Dynamic aggregation management method of virtual power plant edge computing cluster and cloud edge equipment

Abstract

The invention relates to the technical field of virtual power plants, and discloses a dynamic aggregation management method of a virtual power plant edge computing cluster and cloud edge equipment, and the method comprises the steps: carrying out the clock synchronization processing of a virtual power plant cloud control center and an edge computing cluster node; based on communication records of the cloud control center and each edge computing cluster node in historical N control periods, respectively counting an average generation rate of period frames and an average generation rate of event frames; according to the period frame average generation rate, the event frame average generation rate, the preset time slot capacity and the event frame reservation proportion, a fixed time slot is divided for the period frame, and a reserved time slot is divided for the event frame, so that a time slot scheduling table is obtained; and according to the time slot scheduling table, enabling each time slot number to correspond to a switch gating list, and generating window opening time and duration according to a time slot sequence.

Description

Technical Field

[0001] The present invention relates to the technical field of virtual power plants, and more specifically, to a dynamic aggregation management method for a virtual power plant edge computing cluster and cloud-edge devices. Background Art

[0002] As an important form of integrating distributed resources and participating in market-based dispatching, a virtual power plant needs to coordinate and control heterogeneous devices on a millisecond time scale to meet the requirements of complex application scenarios such as rapid regulation, grid interaction, and frequency support.

[0003] To achieve rapid response to dozens to hundreds of devices such as energy storage, electric vehicles, and load controllable terminals, the current mainstream solutions generally adopt an edge computing cluster architecture, deploying edge nodes at the side close to physical devices to sink some control logics and reduce communication latency and improve response speed. This architecture has gradually become a key support platform for future virtual power plants.

[0004] However, the existing virtual power plant edge computing clusters generally face the following technical pain points during operation: 1. The current control strategies mostly adopt simple transceiver mechanisms of non-deterministic networks (such as Ethernet and Wi-Fi), which do not provide strict time constraints. Especially when there are a large number of devices, network congestion, or intense resource competition, there are significant jitters in the transmission order and execution time of control instructions, which are extremely likely to lead to scheduling mismatch.

[0005] 2. Existing technologies often process the control scheduling logic and the underlying communication system separately. Edge nodes only focus on task distribution and do not manage network resources (such as link time slots and radio resource blocks) collaboratively, making it difficult to form a closed-loop control over behaviors such as delay, packet loss, and retransmission, thereby affecting control quality.

[0006] 3. Non-periodic events such as fault alarms and grid disturbances often occur during the operation of virtual power plants, and their response windows have high real-time requirements. However, existing systems often adopt a preemptive coupling queuing mechanism for event frames, making it difficult to guarantee the transmission success rate during peak hours and affecting device action synchronization. Summary of the Invention

[0007] The present invention provides a dynamic aggregation management method for a virtual power plant edge computing cluster and cloud-edge devices to solve the technical problems raised in the background art.

[0008] In a first aspect, the present invention provides a dynamic aggregation management method for a virtual power plant edge computing cluster, including: Step 1, perform clock synchronization processing on the virtual power plant cloud control center and the edge computing cluster nodes; Step 2: Based on the communication records between the cloud control center and each edge computing cluster node in the historical N control cycles, respectively calculate the average generation rate of cycle frames and the average generation rate of event frames; Step 3: According to the average generation rate of cycle frames, the average generation rate of event frames, the preset time slot capacity, and the reserved ratio of event frames, respectively divide fixed time slots for cycle frames and divide reserved time slots for event frames to obtain a time slot scheduling table; Step 4: According to the time slot scheduling table, map each time slot number to the switch gating list, and generate the window opening time and duration in the order of time slots; Step 5: Based on the data load and target reliability rate of each time slot in the time slot scheduling table, reserve resource blocks for each time slot in the 5G-URLLC network slice and set a redundant transmission mechanism; Step 6: Map the device energy scheduling instructions generated by the cloud control center to the corresponding time slots in the time slot scheduling table to generate energy scheduling instructions with timestamps; Step 7: The edge computing cluster node forwards the received energy scheduling instructions to the corresponding devices, and the devices perform actual power output according to the energy scheduling instructions and the maximum power of the devices; Step 8: Record the round-trip delay and reception success rate of the energy scheduling instructions, and compare them with the preset delay index and reliability rate index; when any index does not meet the preset requirements, adjust the reserved ratio of event frames according to the preset adjustment coefficient, and regenerate the time slot scheduling table.

[0009] Further, based on the communication records between the cloud control center and each edge computing cluster node in the historical N control cycles, respectively calculate the average generation rate of cycle frames and event frames, including: The cycle frame represents: in the communication record, the device status monitoring data sent at fixed time intervals; The event frame represents: in the communication record, the non-periodic burst data; The calculation formula for the average generation rate of cycle frames is as follows: The calculation formula for the average generation rate of event frames is as follows: Where, represents the average generation rate of cycle frames, represents the average generation rate of event frames, represents the number of control cycles, represents the duration of each control cycle, represents the number of cycle frames in the nth control cycle, represents the number of event frames in the nth control cycle.

[0010] Further, according to the average generation rate of the periodic frames, the average generation rate of the event frames, the preset time slot capacity, and the reserved ratio of the event frames, fixed time slots are respectively allocated for the periodic frames and reserved time slots are allocated for the event frames to obtain a time slot scheduling table, including: Determine communication resource parameters, where the communication resource parameters include: the preset time slot capacity and the average frame size ; Determine the number of time slots within each control period , where the number of time slots is the ratio of the duration of the control period to the preset time slot duration ; Allocation of fixed time slots for periodic frames, including: Calculate the total number of periodic frames within each control period , ; Calculate the data volume of the periodic frames , , , where \(\overline{F}\) represents the average size of the periodic frames; Determine the number of time slots allocated to the periodic frames , ; Allocation of reserved time slots for event frames, including: Load the reserved ratio \(\zeta\) of the event frames, where \(0\leqslant\zeta\leqslant1\); Determine the number of time slots allocated to the event frames , ; Generate a time slot scheduling table, where the time slot scheduling table includes: fixed time slots for periodic frames and elastic time slots for event frames; among them, the priorities of the fixed time slots for periodic frames and the elastic time slots for event frames decrease in sequence.

[0011] Further, according to the time slot scheduling table, map each time slot number to the switch gating list, and generate the window opening time and duration in the order of the time slots, including: For the \(s\)-th time slot in the time slot scheduling table, multiply the \(s\)-th time slot by the preset time slot duration to obtain the gating window opening offset of the \(s\)-th time slot; Use the preset time slot duration as the gating window duration of the \(s\)-th time slot; If the \(s\)-th time slot is less than the number of time slots allocated to the periodic frames , then map the \(s\)-th time slot to the periodic frame queue mask; otherwise, map the \(s\)-th time slot to the event frame queue mask; Based on the combination of the gating window opening offset, the gating window duration, and the queue mask, form the gating list entry of the switch; All gating list entries are sent and written to the gating control list of the switch in time slot order.

[0012] Furthermore, according to the data load and target reliability rate of each time slot in the time slot scheduling table, resource blocks are reserved for each time slot in the 5G-URLLC network slice and a redundant transmission mechanism is set up, including: Read the predicted data load and preset target reliability rate of each time slot from the time slot scheduling table; Determine the single transmission failure probability of the 5G air interface without redundancy; Based on the target reliability rate and the single transmission failure probability, calculate the number of redundant transmissions; Based on the predicted data load of the time slot and the data volume carried by each resource block, determine the number of resource blocks required for the corresponding time slot; Reserve the required number of resource blocks for each time slot in the 5G-URLLC network slice and execute the redundant transmission mechanism according to the number of redundant transmissions.

[0013] Furthermore, map the device energy scheduling instructions generated by the cloud control center to the corresponding time slots in the time slot scheduling table to generate energy scheduling instructions with timestamps, including: Obtain the time slots corresponding to each device energy scheduling instruction generated by the cloud control center from the time slot scheduling table; Based on the duration of the control period and the gating window opening offset corresponding to the time slot, calculate the transmission timestamp of the device energy scheduling instruction; Attach the transmission timestamp to the device energy scheduling instruction to generate an energy scheduling instruction with a timestamp; Send the energy scheduling instructions with timestamps to the corresponding edge computing nodes in chronological order.

[0014] Furthermore, the edge computing cluster nodes forward the received energy scheduling instructions to the corresponding devices, and the devices perform actual power output according to the energy scheduling instructions and the maximum power of the devices, including: The edge computing cluster nodes forward the received energy scheduling instructions with timestamps to the corresponding devices; When the local clock of the device reaches the transmission timestamp in the energy scheduling instruction with a timestamp, trigger the execution: The device multiplies the power ratio in the energy scheduling instruction with a timestamp by the rated maximum power of the device to calculate the actual output power; the device encapsulates the actual output power and the confirmation timestamp into an execution confirmation message and feeds back the execution confirmation message to the edge computing cluster node.

[0015] Further, record the round-trip delay and reception success rate of the energy scheduling instruction, and compare them with the preset delay index and reliability index; when any index does not meet the preset requirements, adjust the event frame reservation ratio according to the preset adjustment coefficient, and regenerate the time slot scheduling table, including: In each control cycle, record the transmission timestamp of the timestamped energy scheduling instruction and the confirmation timestamp of the device, and count the number of execution confirmation messages received by the cloud control center; Calculate the average round-trip delay and actual reception success rate of the timestamped energy scheduling instruction; Compare the average round-trip delay with the preset maximum allowable round-trip delay, and compare the actual reception success rate with the target reliability; When the average round-trip delay is greater than the maximum allowable round-trip delay, or the actual reception success rate is less than the target reliability, adjust the event frame reservation ratio ζ; Based on the adjusted event frame reservation ratio ζ, return to step 3 to regenerate the time slot scheduling table.

[0016] In a second aspect, a cloud-edge device of a virtual power plant edge computing cluster is applied to the dynamic aggregation management method of any of the above-mentioned virtual power plant edge computing clusters, and is characterized by including: A clock synchronization module for performing clock synchronization processing on the virtual power plant cloud control center and the edge computing cluster nodes; A data acquisition module for respectively counting the average generation rate of cycle frames and the average generation rate of event frames based on the communication records between the cloud control center and each edge computing cluster node in the historical N control cycles; A time slot scheduling module for respectively dividing fixed time slots for cycle frames and dividing reserved time slots for event frames according to the average generation rate of cycle frames, the average generation rate of event frames, the preset time slot capacity, and the event frame reservation ratio, so as to obtain a time slot scheduling table; A time slot matching module for corresponding each time slot number to the switch gating list according to the time slot scheduling table, and generating the window opening time and duration in sequence according to the time slot; A resource allocation module for reserving resource blocks for each time slot in the 5G-URLLC network slice and setting a redundant transmission mechanism according to the data load of each time slot in the time slot scheduling table and the target reliability; An instruction generation module for mapping the device energy scheduling instruction generated by the cloud control center to the corresponding time slot in the time slot scheduling table to generate a timestamped energy scheduling instruction; A device management module for forwarding the received energy scheduling instruction by the edge computing cluster node to the corresponding device, and the device performs actual power output according to the energy scheduling instruction and the maximum power of the device; A parameter optimization module is used to record the round-trip delay and reception success rate of the energy scheduling instruction, and compare them with the preset delay index and reliability index; when any index does not meet the preset requirements, adjust the reserved ratio of the event frame according to the preset adjustment coefficient, and regenerate the time slot scheduling table.

[0017] The beneficial effects of the present invention are as follows: By constructing a time slot scheduling table with unified numbering, the communication resources of the wired TSN and the wireless 5G-URLLC are accurately aligned in the time dimension. Combining the priority separation of the periodic frame and the event frame and the resource redundancy strategy, high-reliability and low-latency energy scheduling control of multiple devices by the virtual power plant within a millisecond-level control cycle is achieved. By collecting the round-trip delay and success rate indicators in real time and dynamically adjusting the reserved ratio of the event frame, a closed-loop adaptive optimization mechanism is formed, thereby continuously ensuring the certainty and stability of the scheduling response in a complex network environment, which is significantly better than the existing solutions that only support static resource configuration of a single network segment. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 is a flowchart of the dynamic aggregation management method of the virtual power plant edge computing cluster of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] Now, the subject matter described herein will be discussed with reference to exemplary embodiments. It should be understood that the discussion of these embodiments is only to enable those skilled in the art to better understand and thus implement the subject matter described herein, and the functions and arrangements of the elements discussed can be changed without departing from the protection scope of the content of this specification. Each example can omit, substitute or add various processes or components as needed. Additionally, the features described relative to some examples can also be combined in other examples.

[0020] Example 1, as Figure 1 shown, the dynamic aggregation management method of the virtual power plant edge computing cluster includes: Step 1, perform clock synchronization processing on the virtual power plant cloud control center and the edge computing cluster nodes; It should be noted that the clock synchronization processing is implemented based on the IEEE1588-PTP protocol.

[0021] Step 2, based on the communication records between the cloud control center and each edge computing cluster node within the historical N control cycles, respectively count the average generation rate of the periodic frame and the average generation rate of the event frame; In an embodiment of the present invention, based on the communication records between the cloud control center and each edge computing cluster node within the historical N control cycles, respectively counting the average generation rate of the periodic frame and the event frame includes: The periodic frame represents: the device status monitoring data sent at fixed time intervals in the communication record; Event frame representation: In a communication record, it represents non-periodic burst data; The calculation formula for the average generation rate of periodic frames is as follows: The calculation formula for the average generation rate of event frames is as follows: Among them, represents the average generation rate of periodic frames, represents the average generation rate of event frames, represents the number of control cycles, represents the duration of each control cycle, represents the number of periodic frames in the nth control cycle, represents the number of event frames in the nth control cycle.

[0022] Specifically, the purpose of step 2 is to quantitatively count the traffic characteristics of periodic frames and event frames, providing data support for subsequent time slot scheduling, as follows: Periodic frame: It refers to the device status monitoring data sent at fixed time intervals in a virtual power plant. The device status monitoring data includes: photovoltaic inverter power, energy storage battery SOC, etc. The device status monitoring data has the characteristics of periodicity, low jitter, and high certainty, and is the basic monitoring data for the stable operation of the system.

[0023] Event frame: It refers to non-periodic burst data. The non-periodic burst data includes: device fault alarms, emergency power adjustment instructions, etc.

[0024] In the formula, is the number of periodic frames in the th period, is the number of historical statistical periods, is the duration of a single period. By accumulating and averaging multiple periods, short-term fluctuations are smoothed, and long-term stable traffic is reflected. For example: If 50 control cycles generate a total of 5000 periodic frames, then frames per second, that is, 100 frames every 20ms.

[0025] The average generation rate of event frames is the same.

[0026] Step 3, according to the average generation rate of the periodic frames, the average generation rate of the event frames, the preset time slot capacity, and the event frame reservation ratio, respectively divide fixed time slots for the periodic frames and divide reserved time slots for the event frames to obtain a time slot scheduling table; In one embodiment of the present invention, fixed time slots are respectively allocated for periodic frames and reserved time slots for event frames according to the average generation rate of periodic frames, the average generation rate of event frames, the preset time slot capacity, and the reserved ratio of event frames, so as to obtain a time slot scheduling table, including: Determine communication resource parameters, where the communication resource parameters include: preset time slot capacity and average frame size ; Determine the number of time slots within each control period , where the number of time slots is the ratio of the duration of the control period to the preset time slot duration ; Allocation of fixed time slots for periodic frames, including: Calculate the total number of periodic frames within each control period , ; Calculate the data volume of the periodic frames , , indicating the average size of the periodic frames; Determine the number of time slots allocated to the periodic frames , ; Allocation of reserved time slots for event frames, including: Load the reserved ratio ζ of event frames, 0 ≤ ζ ≤ 1; Determine the number of time slots allocated to the event frames , ; Generate a time slot scheduling table, where the time slot scheduling table includes: fixed time slots for periodic frames and elastic time slots for event frames; among them, the priorities of the fixed time slots for periodic frames and the elastic time slots for event frames decrease in turn.

[0027] It should be noted that step 3 generates a time slot scheduling table by quantitatively dividing the communication capacity requirements of periodic frames and event frames, and finally forms a time slot scheduling table including fixed time slots for periodic frames and elastic time slots for event frames.

[0028] The system needs to determine communication resource parameters, that is, the maximum number of bytes that each time slot can carry and the average data volume per frame . Among them, depends on the throughput capacity of the underlying physical link (the minimum transmission unit configuration of TSN and 5G-URLLC); is obtained by statistically analyzing the historical frame lengths, reflecting the average message sizes of periodic reports and event reports.

[0029] The system calculates the number of time slots that can be divided within this control period according to the total duration of a control period and the preset time slot duration , ; For periodic frames, the number of periodic frames within a control period is estimated using their average generation rate (frames per second). Subsequently, the number of periodic frames within a control period is multiplied by the average frame size to obtain the total required data volume. Thus, the number of periodic frame time slots allocated to periodic frames is calculated . represents the minimum communication bandwidth required for reporting within a one-time lock control period, avoiding delays or losses of periodic control frames due to resource contention.

[0030] For event frames, the system first pre-sets an event frame reservation ratio ζ, where 0 ≤ ζ ≤ 1. ζ represents the proportion of time slots reserved for randomly arriving event frames in each control period, thereby determining the number of time slots allocated to event frames.

[0031] After obtaining and , a complete time slot scheduling table can be generated: Time slot numbers from 0 to - 1: Used for periodic frame transmission, with the highest priority; Time slot number to : Used for event frame transmission, with a relatively high priority; The remaining time slots (numbered from to ): Used for the elastic preemptible area of event frames, with the lowest priority.

[0032] It should be noted that the preset time slot duration represents the fixed time length of each time slot preset during the system design phase according to the characteristics of the communication physical layer (mini-slot duration of 5GNR, time scheduling granularity of TSN) and service requirements (periodic frame transmission period, upper limit of event frame processing delay).

[0033] For example, if the virtual power plant edge computing cluster adopts 5GNR URLLC technology, then the subcarrier spacing SCS = 75 kHz, and the corresponding mini-slot duration is 125 μs, that is = 125 μs.

[0034] Step 4, according to the time slot scheduling table, map each time slot number to the switch gating list, and generate the window opening time and duration in the time slot order; In an embodiment of the present invention, according to the time slot scheduling table, mapping each time slot number to the switch gating list and generating the window opening time and duration in the time slot order includes: For the s-th time slot in the time slot schedule, by multiplying the s-th time slot by a preset time slot duration , the gating window opening offset of the s-th time slot is obtained; Use the preset time slot duration as the gating window duration of the s-th time slot; If the s-th time slot is less than the number of time slots allocated to the periodic frame , then map the s-th time slot to the periodic frame queue mask; otherwise, map the s-th time slot to the event frame queue mask; Based on the combination of the gating window opening offset, the gating window duration, and the queue mask, form a gating list entry for the switch; Issue all the gating list entries in time slot order and write them into the gating control list of the switch.

[0035] It should be noted that in step 4: configuring the TSN switch gating list, by directly mapping the pre-computed time slot schedule to the IEEE 802.1Qbv gating list (GCL), the wired network segment can release different priority data queues in strict time slot order and on demand, thus achieving end-to-end deterministic communication. The specific logic is as follows: First, for the s-th time slot ( ) in the time slot schedule, the system extracts the sequence number s of this time slot within a control period and multiplies it by the preset time slot duration ; Among them, represents the gating window opening offset after the start point of the control period of the gating window, ensuring that the s-th time slot starts to be released strictly within the s-th gating window.

[0036] Then, the system directly uses as the gating window duration of the s-th time slot, that is to say, the gating window starts from and automatically closes after . Thus, the time window corresponding to the s-th time slot is defined .

[0037] After that, it is necessary to determine the TSN queue corresponding to the time window .

[0038] If s ≤ (that is, the s-th time slot falls within the periodic frame time slots reserved for the periodic frame), then map this time slot to the periodic frame queue mask ; otherwise, map it to the event frame queue mask . Through the mask it is specified which enters the time window The data categories and priorities are such that the periodic frames are always scheduled before the event frames.

[0039] Finally, for the s-th time slot, the system combines the three , and into a gating list entry in the GCL, which can be formally expressed as: All to gating list entries are sorted in order of the size of the gating window opening offset , then sent down and written to the switch to enable the IEEE802.1Qbv time slot scheduling function; In the above manner, the TSN switch can release periodic frames and event frames in each level small window in a fixed priority order, avoiding contention delay inside the queue.

[0040] Step 5: According to the data load and target reliability rate of each time slot in the time slot scheduling table, reserve resource blocks for each time slot in the 5G-URLLC network slice and set a redundant transmission mechanism; In an embodiment of the present invention, according to the data load and target reliability rate of each time slot in the time slot scheduling table, reserving resource blocks for each time slot in the 5G-URLLC network slice and setting a redundant transmission mechanism includes: Read the predicted data load of each time slot and the preset target reliability rate from the time slot scheduling table; Determine the single transmission failure probability of the 5G air interface without redundancy; Based on the target reliability rate and the single transmission failure probability, calculate the number of redundant transmissions; Based on the predicted data load of the time slot and the data volume carried by each resource block, determine the number of resource blocks required for the corresponding time slot; Reserve the required number of resource blocks for each time slot in the 5G-URLLC network slice and execute the redundant transmission mechanism according to the number of redundant transmissions.

[0041] It should be noted that Step 5: Reserving resource blocks in the 5G-URLLC network slice and setting a redundant transmission mechanism is responsible for providing end-to-end high reliability guarantee for each time slot determined by the time slot scheduling table on the wireless side. Specifically as follows: First, for the s-th time slot in the time slot scheduling table, the system reads two predicted data loads and the target reliability rate from the time slot scheduling table; among them, the predicted data load represents the total amount of packets to be transmitted in the s-th time slot; the target reliability rate is a preset value, representing the minimum probability requirement for successful data delivery within the sth time slot.

[0042] Next, the system queries the statistical characteristics of the 5G air interface link to obtain the single transmission failure probability under the condition of no redundant transmission , which is derived from air interface quality measurement or operator slice commitment, such as the packet loss rate when no retransmission is used. To meet the target reliability rate, the same batch of data must be independently transmitted multiple times within the sth time slot to reduce the overall failure probability.

[0043] According to the Bernoulli independent trial model, the minimum number of redundant transmissions required can be calculated as follows: where, represents the minimum number of redundant transmissions, represents the natural logarithm; After determining the minimum number of redundant transmissions, the system then predicts the data load according to the time slot and the data volume that each resource block can carry (bytes) to calculate the total number of resource blocks required within the sth time slot , as follows: where, is the minimum number of resource blocks required for a single transmission, and after multiplying by the number of redundant times, it is the total number of resource blocks required for the sth time slot .

[0044] The system will directly configure and reserve resource blocks within the 5G-URLLC network slice for the total number of resource blocks required calculated for each time slot and initiate multi-path transmission in sequence within the sth time slot according to the number of redundant times . Step 5 not only establishes a wireless extension for the time-sensitive flow on the TSN wired side, but also forms a collaborative mode of time slot one-key drive + redundancy guarantee in the overall end-to-end communication link, ensuring that control instructions and status reports under the virtual power plant edge-cloud architecture reach the device seamlessly and reliably.

[0045] Step 6, map the device energy scheduling instruction generated by the cloud control center to the corresponding time slot in the time slot scheduling table to generate an energy scheduling instruction with a timestamp; In an embodiment of the present invention, mapping the device energy scheduling instruction generated by the cloud control center to the corresponding time slot in the time slot scheduling table to generate an energy scheduling instruction with a timestamp includes: Obtain the time slot corresponding to each device energy scheduling instruction generated by the cloud control center from the time slot scheduling table; Calculate the sending timestamp of the device energy scheduling instruction based on the duration of the control cycle and the gated window opening offset corresponding to the time slot; Adding the sending timestamp to the device energy scheduling instruction to generate an energy scheduling instruction with a timestamp; The energy scheduling instructions with timestamps are sent to the corresponding edge computing nodes in chronological order.

[0046] It should be noted that step 6: mapping the device energy scheduling instructions generated by the cloud control center to the time slot scheduling table and generating energy scheduling instructions with timestamps runs through the two links of scheduling decision and physical issuance, ensuring that each instruction can be strictly issued within the predetermined TSN / 5G time slot.

[0047] First, the system retrieves the time slot corresponding to each device energy scheduling instruction generated by the cloud control center from the time slot scheduling table. For example, the energy scheduling instruction that device i should execute in the τth control cycle is , the energy scheduling instruction is mapped to the number in the time slot scheduling table time slot.

[0048] Next, according to the control cycle length The system calculates the sending time of the energy scheduling instruction based on the gating window opening offset of the sth time slot as follows: in, Indicates the time when the energy scheduling instruction is sent, Position to The starting time of a control cycle, Then the time is accurately shifted to the starting position of the sth time slot; After the time calculation is completed, the system will Energy Scheduling Instructions Encapsulate them together into a control instruction with a timestamp: Finally, the system follows the time These time-stamped instructions are sent to the corresponding edge computing nodes in sequence.

[0049] Step 7: The edge computing cluster node forwards the received energy scheduling instruction to the corresponding device, and the device performs actual power output according to the energy scheduling instruction and the maximum power of the device; In one embodiment of the present invention, the edge computing cluster node forwards the received energy scheduling instruction to the corresponding device, and the device performs actual power output according to the energy scheduling instruction and the maximum power of the device, including: The edge computing cluster node forwards the received energy scheduling instruction with timestamp to the corresponding device; Trigger the execution when the local clock of the device reaches the transmission timestamp in the timestamped energy scheduling instruction: The device multiplies the power ratio in the timestamped energy scheduling instruction by the rated maximum power of the device to calculate the actual output power; the device encapsulates the actual output power and the confirmation timestamp as an execution confirmation message and feeds back the execution confirmation message to the edge computing cluster node.

[0050] It should be noted that after receiving the timestamped energy scheduling instruction sent by the cloud control center, the edge computing cluster node forwards it to the corresponding physical device. After receiving the instruction, the device side continuously monitors the local clock reading. When it reaches , that is, trigger the subsequent energy output task. Once triggered, the device according to the power ratio given in the instruction , ; multiply the power ratio by its own rated maximum output power to calculate the actual target output power.

[0051] After completing the power setting and actual output, the device obtains the actual output power through its built-in measurement devices such as current / voltage sensors , and records the current local clock as the confirmation timestamp . The device will and be encapsulated as an execution confirmation message and sent back to the edge computing cluster node again.

[0052] Step 8, record the round-trip delay and reception success rate of the energy scheduling instruction, and compare them with the preset delay index and reliability index; when any index does not meet the preset requirements, adjust the event frame reservation ratio according to the preset adjustment coefficient and regenerate the time slot scheduling table.

[0053] In an embodiment of the present invention, recording the round-trip delay and reception success rate of the energy scheduling instruction, and comparing them with the preset delay index and reliability index; when any index does not meet the preset requirements, adjusting the event frame reservation ratio according to the preset adjustment coefficient and regenerating the time slot scheduling table includes: In each control cycle, record the transmission timestamp of the timestamped energy scheduling instruction and the confirmation timestamp of the device, and count the number of execution confirmation messages received by the cloud control center; Calculate the average round-trip delay and actual reception success rate of the timestamped energy scheduling instruction; Compare the average round-trip delay with the preset maximum allowable round-trip delay, and compare the actual reception success rate with the target reliability rate; When the average round-trip delay is greater than the maximum allowable round-trip delay, or the actual reception success rate is less than the target reliability rate, then adjust the event frame reservation ratio ζ; Based on the adjusted event frame reservation ratio ζ, return to step 3 to regenerate the time slot scheduling table.

[0054] It should be noted that in order to perform real-time monitoring and closed-loop optimization of the end-to-end communication performance, the system will fully record the timestamped energy scheduling instructions and their feedback processes in each control cycle. Specifically, when the j-th energy scheduling instruction is sent from the edge computing cluster to the device, its sending time is ; the device is triggered to execute when the local clock reaches , and after measuring the actual output power, it will send an execution confirmation message with a confirmation timestamp back to the cloud control center. The cloud control center counts the total number of instructions sent in a cycle as M, and the number of actual received confirmation messages is .

[0055] The system first calculates the average round-trip delay as follows: Among them, represents the average round-trip delay. Dividing the sum of the delays from the issuance to the confirmation of each instruction by the total number of instructions can truly reflect the average delay level of the entire edge → device → edge closed-loop.

[0056] Calculate the actual reception success rate as follows: Among them, represents the actual reception success rate, represents the transmission reliability of the feedback message under the current link and load conditions.

[0057] The system compares the average round-trip delay with the preset maximum allowable round-trip delay , and verifies the actual reception success rate with the target reliability rate . Only when and , the system considers that the time slot allocation meets the performance requirements.

[0058] If any index exceeds the limit, it means that the reservation ratio ζ for the emergency event frame cannot support the necessary delay and reliability guarantees, and it needs to be dynamically adjusted.

[0059] The dynamic adjustment method of the event frame reservation ratio ζ is: according to the pre-set adjustment coefficient κ and the size of the index deviation, the event frame reservation ratio ζ is modified in an incremental or decremental manner, thereby changing the number of reserved time slots allocated to the event frame in the next cycle. After completing the adjustment of ζ, the system automatically returns to step 3, recalculates the number of time slots allocated to the cycle frame and the number of time slots allocated to the event frame according to the new reservation ratio, and generates an updated time slot scheduling table.

[0060] Preferably, in a virtual power plant scenario, the edge computing cluster needs to perform power scheduling for 10 energy storage inverter devices, with a control period of 20ms and a preset time slot length of 125μs. The system first synchronizes the edge and device clocks to the ±100ns level through IEEE1588-PTP, and prepares the minimum transmission unit in each TSN network and 5G-URLLC slice.

[0061] Periodic frame and event frame statistics: The edge nodes measured: the average generation rate of periodic frames is 100 frames / s, the average generation rate of event frames is 10 frames / s; the average frame size is 600B; and the time slot capacity is 1200B.

[0062] Unified time slot table generation: The number of time slots is 160 slots, the data volume of the periodic frame is 1200B, the number of time slots allocated to the periodic frame is 1, the number of time slots allocated to the event frame is 48, and the remaining elastic area time slots are 111.

[0063] This forms the time slot schedule: Slot0 → Cycle frame (highest priority) Slot1…48→Event frame (reserved area) Slot49…159 → Event frame (flexible area, lowest priority) TSNGCL configuration: Map the above 160 time slots to the IEEE802.1Qbv gating list: This mapping ensures that Slot0 always releases periodic frames, and the rest release event frames, and they correspond one-to-one with the 5G timeslot numbers.

[0064] 5G-URLLC slice resource reservation: For each event frame time slot, it is assumed that the average load is 600B, the target reliability is 99.999%, and the probability of a single transmission failure is 0.1%.

[0065] Due to the small traffic volume, only 2 RBs are reserved for the periodic frame time slot to save slice resources. The slice controller sends the redundancy times and RB reservation configuration to achieve time slot-slice-redundancy coordination.

[0066] Issue and Execution: According to the time slot table, the edge attaches timestamps to 10×8 scheduling instructions (8 control cycles) and forwards them to the device. After the local clock of the device reaches the timestamp, precise output is performed and feedback is given.

[0067] Closed-loop adaption: The cloud center counts 8×10 = 80 instructions, the average round-trip delay is 1.3 ms, and the actual success rate is 99.9995%, all of which meet the requirements and there is no need to adjust ζ.

[0068] When a sudden failure occurs, the actual success rate drops to 99.98%. The system automatically adjusts ζ from 0.3 to 0.35 and regenerates a new scheduling table.

[0069] By adopting the same numbered and prioritized time slot scheduling table between TSN and 5G network elements, there is no need for multiple mappings or manual alignment.

[0070] The beneficial effects of this application are as follows: Drive the TSN gating and 5G-URLLC slice redundancy synchronization, and issue time slots and radio resources in one key to ensure strict synchronization and consistent performance between the wired and wireless segments.

[0071] Based on the real-time round-trip delay and success rate indicators, dynamically adjust the event frame reservation ratio ζ, immediately recalculate the time slot table, and balance the certainty of the periodic frame and the elasticity guarantee of the event frame.

[0072] Embodiment 2. The cloud-edge device of the virtual power plant edge computing cluster is applied to the dynamic aggregation management method of the virtual power plant edge computing cluster according to any one of the above. It is characterized by including: A clock synchronization module for performing clock synchronization processing on the virtual power plant cloud control center and the edge computing cluster nodes; A data acquisition module for respectively counting the average generation rate of the periodic frames and the average generation rate of the event frames based on the communication records between the cloud control center and each edge computing cluster node within the historical N control cycles; A time slot scheduling module for respectively dividing fixed time slots for the periodic frames and dividing reserved time slots for the event frames according to the average generation rate of the periodic frames, the average generation rate of the event frames, the preset time slot capacity, and the event frame reservation ratio, so as to obtain a time slot scheduling table; A time slot matching module for corresponding each time slot number to the switch gating list according to the time slot scheduling table, and generating the window opening time and duration in sequence according to the time slot order; A resource allocation module for reserving resource blocks for each time slot in the 5G-URLLC network slice and setting a redundant transmission mechanism according to the data load and the target reliability rate of each time slot in the time slot scheduling table; An instruction generation module, configured to map the device energy scheduling instruction generated by the cloud control center to the corresponding time slot in the time slot scheduling table, and generate an energy scheduling instruction with a timestamp; A device management module, configured to forward the received energy scheduling instruction by the edge computing cluster node to the corresponding device, and the device performs actual power output according to the energy scheduling instruction and the maximum power of the device; A parameter optimization module, configured to record the round-trip delay and reception success rate of the energy scheduling instruction, and compare them with the preset delay index and reliability index; when any index does not meet the preset requirements, adjust the event frame reservation ratio according to the preset adjustment coefficient, and regenerate the time slot scheduling table.

[0073] The above describes the embodiments of this embodiment, but this embodiment is not limited to the above specific implementation manners. The above specific implementation manners are merely illustrative and not restrictive. Under the inspiration of this embodiment, those of ordinary skill in the art can also make many forms, all of which fall within the protection scope of this embodiment.

Claims

1. A dynamic aggregation management method for a virtual power plant edge computing cluster, characterized in that Including: Step 1: Perform clock synchronization processing on the virtual power plant cloud control center and the edge computing cluster nodes; Step 2: Based on the communication records between the cloud control center and each edge computing cluster node within the historical N control cycles, respectively calculate the average generation rate of cycle frames and the average generation rate of event frames; Step 3: According to the average generation rate of cycle frames, the average generation rate of event frames, the preset time slot capacity, and the event frame reservation ratio, respectively allocate fixed time slots for cycle frames and allocate reserved time slots for event frames to obtain a time slot scheduling table; Step 4: According to the time slot scheduling table, map each time slot number to the switch gating list, and generate the window opening time and duration in the order of time slots; Step 5: Based on the data load and target reliability rate of each time slot in the time slot scheduling table, reserve resource blocks for each time slot in the 5G-URLLC network slice and set a redundant transmission mechanism; Step 6: Map the device energy scheduling instructions generated by the cloud control center to the corresponding time slots in the time slot scheduling table to generate energy scheduling instructions with timestamps; Step 7: The edge computing cluster nodes forward the received energy scheduling instructions to the corresponding devices, and the devices perform actual power output according to the energy scheduling instructions and the maximum power of the devices; Step 8: Record the round-trip delay and reception success rate of the energy scheduling instructions, and compare them with the preset delay index and reliability rate index; when any index does not meet the preset requirements, adjust the event frame reservation ratio according to the preset adjustment coefficient and regenerate the time slot scheduling table.

2. The dynamic aggregation management method of the virtual power plant edge computing cluster according to claim 1, wherein, Based on the communication records between the cloud control center and each edge computing cluster node within the historical N control cycles, respectively calculate the average generation rate of cycle frames and event frames, including: The cycle frame represents: in the communication record, the device status monitoring data sent at fixed time intervals; The event frame represents: in the communication record, the non-periodic burst data; The calculation formula for the average generation rate of cycle frames is as follows: The calculation formula for the average generation rate of event frames is as follows: Among them, represents the average generation rate of periodic frames, represents the average generation rate of event frames, represents the number of control cycles, represents the duration of each control cycle, represents the number of periodic frames in the nth control cycle, represents the number of event frames in the nth control cycle.

3. The dynamic aggregation management method of the virtual power plant edge computing cluster according to claim 2, characterized in that According to the average generation rate of cycle frames, the average generation rate of event frames, the preset time slot capacity, and the event frame reservation ratio, respectively allocate fixed time slots for cycle frames and allocate reserved time slots for event frames to obtain a time slot scheduling table, including: Determine communication resource parameters, where the communication resource parameters include: a preset time slot capacity and an average frame size ; Determine the number of time slots within each control period , where the number of time slots is the ratio of the duration of the control period to the preset time slot duration ; Allocation of fixed time slots for cycle frames, including: Calculate the total number of cycle frames in each control period , ; Calculate the data volume of the periodic frame , , indicating the average size of the periodic frame; Determine the number of time slots allocated to a cycle frame , ; Allocation of reserved time slots for event frames, including: Load the event frame reservation ratio ζ, 0 ≤ ζ ≤ 1; Determine the number of time slots allocated to the event frame , ; Generate a time slot scheduling table, and the time slot scheduling table includes: cycle frame time slots and event frame time slot elastic time slots; among them, the priorities of the cycle frame time slots and the event frame time slots decrease in turn.

4. The dynamic aggregation management method of the virtual power plant edge computing cluster according to claim 3, wherein According to the time slot scheduling table, map each time slot number to the switch gating list, and generate the window opening time and duration in the order of time slots, including: For the s-th time slot in the time slot schedule, by multiplying the s-th time slot by a preset time slot duration the gating window opening offset of the s-th time slot is obtained; Use a preset time slot duration as the gating window duration for the s-th time slot; If the s-th time slot is less than the number of time slots allocated to the periodic frame , map the s-th time slot to the periodic frame queue mask; otherwise, map the s-th time slot to the event frame queue mask; Based on the combination of the gating window opening offset, the gating window duration, and the queue mask, form the gating list entries of the switch; Send all the gating list entries in the order of time slots and write them into the gating control list of the switch.

5. The dynamic aggregation management method for the virtual power plant edge computing cluster according to claim 4, characterized in that Based on the data load and target reliability rate of each time slot in the time slot scheduling table, reserve resource blocks for each time slot in the 5G-URLLC network slice and set a redundant transmission mechanism, including: Read the predicted data load and preset target reliability rate for each time slot from the time slot scheduling table; Determine the single - transmission failure probability of the 5G air interface without redundancy; Calculate the redundant transmission times based on the target reliability rate and the single - transmission failure probability; Determine the number of resource blocks required for the corresponding time slot based on the predicted data load of the time slot and the data volume carried by each resource block; Reserve the required number of resource blocks for each time slot in the 5G - URLLC network slice and execute the redundant transmission mechanism according to the redundant transmission times.

6. The dynamic aggregation management method of the virtual power plant edge computing cluster according to claim 5, characterized in that, Map the device energy scheduling instruction generated by the cloud control center to the corresponding time slot in the time slot scheduling table to generate a time - stamped energy scheduling instruction, including: Obtain the time slot corresponding to each device energy scheduling instruction generated by the cloud control center from the time slot scheduling table; Calculate the transmission timestamp of the device energy scheduling instruction based on the duration of the control period and the gating window opening offset corresponding to the time slot; Attach the transmission timestamp to the device energy scheduling instruction to generate a time - stamped energy scheduling instruction; Send the time - stamped energy scheduling instructions to the corresponding edge computing nodes in chronological order.

7. The dynamic aggregation management method of the virtual power plant edge computing cluster according to claim 6, wherein The edge computing cluster nodes forward the received energy scheduling instructions to the corresponding devices, and the devices perform actual power output according to the energy scheduling instructions and the device's maximum power, including: The edge computing cluster nodes forward the received time - stamped energy scheduling instructions to the corresponding devices; When the local clock of the device reaches the transmission timestamp in the time - stamped energy scheduling instruction, trigger the execution: The device multiplies the power ratio in the time - stamped energy scheduling instruction by the device's rated maximum power to calculate the actual output power; the device encapsulates the actual output power and the confirmation timestamp into an execution confirmation message and feeds back the execution confirmation message to the edge computing cluster nodes.

8. The dynamic aggregation management method of the virtual power plant edge computing cluster according to claim 7, characterized in that, Record the round - trip delay and reception success rate of the energy scheduling instruction and compare them with the preset delay index and reliability rate index; When any index does not meet the preset requirements, adjust the event frame reservation ratio according to the preset adjustment coefficient and regenerate the time slot scheduling table, including: In each control period, record the transmission timestamp of the time - stamped energy scheduling instruction and the confirmation timestamp of the device, and count the number of execution confirmation messages received by the cloud control center; Calculate the average round - trip delay and actual reception success rate of the time - stamped energy scheduling instruction; Compare the average round - trip delay with the preset maximum allowable round - trip delay, and the actual reception success rate with the target reliability rate; When the average round - trip delay is greater than the maximum allowable round - trip delay, or the actual reception success rate is less than the target reliability rate, then adjust the event frame reservation ratio ζ; Based on the adjusted event frame reservation ratio ζ, return to step 3 to regenerate the time slot scheduling table.

9. The cloud-edge device of the virtual power plant edge computing cluster is applied to the dynamic aggregation management method of the virtual power plant edge computing cluster according to any one of claims 1-8, and is characterized in that Including: A clock synchronization module for performing clock synchronization processing on the virtual power plant cloud control center and the edge computing cluster nodes; A data acquisition module for respectively counting the average generation rate of cycle frames and the average generation rate of event frames based on the communication records between the cloud control center and each edge computing cluster node in the historical N control periods; A time slot scheduling module, which is used to divide fixed time slots for periodic frames and reserved time slots for event frames respectively according to the average generation rate of the periodic frames, the average generation rate of the event frames, the preset time slot capacity and the reserved ratio of the event frames, so as to obtain a time slot scheduling table; A time slot matching module, which is used to correspond each time slot number with the switch gating list according to the time slot scheduling table, and generate the window opening time and duration in the order of time slots; A resource allocation module, which is used to reserve resource blocks for each time slot and set a redundant transmission mechanism in the 5G-URLLC network slice according to the data load and the target reliability rate of each time slot in the time slot scheduling table; An instruction generation module, which is used to map the device energy scheduling instruction generated by the cloud control center to the corresponding time slot in the time slot scheduling table to generate an energy scheduling instruction with a timestamp; A device management module, which is used to forward the received energy scheduling instruction by the edge computing cluster node to the corresponding device, and the device outputs the actual power according to the energy scheduling instruction and the maximum power of the device; A parameter optimization module, which is used to record the round-trip delay and reception success rate of the energy scheduling instruction, and compare them with the preset delay index and reliability rate index; when any index does not meet the preset requirements, adjust the reserved ratio of the event frames according to the preset adjustment coefficient, and regenerate the time slot scheduling table.

Images