nodes, edge nodes, resource nodes, and directed edges, a task succession graph is obtained.

[0025] In this implementation, the residual recursive BFS algorithm is used to traverse the connection paths in the task connection graph, and the connection paths corresponding to the adjusted tasks are selected to form capability packages, specifically: According to the sorting results in the adjustment task sequence, select adjustment tasks in sequence; read the adjustment direction, segmented adjustment amount, response time limit, duration and task time period of the selected adjustment tasks; The segmented adjustment amount is recorded as the initial adjustment margin, the task node corresponding to the selected adjustment task is written into the queue to be expanded, and the path where the task node is located is recorded as an empty path. Take a task node from the queue to be expanded, and extend it to the resource group node along the directed edge from the task node to the resource group node. When extending to resource group nodes, read the adjustment characteristic tags and capability classification results in the resource group nodes, filter out the paths corresponding to adjustment resource groups that are not bound to adjustment tasks, and retain the paths corresponding to adjustment resource groups that are bound to adjustment tasks. Extend along the directed edge from the resource group node to the edge node, read the edge acquisition node number and acquisition cycle in the edge node, filter out the path corresponding to the edge acquisition node that is not bound to the adjustment resource group, and retain the path corresponding to the edge acquisition node that is bound to the adjustment resource group. Extend along the directed edges from the edge nodes to the resource nodes, and read the capability classification results, resource classification results, adjustable capacity, and sustainable adjustment duration from the resource nodes. When extending to resource nodes, paths corresponding to distributed resources whose capability classification result is non-adjustable are removed, and paths corresponding to distributed resources whose sustainable adjustment duration is less than the duration are also removed. The allowable adjustment amount in the retained path is compared with the adjustment margin, and the allowable adjustment amount is subtracted from the adjustment margin to obtain the updated adjustment margin. If the update adjustment margin is greater than 0, the reserved path, update adjustment margin and untraversed resource nodes are written to the queue to be expanded, and the path corresponding to the same adjustment task continues to be expanded. If the updated adjustment margin is less than or equal to 0, the retained path that participates in the adjustment margin deduction will be used as the receiving path corresponding to the adjustment task. The capability package is formed by aggregating the acceptance path, the corresponding distributed resources, the adjustable capacity, the task time period, the response time limit, the duration, and the capability classification results.

[0026] In this embodiment, the margin recursive BFS algorithm selects adjustment tasks sequentially according to the adjustment task sequence and uses the segmented adjustment amount as the initial adjustment margin. This ensures that the selection of receiving paths revolves around the actual adjustment tasks, avoiding scheduling deviations caused by the mismatch between the theoretical capacity of a single distributed resource and the requirements of the adjustment task. Through the hierarchical expansion of task nodes, resource group nodes, edge nodes, and resource nodes, a clear receiving link is formed between adjustment tasks, adjustment resource groups, edge acquisition nodes, and distributed resources. By filtering out paths that do not meet the binding relationship, capability classification results, and duration requirements at resource group nodes, edge nodes, and resource nodes respectively, invalid receiving paths can be eliminated in advance, reducing the participation of invalid resources in scheduling. By subtracting the available adjustment amount from the adjustment margin and judging the path expansion status based on the updated adjustment margin, multiple distributed resources can jointly undertake the same adjustment task. By aggregating receiving paths, distributed resources, available adjustment amounts, task time periods, response time limits, durations, and capability classification results into capability packages, accurate path receiving basis can be provided for the generation of global scheduling strategies, improving the accuracy and coordination of distributed resource scheduling in virtual power plants.

[0027] In this embodiment, the strategy generation module specifically includes: According to the sorting results in the adjustment task sequence, select the adjustment tasks in order; extract the adjustment direction, segmented adjustment amount, response time limit, duration and task time period from the adjustment tasks; Filter the task time period, response time limit, duration and capability level results from the capability package to find the receiving path that matches the adjustment task. The selected service paths are sorted according to the capability classification results, response time limit, and available adjustment capacity. Select the top-ranked receiving path and use the segment adjustment amount as the remaining segment adjustment amount; Compare the adjustable amount that can be accepted for the accepted path with the remaining segmented adjustable amounts; If the available adjustment amount is less than the remaining segmented adjustment amount, then the available adjustment amount will be used as the allocation adjustment amount corresponding to the receiving path, and the remaining segmented adjustment amount will be reduced by the available adjustment amount. If the available adjustment amount is greater than or equal to the remaining segment adjustment amount, then the remaining segment adjustment amount is used as the allocation adjustment amount corresponding to the receiving path, and the remaining segment adjustment amount is set to 0. If the remaining segment adjustment amount is greater than 0, then continue to select the corresponding receiving path for the same adjustment task from the capability package; If the remaining segmented adjustment amount is equal to 0, then stop selecting the receiving path corresponding to the same adjustment task; The adjustment tasks are assigned to the edge acquisition nodes and distributed resources corresponding to the selected receiving paths; By associating adjustment tasks, adjustment directions, receiving paths, edge acquisition nodes, distributed resources, allocated adjustment amounts, response time limits, durations, and task time periods, a global scheduling strategy is generated.

[0028] In this embodiment, instruction packets are configured according to the global scheduling strategy and sent to the edge acquisition nodes. The edge acquisition nodes then issue resource scheduling instructions to the resource adjustment group based on the instruction packets. Specifically: Extract the adjustment task, adjustment direction, receiving path, edge acquisition node, distributed resources, allocated adjustment amount, response time limit, duration and task time period from the global scheduling strategy; The data extracted from the edge acquisition nodes are aggregated, and the adjustment tasks, adjustment resource groups, distributed resources, allocated adjustment amounts, response time limits, duration, task time periods and receiving paths corresponding to the same edge acquisition node are written into the same instruction package. Write the edge acquisition node number, adjustment resource group number, distributed resource number, adjustment direction, allocated adjustment amount, response time limit, duration, task time period, and receiving path into the instruction packet; The instruction packet is sent to the corresponding edge acquisition node according to the edge acquisition node number; After receiving the instruction packet, the edge acquisition node matches the edge acquisition node number in the instruction packet and retains the instruction packet that matches the edge acquisition node. Parse the retained instruction packets to extract the adjustment resource group number, distributed resource number, adjustment direction, allocated adjustment amount, response time limit, duration and task time period; Distributed resource numbers are arranged according to task time periods, and the distributed resource numbers, adjustment directions, allocated adjustment amounts, response time limits, and durations under the same adjustment resource group number are written into the resource scheduling instructions. Resource scheduling instructions are issued to the resource adjustment group according to the task time period. The resource scheduling instructions include the distributed resource number, adjustment direction, allocation adjustment amount, response time limit and duration.

[0029] In this embodiment, based on the execution result of the resource scheduling instruction, the resource classification result, capability package, and global scheduling policy are updated, and the distributed resource scheduling result is output, specifically as follows: Obtain the execution results of resource scheduling instructions. The execution results include distributed resources, adjustment direction, actual adjustment amount, actual response time, actual duration, and task period. Based on the distributed resources and task time periods, the execution results are mapped to the distributed resources, adjustment direction, allocated adjustment amount, response time limit, duration, and task time periods in the global scheduling policy; the actual adjustment amount is subtracted from the allocated adjustment amount to obtain the adjustment amount deviation; The actual response time is compared with the response time limit to obtain the response time deviation; the actual duration is compared with the duration to obtain the duration deviation; the adjustment direction in the execution result is compared with the adjustment direction in the global scheduling strategy to obtain the adjustment direction deviation; the adjustment deviation, response time deviation, duration deviation, and adjustment direction deviation are aggregated to obtain the execution deviation. If the actual adjustment amount corresponding to the execution deviation is less than the allocated adjustment amount, the actual response time is greater than the response time limit, or the actual duration is less than the duration, then the capability classification result of the corresponding distributed resource in the resource classification result will be reduced. If the actual adjustment amount corresponding to the execution deviation reaches the allocated adjustment amount, the actual response time is less than or equal to the response time limit, and the actual duration is greater than or equal to the duration, then the capability classification result of the corresponding distributed resource in the resource classification result is maintained or improved, and the updated resource classification result is obtained. Based on the updated resource classification results, the adjustable capacity, response time, capability classification results, and access paths of the corresponding distributed resources in the capability package are adjusted to obtain the updated capability package. According to the updated capability package, the receiving paths corresponding to distributed resources whose capability classification results are non-adjustable are removed from the global scheduling strategy, and the edge acquisition nodes, distributed resources and allocation adjustment amounts of the corresponding adjustment tasks are adjusted to obtain the updated global scheduling strategy. The execution results, execution deviations, updated resource classification results, updated capability packages, and updated global scheduling policies are correlated to output distributed resource scheduling results.

[0030] Example 1: To verify the feasibility of this invention in practice, it was applied to a distributed resource scheduling scenario in a virtual power plant within a comprehensive energy park. The park has a 5.8MW photovoltaic power station, a 4MW / 8MWh energy storage system, 620 charging piles, and 3.6MW adjustable load from building air conditioning. These distributed resources are distributed across 9 edge acquisition nodes, with an acquisition cycle of 30 seconds and a scheduling cycle of 15 minutes. The test lasted for 30 consecutive days, during which a total of 286 grid scheduling instructions were received, including 164 peak-shaving instructions, 91 valley-filling instructions, and 31 standby response instructions. The original scheduling method primarily issued scheduling tasks based on resource type and static adjustable capacity, which had issues such as some resources being theoretically adjustable but actually unavailable, mismatch between adjustment tasks and resource response timelines, and unclear target objects for edge acquisition nodes.

[0031] Edge acquisition nodes obtain resource operation data and scheduling execution data of distributed resources. After protocol conversion, acquisition time alignment, field unification, and unit unification, standard data is obtained. The cloud side receives power grid dispatch instructions, divides them into adjustment tasks according to adjustment direction, adjustment amount, response time limit, duration, and execution period, and forms an adjustment task sequence. A two-layer model generates resource classification results based on standard data and adjustment tasks. The resource-level evaluation layer classifies distributed resources into effective adjustment class, potential adjustment class, and non-adjustable class according to response speed, adjustment range, stability, and security. The task-level evaluation layer matches adjustment tasks with the distributed resources corresponding to the effective adjustment class and the potential adjustment class. After resource grouping, adjustment resource groups are obtained. The path assignment module constructs a task assignment graph based on adjustment tasks, adjustment resource groups, edge acquisition nodes, and distributed resources, and uses the margin recursive BFS algorithm to select assignment paths to form capability packages. The strategy generation module selects assignment paths from the capability packages, allocates adjustment tasks to the corresponding edge acquisition nodes and distributed resources, and generates a global scheduling strategy. The closed-loop scheduling module configures instruction packages and issues resource scheduling instructions, and updates resource classification results, capability packages and global scheduling strategies based on the execution results.

[0032] Table 1 Comparison of Resource Identification and Path Selection Results

[0033] As shown in Table 1, the system of this invention has significant advantages in resource identification and path selection. The original scheduling method relied on static adjustable capacity, resulting in the inclusion of non-adjustable resources in the scheduling process, leading to a false participation rate of 14.8% and easily causing adjustment tasks to fail. The system of this invention filters out non-adjustable resources through standard data, capability classification results, and resource classification results, reducing the false participation rate to 2.6%. Simultaneously, the margin recursive BFS algorithm filters the path to the adjustment task layer by layer in the task assignment graph, reducing the error in identifying the actual adjustable capacity from 13.5% to 3.8%, and increasing the resource matching success rate of adjustment tasks to 96.9%. The increased average number of paths per task indicates that multiple distributed resources can collaboratively undertake the same adjustment task, reducing the impact of a single resource failure on the overall scheduling.

[0034] Table 2 Comparison of Scheduling Execution and Closed-Loop Update Effects

[0035] As shown in Table 2, the system of the present invention significantly improves response efficiency and execution stability during the scheduling execution phase. The original scheduling method required multiple manual or platform-side corrections when resource response speed, duration, and actual adjustment volume were insufficient, resulting in 2.8 repeated corrections per task for resource scheduling instructions. The system of the present invention configures instruction packages according to the global scheduling strategy, and edge acquisition nodes issue resource scheduling instructions to the adjustment resource group according to the task time period, reducing the average response time from 92 seconds to 41 seconds. After the resource scheduling instruction execution results are fed back, the system updates the resource classification results, capacity packages, and global scheduling strategy based on the execution deviation, increasing the next day's adjustment task matching success rate to 98.1%. The average deviation of the adjustment volume and the power fluctuation rate decreased simultaneously, indicating that the closed-loop update can continuously correct the judgment of the distributed resource carrying capacity and improve the reliability of the virtual power plant's distributed resource scheduling results.

[0036] This invention can form continuous data links and execution links in the process of accessing multiple types of distributed resources, splitting adjustment tasks, classifying resources, accepting paths, generating strategies, and scheduling closed loops. It solves the problems of inconsistent data, unclear actual adjustability, ambiguous adjustment task acceptance paths, and difficulty in updating and feeding back scheduling execution results in the distributed resource scheduling of virtual power plants, and has good engineering implementation value.

[0037] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A distributed resource scheduling system for virtual power plants based on big data analytics, characterized in that, include: The data acquisition module is used to deploy edge acquisition nodes to acquire resource operation data and scheduling execution data of distributed resources within the virtual power plant, perform protocol conversion and standardization processing, and obtain standard data. The task segmentation module is used to receive power grid dispatch instructions, segment tasks, and obtain regulation tasks; A two-layer evaluation module is used to input standard data and adjustment tasks into a two-layer model, which includes a resource-level evaluation layer and a task-level evaluation layer. The resource-level evaluation layer classifies the adjustment capabilities of distributed resources and generates capability classification results. The task-level evaluation layer matches the adjustment tasks with the capability classification results and generates resource classification results. The resource grouping module is used to group and aggregate distributed resources based on resource classification results, and generate adjustment resource groups. The path receiving module is used to construct a task receiving graph based on adjustment tasks, adjustment resource groups, edge acquisition nodes and distributed resources. It uses the residual recursive BFS algorithm to traverse the receiving paths in the task receiving graph and selects the receiving paths corresponding to the adjustment tasks to form capability packages. The strategy generation module is used to select the receiving path from the capability package according to the adjustment task, and to allocate the adjustment task to the edge acquisition node and distributed resources corresponding to the receiving path, thereby generating a global scheduling strategy. The closed-loop scheduling module is used to configure instruction packages according to the global scheduling strategy and send the instruction packages to the edge acquisition nodes. The edge acquisition nodes send resource scheduling instructions to the adjustment resource group according to the instruction packages. Based on the execution results of the resource scheduling instructions, the module updates the resource classification results, capability packages and global scheduling strategy, and outputs the distributed resource scheduling results.

2. The distributed resource scheduling system for virtual power plants based on big data analysis according to claim 1, characterized in that, The data acquisition module is specifically: Deploy edge acquisition nodes and establish an acquisition mapping between distributed resources and edge acquisition nodes; Edge acquisition nodes acquire resource operation data and scheduling execution data of distributed resources according to the acquisition mapping, perform data integrity checks, remove data with missing fields and data with abnormal acquisition time, and generate valid acquisition data; The valid collected data is subjected to protocol identification, and protocol conversion is performed according to the protocol identification results to generate protocol conversion data; Align the acquisition time of the protocol conversion data to generate aligned data under the same acquisition period; The aligned data is then standardized in terms of fields and units to obtain standard data.

3. The distributed resource scheduling system for virtual power plants based on big data analysis according to claim 1, characterized in that, The task segmentation module specifically comprises: Receive power grid dispatch instructions, verify the instruction format of the power grid dispatch instructions, discard power grid dispatch instructions that lack adjustment direction, adjustment amount, response time limit or duration, and generate valid dispatch instructions; Valid scheduling instructions are parsed to extract the adjustment direction, adjustment amount, response time limit, duration and execution period. Valid scheduling instructions are divided into time periods according to their execution time periods to generate task time periods; The effective scheduling instructions are split into adjustment amounts according to the adjustment amount and task time period, and segmented adjustment amounts corresponding to the task time period are generated. The adjustment direction, segmented adjustment amount, response time limit, duration, and task time period are bound together to obtain adjustment tasks. The adjustment tasks are then sorted according to the response time limit to obtain an adjustment task sequence.

4. The distributed resource scheduling system for virtual power plants based on big data analysis according to claim 1, characterized in that, The two-layer evaluation module is specifically as follows: Standard data is aggregated according to distributed resources, and the running status data and scheduling execution data of distributed resources are extracted. By associating the adjustment direction, segmented adjustment amount, response time limit, duration, and task time period in the adjustment task sequence with distributed resources, resource task association data is obtained. Input the running status data, scheduling execution data, and resource task association data into the resource-level evaluation layer; The resource-level assessment layer classifies the adjustment capabilities of distributed resources according to response speed, adjustment range, stability, and security, and generates capability classification results, which include effective adjustment class, potential adjustment class, and non-adjustable class. Input the capability classification results and adjustment task sequences into the task-level evaluation layer; The task-level evaluation layer matches adjustment tasks with distributed resources corresponding to effective and potential adjustment classes according to adjustment direction, segmented adjustment amount, response time limit, and duration, and eliminates distributed resources corresponding to non-adjustable classes. For the matching distributed resource tags, corresponding adjustment tasks, task time periods, and capability classification results, generate resource classification results.

5. A distributed resource scheduling system for virtual power plants based on big data analysis according to claim 1, characterized in that, The resource grouping module is specifically as follows: Remove distributed resources marked as unadjustable in the resource classification results, and retain distributed resources corresponding to the effective and potential adjustable categories; The reserved distributed resources are marked with adjustment characteristics according to the adjustment direction, segmented adjustment amount, response time limit and duration corresponding to the adjustment task; Based on the adjustment characteristic labels and capability classification results, the retained distributed resources are grouped and aggregated to generate adjustment resource groups; The resource groups are bound to the corresponding adjustment tasks, task time periods, and edge acquisition nodes to obtain the resource grouping results.

6. The distributed resource scheduling system for virtual power plants based on big data analysis according to claim 1, characterized in that, The task assignment diagram based on adjustment tasks, adjustment resource groups, edge acquisition nodes, and distributed resource construction is as follows: Set the adjustment task, adjustment resource group, edge acquisition node, and distributed resource as task node, resource group node, edge node, and resource node, respectively; Directed edges are established sequentially according to the binding relationship between adjustment tasks and adjustment resource groups, the binding relationship between adjustment resource groups and edge acquisition nodes, and the acquisition mapping between edge acquisition nodes and distributed resources. Write the adjustment direction, segmented adjustment amount, response time limit, duration and task time period into the task node, and write the capability classification results and resource classification results into the resource node to obtain the task assignment diagram.

7. A distributed resource scheduling system for virtual power plants based on big data analysis according to claim 1, characterized in that, The process involves using a residual recursive BFS algorithm to traverse the task assignment paths in the task assignment graph, and then filtering and adjusting the assignment paths corresponding to the tasks to form capability packages. Specifically: Select adjustment tasks sequentially according to the sorting results in the adjustment task sequence; Establish a queue to be expanded for the selected adjustment tasks, record the segmented adjustment amount corresponding to the adjustment task as the initial adjustment margin, and add the task node corresponding to the adjustment task to the queue to be expanded. The queue to be expanded is expanded layer by layer according to the hierarchical order of task nodes, resource group nodes, edge nodes and resource nodes. When extending to resource group nodes, paths corresponding to adjustment resource groups that are not bound to adjustment tasks are filtered out; When extending to edge nodes, paths corresponding to edge acquisition nodes that are not bound to the adjustment resource group are filtered out. When extending to resource nodes, paths corresponding to distributed resources whose capability classification results are non-adjustable are excluded. Based on standard data and capacity classification results, determine the adjustable capacity and sustainable adjustment duration corresponding to resource nodes; If the duration of sustainable adjustment is less than the duration of the adjustment task, the corresponding path will be filtered out. Subtract the available adjustment amount from the adjustment reserve to obtain the updated adjustment reserve; if the updated adjustment reserve is greater than 0, the formed path is retained and the untraversed path of the same adjustment task is continued to be expanded; if the updated adjustment reserve is less than or equal to 0, the path participating in the reserve recursion is taken as the corresponding receiving path of the adjustment task. The capability package is formed by aggregating the acceptance path, the corresponding distributed resources, the adjustable capacity, the task time period, the response time limit, and the capability classification results.

8. A distributed resource scheduling system for virtual power plants based on big data analysis according to claim 1, characterized in that, The strategy generation module specifically comprises: According to the sorting results in the adjustment task sequence, select the adjustment tasks in order; extract the adjustment direction, segmented adjustment amount, response time limit, duration and task time period corresponding to the adjustment task; Select and adjust the task time period, response time limit, duration and capability classification results of the task from the capability package to match the receiving path. The segmented adjustment amount is allocated according to the adjustable amount that can be accepted according to the receiving path, and the allocated adjustment amount corresponding to the receiving path is obtained. If the cumulative value of the adjustable amount that can be accepted for the corresponding path is less than the segmented adjustable amount, then continue to select the corresponding path for the same adjustable task from the capability package. If the cumulative value of the adjustable amount that can be accepted for the corresponding path is greater than or equal to the segmented adjustable amount, then stop selecting the corresponding path for the same adjustable task. The adjustment tasks are assigned to the edge acquisition nodes and distributed resources corresponding to the selected receiving paths; By associating adjustment tasks, adjustment directions, receiving paths, edge acquisition nodes, distributed resources, allocated adjustment amounts, response time limits, durations, and task time periods, a global scheduling strategy is generated.

9. A distributed resource scheduling system for virtual power plants based on big data analysis according to claim 1, characterized in that, The process involves configuring instruction packets according to the global scheduling strategy and sending these packets to edge acquisition nodes. The edge acquisition nodes then issue resource scheduling instructions to the resource adjustment group based on these instruction packets. Specifically: The global scheduling strategy is aggregated according to the edge acquisition nodes, and the aggregation results are written into the instruction packet. Write the edge acquisition node, adjustment resource group, distributed resources, adjustment direction, allocated adjustment amount, response time limit, duration, task time period and receiving path into the instruction package; Send the instruction packet to the corresponding edge acquisition node; The edge acquisition node parses the instruction packet to extract the adjustment resource group, distributed resources, adjustment direction, allocated adjustment amount, response time limit, duration and task time period; Edge acquisition nodes issue resource scheduling instructions to the adjustment resource group according to the task time period. The resource scheduling instructions include distributed resources, adjustment direction, allocation adjustment amount, response time limit and duration.

10. A distributed resource scheduling system for virtual power plants based on big data analysis according to claim 1, characterized in that, The process of updating the resource classification results, capability packages, and global scheduling strategy based on the execution results of resource scheduling instructions, and outputting distributed resource scheduling results, specifically involves: Obtain the execution results of resource scheduling instructions. The execution results include distributed resources, adjustment direction, actual adjustment amount, actual response time, actual duration, and task period. The execution results are compared with the distributed resources, adjustment direction, allocation adjustment amount, response time limit, duration and task period in the global scheduling strategy to obtain the execution deviation; The capability classification results in the resource classification results are adjusted according to the execution deviation to obtain the updated resource classification results; Based on the updated resource classification results and execution deviations, the acceptance paths, adjustable quantities, response time limits, and capability classification results in the capability package are adjusted to obtain the updated capability package. Based on the updated capability package, the receiving path, edge acquisition node, distributed resources, and allocation adjustment amount in the global scheduling strategy are adjusted to obtain the updated global scheduling strategy. The execution results, execution deviations, updated resource classification results, updated capability packages, and updated global scheduling policies are correlated to output distributed resource scheduling results.

Images