A Virtual Power Plant Load Aggregation and Dispatch System and Method Based on Cloud-Edge Collaborative Architecture
By using a load aggregation and scheduling system based on a cloud-edge collaborative architecture, the problem of load regulation boundary distortion in virtual power plants under asynchronous edge-cloud conditions is solved, and the consistency of regulation boundaries and scheduling accuracy are improved during the execution phase.
Patent Information
- Application Number
- CN202611243038.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-17
- Publication Date
- 2026-11-17
AI Technical Summary
Under asynchronous edge-cloud conditions, existing virtual power plants suffer from issues such as distorted load regulation boundaries and out-of-bounds command behavior, leading to discrepancies between active power changes at grid connection points and dispatch requests.
The load aggregation and scheduling system adopts a cloud-edge collaborative architecture. Through the loss rate calculation module, execution time determination module, permit generation module, and instruction execution module, it determines the load increase permit, load decrease permit, and certificate version key, ensuring that the cloud computing data and edge execution data are consistent in version, and reducing the boundary deviation caused by outdated capacity and non-reported instructions.
Under the condition of asynchronous operation between edge and cloud, the aggregated scheduling results are adapted to the available adjustment range of the execution stage, reducing the grid-connected active power tracking deviation and improving scheduling accuracy.
Smart Images

Figure CN123121630A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of power dispatching technology, specifically to a virtual power plant load aggregation dispatching system and method based on a cloud-edge collaborative architecture. Background Technology
[0002] Virtual power plants typically connect to distributed air conditioning, charging facilities, industrial auxiliary equipment, and other controllable loads via communication networks. A cloud platform aggregates the adjustable capabilities of each load, generating aggregated control commands for grid dispatch requests. To reduce the complexity of on-site control, existing systems often rely on edge nodes to collect equipment status and upload capacity information. The cloud then performs aggregate calculations based on the received data and sends the control results back to the edge nodes for execution. This approach enables unified management of distributed loads, but the generation, uploading, queuing, command issuance, and on-site triggering of capacity information each consume time, and transmission times may vary across different communication links.
[0003] Between the generation of capacity information and the execution of control actions, the operating point of the controllable load, the allowable power range of the equipment, and existing instructions still in the process of transmission or execution may still change. If the cloud continues to use the capacity at the time of report generation as the available capacity at the actual execution time, or if the data version used for cloud computing is inconsistent with the data version used by the edge nodes for execution, the aggregated instructions may exceed the adjustment range that can be achieved on-site at that time, causing a deviation between the active power changes at the grid connection point and the scheduling request. Therefore, the existing aggregated scheduling process struggles to determine the load adjustment boundary that remains effective during the execution phase under conditions of asynchronous edge-cloud operation and continuously changing load status. Summary of the Invention
[0004] This invention provides a virtual power plant load aggregation and scheduling system and method based on a cloud-edge collaborative architecture to solve the technical problems of load regulation boundary distortion and command overrun during the execution phase of the virtual power plant under asynchronous edge-cloud conditions.
[0005] In a first aspect, the present invention provides a virtual power plant load aggregation and scheduling system based on a cloud-edge collaborative architecture, comprising: The loss rate calculation module is used to collect the load increase capacity sequence, load decrease capacity sequence, capacity time scale sequence and round-trip delay upper limit of the controllable load of the edge node, and calculate the load increase loss rate and load decrease loss rate according to the maximum ratio of the adjacent capacity decrease to the capacity time scale interval within the sliding window. The execution time determination module is used to determine the execution time by summing the cloud lock time, cloud processing time limit, and the upper limit of the maximum round-trip latency. The license generation module is used to determine the execution age by the execution time and capacity timestamp, shift bidirectional capacity according to the in-transit increment, and shrink capacity according to the product of capacity loss rate, execution age and conservative coefficient, and generate load increase license, load decrease license and certificate version key composed of edge node number, capacity timestamp and execution time; The regulation quantity generation module is used to obtain the remaining request quantity by subtracting the sum of the in-transit increments from the grid request quantity, limit the remaining request quantity by the sum of the permitted quantities in the same direction, and generate new regulation quantities according to the proportion of the permitted quantities in the same direction. The instruction execution module is used to issue instruction records containing certificate version keys and new adjustment amounts. The edge node locks the capacity report and writes the new adjustment amount to the active power increment input terminal of the local load aggregation controller at the execution time. It summarizes the in-transit increment and the new adjustment amount to generate the grid connection target increment.
[0006] The virtual power plant load aggregation and scheduling system based on a cloud-edge collaborative architecture provided by this invention determines the load increase loss rate and load decrease loss rate from the load increase capacity sequence and the load decrease capacity sequence, and determines the execution time based on the cloud locking time, cloud processing time limit, and the upper limit of the maximum round-trip delay. It uses in-transit increments to correct the expected execution working point, shrinks capacity based on execution age and a conservative coefficient, and locks the cloud ledger, load increase permits, load decrease permits, instruction records, and edge capacity reports through certificate version keys. Remaining requests are allocated proportionally only within the same direction's permit range, and edge nodes write new adjustment quantities at the execution time. This process reduces boundary deviations caused by stale capacity and unreported instructions, ensuring that the aggregation scheduling results adapt to the available adjustment range during the execution phase, maintaining version consistency between cloud computing data and edge execution data, and correspondingly reducing grid-connected active power tracking deviation under asynchronous edge-cloud operation conditions.
[0007] Secondly, this invention provides a virtual power plant load aggregation and scheduling method based on a cloud-edge collaborative architecture, comprising the following steps: Step S1: Collect the load increase capacity sequence, load decrease capacity sequence, capacity time scale sequence and round-trip delay upper limit of the controllable load of the edge node, and calculate the load increase loss rate and load decrease loss rate according to the maximum ratio of the adjacent capacity decrease to the capacity time scale interval within the sliding window. Step S2: Determine the execution time by summing the cloud lock time, cloud processing time limit, and the upper limit of the maximum round-trip latency; Step S3: Determine the execution age using the execution time and capacity time stamp, shift bidirectional capacity according to the in-transit increment, and shrink capacity according to the product of capacity loss rate, execution age and conservative coefficient, and generate the load increase permit, load decrease permit and certificate version key composed of edge node number, capacity time stamp and execution time; Step S4: Obtain the remaining request quantity by subtracting the sum of the incremental quantities in transit from the grid request quantity. Limit the remaining request quantity with the sum of the permitted quantities in the same direction, and generate the new adjustment quantity according to the proportion of the permitted quantities in the same direction. Step S5: Issue an instruction record containing the certificate version key and the new adjustment amount. The edge node locks the capacity report and writes the new adjustment amount to the active power increment input terminal of the local load aggregation controller at the execution time. The in-transit increment and the new adjustment amount are summarized to generate the grid connection target increment.
[0008] Thirdly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to execute the virtual power plant load aggregation and scheduling method based on cloud-edge collaborative architecture described in the second aspect above. Attached Figure Description
[0009] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.
[0010] Figure 1 This is a schematic diagram of the modules of the virtual power plant load aggregation and dispatching system based on cloud-edge collaborative architecture of the present invention; Figure 2 This is a schematic diagram of the process for generating the license quantity of this invention; Figure 3 This is a schematic diagram illustrating the deployment and data interaction of the virtual power plant load aggregation and dispatching system based on a cloud-edge collaborative architecture according to the present invention; Figure 4 This is a flowchart illustrating the virtual power plant load aggregation and scheduling method based on a cloud-edge collaborative architecture according to the present invention. Detailed Implementation
[0011] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0012] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.
[0013] like Figure 1 As shown, the virtual power plant load aggregation and dispatching system 100 based on a cloud-edge collaborative architecture includes: The loss rate calculation module 110 is used to collect the load increase capacity sequence, load decrease capacity sequence, capacity time scale sequence and round-trip delay upper limit of the controllable load of the edge node, and calculate the load increase loss rate and load decrease loss rate according to the maximum ratio of the adjacent capacity decrease to the capacity time scale interval within the sliding window. The execution time determination module 120 is used to determine the execution time by summing the cloud lock time, the cloud processing time limit, and the upper limit of the maximum round-trip delay. The license generation module 130 is used to determine the execution age by the execution time and capacity time stamp, shift bidirectional capacity according to the in-transit increment, and shrink capacity according to the product of capacity loss rate, execution age and conservative coefficient, and generate load increase license, load decrease license and certificate version key composed of edge node number, capacity time stamp and execution time. The regulation quantity generation module 140 is used to obtain the remaining request quantity by subtracting the sum of the incremental quantities in transit from the grid request quantity, limit the remaining request quantity by the sum of the permitted quantities in the same direction, and generate a new regulation quantity according to the proportion of the permitted quantities in the same direction. The instruction execution module 150 is used to issue instruction records containing certificate version keys and new adjustment amounts. The edge node locks the capacity report and writes the new adjustment amount to the active power increment input terminal of the local load aggregation controller at the execution time. It summarizes the in-transit increment and the new adjustment amount to generate the grid connection target increment.
[0014] In an optional embodiment, the loss rate calculation module 110 is configured to perform: In step S101, each edge node synchronously collects the load increase capacity sequence, load decrease capacity sequence, and capacity time-stamped sequence of the connected controllable load within a sliding window, and obtains the upper bound of the round-trip delay from the link measurement module. The number of window points in the sliding window is an integer not less than two, and the load increase capacity sequence, load decrease capacity sequence, and capacity time-stamped sequence all contain samples corresponding to the number of window points.
[0015] The load capacity sequence is a set of available active power reserves recorded by each edge node at continuous sampling times. Edge nodes read the current active power, allowable active power limit, and control availability status of each controllable load through the device communication interface. For devices in a controllable state, the available difference between the allowable active power limit and the current active power is summarized as the load capacity for the current load increase, and the load capacities obtained from continuous sampling are arranged in chronological order to form the load capacity sequence.
[0016] The load reduction capacity sequence is a set of available active power reserves recorded by each edge node at continuous sampling times. Edge nodes read the current active power, allowable lower limit of active power, and control availability status of each controllable load. For equipment in a controllable state, the available difference between the current active power and the allowable lower limit of active power is summarized as the load reduction capacity for that current operation, and the continuously sampled load reduction capacities are arranged in chronological order to form the load reduction capacity sequence.
[0017] The capacity timestamp sequence is a set of sampling moments corresponding point-by-point to the load increase capacity sequence and the load decrease capacity sequence. Each time bidirectional capacity aggregation is completed, the capacity timestamp is written into the same sampling transaction by the synchronization clock of the edge node. Synchronous acquisition here means that the same edge node forms the corresponding load increase capacity, load decrease capacity, and capacity timestamp within the same sampling transaction. Different edge nodes do not need to complete sampling at exactly the same physical time, but all use a unified time base.
[0018] The window number represents the number of sample points within the sliding window, and is an integer not less than two. The number of adjacent sample pairs formed by these sample points is the window number minus one. The window number is determined through historical playback before deployment. Specifically, sliding calculations are performed on each candidate window number that the historical capacity sequence can support, and the number of times the actual capacity contraction in the next scheduling cycle exceeds the predicted contraction in the verification data is compared. The candidate window number with the fewest such occurrences is selected. In the event of a tie, the candidate window number with the smaller value is selected, and the result is fixed in the loss rate calculation module 110.
[0019] The round-trip time (RTD) upper bound is the latency boundary for this batch of communication links between edge nodes and the cloud, obtained by the link measurement module. The link measurement module periodically sends probe packets with the sending time and sequence number, and the cloud returns a response with the same sequence number. The edge node obtains the RTD duration based on the sending time and response arrival time, and uses the maximum value of valid samples within the current measurement window as the RTD upper bound. Probe packets are matched against responses using unique sequence numbers; duplicate responses are not counted repeatedly, and responses that do not match the sending record are treated as invalid samples. If no valid response is obtained, no valid RTD upper bound is formed for that link.
[0020] Specifically, each edge node triggers a bidirectional capacity aggregation once according to a fixed sampling period. If equipment communication is interrupted, status data is invalid, local protection is locked, or the equipment is not currently accepting external active power regulation, the equipment will not be included in the current load increase and load decrease capacity. After the bidirectional capacity aggregation is completed, a capacity timestamp is written so that the load increase capacity, load decrease capacity, and capacity timestamp at the same location correspond to the same field status. After a new sample is written, the sliding window removes the oldest sample and retains several samples from the most recent window point.
[0021] Round-trip delay measurement and capacity sampling can be performed in parallel, but the capacity report encapsulation reads the link measurement results that have already been completed in this batch. If there are no valid round-trip delay samples in this batch, the edge node will not submit a valid capacity report, and the cloud will not include the edge node in this round of scheduling.
[0022] Step S102: Calculate the ratio of the capacity decrease of adjacent samples in the load increase capacity sequence and the capacity decrease capacity sequence to the capacity time-scale interval. If the adjacent capacity does not decrease, the ratio is set to zero. The load increase loss rate and the load decrease loss rate are determined by the maximum value of the ratio in each direction.
[0023] Specifically, the calculation formulas for the load increase loss rate and the load decrease loss rate are as follows: ; in, Indicates the first Each edge node is marked with a direction. The capacity loss rate at the following conditions; when When the capacity loss rate is the load loss rate; when At that time, the capacity loss rate is the load shedding loss rate. Indicates the first The edge node of the first Samples within a window are marked with orientation. The capacity below, This indicates the capacity of samples within the preceding window labeled in the same direction; when When, the capacity is the increased load capacity; when At that time, the capacity mentioned is the unloaded capacity. Indicates the first The edge node The timescale of the sample size within a window. It represents the capacity timescale of the samples in the preceding window that is adjacent to it. Indicates the sample index within the window. This indicates the number of window points in a sliding window. This indicates positive part operation.
[0024] The capacity decrease represents the magnitude of the reduction in available capacity between adjacent sampling times, and the capacity time-stamp interval represents the elapsed time between corresponding adjacent sampling times. The load increase loss rate represents the maximum observed decrease in capacity per unit time due to load increase within the sliding window, and the load decrease loss rate represents the maximum observed decrease in capacity per unit time due to load decrease within the sliding window; both are measured in kilowatts per second. The calculation results are only used to characterize the unilateral rate of capacity decrease within the window in the corresponding direction.
[0025] For each pair of adjacent samples, the ratio of capacity decrease to capacity timescale interval is calculated only when the latter capacity is less than the former capacity; when the latter capacity is not less than the former capacity, the corresponding ratio is zero. If any adjacent capacity timescale interval is zero or less than zero, the sliding window is marked as invalid time data, the capacity loss rate for that window is not calculated, and no valid capacity report is submitted based on that window. The above maximum value calculation is performed only for sliding windows with strictly increasing capacity timescales. Subsequently, the maximum value is selected from all ratios in both directions to obtain the load increase loss rate and load decrease loss rate. This process ensures that capacity increases do not offset capacity decreases that have already occurred within the same sliding window.
[0026] Step S103: Output a capacity report including the load increase capacity, load decrease capacity, capacity timestamp, load increase loss rate, load decrease loss rate, and round-trip delay upper limit at the end of the sliding window.
[0027] A capacity report is a record of data from the same batch submitted by edge nodes to the cloud. All fields in the capacity report reference the same sliding window end state and are written to the same report record in a single encapsulation operation. When generating a capacity report, the edge node first freezes the sliding window end samples and corresponding link measurement results, and then writes all the aforementioned fields into the report as a whole. After the report is generated, a local version record is added; subsequent sampling only generates new capacity reports without modifying the content of already submitted reports.
[0028] It should be noted that the load increase loss rate and load decrease loss rate are obtained from the decreasing segment of the corresponding capacity sequence, using the maximum decrease rate per unit time within the window. Converting the decrease amplitude into a unit time value using capacity time-stamped intervals avoids judgment bias caused by comparing only the decrease amplitude when sampling intervals are inconsistent. The capacity increase phase is not used to offset the decrease phase because the risk of inward shrinkage of available capacity needs to be assessed between the generation of the capacity report and the execution time, and capacity increase does not constitute a lower bound guarantee for future achievable capacity. The load increase loss rate and load decrease loss rate are combined with the execution age in the corresponding direction, ensuring that subsequent capacity shrinkage retains the different decrease rates exhibited by the two adjustment directions as the equipment operating point changes, and keeping the permissible amount corresponding to the actual adjustment direction.
[0029] In an optional embodiment, the execution timing determination module 120 is configured to execute: Step S201: The cloud locks the capacity reports of each edge node in this batch at the cloud locking time. The capacity report includes the load increase capacity, load decrease capacity, capacity timestamp, load increase loss rate, load decrease loss rate, and round-trip delay upper limit at the end of the sliding window.
[0030] The cloud lock time is the moment recorded by the cloud system clock when the cloud locks the current batch of capacity reports. The cloud first verifies the integrity of the capacity report fields, the validity of the capacity timestamp, and the validity of the round-trip latency upper bound. Then, it selects the latest valid capacity report arriving at each edge node before the cloud lock time, according to the edge node number. After selection, a read-only batch of reports is formed, and the cloud lock time is recorded. Upon reaching the cloud lock time, the edge nodes participating in the calculation for this batch and their capacity report versions remain unchanged. Subsequent capacity reports from the same edge node enter the next scheduling cycle without replacing already locked capacity reports.
[0031] Step S202: Iterate through the upper bounds of round-trip delay carried by each capacity report locked in this batch to determine the upper bound of the maximum round-trip delay.
[0032] The maximum round-trip time (RTT) upper bound is the maximum value among all valid capacity reports in this batch. This maximum RTT upper bound reflects the conservative communication time of the slowest link among the nodes participating in this round of scheduling, excluding edge nodes not included in this batch. The traversal object is consistent with the set of edge nodes that generated new adjustments in this round. Duplicate reports, reports with incomplete fields, and reports that failed time base verification are not included in the traversal.
[0033] Step S203: Add the cloud lock time, cloud processing time limit and maximum round-trip delay upper bound together, and use the resulting time as the execution time of all edge nodes.
[0034] Specifically, the formula for calculating the execution time is: ; in, This indicates the execution time of all edge nodes. This indicates the cloud lock time for this batch of capacity reports. This indicates the cloud processing timeframe covering cloud computing, queuing, instruction encapsulation, clock errors, and edge-triggered jitter. This indicates the number of items included in the scheduling in this batch. The upper bound of the round-trip delay of the link where each edge node is located. This indicates the index of the edge node in the current batch of locked report collection.
[0035] The cloud processing time limit is the upper bound of the total duration covering cloud queuing, scheduling computation, instruction encapsulation, clock error, and edge-triggered jitter, and its value is greater than zero. This parameter is determined through full-load stress testing before deployment. The test uses the maximum number of edge nodes allowed by the platform and the highest report arrival load, recording the total time from the locking of the capacity report set to the completion of cloud queuing, scheduling computation, and instruction encapsulation. This time limit is then combined with the upper bounds of the platform's clock error and edge-triggered jitter tests, and the maximum value from multiple rounds of valid tests is fixed as the cloud processing time limit. If the maximum number of edge nodes allowed by the platform, the computing resource configuration, or the communication processing link configuration changes, causing the existing stress test conditions to no longer cover the current operating conditions, the test is re-executed.
[0036] The execution time is the common point in time when all edge nodes implement the newly added adjustment in this round. Step S203 uses the upper bound of the full round-trip time, instead of fixedly taking half of it. The uplink and downlink times of the actual communication link may be asymmetrical, and estimating the downlink time of the instruction with only half of the round-trip time cannot cover all effective link states. The upper bound of the full round-trip time and the cloud processing time limit together reserve the time from the cloud locking time to the edge node having the execution conditions, so that all participating nodes can use the same execution time.
[0037] The cloud and edge nodes maintain a unified time base through the platform's existing clock synchronization service. When an edge node receives a command record, it checks its local synchronization status. If the synchronization status is invalid, or the execution time has already arrived when the command record is received, the edge node does not execute the command record and returns a rejection status to the cloud. After resynchronizing, the edge node participates in the next scheduling cycle with a new capacity report.
[0038] It should be noted that by traversing only the links corresponding to the current batch of locked report sets, the upper bound of the maximum round-trip latency can be kept consistent with the range of nodes actually involved in the execution. Determining a common execution time based on the cloud locking time, cloud processing time limit, and the aforementioned upper bound of the maximum round-trip latency allows for reserving a unified time boundary for cloud processing, instruction transmission, and edge timed triggering under asynchronous edge-cloud conditions, reducing the likelihood of different execution benchmarks being used by different edge nodes due to varying data arrival times.
[0039] In an optional embodiment, such as Figure 2 As shown, the license quantity generation module 130 is configured to execute: Step S301: Read the load increase capacity, load decrease capacity, capacity timestamp, load increase loss rate, and load decrease loss rate from the capacity reports of each edge node in this batch, and read the in-transit increments of each edge node from the same frozen ledger snapshot. The in-transit increments are the remaining execution increments that have not yet been included in the capacity report, have been confirmed and locked by the edge node, and have taken effect no later than the execution time. The portion that has been implemented is deducted based on the execution receipt. In the same frozen ledger snapshot, multiple incomplete instructions of the same edge node that meet the above inclusion conditions are algebraically summed with load increases being positive and load decreases being negative, forming a net increment.
[0040] In-transit increments are signed residual execution increments read from the same frozen ledger snapshot. Indicates the first The in-transit increments of each edge node are represented by positive values corresponding to load increases and negative values corresponding to load decreases, with units in kilowatts. Execution receipts are execution records returned by the edge node after a control action is implemented. Each execution receipt corresponds to the edge node number, certificate version key, and implemented increment. The cloud updates the ledger accordingly, ensuring that increments already physically reflected in the load operating point are no longer retained as in-transit increments.
[0041] The cloud creates a frozen ledger snapshot after the cloud lock time and reads incomplete instructions from the ledger according to the edge node number. The cloud only retains the portion that has received edge confirmation, has a scheduled effective time no later than the execution time, and has not yet been included in the lock capacity report. Permit generation, remaining request calculation, and grid connection target increment generation all read the frozen ledger snapshot.
[0042] Step S302: Determine the execution age by subtracting the capacity time stamp at the end of the sliding window from the execution time. If the difference is less than zero, the execution age is determined to be zero.
[0043] Specifically, the formula for calculating the execution age is: ; in, Indicates the first The execution age of each edge node. Indicates the first The capacity timescale of each edge node at the end of the sliding window. , and The meaning is the same as the definition above.
[0044] The execution age represents the time elapsed between the capacity timescale at the end of the sliding window and the execution time. The capacity report generation times of different edge nodes can be different, therefore the execution ages of each edge node can be different; all edge nodes still share the same execution time. A capacity timescale later than the execution time usually indicates a clock or version data anomaly. In this case, setting the execution age to zero is only to prevent negative time values; the capacity report still needs to pass the time base verification in step S201.
[0045] Step S303: When the in-transit increment is in the direction of increasing load, decrease the increasing load capacity and increase the decreasing load capacity; when the in-transit increment is in the direction of decreasing load, decrease the decreasing load capacity and increase the increasing load capacity. Subtract the product of the corresponding capacity loss rate, execution age, and conservative coefficient from the adjusted increasing load capacity and decreasing load capacity respectively. If each difference is less than zero, it is determined to be zero, and the increasing load permit quantity and decreasing load permit quantity are generated.
[0046] Specifically, the calculation formulas for the increased load allowance and the reduced load allowance are as follows: ; in, Indicates the first Each edge node is marked with a direction. The following license quantity; when When, the permitted quantity is the load allowance; when At that time, the permitted quantity is the load reduction permitted quantity. This represents the conservative coefficient. The meanings of the remaining mathematical symbols are the same as those defined above.
[0047] The conservatism coefficient is a calibration parameter for capacity shrinkage due to aging, and its value is not less than one. Calibration is performed using validation data independent of the actual operational data before commissioning. The validation data covers the type of equipment to be connected, sampling period, scheduling period, and link status. The predicted shrinkage amount obtained based on the capacity loss rate and execution age is compared with the measured capacity shrinkage amount during the aging period for each sample. The minimum value that ensures the predicted shrinkage amount is not less than the measured shrinkage amount for all validation samples is determined and fixed. If the operating conditions exceed the validation coverage, calibration is re-performed.
[0048] Specifically, a positive in-transit increment indicates that the load will increase before execution, correspondingly occupying the increased load capacity and creating a downward adjustment space from the operating point after the load increase. Therefore, the increased load capacity is reduced and the decreased load capacity is increased. A negative in-transit increment indicates that the load will decrease before execution, correspondingly occupying the decreased load capacity and creating an upward recovery space for the increased load. Therefore, the decreased load capacity is reduced and the increased load capacity is increased. The translation uses the actual kilowatt amplitude of the in-transit increment, without additional directional weighting.
[0049] After completing the in-transit incremental shift, the obsolete capacity reduction amount, determined by the corresponding capacity loss rate, execution age, and conservative coefficient, is deducted from the adjusted increased and decreased load capacities, respectively. Only the increased load loss rate is used for the increased load direction, and only the decreased load loss rate is used for the decreased load direction. If the deduction result is less than zero, it is rounded down to zero to obtain non-negative increased and decreased load allowances. In-transit incremental shift is retained when the execution age is zero, and obsolete capacity reduction is still performed when the in-transit incremental shift is zero; these two adjustments target different physical sources.
[0050] The load increase permit represents the capacity available for new load increase commands at execution time after the load increase capacity at the end of the sliding window has undergone in-transit incremental shifting, obsolete capacity shrinkage, and non-negative constraints. The load decrease permit represents the capacity available for new load decrease commands at execution time after the load decrease capacity at the end of the sliding window has undergone the same type of processing. Both are non-negative values, and the unit is kilowatts.
[0051] Step S304: A certificate version key is constructed using the edge node number, capacity timestamp, and execution time. The load increase and load decrease license quantities are then bound to this certificate version key. The cloud ledger, instruction record, and edge locking record all use the same certificate version key. This certificate version key is locked until the current execution cycle is completed. New capacity reports and new instructions generated during the locking period are included in the next scheduling cycle.
[0052] The edge node number is a unique identifier configured by the platform for each node, used to associate capacity reports, freeze ledger snapshots, license quantities, instruction records, and edge locking records. The certificate version key consists of the edge node number, the capacity timestamp of the locked capacity report, and the common execution time, used to identify the capacity report version and common execution time used by a specific edge node in this round of scheduling. The certificate version key is locked from its generation until the completion of this round of execution. New capacity reports and new instructions generated during the locking period are not written back to the current round's license quantity, nor do they change the already issued execution time.
[0053] It's important to note that in-transit incremental shifting is used to move the work point corresponding to the capacity report to the expected execution work point, while the contraction caused by capacity loss rate and execution age reflects the time effect of potential continued reduction in available capacity after report generation. These two methods target different objects. Implementing signed shifting first allows instructions that haven't yet entered the capacity report but will take effect before execution to occupy the corresponding directional margin, creating a return space in the opposite direction, thus preventing the same adjustment capacity from being repeatedly allocated. The certificate version key binds the edge node, capacity report timestamp, and common execution time to the same data version, ensuring that the capacity state used when generating licenses in the cloud corresponds to the capacity state locked during timed execution by the edge node. New capacity reports and added instructions during the locking period enter the next scheduling cycle, reducing adjustment boundary deviations caused by missing in-transit instructions, concurrent updates, and report version changes.
[0054] In an optional embodiment, the adjustment generation module 140 is configured to perform: Step S401: Subtract the sum of all in-transit increments from the grid request quantity to obtain the remaining request quantity.
[0055] The grid request quantity is the cumulative aggregated active power regulation increment relative to the baseline of this batch of reports, received by the cloud through the grid dispatch interface, and is measured in kilowatts. This indicates the grid request quantity; positive grid request quantities correspond to the load increase direction, and negative grid request quantities correspond to the load decrease direction. The received record includes the batch identifier, ensuring that it uses the same incremental benchmark as the locked capacity report set.
[0056] The remaining request quantity represents the active power regulation that still needs to be implemented by new instructions after deducting the sum of all in-transit increments in this batch from the grid request quantity. The grid request quantity, in-transit increments, and capacity reports use the same reporting baseline for this batch. Since in-transit increments are expected to take effect before the execution time, deducting their contribution first can avoid issuing new regulation quantities repeatedly for the same active power change.
[0057] In step S402, when the remaining request quantity is in the direction of increasing load and the sum of the load increase permit quantities is greater than zero, the remaining request quantity is limited to no more than the sum of the load increase permit quantities; when the remaining request quantity is in the direction of decreasing load and the sum of the load decrease permit quantities is greater than zero, the absolute value of the remaining request quantity is limited to no more than the sum of the load decrease permit quantities; when the remaining request quantity is zero or the sum of the permit quantities in the same direction is zero, the new adjustment quantity of each edge node is determined to be zero.
[0058] When the remaining request volume is in the load-increasing direction, only the sum of the load-increasing permits of each edge node is used as the upper bound, and any portion exceeding this upper bound is limited to that upper bound. When the remaining request volume is in the load-reducing direction, only the sum of the load-reducing permits of each edge node is used as the absolute upper bound, and any portion exceeding this upper bound is limited to that upper bound. Permits in one direction are not used to compensate for insufficient capacity in the other direction.
[0059] In step S403, when the remaining request quantity is in the direction of load increase and the sum of the load increase permit quantities is greater than zero, a positive new adjustment quantity is generated according to the proportion of the load increase permit quantity of each edge node to the sum of the load increase permit quantities; when the remaining request quantity is in the direction of load reduction and the sum of the load reduction permit quantities is greater than zero, a negative new adjustment quantity is generated according to the proportion of the load reduction permit quantity of each edge node to the sum of the load reduction permit quantities.
[0060] Specifically, the formula for calculating the additional adjustment amount is as follows: in, Indicates the first The new adjustment amount for each edge node, a positive new adjustment amount corresponds to the load increase direction, and a negative new adjustment amount corresponds to the load decrease direction. Indicates the first The load capacity of each edge node. Indicates the first The amount of load reduction allowance for each edge node. and The meaning is the same as the definition above. The summation range of each summation term is all edge nodes in this batch of locked report sets.
[0061] The newly added adjustment amount is a signed active power increment instruction generated by the adjustment amount generation module 140 for each edge node, with the unit being kilowatts. Positive newly added adjustment amounts are used for load increase, and negative newly added adjustment amounts are used for load decrease. This amount is an instruction value written to the edge node control interface, and is not a proportional coefficient, score, or priority.
[0062] Proportional allocation is performed when the sum of permits in the same direction is greater than zero. For load increases, the allocation is based on the proportion of each edge node's load increase permit to the total load increase permit; for load decreases, the allocation is based on the proportion of each edge node's load decrease permit to the total load decrease permit. The set of edge nodes participating in the summation and allocation is consistent with the set of locked reports in this batch; therefore, the sum of the new adjustment amounts for each edge node equals the remaining request amount after permit boundary restrictions.
[0063] When the remaining request quantity is zero, the new adjustment quantity of each edge node is set to zero. When the sum of the permit quantities in the same direction is zero, there is no permit capacity available for allocation in that direction, and the new adjustment quantity of each edge node is also set to zero, thus no division is performed. The above conditions constitute a definite input-output relationship, no candidate allocation scheme is set, and no control instruction is selected based on the scoring results.
[0064] It should be noted that first deducting the in-transit increment from the grid request volume, and then selecting the same-direction permitted volume according to the direction of the remaining request volume and implementing boundary restrictions, can avoid the duplicate allocation of active power changes that have already been anticipated to take effect, and can also prevent the use of the opposite-direction permitted volume to compensate for the current direction's capacity shortage. Subsequently, deterministic allocation is performed according to the proportion of the same-direction permitted volume, so that the sum of the new regulation volume of each edge node is consistent with the remaining request volume after restriction, and the new regulation volume of a single edge node does not exceed its corresponding direction's permitted volume.
[0065] In an optional embodiment, the instruction execution module 150 is configured to execute: Step S501: The cloud sends instruction records consisting of certificate version keys and new adjustment values to each edge node.
[0066] The cloud generates instruction records based on the edge node number. Each instruction record consists of the certificate version key of that edge node and the newly added adjustment amount. Both fields are written and integrity verified in the same encapsulation. The cloud only sends the instruction record to the edge node corresponding to the certificate version key, and does not concatenate fields from different edge nodes or different scheduling rounds.
[0067] In step S502, the edge node reads the capacity timestamp and execution time from the certificate version key, locks the capacity report corresponding to the capacity timestamp, and writes the new adjustment amount to the active power increment input terminal of the local load aggregation controller at the execution time.
[0068] Upon receiving the instruction record, the edge node reads the edge node number, capacity timestamp, and execution time from the certificate version key. The edge node first verifies the node number, then locates and locks the local capacity report corresponding to the capacity timestamp. If the certificate version key does not match, the corresponding capacity report does not exist, or the execution time has expired, the edge node refuses to write the new adjustment and returns the corresponding status. After successful verification, the new adjustment is saved along with the instruction record in the scheduled execution queue.
[0069] When the execution time arrives, the edge node writes the new adjustment amount to the active power increment input of the local load aggregation controller. The write operation uses only locked instruction records; new measurements and new capacity reports after the certificate version key is formed do not modify the input values for this round. The active power increment input is the control interface for the local load aggregation controller to receive active power change instructions relative to the capacity report baseline. The input unit is kilowatts; a positive value indicates an increase in active power consumption by the load, and a negative value indicates a decrease in active power consumption by the load. System 100 provides the aggregated new adjustment amount to the interface without changing the existing device-level constraint processing of the local load aggregation controller. The local load aggregation controller implements device-level control based on the existing device interface and local protection constraints.
[0070] In step S503, the cloud adds the sum of each in-transit increment to the sum of each newly added adjustment to generate the grid connection target increment relative to the baseline of this batch of reports.
[0071] Specifically, the formula for calculating the incremental grid connection target is as follows: ; in, This represents the incremental grid connection target relative to the baseline of this batch of reports. The meanings of the remaining mathematical symbols are the same as those defined above, and the summation range of each term is all edge nodes included in the scheduling of this batch.
[0072] The baseline of this batch of reports is the grid-connected active power reference state corresponding to the locked capacity report of this batch, and is not equivalent to the measured absolute power of the grid connection point at any given time. The grid connection target increment is the change in the grid connection point's active power target obtained by summing all in-transit increments and all new regulation quantities in this batch, in kilowatts. Grid requests, in-transit increments, new regulation quantities, and grid connection target increments are all expressed as increments using the baseline of this batch of reports. The cloud uses the same frozen ledger snapshot generated as the permitted quantity to summarize all in-transit increments and all new regulation quantities. When the remaining requested quantity is within the range formed by the aggregated load increase permit and aggregated load decrease permit, the grid connection target increment is consistent with the grid request quantity; when the request exceeds the permitted range in the corresponding direction, the grid connection target increment is limited to the permitted boundary along the request direction. The grid connection target increment is not used as the measured absolute power of the grid connection point.
[0073] After each edge node completes the write operation, it returns the execution status corresponding to the certificate version key. Upon receiving the execution status of the nodes in this round, the cloud closes the corresponding instruction record and updates the ledger when the execution receipt proves that the increment has been implemented. After the certificate version key is unlocked, new capacity reports and new instructions generated during the locking period enter the next scheduling cycle, thus forming continuous but isolated scheduling rounds.
[0074] It should be noted that the instruction record binds the certificate version key and the newly added adjustment amount in the same encapsulation. Before execution, the edge node verifies the node number, capacity timestamp, and execution time, ensuring that the capacity report locked by the edge execution corresponds to the report version used when calculating the license amount in the cloud. The grid connection target increment uses the same frozen ledger snapshot generated with the license amount to summarize the in-transit increment and the newly added adjustment amount. This allows the target amount and instruction allocation to use the same incremental benchmark, reducing grid connection active power tracking deviations caused by report version changes, missing in-transit instructions, and updates across scheduling rounds.
[0075] It should be noted that, as Figure 3As shown, this system can be deployed in a virtual power plant consisting of a cloud-based scheduling platform and multiple park edge nodes. Taking the connection of three edge nodes to air conditioning load, charging facilities, and industrial auxiliary load as an example, each edge node forms an increase capacity sequence, a decrease capacity sequence, and a capacity time-scale sequence with a sampling interval of 1 second, and the number of window points is 5. In a certain batch, the increase capacity at the end of the sliding window of the three edge nodes are 420 kW, 280 kW, and 500 kW, respectively, and the decrease capacity are 310 kW, 360 kW, and 450 kW, respectively, with round-trip delay upper limits of 0.12 seconds, 0.18 seconds, and 0.15 seconds, respectively. The cloud processing time limit was determined to be 0.30 seconds after stress testing, and the in-transit increments recorded in the frozen ledger snapshots are 40 kW, -20 kW, and 0 kW, respectively. After receiving a 700 kW load increase request from the grid, the cloud determines the joint execution time. Based on the load loss rate, execution age, conservative coefficient, and in-transit increments for each node, it generates a load increase permit. Then, it deducts all in-transit increments and allocates the new regulation capacity according to the proportion of the load increase permit. Each edge node verifies the certificate version key and writes the new regulation capacity to the active power increment input at the execution time. The cloud then aggregates the in-transit increments and the new regulation capacity to form the grid connection target increment. After execution, each edge node returns an execution receipt, which the cloud uses to deduct the implemented increments and includes the new capacity report generated during the lockout period in the next scheduling cycle.
[0076] This invention also provides a virtual power plant load aggregation and scheduling method based on a cloud-edge collaborative architecture, such as... Figure 4 As shown, it includes the following steps: Step S1: Collect the load increase capacity sequence, load decrease capacity sequence, capacity time scale sequence and round-trip delay upper limit of the controllable load of the edge node, and calculate the load increase loss rate and load decrease loss rate according to the maximum ratio of the adjacent capacity decrease to the capacity time scale interval within the sliding window. Step S2: Determine the execution time by summing the cloud lock time, cloud processing time limit, and the upper limit of the maximum round-trip latency; Step S3: Determine the execution age using the execution time and capacity time stamp, shift bidirectional capacity according to the in-transit increment, and shrink capacity according to the product of capacity loss rate, execution age and conservative coefficient, and generate the load increase permit, load decrease permit and certificate version key composed of edge node number, capacity time stamp and execution time; Step S4: Obtain the remaining request quantity by subtracting the sum of the incremental quantities in transit from the grid request quantity. Limit the remaining request quantity with the sum of the permitted quantities in the same direction, and generate the new adjustment quantity according to the proportion of the permitted quantities in the same direction. Step S5: Issue an instruction record containing the certificate version key and the new adjustment amount. The edge node locks the capacity report and writes the new adjustment amount to the active power increment input terminal of the local load aggregation controller at the execution time. The in-transit increment and the new adjustment amount are summarized to generate the grid connection target increment.
[0077] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded over a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the virtual power plant load aggregation and scheduling method based on a cloud-edge collaborative architecture shown in the above embodiments is implemented.
[0078] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.
Claims
1. A virtual power plant load aggregation and dispatching system based on a cloud-edge collaborative architecture, characterized in that: The system includes: The loss rate calculation module is used to collect the load increase capacity sequence, load decrease capacity sequence, capacity time scale sequence and round-trip delay upper limit of the controllable load of the edge node, and calculate the load increase loss rate and load decrease loss rate according to the maximum ratio of the adjacent capacity decrease to the capacity time scale interval within the sliding window. The execution time determination module is used to determine the execution time by summing the cloud lock time, cloud processing time limit, and the upper limit of the maximum round-trip latency. The license generation module is used to determine the execution age by the execution time and capacity timestamp, shift bidirectional capacity according to the in-transit increment, and shrink capacity according to the product of capacity loss rate, execution age and conservative coefficient, and generate load increase license, load decrease license and certificate version key composed of edge node number, capacity timestamp and execution time; The regulation quantity generation module is used to obtain the remaining request quantity by subtracting the sum of the in-transit increments from the grid request quantity, limit the remaining request quantity by the sum of the permitted quantities in the same direction, and generate new regulation quantities according to the proportion of the permitted quantities in the same direction. The instruction execution module is used to issue instruction records containing certificate version keys and new adjustment amounts. The edge node locks the capacity report and writes the new adjustment amount to the active power increment input terminal of the local load aggregation controller at the execution time. It summarizes the in-transit increment and the new adjustment amount to generate the grid connection target increment.
2. The virtual power plant load aggregation and dispatching system based on cloud-edge collaborative architecture according to claim 1, characterized in that, The loss rate calculation module is configured to perform: Each edge node synchronously collects the load increase capacity sequence, load decrease capacity sequence, and capacity timestamp sequence of the controllable load connected within the sliding window, and obtains the upper bound of the round-trip delay from the link measurement module; Calculate the ratio of the capacity decrease of adjacent samples in the two types of capacity sequences to the capacity time interval. When the adjacent capacity does not decrease, the ratio is set to zero. The maximum value of the ratio in each direction is used to determine the load increase loss rate and the load decrease loss rate. The output includes a capacity report showing the increased load capacity, decreased load capacity, capacity timestamp, increased load loss rate, decreased load loss rate, and upper bound of round-trip delay at the end of the sliding window.
3. The virtual power plant load aggregation and dispatching system based on a cloud-edge collaborative architecture according to claim 1, characterized in that, The execution timing determination module is configured to execute: The cloud locks the capacity reports of each edge node in this batch at the cloud lock time, determines the upper limit of the maximum round-trip latency in the capacity report, and adds the cloud lock time, cloud processing time limit and the upper limit of the maximum round-trip latency together, and uses the resulting time as the execution time of all edge nodes.
4. The virtual power plant load aggregation and dispatching system based on cloud-edge collaborative architecture according to claim 1, characterized in that, The license generation module determines the execution age and generates load increase and load decrease licenses, including: Read the in-transit increments of each edge node from the same frozen ledger snapshot. The in-transit increments are the remaining execution increments that have not yet been included in the capacity report, have been confirmed and locked by the edge node, and have taken effect no later than the execution time. The implemented portion is deducted based on the execution receipt. Continuous instructions from the same edge node are merged into a net increment. The execution age is determined by subtracting the capacity time stamp at the end of the sliding window from the execution time; if the difference is less than zero, the execution age is determined to be zero. When the incremental load is in the direction of increasing load, reduce the increasing load capacity and increase the decreasing load capacity; when the incremental load is in the direction of decreasing load, reduce the decreasing load capacity and increase the increasing load capacity. The product of the corresponding capacity loss rate, execution age, and conservative coefficient is deducted from the adjusted increased and decreased load capacities, respectively. All differences less than zero are set to zero, thus generating the increased and decreased load permits.
5. The virtual power plant load aggregation and dispatching system based on cloud-edge collaborative architecture according to claim 1, characterized in that, The certificate version key, which is composed and locked in the license quantity generation module, includes: The certificate version key is composed of edge node number, capacity timestamp and execution time, and the increase and decrease license amounts are bound to the certificate version key; The cloud ledger, instruction record, and edge lock record all use the same certificate version key. The certificate version key is locked until the current round of execution is completed. New capacity reports and new instructions generated during the locking period are included in the next scheduling cycle.
6. The virtual power plant load aggregation and dispatching system based on cloud-edge collaborative architecture according to claim 1, characterized in that, The adjustment generation module is configured to execute: The remaining request quantity is obtained by subtracting the sum of the in-transit increments from the grid request quantity. When the remaining request quantity is in the direction of load increase and the sum of the load increase permit quantities is greater than zero, the remaining request quantity is limited to no more than the sum of the load increase permit quantities, and a positive new regulation quantity is generated according to the proportion of each load increase permit quantity to the sum of the load increase permit quantities. When the remaining request quantity is in the load reduction direction and the sum of the load reduction permit quantities is greater than zero, the absolute value of the remaining request quantity is limited to no more than the sum of the load reduction permit quantities, and a negative new adjustment quantity is generated according to the proportion of each load reduction permit quantity to the sum of the load reduction permit quantities; when the remaining request quantity is zero or the sum of the permit quantities in the same direction is zero, the new adjustment quantity is determined to be zero.
7. The virtual power plant load aggregation and dispatching system based on cloud-edge collaborative architecture according to claim 1, characterized in that, The instruction execution module includes instruction issuance and execution, including: The cloud sends instruction records consisting of certificate version keys and new adjustment values to each edge node; The edge node reads the capacity timestamp and execution time from the certificate version key, locks the capacity report corresponding to the capacity timestamp, and writes the new adjustment amount to the active power increment input terminal of the local load aggregation controller at the execution time.
8. The virtual power plant load aggregation and dispatching system based on cloud-edge collaborative architecture according to claim 1, characterized in that, The instruction execution module generates the grid connection target increment, including: The cloud adds the sum of all ongoing increments to the sum of all newly added adjustments to generate the grid connection target increment relative to the baseline of this batch of reports.
9. A virtual power plant load aggregation and scheduling method based on a cloud-edge collaborative architecture, characterized in that, The virtual power plant load aggregation and dispatching system based on a cloud-edge collaborative architecture as described in any one of claims 1 to 8 includes the following steps: Step S1: Collect the load increase capacity sequence, load decrease capacity sequence, capacity time scale sequence and round-trip delay upper limit of the controllable load of the edge node, and calculate the load increase loss rate and load decrease loss rate according to the maximum ratio of the adjacent capacity decrease to the capacity time scale interval within the sliding window. Step S2: Determine the execution time by summing the cloud lock time, cloud processing time limit, and the upper limit of the maximum round-trip latency; Step S3: Determine the execution age using the execution time and capacity time stamp, shift bidirectional capacity according to the in-transit increment, and shrink capacity according to the product of capacity loss rate, execution age and conservative coefficient, and generate the load increase permit, load decrease permit and certificate version key composed of edge node number, capacity time stamp and execution time; Step S4: Obtain the remaining request quantity by subtracting the sum of the incremental quantities in transit from the grid request quantity. Limit the remaining request quantity with the sum of the permitted quantities in the same direction, and generate the new adjustment quantity according to the proportion of the permitted quantities in the same direction. Step S5: Issue an instruction record containing the certificate version key and the new adjustment amount. The edge node locks the capacity report and writes the new adjustment amount to the active power increment input terminal of the local load aggregation controller at the execution time. The in-transit increment and the new adjustment amount are summarized to generate the grid connection target increment.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to execute the virtual power plant load aggregation and scheduling method based on cloud-edge collaborative architecture as described in claim 9.
