A Virtual Power Plant Optimal Scheduling and Revenue Allocation Method for Multi-Source Demand-Side Resources

Abstract

This invention discloses a method for optimized scheduling and revenue distribution in virtual power plants oriented towards multi-source demand-side resources, applicable to regional virtual power plants containing industrial load nodes, building load nodes, and distributed photovoltaic energy storage nodes. The method first collects real-time operating data from each node, generates a node operating status table, and classifies nodes based on historical performance. A scheduling negotiation table is constructed through the pairing relationship between active and lagging nodes, determining the negotiation round and instruction priority in the event of resource sharing conflicts. The system generates a joint instruction set based on the scheduling negotiation table and the status table. After the scheduling cycle ends, the execution results are collected, and a lagging resource incentive reserve pool is established to incentivize nodes with adjustment intentions but failed to respond. Finally, node revenue is generated based on the completion status of adjustment tasks and the contents of the incentive reserve pool, and the scheduling priority for the next cycle is dynamically adjusted accordingly, achieving closed-loop optimization control driven by revenue.

Description

Technical Field

[0001] This invention relates to the field of power system dispatching technology, and in particular to a method for optimized dispatching and revenue distribution of virtual power plants oriented towards multi-source demand-side resources. Background Technology

[0002] With the rapid development of new power systems and distributed energy resources, virtual power plants (VPS), as an important platform for aggregating and regulating distributed power sources, controllable loads, and energy storage resources, have become a crucial component of the power dispatching system. Existing VPS dispatching methods mostly employ centralized optimization strategies, generating regulation commands by combining real-time market prices and load forecasts, and then distributing the dispatch plan to each resource node through a remote control system. Regarding resource incentives, existing technologies generally allocate revenue based on regulation power or the number of participations. Some systems introduce performance scoring mechanisms to evaluate response accuracy, thereby enhancing the user-side incentive to regulate resources.

[0003] However, existing technologies generally suffer from several problems: First, insufficient consideration is given to the historical performance capabilities of nodes, making it difficult to effectively distinguish differences in scheduling reliability, resulting in significant deviations in the execution of control plans; second, during resource allocation, there is a lack of coordination mechanisms for resource contention conflicts, which may cause multiple schedulers to repeatedly rely on the same redundant resources; third, nodes that fail to be scheduled are often directly removed, failing to consider their willingness to participate and room for improvement, which can easily lead to resource loss; fourth, the revenue distribution rules are generally static, failing to form a closed-loop incentive relationship with the actual performance of nodes, making it difficult to support continuous optimization of scheduling quality.

[0004] To address the aforementioned issues, this invention proposes a virtual power plant optimization scheduling and revenue distribution method oriented towards multi-source demand-side resources. Summary of the Invention

[0005] This application provides a method for optimizing the scheduling and revenue distribution of virtual power plants for multi-source demand-side resources, so as to improve the success rate of virtual power plant multi-source demand-side resource scheduling tasks.

[0006] This application provides a method for optimized scheduling and revenue allocation of virtual power plants for multi-source demand-side resources, including:

[0007] Acquire the current operating data of each industrial load node, building load node, and distributed photovoltaic energy storage node, and generate a node operating status table;

[0008] Based on the node running status table, the node performance is classified, a scheduling negotiation table and a joint instruction set are constructed, and the joint instruction set is issued to each node to obtain the node execution result table.

[0009] Update the performance record based on the node execution result table, and construct a delayed resource incentive reserve pool;

[0010] Based on the incentive reserve pool table and the power adjustment instructions in the joint instruction set, the instant settlement revenue of each node is generated; and the priority of redundancy contest in the next cycle is adjusted according to the revenue level to obtain the performance penalty table and node revenue allocation table for the next cycle.

[0011] The beneficial effects of the technical solution provided in this application include:

[0012] (1) By introducing a node classification mechanism based on historical performance success rate, active nodes and lagging nodes are effectively identified, and the impact of unreliable resources is eliminated or weakened before scheduling, significantly reducing scheduling deviation and failure rate. (2) By using a scheduling negotiation table mechanism, the resource sharing conflict caused by multiple active nodes competing for the redundant adjustment capacity of the same lagging node is resolved, realizing coordinated scheduling among multiple nodes and improving the system's resource utilization efficiency. (3) By establishing a lagging resource incentive reserve pool, lagging nodes that have not fully performed but have the willingness to participate in adjustment can obtain retention benefits, supporting their continued participation in subsequent scheduling and effectively preventing resources from leaving the aggregation system prematurely. (4) Real-time benefits are dynamically generated based on the node execution results and used to adjust the redundancy priority ranking of the next cycle, forming an incentive feedback mechanism that links scheduling performance with benefit distribution, continuously improving the overall performance of the scheduling system. Attached Figure Description

[0013] Figure 1 This is a flowchart of a virtual power plant optimization scheduling and revenue distribution method for multi-source demand-side resources provided in the first embodiment of this application. Detailed Implementation

[0014] Many specific details are set forth in the following description to provide a full understanding of this application. However, this application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this application; therefore, this application is not limited to the specific embodiments disclosed below.

[0015] The first embodiment of this application provides a method for optimized scheduling and revenue distribution of virtual power plants oriented towards multi-source demand-side resources. Please refer to... Figure 1 This figure is a schematic diagram of the first embodiment of this application. The following is in conjunction with... Figure 1 The first embodiment of this application provides a method for optimized scheduling and revenue distribution of virtual power plants oriented towards multi-source demand-side resources.

[0016] Step S101: Obtain the current operating data of each industrial load node, building load node, and distributed photovoltaic energy storage node, and generate a node operating status table; wherein, the operating data includes instantaneous power value, remaining regulation capacity, load adjustable time period, and historical performance status.

[0017] In step S101, it is first necessary to clarify the type and scope of resource nodes. The resource nodes described in this invention include, but are not limited to, three main categories: industrial load nodes, building load nodes, and distributed photovoltaic energy storage nodes. Industrial load nodes typically refer to industrial user-side equipment with adjustable power loads, such as large machining equipment, frequency converter-controlled production lines, and electric heating devices. Building load nodes include power-consuming units with a certain degree of load flexibility, such as HVAC systems, elevators, and lighting systems in office buildings and commercial buildings. Distributed photovoltaic energy storage nodes refer to power-side resources that integrate distributed photovoltaic power generation devices and supporting energy storage battery systems; they can both feed power to the grid and absorb and regulate loads as needed.

[0018] To comprehensively collect operational data from the aforementioned nodes, four key types of operational data need to be collected in real time by edge controllers (such as smart meters, power monitoring modules, and BMS energy storage management systems) deployed on each node side, and then uploaded to the virtual power plant main control platform in a structured format. The first type of data is instantaneous power values, typically recorded at a time granularity of one second or one minute, showing the actual power consumption (for load nodes) or output power (for photovoltaic energy storage nodes) of the current node, in kilowatts (kW) or watts (W). This value can be obtained by sampling voltage and current and performing power calculations, and should support periodic timestamps to construct a continuous power curve.

[0019] The second type of data is residual adjustment capacity, which reflects the maximum power range that the current node can adjust upwards or downwards without affecting the user's basic operation. For example, for air conditioning loads, the maximum adjustable power may be limited by the indoor temperature setpoint, comfort limits, etc. This data should include two directional parameters: the adjustable power value and the adjustable power value, which can be calculated in real time by combining the load forecasting model with the current equipment operating status (such as inverter power, load rate, etc.).

[0020] The third type of data is the load adjustable time period, which refers to the time window during which each node is allowed to participate in adjustment operations within a future scheduling cycle. For example, an industrial node operates between 9:00 and 17:00 on weekdays, and its adjustment participation period may be limited to two windows: 10:00 to 12:00 and 14:00 to 16:00; building nodes may need to avoid peak elevator hours or air conditioning start-up times. This data should be stored in timestamp pairs, for example, indicating whether adjustment is allowed at 15-minute intervals, to support subsequent window overlap determination.

[0021] The fourth category of data is historical performance data, reflecting the actual completion status of nodes after receiving adjustment commands in the past several scheduling cycles. This may include indicators such as response success rate, average response latency, and execution accuracy deviation. For example, the performance success rate can be calculated by dividing the number of successfully completed adjustment tasks in the last 30 scheduling cycles by the total number of tasks received, which serves as a basis for subsequent classification. This data should be extracted from the historical records of the virtual power plant dispatch control platform, or periodically transmitted back and stored in the database by the edge controller.

[0022] The above four types of data are integrated into a node operation status table, indexed by node ID. This table can adopt a two-dimensional structure, with each row corresponding to a node and each column corresponding to an operation indicator field, along with an update timestamp. The node operation status table can be built using database systems such as MySQL, or it can be combined with a time-series database for higher-frequency data processing.

[0023] After constructing the node running status table, the system needs to set the data refresh cycle (e.g., 5 minutes or 15 minutes) and fault tolerance mechanism. For example, nodes with delayed uploads or missing data should be temporarily marked as "incomplete status" and processed or downgraded in subsequent steps.

[0024] By implementing this step, the virtual power plant platform can accurately perceive the operational capabilities of various types of demand-side resources, providing a detailed, measurable, and comparable data foundation for subsequent scheduling optimization, and significantly improving the reliability of scheduling decisions and the adaptability of the entire system.

[0025] Step S102: Classify node performance based on the node running status table, construct a scheduling negotiation table and a joint instruction set, and issue the joint instruction set to each node to obtain the node execution result table.

[0026] The process involves classifying node performance based on the node running status table, constructing a scheduling negotiation table and a joint instruction set, issuing the joint instruction set to each node, and obtaining a node execution result table, including:

[0027] Step S102-1: Based on the node running status table, select nodes with a historical performance success rate lower than a preset threshold as lagging nodes, and the rest as active nodes to obtain the node performance classification results;

[0028] Step S102-2: Based on the node performance classification results, according to spatial proximity and adjustable window overlap, assign up to two lagging nodes to each active node, and establish a scheduling negotiation table;

[0029] Step S102-3: Based on the scheduling negotiation table and the node running status table, generate a joint instruction set that includes the target adjustment power of active nodes, the share of transferable adjustment tasks of lagging nodes, the redundancy sharing ratio, and the conflict resolution identifier; issue the joint instruction set to each node, and collect the execution data at the end of the scheduling cycle to obtain the node execution result table.

[0030] Step S102-1: Based on the node operation status table, select nodes with a historical performance success rate lower than a preset threshold as lagging nodes, and the rest as active nodes to obtain the node performance classification results.

[0031] In practical implementation, a clear calculation period for the success rate of contract fulfillment needs to be established first. For example, the past 30 consecutive scheduling cycles can be selected as the statistical window. For each node, the system extracts the total number of times it received adjustment commands within these 30 cycles, as well as the number of times it actually responded successfully. The criteria for judging a successful response should be set in advance according to the nature of the adjustment task and the system tolerance requirements. For example, the response start time delay should not exceed 30 seconds, the adjustment power error should not exceed a set threshold (such as ±10%), and the response duration should not be less than 80% of the target time. Only when these criteria are met can an adjustment be counted as "successfully fulfilled".

[0032] Based on the above criteria, the system can calculate the fulfillment success rate for each node, which is the number of successful tasks divided by the total number of tasks, expressed as a percentage. Subsequently, the system performs preliminary screening of all nodes based on a pre-defined fulfillment threshold, such as 90%. Nodes with a fulfillment success rate higher than this threshold are marked as "active nodes," indicating reliable adjustment capabilities and stable response behavior, making them suitable for priority in undertaking adjustment tasks. Nodes with a fulfillment success rate lower than the threshold are marked as "lagging nodes," indicating some uncertainty in their adjustment or a high number of historical response failures, requiring them to be limited in subsequent scheduling or have other nodes cooperate in undertaking adjustment tasks. The aforementioned threshold can be preset based on historical operating experience or dynamically adjusted during actual system operation.

[0033] In practical engineering implementation, to prevent nodes from being misjudged as lagging nodes due to occasional failures or temporary communication interruptions, the system can introduce certain smoothing or compensation mechanisms. For example, a minimum task quantity threshold can be set, and the performance rate will only be evaluated when a node receives more than 10 tasks within the calculation cycle, preventing misjudgments due to insufficient statistical samples. At the same time, a "attempted response" weighting mechanism can be introduced, that is, for nodes that have not fully achieved the adjustment target but have indeed made adjustment actions, their failure counts are included by a certain reduction ratio to reflect the node's responsiveness and potential for improvement.

[0034] After classification, the system combines the unique identifier of each node with its classification result to form a "Node Performance Classification Result," which can be stored in a table or dictionary structure, such as key-value pairs: {Node ID → Classification Tag}, where the classification tag can be "active" or "delayed." This classification result will serve as the basis for the construction of the scheduling negotiation table and the task matching logic in subsequent steps, and will be dynamically updated after the scheduling execution results are fed back, realizing the system's rolling self-optimization.

[0035] Through the implementation of this step, the system can accurately identify high-reliability nodes and nodes with insufficient performance capabilities, realize hierarchical management of regulation resources, provide a structural premise for joint regulation and scheduling mechanism, and establish a performance basis for scheduling results and revenue distribution, thereby improving the scheduling execution efficiency and control stability of the entire virtual power plant platform.

[0036] Step S102-2: Based on the node performance classification results, according to spatial proximity and adjustable window overlap, assign up to two lagging nodes to each active node, and establish a scheduling negotiation table. The scheduling negotiation table is used to record the negotiation rounds and instruction priorities when multiple active nodes and the same lagging node have resource sharing conflicts.

[0037] First, while acquiring the node performance classification results, the system also combines the node operation status table generated in step S101 to extract the geographical location information and adjustable time period of each node. The geographical location information is typically provided by the GIS module; each node has precise coordinates or can be mapped to a unified geographic raster number. In this embodiment, the system defines "spatial proximity" as two nodes belonging to the same sub-region in geographic space, or their physical distance being less than a certain set threshold (e.g., 2 kilometers). This distance threshold can be flexibly adjusted based on actual engineering constraints such as communication network coverage and response speed adjustment.

[0038] Secondly, in the time dimension, for each active node, the system analyzes the specific time period during which it can perform adjustments within the future scheduling cycle (e.g., 9:00 AM to 11:30 AM) and calculates the overlap window between this time period and the adjustable time periods of all lagging nodes. Specifically, the "window overlap" is obtained by the ratio of the intersection interval length to the active node's own window length. If the overlap exceeds a certain threshold (e.g., 60%), the lagging node is considered to be temporally compatible. Under the premise of satisfying both spatial proximity and temporal window overlap conditions, the system assigns a maximum of two lagging nodes to each active node, prioritizing lagging nodes with high overlap and close geographical distance.

[0039] To avoid resource duplication or response conflicts caused by multiple active nodes simultaneously matching the same lagging node, the system explicitly marks the conflicting node group in such cases when establishing the scheduling negotiation table, and generates a "negotiation round" and "instruction priority" for each conflict scenario. The negotiation round can be automatically sorted based on factors such as the order of scheduling request submission, the node's historical contribution record, or the urgency of the task in the current cycle, while the instruction priority is used to control the execution priority when subsequent scheduling instructions conflict. For example, if both active node A and active node B match lagging node C, and the system predicts that C's adjustment capacity is insufficient to support the needs of both, then the negotiation table records the following information: Lagging node C → Conflicting active node {A, B}, negotiation round = A priority, instruction priority = A high, B low.

[0040] The scheduling negotiation table typically employs a multi-level mapping structure. The primary key is the active node ID, and the value range is the set of its corresponding lagging node IDs. Each active-lagging mapping is accompanied by a negotiation information structure, including fields such as matching score, whether a conflict exists, the set of conflicting node IDs, round number, and instruction priority identifier. The system can periodically refresh this table to adapt to changes in node status or updates to fulfillment classifications.

[0041] Through the complete execution of the above process, this step enables the structured embedding of unstable resources into the stable resource scheduling process, and the prediction and management of resource usage conflicts through the scheduling negotiation table. This makes the generation of subsequent joint regulation instructions more coordinated and executable, significantly reduces the risk of failure caused by resource competition during the scheduling phase, and improves the flexibility and regulation accuracy of the entire virtual power plant under the background of multi-source load.

[0042] For example, during the operation cycle of a virtual power plant, the system acquires the status information of six nodes, including three active nodes A1, A2, and A3, and three lagging nodes B1, B2, and B3. Node A1 is located at coordinates (120.10, 30.25) in the X region, with an adjustable time window from 9:00 to 11:00; node A2 is located at coordinates (120.13, 30.27) in the X region, with an adjustable time window from 10:00 to 12:30; and node A3 is located at coordinates (121.50, 31.20) in the Y region, with a time window from 14:00 to 16:00. The lagging nodes B1, B2, and B3 are located in region X (120.12, 30.26), region X (120.11, 30.23), and region Y (121.48, 31.18), respectively, with time windows of 9:30 to 11:30, 10:00 to 11:00, and 13:30 to 15:30.

[0043] The system first calculates the spatial proximity between each active node and the lagging node, setting a spatial proximity threshold of 3 kilometers. Using the GIS calculation module, the Euclidean geographical distance between all active and lagging nodes is calculated. The results show that the distances between A1 and B1 and B2 are both less than 3 kilometers, A2 is also within the proximity range of B1 and B2, while A3 only satisfies spatial proximity with B3.

[0044] Next, the system calculates the overlap of adjustment windows between each active node and its neighboring lagging nodes. For example, A1's window is 9:00 to 11:00, and B1's window is 9:30 to 11:30. Their intersection is 9:30 to 11:00, totaling 1.5 hours, accounting for 75% of A1's window, which is greater than the system's set threshold of 60%. The intersection of A1 and B2's windows is 10:00 to 11:00, accounting for 50% of A1's window, which is below the threshold and is not included in the pairing. The intersection of A2 and B1 is 10:00 to 11:30, accounting for 60% of A2's window; the intersection of A2 and B2 is 10:00 to 11:00, accounting for 40% of A2's window, which does not meet the requirements. The intersection of A3 and B3 is 14:00 to 15:30, accounting for 75% of A3's window.

[0045] Based on the matching results, the system assigned B1 to A1, B1 to A2 (B2 did not meet the overlap threshold), and B3 to A3. Since both A1 and A2 matched with B1, the system determined a resource sharing conflict and triggered the scheduling negotiation table construction mechanism. The system analyzed the historical fulfillment rates of A1 and A2; A1's rate was 95%, and A2's was 88%. Therefore, A1 gained scheduling priority. The scheduling negotiation table records: Conflicting lagging node = B1, Conflicting active node group = {A1, A2}, Negotiation round = 1, Priority allocation = {A1>A2}.

[0046] The final generated scheduling negotiation table is as follows:

[0047] A1 → [B1] (Priority: High)

[0048] A2 → [B1] (Priority: Low, needs to wait)

[0049] A3 → [B3] (No conflict)

[0050] Furthermore, based on the node performance classification results, and according to spatial proximity and adjustable window overlap, each active node is assigned at most two lagging nodes, and a scheduling negotiation table is established, including:

[0051] Based on the node performance classification results, the load adjustable time period of each active node and the load adjustable time period of the lagging node are extracted, and the ratio of their intersection length to union length is calculated as the adjustable window overlap index; only when the ratio is greater than a set threshold, the lagging node is included in the candidate set to obtain a preliminary adjustable point pair set.

[0052] Based on the aforementioned set of adjustable point pairs, the equivalent resistance distance in the power grid topology is used as the spatial proximity metric. The lagging nodes associated with each active node are sorted from smallest to largest, and the top two lagging nodes are selected as the scheduling matching objects for the active node to obtain the node allocation list.

[0053] Based on the historical performance stability index of each active node in the node allocation list, a negotiation priority weight is generated. In the case of multiple active nodes competing for the same lagging node, the scheduling control sequence is determined according to the rule that the higher priority node takes precedence and subsequent nodes make back-off adjustments based on the remaining adjustable capacity.

[0054] A scheduling negotiation table is constructed based on the scheduling control sequence. The scheduling negotiation table is used to record the negotiation priority, scheduling instruction order and conflict resolution round information between each active node and its corresponding lagging node.

[0055] First, after obtaining the node performance classification results, all target objects classified as "active nodes" and candidate objects classified as "lagging nodes" are processed separately. For each active node, the system extracts its current and future adjustable load time periods, such as the adjustment start and end times recorded in hours. Simultaneously, the adjustable time periods are extracted from the lagging nodes. Next, the system calculates the overlap of the adjustable windows for each pair of active and lagging nodes. Specifically, the intersection of their adjustable time periods is taken to obtain the common adjustable duration, denoted as . Then, the union of the two adjustable time periods is taken to obtain the total feasible adjustment duration, denoted as . Then calculate their ratio. This ratio reflects the degree of adjustment matching between the two nodes over time. The system sets a dynamic threshold. (If the default value is 0.5, it can be adjusted in real time according to load forecasting). If the ratio... If a lagging node is deemed capable of coordinating with the active node to perform the adjustment task in time, it is retained; otherwise, it is discarded. This process is repeated for all node combinations to obtain a preliminary set of all adjustable point pairs that meet the conditions.

[0056] After completing the time window matching, the system further filters the aforementioned node pairs spatially based on the power grid topology. Specifically, the spatial proximity between each active node and its candidate lagging nodes is not measured by physical straight-line distance, but rather by the equivalent resistance distance between them. The equivalent resistance distance refers to the equivalent impedance length along the electrical connection path from the feeder of an active node to the feeder of a lagging node in the power system topology diagram. Considering factors such as voltage level, wiring method, and transformer impedance, it can be calculated using topology graph algorithms or matrix methods. For each active node, the system sorts them according to their equivalent resistance distance from smallest to largest, retaining the top two lagging nodes as its final scheduling matching targets, forming the node allocation list. This filtering method not only ensures efficient matching of power regulation flow but also effectively reduces control delays and execution errors caused by distant nodes.

[0057] After node allocation is completed, the system further considers potential contention during scheduling. Since multiple active nodes may select the same lagging node as a collaboration partner, the system introduces a scheduling priority control mechanism to resolve resource conflicts. Specifically, the scheduling negotiation priority of each active node is calculated based on its historical performance stability. For example, a negotiation priority weight is synthesized using dimensions such as success rate, response time, and adjustment error rate from the last N scheduling operations. A higher value indicates better stability. Before scheduling execution, if multiple active nodes are detected to contain the same lagging node in their node allocation lists, the system first determines the node with the highest negotiation priority as the preferred scheduler, allowing it to issue adjustment instructions to the lagging node first. Other active nodes then readjust their target adjustment power based on the remaining adjustment capacity of the lagging node. If the adjustment capacity is insufficient, a fallback adjustment strategy is executed, such as switching to a backup lagging node or lowering the adjustment expectation.

[0058] Based on the aforementioned scheduling control sequence, a scheduling negotiation table is ultimately established. This table is a structured mapping dataset used to define scheduling priorities. It records the lagging nodes corresponding to each active node, the scheduling priorities between them, the order initiation of instructions, and the number of negotiation rounds required in resource conflict situations. The table may also include conflict resolution identifiers for the current scheduling cycle, used to automatically identify and resolve scheduling interference in dynamic environments. This scheduling negotiation table will be directly input into the subsequent joint adjustment instruction generation process to ensure the consistency, continuity, and feasibility of overall scheduling instructions.

[0059] In summary, the entire allocation and negotiation process not only considers the three dimensions of matching time, space, and behavioral stability, but also establishes a collaborative priority mechanism and a fallback adjustment strategy.

[0060] For example, in a given scheduling cycle, the system first obtains the current node performance classification results, identifying nodes A, B, and C as active nodes, and nodes X, Y, Z, and W as lagging nodes. For node A, its adjustable load period is from 9:00 to 12:00; for node X, it is from 10:00 to 13:00; for node Y, it is from 8:30 to 9:30; and for node Z, it is from 9:30 to 11:30. The adjustable window intersection of nodes A and X is from 10:00 to 12:00, a total of 2 hours, and the union of the time periods is from 9:00 to 13:00, a total of 4 hours. Therefore, the adjustable window overlap is 2 / 4 = 0.5, which is exactly equal to the set threshold, so it is retained. The intersection of nodes A and Y is from 9:00 to 9:30, a length of 0.5 hours, and the union is from 8:30 to 12:00, a total of 3.5 hours. The overlap is 0.5 / 3.5 ≈ 0.14, which is far below the threshold of 0.5, so it is discarded. The intersection of nodes A and Z is from 9:30 to 11:30, a total of 2 hours, and the union is from 9:00 to 12:00, a total of 3 hours. The overlap is 2 / 3 ≈ 0.67, which is greater than the threshold, so it is retained. Therefore, the adjustable point pairs for node A are initially selected as AX and AZ.

[0061] After initial node pair screening, the system evaluates the spatial proximity of node A with X and Z based on the equivalent resistance distance of the power grid topology. The calculated equivalent resistance distance between node A and X is 0.23 ohms, and between node A and Z is 0.17 ohms. Sorted from smallest to largest, the results are Z and X. Therefore, node A is ultimately matched with two lagging nodes, Z and X, forming the node allocation list for node A.

[0062] Subsequently, the system evaluates the historical performance stability of node A. Assuming it successfully responded 18 times, experienced one delay, and failed once in the last 20 scheduling attempts, the system assigns a performance stability score of 0.93, which is also set as its negotiation priority weight. Assuming node B also selects the lagging node Z with a performance score of 0.85, and node C selects X with a score of 0.78, then in scheduling conflict resolution, node A's instructions to Z have a higher priority than node B's, and node A's instructions to X also have a higher priority than node C's. Therefore, the system determines that when issuing scheduling instructions, node A prioritizes scheduling Z and X. Node B must attempt to schedule other candidate nodes W. If W's time matching and spatial distance do not meet the requirements, node B reduces its regulation power target from 30kW to 15kW and regenerates the scheduling request. Node C also performs a similar rollback operation.

[0063] Finally, the scheduling negotiation table is constructed and stored, which includes that the scheduling priority of nodes A and Z is 1, the priority of nodes A and Z is 1, the priority of node B to Z is 2 and the conflict resolution round is marked as 1 backoff, the priority of node C to X is 2 and the conflict resolution round is also 1. Based on this, the system ensures that the scheduling process is stable and efficient.

[0064] Step S102-3: Based on the scheduling negotiation table and the node operation status table, generate a joint instruction set containing the target adjustment power of active nodes, the share of transferable adjustment tasks for lagging nodes, the redundancy sharing ratio, and the conflict resolution identifier; issue the joint instruction set to each node, and collect the execution data at the end of the scheduling cycle to obtain a node execution result table; wherein, the execution data includes the adjustment start and end time, actual adjustment power, and execution delay status of each node.

[0065] In implementing step S102-3, the system needs to integrate the previously established scheduling negotiation table and node operating status table to generate a set of joint instructions for multi-source load coordinated response. First, the system performs node-by-node calculations based on the target adjustment power of each active node in the current scheduling cycle. This target adjustment power can be obtained by analyzing the real-time load forecasting model, the performance of the previous cycle, and the system balancing requirements, ensuring that the adjustment tasks undertaken by each node both meet the overall scheduling strategy and do not exceed its remaining adjustment capacity.

[0066] Subsequently, based on the lagging node information matched in step S102-2, the system determines the transferable adjustment task share for each lagging node. The calculation of this share considers not only the maximum adjustment capacity of the lagging node, but also its willingness to participate, the dynamic penalty coefficient of its historical failure rate, and the current adjustment redundancy of the matched active nodes. To ensure adjustment stability, the system sets a minimum response ratio threshold for the task share; if the response ratio is below this threshold, the lagging node is not assigned a task share.

[0067] Next, the system generates a redundancy sharing ratio for each active node, which is the degree of redundancy to which the node can call upon the capabilities of lagging nodes. This ratio is dynamically adjusted based on the current load fluctuation of the system and the degree of overlap of time windows between nodes. If the overlap between an active node and a lagging node is high, the ratio can be set between 0.8 and 1.0; if the overlap is low, it can be set between 0.3 and 0.6 to avoid adjustment failures caused by scheduling uncertainties.

[0068] Meanwhile, to resolve potential conflicts when multiple active nodes compete for the same lagging node resource, the system also needs to add a conflict resolution flag to the joint instruction set. This flag is dynamically generated based on the negotiation rounds and priorities recorded in the scheduling negotiation table, indicating which active node has the right to adjust and assist the lagging node, as well as the switching threshold and instruction coverage order under the backup plan, ensuring that the execution logic among multiple nodes does not conflict or overlap.

[0069] All the above parameters are uniformly encapsulated into a structured data packet, forming a joint instruction set. This set includes the target regulation power value, redundancy sharing ratio, regulation task transfer quota for the corresponding lagging node, and conflict resolution strategy number for each active node. This instruction set will be distributed to all participating nodes through edge control nodes or the virtual power plant master control platform, and all nodes will execute regulation tasks according to the preset timestamps in the instructions.

[0070] At the end of the scheduling cycle, the system collects the actual execution data of each node, mainly including the task start and end times, the actual output power value recorded during the adjustment process, and whether execution delay occurred. The execution delay status is automatically determined by comparing the node's local clock with the command issuance time of the central controller, and combined with the set fault tolerance window to determine whether a breach of contract has occurred. The adjustment execution results of all nodes are uniformly recorded in the node execution result table, providing a basis for subsequent performance updates and revenue distribution.

[0071] In summary, steps S102-3 not only achieve multi-node collaborative allocation of regulation tasks, but also ensure efficient execution of scheduling tasks and optimal utilization of regulation resources through multi-parameter joint control and conflict resolution mechanisms, thereby providing technical support for the stability and economy of the entire virtual power plant system. This step is particularly crucial in multi-source heterogeneous demand response scenarios, serving as a core link in ensuring system robustness and fairness of benefits.

[0072] In practical applications, to more clearly illustrate the generation method of the joint instruction set, a specific example is given to detail the execution process of step S104. Assume the current scheduling period is from 10:00 AM to 10:15 AM, a total of 15 minutes. The system contains three active nodes A1, A2, and A3, and two lagging nodes B1 and B2. The running status table of each node records the following information:

[0073] A1: The remaining regulating capacity is 20kW, the adjustable time period is 10:00–10:15, the historical fulfillment success rate is 95%, and the instantaneous power value is 150kW;

[0074] A2: The remaining regulation capacity is 25kW, the adjustable time period is 10:05–10:15, the historical performance success rate is 92%, and the instantaneous power value is 140kW.

[0075] A3: The remaining regulating capacity is 15kW, the adjustable time period is 10:00–10:10, the historical fulfillment success rate is 90%, and the instantaneous power value is 160kW;

[0076] B1: The remaining regulating capacity is 18kW, the adjustable time period is 10:00–10:15, the historical fulfillment success rate is 65%, and the instantaneous power value is 120kW;

[0077] B2: The remaining regulation capacity is 12kW, the adjustable time period is 10:00–10:10, the historical performance success rate is 60%, and the instantaneous power value is 130kW.

[0078] First, according to step S103, the system has established a scheduling negotiation table, determining that A1 matches B1 and B2, A2 matches B1, and A3 matches B2. The conflict negotiation rounds are marked as follows: B1 is selected by both A1 and A2, the negotiation round is 1, A1 priority is 1, and A2 priority is 2; B2 is selected by both A1 and A3, the negotiation round is 1, A3 priority is 1, and A1 priority is 2.

[0079] The system then determines the target regulation power based on the load forecast trend of each active node and the system load reduction target. Assume that the system requires A1, A2, and A3 to undertake target regulation power of 18kW, 22kW, and 13kW respectively, of which at least 40% must be completed by the active nodes themselves, and the surplus can be transferred to the lagging nodes.

[0080] A1 needs to assume its own regulation capacity = 18kW × 0.4 = 7.2kW (actually set at 8kW), and needs to transfer 10kW to the lagging node;

[0081] A2 needs to assume its own regulation capacity = 22kW × 0.4 = 8.8kW (set to 9kW), and transfer 13kW;

[0082] A3 needs to assume self-regulating capacity = 13kW × 0.4 = 5.2kW (set to 6kW), and transfer 7kW.

[0083] Considering the remaining capacity of the lagging nodes, B1 can handle a maximum of 18kW, and B2 a maximum of 12kW. The system needs to assess the degree of overlap in time periods and spatial proximity, and calculate the redundancy sharing ratio for each pair of nodes. Assuming that A1 and B1 completely overlap in time periods and are less than 100 meters apart, their redundancy sharing ratio is set to 0.85; A2 and B1 only overlap from 10:05 to 10:15, set to 0.65; A1 and B2 overlap from 10:00 to 10:10, set to 0.75; A3 and B2 completely overlap but are far apart, set to 0.70.

[0084] Based on the conflict priority, B1's resources are preferentially allocated to A1, so A1 obtains a 10kW adjustment share from B1, satisfying its target; although A2 needs 13kW, it only obtains the remaining 8kW (18-10) from B1, leaving 5kW unsatisfied, and its redundancy sharing ratio is reduced to 0.5, recording the conflict resolution flag as "2-compensation scheduling wait"; A3 obtains 7kW from B2, which is fully satisfied because its priority is 1.

[0085] Finally, the system encapsulates the above calculation results into a unified instruction set, as shown in the following example:

[0086] A1 instruction: Target power 18kW, self-sufficient 8kW, B1 adjustment 10kW, redundancy ratio 0.85, conflict resolution flag 1;

[0087] A2 instruction: Target power 22kW, self-sufficient 9kW, B1 adjustment 8kW, redundancy ratio 0.65, conflict resolution indicator 2;

[0088] A3 instruction: Target power 13kW, self-sufficient 6kW, B2 adjustment 7kW, redundancy ratio 0.70, conflict resolution flag 1;

[0089] B1 instruction: Transfer the regulation task of 18kW to A1 and A2;

[0090] B2 instruction: Transfer the regulation task of 7kW to A3.

[0091] All instructions are encapsulated into a scheduling instruction package and transmitted to the corresponding nodes through the control system. The nodes trigger the adjustment action at 10:00. After the scheduling cycle ends, the start and end times of the adjustment, the power response curve per minute, and the response delay (if any) are collected and written into the execution result table to provide a data basis for subsequent performance judgment and revenue distribution.

[0092] After generating the joint instruction set, the system will synchronously issue scheduling instructions to all active and lagging nodes through the scheduling control bus of the virtual power plant platform. The scheduling instructions are encapsulated in a unified structure format as JSON or XML data packets, specifically including information such as the adjustment task target value of each node in the current scheduling cycle, its own share, the identifier of the connected cooperating node, the start and end times of adjustment, the redundancy sharing ratio, the conflict resolution identifier, and whether it has an abnormal interruption protection flag.

[0093] Continuing with the example above, the system encapsulates the adjustment command for A1 as follows:

[0094] {

[0095] "node_id": "A1",

[0096] "target_power": 18000,

[0097] "self_power": 8000,

[0098] "partner_node": "B1",

[0099] "partner_power": 10000,

[0100] "time_start": "10:00",

[0101] "time_end": "10:15",

[0102] "redundancy_ratio": 0.85,

[0103] "conflict_flag": "resolved",

[0104] "fallback_enabled": true

[0105] }

[0106] The platform connects to the edge gateway control module deployed on node A1 via a scheduling bus. Upon receiving the command packet, this module immediately enters the adjustment preprocessing stage to confirm whether the power adjustment capability is sufficient and whether the time window is feasible, and returns an acknowledgment signal to the platform to confirm that the command has been successfully received. Throughout the entire command issuance process, the platform asynchronously polls the response status of all active and lagging nodes to ensure that all critical nodes are ready for scheduling.

[0107] Once the 10:00 scheduling cycle officially begins, each node initiates its adjustment operation. Active node A1 reduces its own load from 150kW to 142kW to achieve an 8kW autonomous adjustment, and simultaneously issues a remote call request, instructing coordinating node B1 to further reduce its own load from 120kW to 110kW, achieving a 10kW transfer adjustment target. The platform's real-time power acquisition module continuously collects the minute-by-minute power change curve, the switching behavior records of the node load control modules, fault status indicators, and equipment feedback codes over a 15-minute period.

[0108] At 10:15, the end of the scheduling cycle, the platform automatically triggers a data collection and aggregation program, encapsulating the adjustment logs from all nodes into a node execution result table. Taking A1 as an example, its execution data record is as follows:

[0109] Node ID: A1;

[0110] Actual adjustment start time: 10:00:05 (5 seconds later than the target time);

[0111] Actual adjustable power: 8200W;

[0112] Execution delay status: There is a slight delay, recorded as "L1 level";

[0113] Task completion rate: 100%;

[0114] Interaction status with B1: Normal;

[0115] Fault indicator: None;

[0116] Environmental disturbance record: Temperature increased by 2.3°C, but did not affect load response.

[0117] The execution data of node B1 is also recorded synchronously:

[0118] Node ID: B1;

[0119] Actual adjustment start time: 10:00:07;

[0120] Actual adjustable power: 9800W;

[0121] Execution delay status: L1 level;

[0122] Matching rate with A1 task: 98%;

[0123] Fault indicator: None;

[0124] Cumulative compensation mechanism triggered: No.

[0125] Due to insufficient resources from B1, the 13kW transfer task in node A2's instruction was not fully executed. The platform automatically marked this situation as "conflict compensation failure," recorded it as a partially incomplete state, and used it to trigger the delayed resource reserve mechanism in subsequent steps.

[0126] The node execution result table is stored in the distributed scheduling database, providing a standard data foundation for subsequent performance updates and revenue settlement in step S105. All data transmission throughout the entire process is conducted under encrypted protocols, using MQTT+TLS or HTTPS to ensure real-time performance and security.

[0127] Step S103: Update the performance record based on the node execution result table and construct a delayed resource incentive reserve pool. The delayed resource incentive reserve pool is used to retain part of the adjustment income of nodes that have failed to respond but have attempted to perform adjustment tasks in the delayed nodes, which is used to offset the penalty for subsequent scheduling task failures.

[0128] First, the system iterates through the adjustment records of each node in the execution results table, comparing the target power issued in the adjustment instruction with the actual adjusted power. If a node fails to reach the target power but still has a clear intention to perform, such as having initiated adjustment behavior in a timely manner within the adjustment start time and responding as best as possible within the load constraint range, then the node is identified as a "failed attempt to fulfill obligations" lagging node. The system classifies different degrees of incomplete fulfillment behavior into several levels by setting multiple fulfillment threshold levels (such as 80%, 60%, and 30%), and allocates a portion of the retainable adjustment benefits according to the level. For example, if a lagging node's target adjustment power in a certain period is 10kW, but the actual adjustment power is 7.5kW, with a fulfillment rate of 75%, then the node is classified into the "Level II Attempted Fulfillment" category, and 60% of its original benefits will be allowed to be temporarily stored in the incentive reserve pool as a deduction amount in future scheduling.

[0129] The system maintains an incentive reserve pool record table. Each record corresponds to information such as the reward retention amount, retention time window, trigger conditions, and deduction limit for a lagging node in a specific scheduling cycle. This retention amount will automatically offset part of the penalty amount if the node fails to complete a scheduling task again in the future. For example, if the node incurs a penalty of 100 yuan for another response failure in the next cycle, and the previously retained incentive amount is 60 yuan, the system will automatically transfer 60 yuan from the reserve pool for deduction, with only 40 yuan actually being deducted.

[0130] Furthermore, to prevent lagging nodes from relying on incentive reserves for extended periods and losing their motivation to fulfill their obligations, the system sets an expiration date for reserve quotas, typically 3 to 5 scheduling cycles. A dynamic parameter model, combined with the node's historical performance fluctuation trends and changes in its adjustment capabilities, is used to periodically assess whether its reserve quota has expired, should be cleared, or its usage weight adjusted. Simultaneously, the system links the incentive reserve mechanism to the performance credit rating system. If a node is in a "failed attempt to fulfill obligations" state for multiple consecutive cycles but is covered by incentive reserves, manual inspection or mandatory downgrading will be triggered to ensure the fairness and responsiveness of the overall virtual power plant operation.

[0131] Through the above mechanism, the incentive reserve pool achieves the rational utilization and incentive guidance of resources for those with a positive willingness to respond in lagging nodes without disrupting the balance of revenue distribution. This enables the virtual power plant system to better coordinate the differences in performance behavior among different types of nodes during the scheduling process, thereby improving the overall response success rate and scheduling stability.

[0132] For example, in a certain actual scheduling cycle, assume that active nodes A1 and A2 have completed their target adjustment tasks, while lagging nodes B1 and B2 have experienced varying degrees of response failure. Specifically, A1 and A2 had target adjustment powers of 8kW and 12kW respectively, and their actual completed powers were 8kW and 12kW respectively; the system records them as fully compliant nodes. Lagging node B1 had a target adjustment task of 5kW and an actual adjustment power of 3kW, while B2 had a target adjustment task of 4kW and an actual adjustment power of 0kW. Based on the execution data, the system generates the following node execution result table:

[0133] Node number Target regulating power (kW) Actual regulated power (kW) Response latency (s) Adjust duration (min) Performance status A1 8.0 8.0 0 15 Full performance A2 12.0 12.0 5 15 Full performance B1 5.0 3.0 12 15 Partial failure B2 4.0 0.0 - 0 Complete failure

[0134] Based on the execution result table, the system constructs an incentive reserve pool for lagging nodes. First, it determines whether there are any nodes that "attempted to fulfill their obligations but did not fully achieve their goals." Although B1 did not complete all tasks, its actual adjustment power reached 60% of the target power, and its response delay was within the specified threshold of 30 seconds. Therefore, it meets the condition of "partial response failure but attempted to fulfill the adjustment task." The system retains a portion of its adjustment revenue in the incentive reserve pool according to the set incentive retention ratio (e.g., 60%). B2 showed no response behavior and is considered a complete failure, therefore it is not included in the incentive pool.

[0135] Assuming the basic revenue per kilowatt-hour of regulation task is 0.5 yuan / kWh, and the regulation time is 15 minutes (i.e., 0.25 hours), then the theoretical regulation revenue of B1 is:

[0136] 3.0 kW × 0.25 h × 0.5 yuan / kWh = 0.375 yuan.

[0137] Based on the system's set partial failure incentive retention ratio of 60%, the system allocates 60% of this as retainable adjustment income to the incentive reserve pool, calculated as follows:

[0138] 0.375 yuan × 60% = 0.225 yuan (recorded into the incentive reserve pool)

[0139] The remaining 40% will not be settled and will be treated as a penalty for scheduling failure.

[0140] The system records the incentive amount in the incentive reserve pool structure, as shown in the following table:

[0141] Lagging node Current period response ratio Adjusting the value of income due Incentive retention ratio Reserve pool recorded amount B1 60% 0.375 yuan 60% 0.225 yuan

[0142] This reserve amount will be used to offset the failure penalty amount that B1 should bear if it fails to respond in the next cycle. For example, if B1 fails to schedule again in the next cycle and the penalty amount is 0.3 yuan, the system will automatically deduct 0.225 yuan from the reserve pool, and the remaining 0.075 yuan will be borne by the node and reflected in the performance penalty table of the next cycle.

[0143] Furthermore, the system synchronizes the execution result to the node's fulfillment history, updating node B1's fulfillment success rate. For example, if B1 succeeded 6 out of the first 9 times, then this partial failure is counted as 0.5 successes, and the updated total number of successes is 6.5, resulting in a fulfillment success rate of:

[0144] (6.5 / 10) × 100% = 65%

[0145] If the system sets the performance layering threshold to 70%, then B1 will still be classified as a lagging node and will continue to participate in redundant collaborative scheduling in the next cycle, with a lower priority.

[0146] Through the above calculation and processing flow, step S105 not only realizes the recording and classification of the scheduling execution status, but also constructs a dynamic and deductible incentive reserve mechanism, which not only reflects the recognition of the efforts of lagging nodes to fulfill their obligations, but also maintains the stability of system scheduling and incentive fairness through cross-cycle correlation, thus ensuring the controllability and economy of the virtual power plant platform operation.

[0147] Furthermore, the step of updating the performance record based on the node execution result table and constructing a delayed resource incentive reserve pool includes:

[0148] Based on the adjustment start and end time, actual adjustment power and execution delay status of each lagging node in the node execution result table, the response performance index of the lagging node in the current scheduling cycle is calculated. The response performance index is obtained by weighting the degree of overlap between the successful response time and the delay tolerance interval and the ratio of the actual execution power to the target adjustment power.

[0149] The response performance index is jointly analyzed with the response trend change rate of the lagging node in the last three scheduling cycles. If the trend is improving, the proportion of the lagging node used to retain part of the adjustment benefits is increased; otherwise, it is decreased. The retention ratio is dynamically adjusted within a preset threshold range, and the adjusted incentive retention weight is obtained.

[0150] Apply the incentive retention weight to the actual adjustment power value in the node execution result table to obtain the amount of adjustment revenue that can be retained in the current period, and register it in the incentive reserve entry according to the node's unique identifier;

[0151] Based on the incentive reserve entries, a delayed resource incentive reserve pool is constructed. The delayed resource incentive reserve pool uses node identifiers as key indexes to record its deductible amount, frozen status, and trigger conditions, which are used to match the penalty deduction path of the adjustment failure node in subsequent scheduling cycles.

[0152] In this invention, the incentive strategy for lagging nodes not only focuses on the execution results of the current scheduling cycle but also introduces trend judgment and dynamic adjustment mechanisms to enhance the participation enthusiasm of lagging nodes and suppress malicious response behavior. To implement this strategy, the adjustment behavior data recorded in the node execution result table must be fully utilized. The adjustment start and end times, actual adjustment power, and execution delay status of each lagging node are extracted as basic inputs. Based on this data, the system evaluates the node's response performance within the current scheduling cycle. This performance indicator not only focuses on whether the adjustment task is completed but also introduces a more refined trade-off mechanism. For example, the degree of overlap between the duration of a successful response and the system's allowed delay tolerance range is used as the first evaluation factor, while the ratio between the actual execution power and the target adjustment power specified in the joint instruction set is used as the second factor. These two factors are combined using a linear or non-linear weighted approach to form a comprehensive performance indicator that reflects the quality of the adjustment response.

[0153] Subsequently, the system introduces a time dimension, combining the response performance indicators of the lagging node in the previous three consecutive scheduling cycles to construct a response trend change rate indicator. This indicator reflects whether the node's response behavior is evolving in a stable and positive direction. If the trend shows improvement, such as performance values ​​increasing periodically or fluctuations converging to a high value range, the system automatically determines that the node's behavior reliability has increased and dynamically adjusts its incentive retention ratio accordingly. Conversely, if the trend is downward or shows sharp fluctuations, it indicates uncertainty or game risk in its adjustment behavior, and the system will appropriately lower its incentive retention ratio to prevent the incentive mechanism from being maliciously exploited. The aforementioned incentive retention ratio is set to fluctuate within a predefined maximum and minimum ratio range, with a typical range of 10% to 80%, ensuring adjustable control and flexibility.

[0154] Once the adjusted incentive retention weight is determined, the system directly applies this weight to the actual adjustment power value completed by the lagging nodes in the current period, thereby calculating the amount of adjustment revenue that the node can retain in this period. This amount is then combined with the price weight corresponding to the unit adjustment power (which can be a dynamically set adjustment unit price by the system) to finally obtain an economic quantification value. This value is indexed using the node's unique identifier as the key and written into the incentive reserve entry. In addition to the amount of revenue that can be retained, the entry also includes fields such as whether the record is frozen (i.e., frozen status), the trigger conditions for unfreezing or transferring, and summary information of the adjustment behavior, forming complete incentive account information.

[0155] Based on all written incentive reserve entries, the system further constructs a delayed resource incentive reserve pool. This pool is organized in dictionary form, with keys serving as unique node identifiers and values ​​representing the corresponding incentive record content. Subsequently, during the scheduling cycle, if a node fails to respond again and meets certain conditions (such as non-subjective evasion of response, the existence of explainable external disturbances, etc.), the system can prioritize using the incentive amount in its reserve pool for deduction, reducing economic penalty pressure and encouraging continued participation in the future. The frozen status and triggering conditions recorded in the reserve pool support linkage with the fault diagnosis system, further improving the fairness and transparency of the strategy. After the entire process is completed, the incentive reserve pool can be used by the system for querying, scheduling, or liquidation at any time, realizing the flexible management and performance-oriented incentive mechanism for delayed resources in this invention. The introduction of this mechanism breaks the traditional linear logic of "reward for response, penalty for failure," introducing performance trends and game-theoretic suppression methods, significantly improving the elasticity and economy of the virtual power plant scheduling system.

[0156] Here's an example. Suppose that in a certain scheduling period T, there is a lagging node A in the system, whose target regulation power assigned in the joint instruction set is 60kW, and the instruction regulation period is from 14:00 to 14:30, a total of 30 minutes. During the actual scheduling execution, node A's monitored regulation behavior is as follows: regulation begins at 14:05 and ends at 14:28, with an actual continuous regulation time of 23 minutes and an average actual regulation power of 54kW. The system's set delay tolerance interval is ±5 minutes, meaning that regulation behavior is allowed to begin between 13:55 and 14:05, which is considered "not significantly delayed".

[0157] The first step involves the system calculating the response performance index of node A within period T based on the aforementioned scheduling instructions and actual execution data. First, the timeliness of the node is evaluated. Since the node began execution at 14:05, which falls within the upper limit of the delay tolerance interval, it is considered a boundary-effective response with a response time of 23 minutes. Given the expected adjustment time of 30 minutes, the time percentage is 23 / 30 ≈ 0.7667. Second, the actual power completion rate is evaluated. The actual adjusted power of the node is 54kW, compared to the target of 60kW, resulting in a power completion rate of 54 / 60 = 0.9. The system uses weighting coefficients of 0.6 and 0.4, assigning them to the time response and power response dimensions respectively, yielding the performance index P as follows:

[0158] P = 0.6 × 0.7667 + 0.4 × 0.9 = 0.4600 + 0.3600 = 0.8200

[0159] The second step involves the system retrieving the performance indicators of node A in the first two scheduling cycles (T−1 and T−2), which are 0.61 and 0.72 respectively. The performance trend change rate ΔP is calculated as follows:

[0160] ΔP = (P − P (t-1) ) + (P (t₋1) - P (t₋2) )

[0161] = (0.82 − 0.61) + (0.61 − 0.72)

[0162] = 0.21 − 0.11 = 0.10

[0163] Since ΔP is positive and its value is greater than the set trend improvement threshold of 0.05, the system determines that the node's response trend is in an improving state. The incentive system sets a base retention ratio of 30%, and when the trend improves, each 0.05 improvement increment increases the retention ratio by 5%. Therefore, the final incentive retention ratio R for node A is:

[0164] R = 30% + (0.10 / 0.05) × 5% = 30% + 10% = 40%

[0165] Thirdly, the system calculates the amount of regulation revenue that node A can retain in this period T based on the incentive retention ratio R. The actual regulation power of node A is 54kW, the regulation duration is 23 minutes (i.e., 23 / 60 hours), and the system's uniform regulation compensation unit price is set at 1.2 yuan per kilowatt-hour. Therefore, the total regulation amount Q of node A is:

[0166] Q = 54 × (23 / 60) ≈ 20.7kWh Total adjustment revenue S = Q × 1.2 ≈ 24.84 yuan

[0167] The incentive retention limit S' is:

[0168] S' = S × R = 24.84 × 40% ≈ 9.94 yuan

[0169] The fourth step is to construct an incentive reserve entry in the system, registering the aforementioned incentive amount S' into the incentive reserve pool. This entry is recorded as follows:

[0170] Node ID: A

[0171] Period identifier: T

[0172] Retainable earnings: 9.94 yuan

[0173] Incentive retention rate: 40%

[0174] Freeze status: Not frozen

[0175] Unfreeze trigger condition: When the response in period T+1 fails, and the scheduling anomaly is not due to a subjective fault.

[0176] Update timestamp: End time of period T

[0177] Fifth, the system writes the entry into the incentive reserve pool data structure. This structure uses the node ID as the primary key and the period as the secondary key to construct a two-level index dictionary. The system then completes the update of the lagging resource incentive reserve pool. Subsequently, if node A fails to respond in period T+1, the system will trigger a deduction judgment and call the reserve amount S' in period T to cover part of the economic penalty, thereby achieving the goal of dynamic flexible penalty exemption.

[0178] Step S104: Generate the real-time settlement revenue for each node based on the incentive reserve pool table and the power adjustment instructions in the joint instruction set; and adjust the priority of redundancy contest in the next cycle according to the revenue level to obtain the performance penalty table and node revenue allocation table for the next cycle.

[0179] After completing the construction of the delayed resource incentive reserve pool, the system will enter the final stage of the scheduling cycle, which is to calculate the real-time settlement revenue of each node in the current cycle based on the incentive reserve pool table and the power adjustment instructions in the joint instruction set, and update the redundancy right contention priority for the next cycle accordingly, and finally generate the performance penalty table and node revenue allocation table for subsequent scheduling.

[0180] First, the system reads the execution data table of each node within the scheduling cycle, including parameters such as its actual adjustment power, execution start and end times, and execution delay status, and compares them one by one with the originally assigned target power, adjustment window requirements, and task allocation weights. For example, if a node is assigned a target adjustment power of 8kW, and its actual execution is 8.1kW, and the adjustment start and end times fall entirely within the required adjustment window, the system will determine that the node has fully performed its obligations and will award it full settlement revenue. The amount can be calculated by multiplying the power adjustment standard price by the actual execution time, multiplying by the node's contribution weight, and the market adjustment factor.

[0181] For nodes with partial deviations or slight delays, the system will deduct their rewards accordingly, based on the rules for classifying delay levels. For example, if a node only achieves 90% of the target power and has a response delay of less than one minute, the system can make immediate adjustments according to a pre-agreed percentage (e.g., deducting 10% of the rewards). If the delay exceeds the adjustment window or the power completion rate is less than 80%, the system will significantly reduce or even completely deduct the node's rewards, depending on the scheduling mechanism settings. If the deducted portion meets the conditions for offsetting in the incentive reserve pool, the corresponding reserve balance will be automatically used to offset a portion of the penalty, reflecting the role of the fault-tolerant incentive mechanism.

[0182] After calculating the revenue of each node, the system further dynamically adjusts its "redundancy priority" for the next cycle based on the node's actual revenue during the scheduling cycle and its share of redundant tasks in the joint instruction set. Priority is essentially a scheduling resource bidding weight; high-revenue nodes indicate strong adjustment capabilities and stable responses in the previous cycle, increasing the probability of them being prioritized for redundant adjustment tasks in the next cycle. Conversely, low-revenue or poorly performing nodes will be prioritized for passive support roles in the next cycle, or will relinquish priority in case of conflicts. This mechanism achieves a dynamic balance between scheduling efficiency and fairness by constructing a weighted ranking list and assigning adjustment coefficients.

[0183] Finally, based on the settlement and priority adjustment results, the system generates two key documents: one is the "Next Cycle Performance Penalty Table," which records the penalty amount and compensation method that all nodes must bear when their scheduling performance is poor, as part of the scheduling incentive for the next cycle; the other is the "Node Revenue Allocation Table," which clearly lists the total amount of immediate revenue obtained by each node, the deduction status of the reserve pool, and the final amount received, and is used to synchronously update the virtual power plant revenue account and the node settlement account.

[0184] Through the above steps, the system achieves closed-loop revenue management for multi-source loads and energy storage nodes. This not only reflects a refined scheduling strategy that combines incentives and penalties, but also forms a positive feedback mechanism at the economic level, guiding various nodes to continuously optimize their self-regulation behavior, ultimately improving the overall response quality and market competitiveness of the virtual power plant.

[0185] Taking a joint instruction set issued by the system within a scheduling cycle as an example, assuming active nodes A1 and A2 are assigned to regulate target power of 10kW and 12kW respectively, and lagging node B1 is assigned a transferable regulation task share of 6kW, with the system setting a redundancy sharing ratio of 0.3, meaning that up to 30% (i.e., 1.8kW) of the 6kW task undertaken by B1 can be redundantly shared. After the scheduling cycle ends, the system collects the execution data of all nodes and obtains the following node execution result table:

[0186] The actual regulating power of A1 is 10.2kW, the regulating duration is 15 minutes, and the response delay is 0 seconds.

[0187] The actual regulating power of A2 is 11.5kW, the regulating duration is 15 minutes, and the response delay is 10 seconds.

[0188] The actual regulating power of B1 is 4.0kW, the regulating duration is 15 minutes, and the response delay is 20 seconds.

[0189] First, revenue calculation is performed. The system sets the market adjustment reward to 0.6 yuan / kWh, and the node contribution weight is uniformly set to 1.0. Therefore, A1's settlement revenue is:

[0190] 10.2kW × 0.25h × 0.6 yuan / kWh = 1.53 yuan (full performance, no delay, full payment).

[0191] The A2's regulating power is 11.5kW, lower than the target value of 12kW, but within an acceptable range (completion rate 95.83%). The 10-second delay is also within the tolerable threshold (<30 seconds). Therefore, the 5% revenue deduction rule applies:

[0192] 11.5kW × 0.25h × 0.6 yuan / kWh × 0.95 = 1.64 yuan

[0193] While B1's target was 6kW, only 4.0kW was achieved, representing 66.67% of the target. The 20-second response delay is still within acceptable limits, but the completion rate is insufficient. Assuming a completion rate below 80% will be considered a partial response failure, only 50% of the base revenue will be settled. The system calculation is as follows:

[0194] 4.0kW × 0.25h × 0.6 yuan / kWh × 0.5 = 0.30 yuan

[0195] However, B1 is a lagging node, and the execution log shows that its response did indeed occur. The system allows a portion of its adjusted revenue to enter the incentive reserve pool. Assuming the current period's lagging incentive ratio is set at 60%, B1 will only settle 40% of the 0.30 yuan, i.e., 0.12 yuan, while the remaining 0.18 yuan will be transferred to the incentive reserve pool for penalty deduction in subsequent periods.

[0196] Subsequently, the system adjusts the priority of redundancy claiming based on the node's real-time revenue. A1, which has fully fulfilled its obligations and has higher revenue, has its priority increased by one level; A2's priority is slightly reduced, but remains unchanged; B1, due to partial response failure, has its priority decreased by one level and will be prioritized as a controlled redundant node rather than a primary scheduling node in the next cycle.

[0197] Ultimately, the system generates two tables: the "Next Cycle Performance Penalty Table" records B1's failures and pre-sets a deduction mechanism for its next unfinished scheduling task. If a future response fails, a maximum of 0.18 yuan can be deducted from the aforementioned incentive reserve pool. The "Node Revenue Distribution Table" lists:

[0198] Settlement for A1: 1.53 yuan, fully credited to account;

[0199] A2 settlement: 1.64 yuan, 95% received;

[0200] B1 settlement: 0.12 yuan received, 0.18 yuan credited to the incentive reserve pool.

[0201] Through the aforementioned calculation process and rule setting, the system ensures that the performance of each node is strongly correlated with economic incentives and can continuously guide its behavior optimization across cycles, achieving a closed-loop control mechanism for dynamic scheduling of multi-source demand-side resources and fair distribution of benefits. This method can adapt to actual virtual power plant environments with large-scale heterogeneous node access, ensuring compatibility between scheduling efficiency, performance stability, and economic incentives.

[0202] Furthermore, the step of adjusting the priority of redundancy contest in the next cycle based on the level of revenue, and obtaining the performance penalty table and node revenue distribution table for the next cycle, includes:

[0203] Based on the actual adjustment power and incentive reserve amount registered by each node in the node execution result table in the previous scheduling cycle, calculate the total revenue value actually obtained by each node in the previous cycle, and use it as the revenue evaluation benchmark to construct the current revenue level classification result table.

[0204] Based on the revenue level classification result table, for edge nodes whose total revenue value is lower than the preset participation threshold for two or more consecutive periods, a one-time redundancy priority inversion mechanism is set up, that is, the scheduling priority of their redundancy adjustment task is increased in the next period to obtain the initial sequence of redundancy competition.

[0205] Based on the initial sequence of redundancy contention, a weighted revenue smoothing function is constructed by combining the actual revenue value of each node in the last five scheduling cycles. The weighted revenue smoothing function is constructed by using time-decreasing weights arranged in reverse order of the cycle to generate a node stability revenue index sequence, which serves as the basis for sorting the revenue allocation factors in the next cycle.

[0206] The node stability benefit index sequence and the initial sequence of redundancy right contention are jointly analyzed node by node to construct the performance penalty table and node benefit allocation table for the next period. The performance penalty table is used to record the deviation between the target response value and the actual response value of each node and the corresponding penalty amount. The node benefit allocation table is used to record the benefit base, allocation weight, incentive reserve deduction path and pre-allocated amount, thus obtaining the performance penalty table and node benefit allocation table for the next period.

[0207] In the virtual power plant optimization scheduling and revenue distribution method for multi-source demand-side resources described in this invention, a dynamic adjustment mechanism for scheduling fairness and revenue incentives is constructed through a series of specific steps to improve the long-term stability of node participation and the overall resource allocation efficiency of the system.

[0208] First, the system performs preliminary revenue statistics based on the execution results data of each node in the previous scheduling cycle. Specifically, it extracts the actual adjustment power value and the corresponding incentive reserve amount for each node from the node execution result table. These two values ​​together constitute a comprehensive reflection of the node's adjustment performance and the degree of incentive. Based on this, the system performs a weighted summation to obtain the actual total revenue value of each node in the current cycle. This revenue value is then normalized and used as the revenue benchmark for the next evaluation step. The system classifies nodes into revenue levels based on this benchmark and forms a revenue level classification result table for the current cycle, distinguishing between high-revenue nodes, intermediate-revenue nodes, and marginal nodes.

[0209] Next, the system takes special measures for edge nodes whose total revenue value is below the system's preset participation threshold for two or more consecutive cycles. To prevent these nodes from entering a persistent marginalization state due to insufficient resource allocation, the system sets up a one-time redundancy priority inversion mechanism for them in the next scheduling cycle. That is, when constructing the priority sequence for redundancy adjustment tasks, the scheduling priority of these nodes is temporarily increased, allowing them to be matched with resources from lagging nodes first when competing for task allocation. This mechanism can improve their success rate in adjusting tasks and indirectly boost their revenue level. After this process, the system obtains the initial sequence for redundancy contention, which is used as the priority basis for subsequent sorting and allocation.

[0210] To further enhance the stability and fairness of the allocation, the system incorporates historical revenue data for smoothing. Specifically, the system calculates a weighted average of the actual revenue values ​​of each node over the most recent five scheduling periods, with weights decreasing over time (higher weights for more recent periods) to reflect the representativeness of the node's current state. This weighting process constructs a revenue smoothing function, ultimately outputting a sequence of stability revenue indicators for each node. This sequence can be considered a comprehensive score of the continuity and stability of a node's behavior in revenue allocation.

[0211] Finally, the system performs node-by-node joint analysis of the node stability benefit index sequence and the initial sequence of redundancy contention. This involves comprehensively considering historical benefit trends and marginal adjustment results during sorting to determine the scheduling priority and economic benefit quota for each node in the next cycle. Based on this result, the system generates a performance penalty table and a node benefit allocation table. The performance penalty table records the deviation between the target response value and the actual response value for each node and allocates corresponding performance penalty amounts accordingly. The node benefit allocation table, based on the node's benefit base, allocation weight, and the incentive reserve deduction path formed in the previous cycle, provides its pre-allocated benefit amount for the next cycle.

[0212] Through the refined modeling and multi-source data-driven mechanism of the above integrated process, the system can achieve dynamic support for edge nodes and continuous incentives for highly stable nodes without compromising overall scheduling efficiency and fairness of benefits. Ultimately, it can construct and distribute the performance penalty table and node benefit distribution table for the next cycle, thereby supporting the sound operation of the entire virtual power plant scheduling closed loop.

[0213] Taking 12 multi-source demand-side nodes dispatched by a virtual power plant in a certain region during dispatch cycle T as an example, the node types include industrial load nodes, building load nodes, and distributed photovoltaic energy storage nodes. The system first extracts the actual regulated power value (denoted as ) registered by each node in the node execution result table. ), in kW; combined with the incentive retention weight registered by this node in the incentive reserve pool (denoted as The range of values ​​is ), calculate the total profit value of each node in period T. The calculation formula is as follows:

[0214]

[0215] in, The basic yield coefficient is set at 1 yuan / kWh. The incentive benefit factor is set at 0.5 yuan / kWh. In the above formula, the first term represents the basic adjustment benefit obtained by the node through actual adjustment behavior, and the second term represents the additional benefit obtained with the support of the incentive reserve pool.

[0216] Subsequently, the system based on all nodes Construct a table of profit level classification results for the current period, and filter out two consecutive periods. and Total returns were all below the participation threshold. The nodes of a given element are denoted as the set of edge nodes. For each belonging to For nodes, the system sets up a one-time redundancy priority inversion mechanism, that is, in the next cycle T+1, the initial scheduling order of the nodes competing for redundancy tasks is manually adjusted.

[0217] At the same time, the system constructs a weighted profit smoothing function for all nodes. To improve the fairness of distribution and the sustainability of response incentives, its expression is as follows:

[0218]

[0219] in, Indicates the node in the past... Total return over a period of time Let the time-reverse weighting coefficient be set as follows: ,satisfy This smoothing function prioritizes near-term returns and secondarily considers long-term returns, effectively avoiding the impact of short-term abnormal fluctuations on return ranking. For the first The stability and benefit metrics for each node.

[0220] The system then calculates the stability return index using the aforementioned return smoothing function. The priority order in the initial sequence is sorted in parallel with the redundancy weight competition, and a dual-weight fusion mechanism is used to generate a comprehensive priority sequence for redundancy competition.

[0221]

[0222] in, The weighting is based on the recent earnings priority. This indicates the ranking obtained by sorting the values ​​in reverse order from smallest to largest. This method ensures a comprehensive consideration of the node's real-time adjustment capabilities and long-term contributions. This is a comprehensive scheduling priority indicator.

[0223] Represents a node The ranking position of the redundancy adjustment right contention order preset for the next scheduling cycle in the system. Assume there are 5 nodes in total. Nodes D and E are edge nodes. The system generates the initial sequence for redundancy contention according to the following rules:

[0224] Prioritize and increase the priority of D and E;

[0225] The remaining nodes are sorted from highest to lowest based on their returns in the previous period: .

[0226] The initial priority sequence is obtained as follows:

[0227]

[0228] but: These rankings will be used as the initial order for redundancy contests in the calculation, for example, as the first part of the fusion weights.

[0229] Finally, the system according to Generate the first The scheduling cycle includes a performance penalty table and a node reward allocation table. The performance penalty table records the target response value of each node. Actual deviation value The corresponding penalty amount ,in The penalty factor is set to 2 yuan / kW by default. The node revenue allocation table includes: basic revenue amount, scheduling priority weight, incentive reserve pool deduction path, and final allocable amount.

[0230] Through the above process, it can be ensured that the system can reasonably adjust the order of redundancy contention and the proportion of revenue distribution based on multi-period dynamic revenue data and scheduling behavior, taking into account efficiency, fairness and a continuous incentive mechanism.

[0231] Furthermore, the step of generating the real-time settlement revenue for each node based on the incentive reserve pool table and the power adjustment instructions in the joint instruction set includes:

[0232] Based on the target adjustment power of each node in the joint instruction set and the actual adjustment power recorded in the node execution result table, the actual response deviation of each node is calculated, and a response deviation result table is generated.

[0233] Based on the actual response deviation and performance record of each node recorded in the response deviation result table, update the adjustment failure penalty deduction item in the incentive reserve pool table, and obtain the incentive reserve amount that each node can retain.

[0234] Based on the updated incentive reserve amount in the incentive reserve pool table, and combined with the power adjustment instructions and resource types of each node in the joint instruction set, an adjustment response intensity evaluation factor for the current period is generated.

[0235] Based on the adjustment response intensity evaluation factor and the incentive revenue parameter corresponding to each node in the joint instruction set, the periodic revenue value that each node should obtain is calculated.

[0236] The periodic expected income value of each node is integrated with the updated incentive reserve amount in the incentive reserve pool table to generate the instant settlement income of each node.

[0237] In the virtual power plant optimization scheduling and revenue distribution method described in this invention, in order to ensure that various multi-source demand-side nodes can obtain reasonable immediate revenue distribution based on their performance after scheduling is completed, it is necessary to systematically calculate and process the node's execution performance, resource type, incentive reserve status, and preset revenue rules.

[0238] After the scheduling cycle is completed, the system first needs to compare and analyze the target adjustment power and the actual adjustment result for each node. The target adjustment power is the specific power adjustment task issued to each node by the system at the beginning of the cycle through a joint instruction set, which records the positive or negative power adjustment value that each node should complete. The node execution result table contains the power value actually achieved by each node in the cycle. The system calculates the difference between the target adjustment power and the actual adjustment power for each node to obtain the actual response deviation of that node. If the deviation value is zero, it indicates full compliance; if the deviation value is positive or negative, it means that the actual response capability is lower or higher than the expected requirement. Based on these calculation results, the system constructs a response deviation result table to summarize the adjustment execution difference data of all nodes.

[0239] After obtaining the response deviation result table, the system needs to further determine whether the adjustment failure of each node triggers the penalty deduction mechanism. To this end, the system retrieves the historical performance records of each node, determines whether its performance over the past several periods has shown continuous failures or frequent deviations, and determines whether penalty deduction should be applied based on the set tolerance threshold. If a node fails to meet the performance requirements but demonstrates a clear adjustment attempt during the period, even if it fails, a portion of the incentive amount can be retained according to the incentive mechanism of this invention. The upper limit, retention ratio, and deduction factor of this incentive amount can be pre-configured by the management and registered in the incentive reserve pool table. When updating the incentive reserve pool table, the system will deduct the incentive amount to be penalized in the current period, while retaining a portion of the incentive amount for eligible nodes, which can be used to offset losses in subsequent settlements or converted into a certain proportion of revenue compensation.

[0240] After the above steps are completed, the system begins to comprehensively analyze the response performance and resource characteristics of each node in the current cycle to construct a regulation response intensity assessment factor. The core of the assessment factor is to reflect the initiative and capability demonstrated by a node in its actual response. Its construction depends not only on the node's response deviation and performance trend but also on the resource type and regulation role assigned to it in the joint scheduling instruction set. Different types of nodes, such as industrial loads, building loads, or distributed photovoltaic energy storage, have different regulation capabilities and risk exposures, and their contributions to the system also differ. Therefore, the system incorporates resource type into the factor construction process, using weighted logic to assign higher intensity scores to nodes with high response accuracy, large response amplitude, or important roles. The final generated regulation response intensity assessment factor will be used as the weighted basis for subsequent revenue calculations.

[0241] Based on the aforementioned regulation response intensity assessment factors, the system, in conjunction with pre-set incentive revenue parameters in the joint instruction set, calculates the periodic expected revenue value for each node. Incentive revenue parameters may include the base rate of return, response level bonus coefficient, and regulation depth coefficient, which together determine the value assessment benchmark for the node's response. For example, if a node completes a high-depth, high-power, low-deviation regulation task, and its resource type is flexible energy storage, its corresponding periodic expected revenue value will typically be significantly higher than that of a typical industrial load node. The calculation results will be reflected as the expected revenue amount per unit yuan per cycle.

[0242] Finally, the system integrates the periodic expected revenue value of each node calculated above with its updated incentive reserve amount recorded in the incentive reserve pool table to generate the final instant settlement revenue. The integration logic considers whether the incentive reserve amount can be used to offset revenue losses caused by response deviations, whether it can be converted into positive incentive compensation, and whether penalties for incomplete tasks need to be deducted proportionally. All these factors will be processed item by item in the system according to the set rules to ensure that the instant settlement revenue obtained by each node is based on its actual adjustment contribution and historical incentive situation, taking into account fairness, incentives, and operability.

[0243] Through the above steps, the system not only realizes closed-loop settlement after scheduling behavior, but also closely integrates the compensation mechanism for lagging nodes, resource differentiation incentive strategy and dynamic performance evaluation method, providing a solid and reliable basis for revenue distribution for the distributed optimization scheduling of virtual power plants.

[0244] Taking a specific scheduling cycle as an example, assume the virtual power plant scheduling platform issues a joint instruction set to nodes A, B, and C. Node A's target adjustment power is +20kW (i.e., increased load output), node B's is -15kW (reduced load output), and node C's is +10kW. Their resource types are industrial load, building load, and distributed photovoltaic energy storage, respectively. At the end of the cycle, the platform collects the following node execution results: Node A's actual adjustment is +18kW, node B's is -5kW, and node C's is +10kW.

[0245] The first step is to compare the target regulation power with the actual regulation power at each node and calculate the response deviation. The response deviation for node A is 2kW (20 − 18), for node B it is 10kW (|−15 − (−5)|), and for node C it is 0kW (10 − 10). Based on these differences, a response deviation result table is constructed, indicating the direction and absolute value of the deviation, which serves as the basis for subsequent benefit assessment and penalty judgment.

[0246] The second step involves the system determining, based on the aforementioned deviation and the performance records of previous cycles, whether a node has triggered a penalty deduction for adjustment failure. Taking node B as an example, its current deviation value is relatively large (10kW). The platform further confirms that it also had records of insufficient response in the past two cycles, determining it as a continuous failure and triggering the penalty mechanism. However, according to system rules, if a node did attempt an adjustment task in this cycle (e.g., performing a partial adjustment of −5kW), its incentive quota can be partially retained. Therefore, the system allocates 50 yuan of the original 100 yuan incentive quota configured for node B to the "retainable incentive reserve," and the remaining 50 yuan is deducted as a penalty, updating the relevant fields in the incentive reserve pool table.

[0247] Third, based on the updated incentive reserve pool table, the platform begins constructing a regulatory response strength evaluation factor. This factor aims to comprehensively evaluate the regulatory contribution strength of a node and provide a weighting basis for revenue calculation. Specifically, the evaluation factor comprehensively considers the following three types of data:

[0248] Adjusting response accuracy: This is the inverse ratio of response deviation. The deviation at node A is 2kW, accounting for 10% of the target value, at node B it is 66.7%, and at node C it is 0%. The system sets the base scores based on this indicator: A: 90 points, B: 33 points, and C: 100 points.

[0249] Resource type coefficient: The system assigns differentiated bonuses to different resource types. For example, distributed photovoltaic energy storage has higher response flexibility and adjustment costs, so it is assigned a weight coefficient of 1.2, building load is 1.0, and industrial load is 0.8. Thus, the weighted score of node A is 90 × 0.8 = 72, node B is 33 × 1.0 = 33, and node C is 100 × 1.2 = 120.

[0250] Performance Trend Correction Factor: If a node's performance has been stable over the past three periods, the system will award a 5% bonus; if there are fluctuations or consecutive failures, the score will be reduced proportionally. For example, if nodes A and C both have stable performance, the score will increase by 5%; if node B has experienced consecutive failures, the score will be reduced by 10%.

[0251] Ultimately, the evaluation factors are as follows:

[0252] A: 72 × 1.05 = 75.6;

[0253] B: 33 × 0.9 = 29.7;

[0254] C: 120 × 1.05 = 126.

[0255] Fourth, the system calculates the periodic expected revenue based on the evaluation factor and the preset incentive revenue parameters for each node in the joint instruction set. For example, if the system sets the incentive revenue base to 1 yuan per unit of response points, node A should receive 75.6 yuan, node B should receive 29.7 yuan, and node C should receive 126 yuan.

[0256] Fifth, the system integrates the aforementioned periodic expected earnings with the reserve amount recorded in the incentive reserve pool. Node B's original reserve was 100 yuan, and it is allowed to retain 50 yuan this period to offset losses in this period. Therefore, its actual expected earnings are 29.7 yuan (current earnings) + 50 yuan (retainable incentive) = 79.7 yuan. Nodes A and C do not need to use their reserves; their expected earnings are directly recognized.

[0257] Ultimately, the platform generates an instant settlement revenue table: Node A receives 75.6 yuan, Node B receives 79.7 yuan, and Node C receives 126 yuan. This table will serve as input for end-of-cycle settlement and subsequent revenue distribution, supporting the platform in achieving closed-loop management of scheduling response and revenue incentives.

[0258] Furthermore, the step of calculating the actual response deviation of each node based on the target adjustment power of each node in the joint instruction set and the actual adjustment power recorded in the node execution result table, and generating a response deviation result table, includes:

[0259] Based on the target adjustment power and adjustment start and end time of each node recorded in the joint instruction set, a target power time series is constructed to characterize the theoretical power trajectory of the adjustment task in the current scheduling cycle.

[0260] Based on the actual adjustment power recorded in the node execution result table, an actual power response sequence with the same sampling frequency is generated to ensure that it completely corresponds to the target power time sequence in the time dimension.

[0261] The target power time series is compared with the actual power response series at each time point to obtain the deviation value at each time point, forming a node adjustment deviation time series, and statistical indicators including peak deviation, average deviation and standard deviation are extracted based on the series.

[0262] The statistical indicators of the node adjustment deviation time series are combined with the associated information, including node identifier, adjustment start and end time, target adjustment power and actual adjustment power, into structured entries to generate a response deviation result table.

[0263] First, the system acquires the joint instruction set issued within the current scheduling period. Each node's regulation task explicitly includes a target regulation power value and specific start and end times. Based on this, the system constructs a power time series for the theoretical regulation task of each node. This target power time series is divided into time periods according to a preset uniform sampling frequency, starting from the start time and ending at the end time of the regulation task. The corresponding target regulation power value is then filled in at each time point, forming a continuous and ordered power trajectory array. This time series serves as a reference benchmark for the node's theoretical response.

[0264] Subsequently, the system retrieves the actual regulation power information recorded in the node execution result table. To ensure the rigor and accuracy of the comparison, the system first performs sampling frequency verification and time synchronization processing on the actual power data. That is, the system interpolates or resamples the original actual power records according to the time axis of the target power time series to ensure that its time axis corresponds one-to-one with the target power time series. After completing this correction process, an actual power response sequence with the same number of sampling points and time dimension as the target power time series can be generated.

[0265] After obtaining the time-synchronized target power time series and actual power response series, the system compares the numerical differences between the two series point by point. At each time point, the difference between the actual power value and the target power value is calculated, which is the response deviation at that time point. By continuously processing each time point, a complete regulation deviation time series for that node is constructed. This series reflects the degree of deviation of the node from the target response at each moment during the entire regulation task.

[0266] Based on the aforementioned adjustment deviation time series, the system further performs statistical indicator extraction. First, it calculates the maximum deviation value in the series, i.e., the peak value of the response error of that node during the adjustment process; second, it calculates the arithmetic mean of the deviations over the entire period, reflecting the accuracy of the overall response; and third, it calculates the standard deviation of the series to measure the stability and volatility of the response. These statistical indicators collectively constitute important characteristics of the node's response accuracy.

[0267] After the statistical results are extracted, the system integrates these indicators with other relevant information about the nodes in a structured manner. This relevant information includes, but is not limited to, the node's unique identifier, the corresponding target regulation power value, the actual average regulation power value, and the start and end times of the regulation task. The system combines this information into structured data entries, where each entry corresponds to a regulation node, and the fields are complete, the format is consistent, and the entry has the ability to be sorted, filtered, and analyzed in relation to other data.

[0268] Finally, the structured entries of all nodes are summarized into a response deviation result table. This result table serves as the basic input for subsequent adjustment response intensity assessment, incentive benefit calculation, and performance determination, providing the virtual power plant dispatching system with high-precision, traceable, and auditable response performance records, ensuring the quantification, controllability, and data-driven attributes of the entire dispatching process.

[0269] Furthermore, based on the actual response deviation and performance record of each node recorded in the response deviation result table, the adjustment failure penalty deduction item in the incentive reserve pool table is updated, and the incentive reserve amount that each node can retain is obtained, including:

[0270] Extract the average response deviation and peak response deviation of each node recorded in the response deviation result table, and map them one by one with the node identifier to construct a response deviation index dictionary, which is used to unify the input for subsequent penalty ratio calculation;

[0271] The response deviation index dictionary is associated with the adjustment task response success rate, historical trigger penalty count and cumulative performance deviation magnitude of each node in the performance record in the previous several scheduling cycles, and the node adjustment failure penalty factor in the current cycle is calculated based on the preset deviation tolerance threshold.

[0272] The adjustment failure penalty factor and the penalty deduction amount registered by the node in the incentive reserve pool table are multiplied to obtain the incentive reserve amount to be deducted in this period. The updated adjustment failure penalty deduction item is then generated by comparing it with the current deductible amount of the corresponding node in the incentive reserve pool table.

[0273] The incentive reserve amount that can be retained for each node is obtained by deducting the updated adjustment failure penalty deduction item from the initial incentive reserve amount registered in the incentive reserve pool table.

[0274] The system first extracts the average and peak response deviation data for all nodes participating in the scheduling from the response deviation result table, and then combines these deviation values ​​with the corresponding unique node identifiers to form a dictionary structure. This structure allows the system to call the latest response deviation index for each node in subsequent steps without multiple searches, improving data processing efficiency. This response deviation index dictionary is used to standardize the input format for all nodes in the calculation of adjustment failure penalties and provides a standardized basis for evaluating penalty factors.

[0275] Next, the system integrates the constructed response deviation index dictionary with the performance record data. Specifically, the system retrieves the adjustment task completion status of each node over several past scheduling cycles, including information such as whether the adjustment response was successful, whether the delayed response was within the tolerance range, and whether the adjustment task was completely interrupted, and calculates the response success rate of the adjustment task accordingly. Simultaneously, the system also calculates the total number of times the penalty mechanism has been triggered for each node historically, as well as the cumulative response deviation magnitude in each historical cycle. This historical data reflects the overall stability and performance of the nodes.

[0276] The system compares the average and peak response deviations of the current period with a predefined response deviation tolerance threshold to determine whether the current adjustment behavior has exceeded the system's expected acceptable range. If the response deviation exceeds the threshold, the system will increase the adjustment failure penalty factor for that node in the current period based on the degree of exceedance; otherwise, it will decrease it. This penalty factor is a dynamic value obtained by weighting response accuracy, historical performance, and preset rules, used to measure the degree of incentive penalty that the node should bear in the current period.

[0277] The calculated adjustment failure penalty factor will be used to calculate the deduction of the current period's incentive reserve amount. The system retrieves the deductible amount registered for this node from the incentive reserve pool table as the base amount, and multiplies it by the current penalty factor to obtain the incentive amount to be deducted. The system compares the remaining portion of the original incentive reserve amount that can be used for deduction, performs the actual deduction operation, and updates the incentive reserve pool table, overwriting the original deduction item, to ensure that the incentive penalty result for this period takes effect immediately.

[0278] Finally, based on the updated adjustment failure penalty deduction item, the system deducts it from the initial incentive reserve amount registered by the node in the incentive reserve pool table, calculating the remaining incentive reserve amount of the node at the end of the current cycle. This remaining amount is the maximum amount that the node can use for deduction or incentive in the next scheduling cycle. The system records this value in the incentive reserve pool table and marks it as "currently retainable incentive amount", providing a basic parameter for subsequent scheduling revenue calculation.

[0279] Furthermore, the step of generating a regulation response intensity evaluation factor for the current period based on the updated incentive reserve amount in the incentive reserve pool table, combined with the power regulation instructions and resource types of each node in the joint instruction set, includes:

[0280] Extract the available incentive value of each node from the updated incentive reserve amount in the incentive reserve pool table, and pair it with the power adjustment instruction corresponding to that node in the joint instruction set to construct a node incentive adjustment association table;

[0281] Based on the ratio between the available excitation value and the power regulation command amplitude of each node in the node excitation regulation association table, the basic response capability index is calculated, and a type correction factor is set according to the resource type to reflect the performance heterogeneity of different resources in the regulation task; wherein, the resource type includes industrial load nodes, building load nodes and distributed photovoltaic energy storage nodes;

[0282] The basic response capability index is weighted and fused with the type correction factor to generate the node initial response strength score. Based on the node's historical response stability record, a historical adjustment fluctuation compensation term is introduced to dynamically correct the node initial response strength score.

[0283] The standardized score of each node's response intensity is used as the evaluation factor for the adjustment response intensity of the current cycle.

[0284] In this implementation, in order to generate the adjustment response intensity assessment factor for the current period based on the updated incentive reserve amount in the incentive reserve pool table and the power adjustment command and resource type of each node in the joint instruction set, the system needs to rely on the incentive amount update results and scheduling commands completed in the previous period, fully integrate the capabilities and characteristics of various nodes, and construct a representative intensity assessment index through a series of orderly processing steps for subsequent revenue calculation and node behavior incentives.

[0285] The system first reads the available incentive amount for each node in the current period from the incentive reserve pool table. This amount is the result of dynamic deductions based on response deviations and performance records during the preceding processing, reflecting the node's remaining "credit value" or incentive weight at the incentive level. The system then pairs and integrates this available incentive value with the power adjustment instructions issued to that node in the joint instruction set, forming a structured list that includes the node identifier, available incentive value, and target adjustment power. This list, known as the node incentive adjustment association table, serves as the basic input for all subsequent calculations.

[0286] After obtaining the node excitation adjustment correlation table, the system reads the data of each node one by one and calculates its basic response capability index. This index is obtained by dividing the available excitation value by the adjustment command amplitude, logically reflecting the amount of excitation resources corresponding to a unit of adjustment power, or it can be understood as the excitation capability load level undertaken by the node to complete a unit of adjustment task. Based on this, in order to reflect the differences in response capability, control accuracy, execution speed, etc., of different types of resources, the system presets a correction factor for each resource type. Resource types include, but are not limited to, industrial load nodes, building load nodes, and distributed photovoltaic energy storage nodes. These three types of resources have significant differences in performing adjustment tasks. For example, industrial loads respond quickly but fluctuate greatly, building loads respond steadily but with limited amplitude, and energy storage nodes respond rapidly but have limited energy. The system sets correction factors according to these characteristics, explicitly incorporating the physical differences of resources into the calculation process.

[0287] The basic response capability index and the resource type correction factor are not simply added together, but synthesized through a weighted fusion method. The weighting coefficients can be set based on empirical rules, expert settings, or historical scheduling performance to ensure that the fusion result is closer to the actual performance. The fused result is the initial response strength score of the node in the current scheduling cycle, representing its relative ability to assume adjustment responsibility in the current task.

[0288] To further improve the dynamic accuracy of the evaluation results, the system introduces a historical response fluctuation compensation term. This compensation term is derived from the fluctuation range and frequency of the node's response power curve over multiple historical periods. By calculating the trend of fluctuation and superimposing it with the current initial score, the system corrects the error, enhancing its ability to identify nodes that have historically performed stably but have lower excitation in the current period. This prevents changes in excitation amount from misleading the assessment of the adjustment response intensity.

[0289] Finally, the system standardizes the corrected response strength scores of all nodes, ensuring that the adjustment response strength evaluation factors of all nodes are under a unified measurement system. This standardization process can employ methods such as range standardization, Z-score standardization, or normalization to ensure that the factor values ​​have consistent explanatory power when compared horizontally. The standardized factor results will serve as one of the most critical input parameters in subsequent revenue calculations, measuring the contribution weight of a node's ability to complete the task in the current cycle.

[0290] Through the above process, the system can effectively take into account the remaining incentive resources, target task requirements, resource physical attributes and historical behavior characteristics of each node while ensuring fairness, rationality and feasibility, and construct a stable and reliable adjustment response intensity evaluation factor, providing accurate quantitative basis for the entire virtual power plant scheduling and revenue distribution mechanism.

[0291] Furthermore, the calculation of the periodic due revenue value of each node based on the adjustment response intensity evaluation factor and the incentive revenue parameter corresponding to each node in the joint instruction set includes:

[0292] Extract the incentive benefit parameters corresponding to each node in the joint instruction set and match them with the adjustment response intensity evaluation factor of that node to form the adjustment incentive comparison data item for each node.

[0293] The adjustment response intensity evaluation factor of each node is multiplied by its corresponding incentive benefit parameter to obtain the initial benefit score of the node. The initial benefit score reflects the response quality and incentive capability of the node in the current scheduling cycle.

[0294] Based on the resource type of each node, the initial revenue score is adjusted according to the preset resource type weight to obtain the revenue score after resource type adjustment, so as to take into account the differences in contribution of different types of nodes in the adjustment task.

[0295] The revenue score after resource type adjustment is processed by smoothing the response based on the time average of the response stability of each node in the current scheduling cycle to eliminate the impact of occasional response deviations. The processing result is then limited to a preset incentive range to finally determine the periodic revenue value that the node should receive and record it in the periodic revenue allocation table.

[0296] To ensure the accuracy and fairness of periodic revenue distribution, after constructing the adjustment response intensity evaluation factor, the system continues to calculate the revenue value that each node should receive in the current scheduling cycle based on the incentive revenue parameters preset in the joint instruction set. This process not only considers the node's current execution capability but also integrates its preset incentive targets, resource attributes, and actual response behavior, thereby achieving multi-dimensional monetization of adjustment capabilities.

[0297] First, the system extracts the incentive benefit parameters corresponding to each node from the joint instruction set. These parameters are typically in the form of incentive amounts or coefficients corresponding to unit response intensity, preset during the instruction issuance phase, and set based on the node's historical performance, the importance of the adjustment task, and the budget strategy. Subsequently, the system pairs these parameters one by one with the adjustment response intensity evaluation factors calculated in the previous phase, generating adjustment incentive comparison data items. This involves binding the intensity factor of the same node's response capability in the current period with the reference parameter of the incentive value it can obtain.

[0298] Next, the system uses the adjustment response intensity evaluation factor and its incentive benefit parameter of each node as multiplication factors to perform a product operation, obtaining the initial benefit score of that node in the current scheduling cycle. This score directly reflects the degree of quantification of the node's response quality and the degree of integration between its incentive capability and performance in the current cycle, serving as the basis for subsequent benefit calculations. At this stage, even if nodes with different response capabilities have the same incentive benefit parameter, their initial score will differ due to variations in response performance, ensuring differentiated benefit allocation.

[0299] To account for the inherent heterogeneity of different resource types in regulation tasks, such as response speed, capacity limitations, and cost structure, the system further introduces a resource type correction mechanism. Each node adjusts its initial revenue score based on the weight parameters corresponding to its resource type. These weight parameters are determined based on actual operational experience or industry standards. For example, industrial load nodes typically have large response volumes but high variability, building load nodes are relatively stable but have small capacities, and distributed energy storage may have fast responses but is limited by available power. Therefore, different coefficients need to be set for each type to ensure a reasonable reflection of resource attributes in revenue allocation. The adjusted score more closely reflects the node's actual value performance during the regulation process.

[0300] Subsequently, the system continues to perform response stability correction on the revenue score after resource type adjustment. This process aims to eliminate non-systematic biases caused by occasional events, sudden fluctuations, or one-off adjustment errors. The system retrieves the response records of nodes at each time point within the current period, calculates the mean or coefficient of variation of their response capability over time, further evaluates the stationarity of the response curve, and smooths the revenue score based on the result. This suppresses the interference of extreme response behavior on the final revenue, ensuring that the revenue results are more representative and continuous.

[0301] Ultimately, the system maps the smoothed revenue score to a predefined incentive range, ensuring all revenue allocations remain within budget and do not exceed the funding constraints of the scheduling cycle. The mapping process can employ linear compression or normalization, yielding the periodic revenue value each node is entitled to in the current scheduling cycle. This value is recorded in the periodic revenue allocation table as the basis for settlement, providing data support for subsequent immediate settlements or cumulative payments.

[0302] Through the above processing flow, the system realizes a closed-loop calculation of the entire process from the adjustment capability assessment factor to the monetized income value, ensuring that the income distribution results achieve the optimal balance between fairness, rationality and incentive effectiveness.

[0303] A second embodiment of this application provides an electronic device, the electronic device comprising:

[0304] processor;

[0305] The memory is used to store a program, which, when read and executed by the processor, executes a virtual power plant optimization scheduling and revenue distribution method for multi-source demand-side resources provided in the first embodiment of this application.

[0306] The third embodiment of this application provides a computer-readable storage medium storing a computer program thereon. When the program is executed by a processor, it executes a virtual power plant optimization scheduling and revenue distribution method for multi-source demand-side resources provided in the first embodiment of this application.

[0307] Although this application discloses preferred embodiments as described above, it is not intended to limit this application. Any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of this application. Therefore, the scope of protection of this application should be determined by the scope defined in the claims of this application.

Claims

1. A method for optimized scheduling and revenue distribution of virtual power plants oriented towards multi-source demand-side resources, characterized in that, include: Acquire the current operating data of each industrial load node, building load node, and distributed photovoltaic energy storage node, and generate a node operating status table; Based on the node running status table, the node performance is classified, a scheduling negotiation table and a joint instruction set are constructed, and the joint instruction set is issued to each node to obtain the node execution result table. Update the performance record based on the node execution result table, and construct a delayed resource incentive reserve pool; Based on the incentive reserve pool table and the power adjustment instructions in the joint instruction set, the instant settlement revenue of each node is generated; and the priority of the redundancy right competition in the next cycle is adjusted according to the revenue level, so as to obtain the performance penalty table and node revenue distribution table for the next cycle. The process of classifying node performance based on the node running status table, constructing a scheduling negotiation table and a joint instruction set, issuing the joint instruction set to each node, and obtaining a node execution result table includes: Based on the node operation status table, nodes with a historical performance success rate lower than a preset threshold are selected as lagging nodes, and the rest are selected as active nodes to obtain the node performance classification results. Based on the node performance classification results, according to spatial proximity and adjustable window overlap, at most two lagging nodes are assigned to each active node, and a scheduling negotiation table is established. Based on the scheduling negotiation table and the node running status table, a joint instruction set is generated, which includes the target adjustment power of active nodes, the share of transferable adjustment tasks of lagging nodes, the redundancy sharing ratio, and the conflict resolution identifier; the joint instruction set is issued to each node, and the execution data at the end of the scheduling cycle is collected to obtain the node execution result table; Specifically, based on the node performance classification results, and according to spatial proximity and adjustable window overlap, each active node is assigned at most two lagging nodes, and a scheduling negotiation table is established, including: Based on the node performance classification results, the load adjustable time period of each active node and the load adjustable time period of the lagging node are extracted, and the ratio of their intersection length to union length is calculated as the adjustable window overlap index; only when the ratio is greater than a set threshold, the lagging node is included in the candidate set to obtain a preliminary adjustable point pair set. Based on the aforementioned set of adjustable point pairs, the equivalent resistance distance in the power grid topology is used as the spatial proximity metric. The lagging nodes associated with each active node are sorted from smallest to largest, and the top two lagging nodes are selected as the scheduling matching objects for the active node to obtain the node allocation list. Based on the historical performance stability index of each active node in the node allocation list, a negotiation priority weight is generated. In the case of multiple active nodes competing for the same lagging node, the scheduling control sequence is determined according to the rule that the higher priority node takes precedence and subsequent nodes make back-off adjustments based on the remaining adjustable capacity. A scheduling negotiation table is constructed based on the scheduling control sequence. The scheduling negotiation table is used to record the negotiation priority, scheduling instruction sequence and conflict resolution round information between each active node and its corresponding lagging node. The step of updating the performance record based on the node execution result table and constructing a delayed resource incentive reserve pool includes: Based on the adjustment start and end time, actual adjustment power and execution delay status of each lagging node in the node execution result table, the response performance index of the lagging node in the current scheduling cycle is calculated. The response performance index is obtained by weighting the degree of overlap between the successful response time and the delay tolerance interval and the ratio of the actual execution power to the target adjustment power. The response performance index is jointly analyzed with the response trend change rate of the lagging node in the last three scheduling cycles. If the trend is improving, the proportion of the lagging node used to retain part of the adjustment benefits is increased; otherwise, it is decreased. The retention ratio is dynamically adjusted within a preset threshold range, and the adjusted incentive retention weight is obtained. Apply the incentive retention weight to the actual adjustment power value in the node execution result table to obtain the amount of adjustment revenue that can be retained in the current period, and register it in the incentive reserve entry according to the node's unique identifier; Based on the incentive reserve entry, a delayed resource incentive reserve pool is constructed. The delayed resource incentive reserve pool uses the node identifier as the key index to record its deductible amount, frozen status and triggering conditions, which are used to match the penalty deduction path of the adjustment failure node in the subsequent scheduling cycle. The step of adjusting the priority of redundancy contest in the next cycle based on the level of revenue, and obtaining the performance penalty table and node revenue distribution table for the next cycle, includes: Based on the actual adjustment power and incentive reserve amount registered by each node in the node execution result table in the previous scheduling cycle, calculate the total revenue value actually obtained by each node in the previous cycle, and use it as the revenue evaluation benchmark to construct the current revenue level classification result table. Based on the revenue level classification result table, for edge nodes whose total revenue value is lower than the preset participation threshold for two or more consecutive periods, a one-time redundancy priority inversion mechanism is set up, that is, the scheduling priority of their redundancy adjustment task is increased in the next period to obtain the initial sequence of redundancy competition. Based on the initial sequence of redundancy contention, a weighted revenue smoothing function is constructed by combining the actual revenue value of each node in the last five scheduling cycles. The weighted revenue smoothing function is constructed by using time-decreasing weights arranged in reverse order of the cycle to generate a node stability revenue index sequence, which serves as the basis for sorting the revenue allocation factors in the next cycle. The node stability benefit index sequence and the initial sequence of redundancy right contention are jointly analyzed node by node to construct the performance penalty table and node benefit allocation table for the next period. The performance penalty table is used to record the deviation between the target response value and the actual response value of each node and the corresponding penalty amount. The node benefit allocation table is used to record the benefit base, allocation weight, incentive reserve deduction path and pre-allocated amount, thus obtaining the performance penalty table and node benefit allocation table for the next period.

2. The virtual power plant optimization scheduling and revenue distribution method for multi-source demand-side resources according to claim 1, characterized in that, The process of generating real-time settlement revenue for each node based on the incentive reserve pool table and the power adjustment instructions in the joint instruction set includes: Based on the target adjustment power of each node in the joint instruction set and the actual adjustment power recorded in the node execution result table, the actual response deviation of each node is calculated, and a response deviation result table is generated. Based on the actual response deviation and performance record of each node recorded in the response deviation result table, update the adjustment failure penalty deduction item in the incentive reserve pool table, and obtain the incentive reserve amount that each node can retain. Based on the updated incentive reserve amount in the incentive reserve pool table, and combined with the power adjustment instructions and resource types of each node in the joint instruction set, an adjustment response intensity evaluation factor for the current period is generated. Based on the adjustment response intensity evaluation factor and the incentive revenue parameter corresponding to each node in the joint instruction set, the periodic revenue value that each node should obtain is calculated. The periodic expected income value of each node is integrated with the updated incentive reserve amount in the incentive reserve pool table to generate the instant settlement income of each node.

3. The virtual power plant optimization scheduling and revenue distribution method for multi-source demand-side resources according to claim 2, characterized in that, The actual response deviation of each node is calculated based on the target adjustment power of each node in the joint instruction set and the actual adjustment power recorded in the node execution result table, and a response deviation result table is generated, including: Based on the target adjustment power and adjustment start and end time of each node recorded in the joint instruction set, a target power time series is constructed to characterize the theoretical power trajectory of the adjustment task in the current scheduling cycle. Based on the actual adjustment power recorded in the node execution result table, an actual power response sequence with the same sampling frequency is generated to ensure that it completely corresponds to the target power time sequence in the time dimension. The target power time series is compared with the actual power response series at each time point to obtain the deviation value at each time point, forming a node adjustment deviation time series, and statistical indicators including peak deviation, average deviation and standard deviation are extracted based on the series. The statistical indicators of the node adjustment deviation time series are combined with the associated information, including node identifier, adjustment start and end time, target adjustment power and actual adjustment power, into structured entries to generate a response deviation result table.

4. The virtual power plant optimization scheduling and revenue distribution method for multi-source demand-side resources according to claim 2, characterized in that, The process involves updating the adjustment failure penalty deduction item in the incentive reserve pool table based on the actual response deviation and performance record of each node recorded in the response deviation result table, and obtaining the incentive reserve amount that each node can retain, including: Extract the average response deviation and peak response deviation of each node recorded in the response deviation result table, and map them one by one with the node identifier to construct a response deviation index dictionary, which is used to unify the input for subsequent penalty ratio calculation; The response deviation index dictionary is associated with the adjustment task response success rate, historical trigger penalty count and cumulative performance deviation magnitude of each node in the performance record in the previous several scheduling cycles, and the node adjustment failure penalty factor in the current cycle is calculated based on the preset deviation tolerance threshold. The adjustment failure penalty factor is multiplied by the penalty deduction amount registered by the node in the incentive reserve pool table to obtain the incentive reserve amount to be deducted in this period. The updated adjustment failure penalty deduction item is then generated by comparing it with the current deductible amount of the corresponding node in the incentive reserve pool table. The incentive reserve amount that can be retained for each node is obtained by deducting the updated adjustment failure penalty deduction item from the initial incentive reserve amount registered in the incentive reserve pool table.

5. The virtual power plant optimization scheduling and revenue distribution method for multi-source demand-side resources according to claim 2, characterized in that, The process of generating a regulation response intensity assessment factor for the current period based on the updated incentive reserve amount in the incentive reserve pool table, combined with the power regulation instructions and resource types of each node in the joint instruction set, includes: Extract the available incentive value of each node from the updated incentive reserve amount in the incentive reserve pool table, and pair it with the power adjustment instruction corresponding to that node in the joint instruction set to construct a node incentive adjustment association table; Based on the ratio between the available excitation value and the power regulation command amplitude of each node in the node excitation regulation association table, the basic response capability index is calculated, and a type correction factor is set according to the resource type to reflect the performance heterogeneity of different resources in the regulation task; wherein, the resource type includes industrial load nodes, building load nodes and distributed photovoltaic energy storage nodes; The basic response capability index is weighted and fused with the type correction factor to generate the node initial response strength score. Based on the node's historical response stability record, a historical adjustment fluctuation compensation term is introduced to dynamically correct the node initial response strength score. The standardized score of each node's response intensity is used as the evaluation factor for the adjustment response intensity of the current cycle.

6. The virtual power plant optimization scheduling and revenue distribution method for multi-source demand-side resources according to claim 2, characterized in that, The calculation of the periodic payable value for each node based on the adjustment response intensity evaluation factor and the incentive payable parameter corresponding to each node in the joint instruction set includes: Extract the incentive benefit parameters corresponding to each node in the joint instruction set and match them with the adjustment response intensity evaluation factor of that node to form the adjustment incentive comparison data item for each node. The adjustment response intensity evaluation factor of each node is multiplied by its corresponding incentive benefit parameter to obtain the initial benefit score of the node. The initial benefit score reflects the response quality and incentive capability of the node in the current scheduling cycle. Based on the resource type of each node, the initial revenue score is adjusted according to the preset resource type weight to obtain the revenue score after resource type adjustment, so as to take into account the differences in contribution of different types of nodes in the adjustment task. The revenue score after resource type adjustment is processed by smoothing the response based on the time average of the response stability of each node in the current scheduling cycle to eliminate the impact of occasional response deviations. The processing result is then limited to a preset incentive range to finally determine the periodic revenue value that the node should receive and record it in the periodic revenue allocation table.

Images