Edge data aggregation and real-time uploading scheduling system for virtual power plant monitoring terminal

Abstract

The application provides an edge data aggregation and real-time uploading scheduling system for a virtual power plant monitoring terminal, and belongs to the technical field of power system automation. The application exchanges and collects data through short-distance communication between terminals, carries out mutual verification based on physical constraint rules such as power conservation, paired deviation and power flow direction, generates a comprehensive credibility level, realizes collaborative credibility grading of the collected data content itself, and avoids the limitation of single dependence on equipment identity authentication.

Description

Technical Field

[0001] This invention belongs to the field of power system automation technology, specifically relating to an edge data aggregation and real-time uploading scheduling system for virtual power plant monitoring terminals. Background Technology

[0002] Virtual Power Plants (VPPs) participate in grid dispatch by aggregating heterogeneous resources such as distributed power sources, energy storage systems, and controllable loads. Their large-scale development highly depends on a cloud-edge-device collaborative architecture. On the device side, massive monitoring terminals need to periodically collect measurement data such as voltage, current, and power and upload them to edge nodes; on the edge side, edge nodes complete data aggregation and upload it to the cloud. With the surge in the number of connected terminals, large-scale concurrent data transmission and highly reliable data aggregation have become key technical bottlenecks restricting system reliability.

[0003] In existing technologies, such as patent application CN121462208A (Data Edge Access Method and System, Computing Device, and Storage Medium), the edge computing device performs trustworthiness detection on the sensing device, establishing a secure data channel when the identity is trustworthy and the network system is intact. This technical solution relies on one-way authentication from the edge node to the terminal for trustworthiness detection, focusing on verifying the device's identity legitimacy. It fails to address the issue of data quality verification before the terminal-collected data is transmitted to the edge node. When the terminal device itself experiences abnormal data collection or the data is contaminated at the terminal, the one-way authentication by the edge node cannot perceive the reliability of the data content.

[0004] For example, patent application CN121727049A, "Multi-dimensional Resource Dynamic Aggregation and Optimization Control Method for Cloud-Edge Collaboration in Virtual Power Plants," establishes a multi-dimensional feature index system and aggregates complementary resources into virtual units based on multi-objective optimization. The aggregation object of this technical solution is virtual power plant resources (such as distributed energy sources and energy storage devices), and the aggregation is based on the static attributes and dynamic response characteristics of the resources, belonging to the resource management level of aggregation. However, this technical solution does not address issues such as the reliability classification of terminal-collected data and conflicts arising from concurrent uploads by multiple terminals. Summary of the Invention

[0005] The purpose of this section is to outline some aspects of the embodiments of the present invention and to briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section, as well as in the abstract and title of the present application, to avoid obscuring the purpose of this section, the abstract and title of the invention. Such simplifications or omissions shall not be used to limit the scope of the present invention.

[0006] In view of the aforementioned existing problems, the present invention is proposed.

[0007] Therefore, the technical problem solved by this invention is that the virtual power plant monitoring terminal lacks the technical capability to conduct a credibility-level assessment of the collected data content itself during the data acquisition and uploading process.

[0008] To address the aforementioned technical problems, the present invention provides the following technical solution: An edge data aggregation and real-time upload scheduling system for virtual power plant monitoring terminals includes: a terminal-side mutual verification module, used to exchange and verify collected data with adjacent terminals via short-range communication, obtain mutual verification marks, and generate a comprehensive credibility level based on the number and level of the mutual verification marks; a terminal-side backoff scheduling module, used to calculate a backoff base value based on the comprehensive credibility level, and upload collected data packets to the edge nodes after superimposing the backoff base value on a unified reference time base point indicated by the start trigger beacon broadcast by the edge nodes, wherein the start trigger beacon contains a round number; and an edge-side classification and aggregation module, used to receive the collected data packets and bind each data packet to its corresponding round number, assigning data belonging to the same round number to the same aggregation period. The system categorizes data into trusted, questionable, and untrusted data based on a comprehensive credibility level and communication quality flags. Data carrying compensation channel identifiers is not subject to the comprehensive credibility level and is directly classified as questionable data. The trusted data is aggregated to generate a baseline aggregation result, and the boundary floating range is calculated for the questionable data. The edge-side feature extraction module extracts features from the aggregation results of multiple consecutive aggregation cycles along a time window sequence, retains the aggregation values ​​of the start and end cycles, extracts the maximum upward and downward ramp values ​​between adjacent cycles, and superimposes the boundary floating ranges of the two cycles corresponding to each ramp value as the floating range of that ramp value. The edge-side upload module uploads the extracted feature tuples to the cloud.

[0009] As a preferred embodiment of the present invention, the terminal-side mutual verification module includes: a tag acquisition unit, used to exchange collected data with neighboring terminals through short-range communication, verify the data of neighboring terminals according to preset physical constraint rules, and obtain mutual verification tags returned by each neighboring terminal, wherein the levels of the mutual verification tags include trusted, questionable, and untrustworthy; and a level determination unit, used to determine the overall trustworthiness level of the terminal itself based on the number and level of the mutual verification tags; when the number of mutual verification tags is three, the level determination unit determines the overall trustworthiness level according to the majority principle; when the number of mutual verification tags is two, if both mutual verification tags are trusted, it is determined to be a high level; if one is trusted and one is questionable, it is determined to be a medium level; otherwise, it is determined to be a low level; when the number of mutual verification tags is less than two, it is forcibly determined to be a low level and an isolated status identifier is reported to the edge node.

[0010] As a preferred embodiment of the present invention, the terminal-side backoff scheduling module includes: a backoff base value calculation unit, used to calculate a backoff base value based on the comprehensive confidence level, wherein a high comprehensive confidence level corresponds to a first backoff interval, a medium level corresponds to a second backoff interval, and a low level corresponds to a third backoff interval, wherein the upper limit of the first backoff interval is less than the lower limit of the second backoff interval, and the upper limit of the second backoff interval is less than the lower limit of the third backoff interval; a micro-jitter overlay unit, used to add random micro-jitter values ​​to the backoff base value; and an upload trigger unit, used to, after receiving the start trigger beacon broadcast by the edge node, use the unified reference time base point indicated by the start trigger beacon as the timing start point, wait for the total duration of the backoff base value and the micro-jitter value overlay, and then upload the acquisition data packet to the edge node; the edge node broadcasts the start trigger beacon at a fixed period, wherein the lower limit of the fixed period is determined by the sum of the regular receiving window duration, the compensation micro-slot duration, and the protection interval duration.

[0011] As a preferred embodiment of the present invention, the terminal-side backoff scheduling module further includes a ready flag management unit, which is used to set the ready flag to a valid state after the comprehensive trust level calculation is completed, and to clear the ready flag after receiving the link layer confirmation response returned by the edge node; if the link layer confirmation response is not received within a preset timeout period, the ready flag remains valid and the process switches to the compensation upload mode; the compensation upload mode is as follows: within the compensation micro-slot opened by the edge node, the upload is retried using a carrier sense multiple access and collision avoidance mechanism; if the link layer confirmation response returned by the edge node is still not received, the ready flag is forcibly cleared at the end of the compensation micro-slot.

[0012] As a preferred embodiment of the present invention, the edge-side classification and aggregation module includes: a window alignment unit, used to uniformly assign all terminal data belonging to the same round number to the same aggregation period, ignoring the upload time deviation of each terminal data within the period; a data classification unit, used to divide the window-aligned data into three categories: reliable data, questionable data, and unreliable data according to the comprehensive reliability level, communication quality flag, and compensation channel identifier; data carrying the compensation channel identifier is not restricted by the comprehensive reliability level and is directly classified into the questionable data category; and a benchmark aggregation unit, used to directly incorporate all reliable data into the calculation to generate a benchmark aggregation result. The floating range calculation unit retrieves historical fluctuation statistics from the terminal, uses the current reading of the questionable data as a benchmark, and calculates the maximum positive and maximum negative deviations of the data from the benchmark aggregation result to form a boundary floating range. The unreliable data processing unit, when the physical conservation condition is met, calculates the energy balance difference using reliable data from the same bus or transformer area according to Kirchhoff's current law to obtain a substitute value for the unreliable data, and classifies the substitute value into the questionable data category to participate in the boundary floating range calculation. When the physical conservation condition is not met, the unreliable data is included in the uncovered quantity field, which is appended to the benchmark aggregation result.

[0013] In a preferred embodiment of the present invention, the historical fluctuation statistics retrieved by the floating range calculation unit are the statistical values ​​of the upper limit of the fluctuation of the terminal's historical valid data; the historical valid data includes historical readings that are classified as reliable data and questionable data, but does not include historical readings that are classified as unreliable data.

[0014] As a preferred embodiment of the present invention, the edge-side feature extraction module includes: an anchor point extraction unit, used to extract the starting period aggregation value and the ending period aggregation value from a sequence of multiple consecutive aggregation periods, and bind the corresponding boundary floating intervals respectively; a ramp calculation unit, used to calculate the difference between adjacent period aggregation values, extract the maximum rising ramp value and the maximum falling ramp value according to the absolute value of the difference, and record the ending period number corresponding to each maximum value; an interval propagation unit, used to superimpose the boundary floating intervals of the two source periods corresponding to each ramp value, and use the superimposed interval as the floating interval of the ramp value; when there is a low confidence aggregation result in the two source periods, the superimposed interval is expanded outward to compensate, and when both source periods are low confidence, the superimposed interval is set to the full range interval; and a feature encapsulation unit, used to combine the starting aggregation value and floating interval, the ending aggregation value and floating interval, the maximum rising ramp value and floating interval and the ending period number, and the maximum falling ramp value and floating interval and the ending period number to form a feature tuple.

[0015] In a preferred embodiment of the present invention, the interval propagation unit adds the negative absolute values ​​of the boundary floating intervals of the two source periods as the negative boundary of the superimposed interval, and adds the positive absolute values ​​of the boundary floating intervals of the two source periods as the positive boundary of the superimposed interval; the edge-side feature extraction module further includes an interval merging unit, used to perform horizontal comparison of floating intervals of different aggregation groups; when the upper and lower bound deviation values ​​of the two floating intervals do not exceed a preset percentage threshold of the larger interval value, they are determined to be of the same magnitude and merged, and the merged interval is used as a general floating interval field, along with an index list of aggregation groups sharing the interval.

[0016] In a preferred embodiment of the present invention, the duration of the compensation micro-slot is dynamically adjusted by the edge node according to the collision rate; when the collision rate continuously exceeds a preset probability threshold, the terminal is instructed in the next round of broadcast beacon to extend the duration of the compensation micro-slot to a preset extended duration.

[0017] In a preferred embodiment of the present invention, if the ready flag is detected as invalid when the next round of start trigger beacon arrives, the terminal-side backoff scheduling module triggers the terminal-side mutual authentication module to perform a new round of data collection and mutual authentication; if the ready flag is detected as valid, the terminal-side mutual authentication module is forcibly reset and then re-triggered.

[0018] The beneficial effects of this invention are as follows: Compared with the prior art, the technical effects of this invention are as follows: This invention collects data through short-range communication between terminals, performs mutual verification based on physical constraints such as power conservation, pairwise deviation, and power flow direction, and generates a comprehensive credibility level. This achieves collaborative credibility grading of the collected data content itself, avoiding the limitations of relying solely on device identity authentication. Furthermore, based on the comprehensive credibility level, differentiated backoff scheduling is implemented for terminals, with high-credibility data occupying the channel first to reduce the collision rate. At the same time, the edge side calculates the boundary fluctuation range for questionable data and extracts feature tuples such as the maximum ramp value along the time window and uploads them to the cloud, significantly compressing the uplink data volume while retaining the key characteristics of load fluctuation. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the terminal-side mutual verification process shown in this invention.

[0020] Figure 2 This is a flowchart of the terminal-side backoff scheduling process shown in this invention.

[0021] Figure 3 This is a flowchart of edge-side data classification and feature extraction as shown in this invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of this invention. The embodiments described in this application are merely some embodiments of this invention, and not all embodiments. Based on the spirit of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of this invention.

[0023] like Figures 1-3 As shown, the edge data aggregation and real-time upload scheduling system for virtual power plant monitoring terminals of the present invention includes: a terminal-side mutual verification module, used to exchange and verify collected data with adjacent terminals through short-range communication, obtain mutual verification marks, and generate a comprehensive credibility level based on the number and level of mutual verification marks; a terminal-side backoff scheduling module, used to calculate a backoff base value based on the comprehensive credibility level, and upload the collected data packets to the edge nodes after superimposing the backoff base value on the unified reference time base point indicated by the start trigger beacon broadcast by the edge nodes, wherein the start trigger beacon contains a round number; and an edge-side classification and aggregation module, used to receive the collected data packets and bind each data packet to its respective round number, and assign data belonging to the same round number to the same aggregation period. Based on the comprehensive credibility level and communication quality flag, the data is divided into credible data, questionable data, and untrustworthy data. Data carrying compensation channel identifiers is not restricted by the comprehensive credibility level and is directly classified as questionable data. The credible data is aggregated to generate a benchmark aggregation result, and the boundary floating range is calculated for the questionable data. The edge-side feature extraction module is used to extract features from the aggregation results of multiple consecutive aggregation cycles along the time window sequence, retain the aggregation values ​​of the start and end cycles, extract the maximum upward and downward ramp values ​​between adjacent cycles, and superimpose the boundary floating ranges of the two cycles corresponding to each ramp value as the floating range of that ramp value. The edge-side upload module is used to upload the extracted feature tuples to the cloud.

[0024] In some embodiments, the terminal-side mutual authentication module is triggered at the beginning of the acquisition cycle of each monitoring terminal. This module establishes a temporary point-to-point communication link with adjacent terminals in the same area or on the same feeder via short-range communication protocols (including but not limited to ZigBee, BLEMesh, or LoRa) to exchange their current acquisition data. The short-range communication and the uplink communication from the terminal to the edge node are performed in a time-division multiplexing manner, and the mutual authentication communication is only performed in the idle window after the regular receiving window of the uplink communication is closed and before the next round of triggering beacon broadcast.

[0025] The exchanged data includes at least: terminal identifier, timestamp of the acquisition time, physical measurement values ​​(such as power, current, and voltage amplitude), and quality flag bits of the measurement values. After mutual verification is completed, the terminal generates a comprehensive trust level based on the number and level of the received mutual verification flags, which is used for subsequent backoff scheduling and edge-side classification and aggregation.

[0026] The terminal-side mutual verification module includes a tag acquisition unit and a level determination unit.

[0027] In some embodiments, the tag acquisition unit sends mutual verification request frames to several terminals that are physically adjacent to the terminal, located in the same transformer substation, and connected to the same bus branch node (the same electrical connection point) according to the communication neighborhood table issued by the edge node. The mutual verification request frame carries the collected data of the terminal. The communication neighborhood table is generated by the edge node according to the transformer substation topology during system initialization and is dynamically updated and reissued to each terminal when the topology changes.

[0028] Among them, the branch nodes of the same bus are uniformly determined by the edge nodes during system initialization based on the transformer area topology (including but not limited to the parent node number, connection phase, and feeder number of each terminal) and distributed to each terminal as the basis for mutual verification neighbor selection. When a terminal has no adjacent terminals under the same bus branch node, it is regressed to the adjacent terminals under the same phase (one of the three phases A / B / C) as the mutual verification object, and a phase matching identifier is added to the mutual verification mark. When there are no adjacent terminals under the same phase, the mark acquisition unit skips the mutual verification verification and directly sets the overall credibility level of the terminal to a low level, and reports the isolated status identifier to the edge nodes. The terminal performs uploading at a low level in the backoff scheduling.

[0029] Upon receiving the request, the neighboring terminal compares its own data with its own data, generates a mutual verification tag based on preset physical constraint rules, and returns it to the requesting terminal.

[0030] In some embodiments, the preset physical constraint rules are verified based on the principle that the algebraic sum of the data power within the same distribution area should equal the total incoming line power (an engineering approximation of Kirchhoff's current law under steady state). Specifically, let the measured value of this terminal be... The measurement values ​​of adjacent terminals The total incoming line measurement value of this transformer area is All the above power measurements use the same power unit, such as watts (W) or kilowatts (kW). The specific unit is configured uniformly by the edge nodes. All power values ​​involved in the same formula calculation are converted to the same unit by the edge nodes before the calculation. The absolute value of the deviation between the algebraic sum of all terminal measurements in this transformer area and the total incoming line measurement should meet the following requirements: ; in, This represents the total number of terminals participating in the mutual verification within this area. For the first Measurement values ​​of each terminal, The preset power conservation tolerance threshold is taken as the greater of 2% of the rated capacity of the transformer in the distribution area and three times the combined standard uncertainty of all participating terminal measuring devices. The inherent error of each measuring device is calculated as the standard uncertainty corresponding to the nominal accuracy class on the device nameplate (maximum permissible error divided by...). The value is taken as a uniformly distributed value. The combined standard uncertainty is obtained by taking the square root of the sum of the squares of the standard uncertainties of each terminal measuring device. If the above formula holds, the current distribution area is determined to be in a power balance state.

[0031] In some embodiments, the logic for determining the data of a neighboring terminal on this terminal is as follows: If If the measured value of this terminal does not disrupt the balance after being substituted into the above conservation check, then a reliable flag is returned; where the total incoming line measured value of the transformer area required to participate in the conservation check is... The total number of terminals N in the distribution area is obtained by the adjacent terminals from the system parameters broadcast by the edge node in the previous round. At the beginning of each aggregation period, the edge node sends the parameters to all terminals in the distribution area through the downlink broadcast channel. The total incoming line measurement value is obtained by the edge node directly from the metering device at the total incoming line of the distribution area through the local data interface, without relying on the aggregation result of the data reported by the terminals.

[0032] like Then, it is further determined whether the deviation direction is consistent with the known power flow direction of the transformer area. The determination method is: using the total incoming power direction flag of the transformer area broadcast by the edge node. Based on this, +1 indicates that power flows from the bus to the feeder, and -1 indicates that power flows from the feeder to the bus. If the directions are consistent, the direction is considered to be the same; otherwise, it is considered to be inconsistent. If the directions are consistent, a question mark is returned; otherwise, an untrusted mark is returned. The deviation threshold between paired terminals is set to three times the combined standard uncertainty of the two terminal measuring devices. The combined standard uncertainty is obtained by combining the measurement uncertainty based on the accuracy class indicated on the nameplate of each terminal measuring device and the measurement uncertainty given in the calibration certificate, according to the uncertainty propagation law. Before the system goes online, it is uniformly configured by the edge node and distributed to each terminal.

[0033] In some embodiments, the mutual authentication token is encoded in two binary codes and encapsulated in a short-range communication response frame, where trusted corresponds to code 00, doubtful corresponds to code 01, and untrusted corresponds to code 10. After receiving the response frame, the terminal parses out the mutual authentication tokens and their levels returned by each neighboring terminal.

[0034] In some embodiments, the level determination unit first counts the total number of received mutual verification tags. and according to The following decision logic is executed for each value: (1) When When determining the overall credibility level, the majority rule applies. Let the number of times the three mutual verification markers appear credible be . The number of times unreliable information appeared was... The number of times doubtful was found was And the sum of the three is 3. Then the overall credibility level L is determined as follows: If If so, then L is a higher level; If the condition is met, then L is a low level; otherwise, L is a medium level. These other conditions include... , or , or , Such situations.

[0035] (2) When M=2, the judgment rules corresponding to the two mutual verification mark level combinations are as follows: Table 1 Judgment Rules:

[0036] Trustworthy Trustworthy High level Trustworthy Questionable Intermediate level Questionable Questionable low level Trustworthy Unreliable low level Questionable Unreliable low level Unreliable Unreliable low level

[0037] Equivalently, the overall credibility level L can be expressed as follows: if both labels are credible, then L is high; if one is credible and the other is questionable, then L is medium; if all other combinations are true, then L is low.

[0038] The judgment rule adopts a conservative strategy when the amount of mutual verification information is limited (only two neighbors). That is, as long as there is an untrustworthy or two questionable information, it is directly judged as low-level to reduce the probability of low-quality data entering the edge-side benchmark aggregation. At the same time, it retains a moderate trust in some effective information through a unique intermediate-level transition combination (trustworthy + questionable).

[0039] (3) When M<2, i.e., a mutual authentication flag is received or no mutual authentication flag is received, it is judged as low level and an isolated state identifier is generated. The isolated state identifier is a status report frame containing the terminal ID, the area number where it is located, and a count of not receiving enough mutual authentication responses. The count starts from zero and increments by 1 for each round where M<2. When the count exceeds the preset upper limit (e.g., 5 consecutive rounds), the terminal actively reports a warning message to the edge node, indicating that the terminal may be at the network coverage edge or in a communication module failure state.

[0040] It should be noted that, in normal operating mode, the number of mutually verifiable terminals within the short-range communication neighborhood of the terminal-side mutual authentication module in this embodiment of the invention is typically 2 to 3 (limited by the physical topology and communication coverage within the same area). Therefore, the above-mentioned determination rules are fully disclosed using the three typical cases of M=3, M=2, and M<2 as preferred implementation methods. When M>3 occasionally occurs in engineering, those skilled in the art can fully extend the determination based on the majority decision principle (such as using a weighted scoring rule), which is a conventional variation under the design concept of this invention and does not affect the full disclosure of the core solution of this invention.

[0041] In some embodiments, the overall trust level is stored in the terminal local register as an enumeration type: high level is mapped to the numeric value 2, medium level to 1, and low level to 0. This mapping value is directly read by the subsequent backoff base value calculation unit as an index parameter for backoff interval selection.

[0042] In some embodiments, the terminal-side backoff scheduling module is built into the communication coprocessor of each monitoring terminal, and interacts with the terminal-side mutual verification module through a shared register to exchange comprehensive trust level data.

[0043] The terminal-side backoff scheduling module and the terminal-side mutual authentication module follow the following timing relationship: During the idle time window between the closing of the previous round's regular reception window and the broadcast of the next round's start trigger beacon, the terminal-side mutual authentication module is triggered to perform data exchange and mutual authentication with neighboring terminals, and to calculate the comprehensive credibility level. After the comprehensive credibility level calculation is completed, the terminal-side backoff scheduling module sets the ready flag to an active state and writes the acquired data packets into the transmission buffer. Subsequently, the terminal continuously waits for and receives the next round's start trigger beacon broadcast by the edge node. This beacon only provides a unified reference time base point as the starting reference for transmission timing and does not trigger new acquisition actions. That is, the mutual authentication process and beacon reception are executed serially on the timeline, with mutual authentication completed before the beacon arrives, and only the backoff timer starting after the beacon arrives.

[0044] This module uses the timestamp carried by the start trigger beacon periodically broadcast by the edge nodes as the global synchronization benchmark, and adopts a scheduling strategy that combines hierarchical backoff intervals with random micro-jitter to control the transmission timing of each terminal within the normal receiving window.

[0045] The terminal-side backoff scheduling module includes a backoff base value calculation unit, a micro-jitter overlay unit, an upload trigger unit, and a ready flag management unit.

[0046] In some embodiments, the backoff baseline calculation unit maps to three sets of non-overlapping backoff intervals based on the overall confidence level. Let the minimum backoff time slot be... (Unit: milliseconds), then the higher level corresponds to the first retreat interval. The medium level corresponds to the second retreat zone. The lower level corresponds to the third retreat zone. ,in, , All an integer multiple of, and satisfying As a preferred example, When the value is 1 millisecond, , The values ​​are 2, 5, and 10 respectively, and these values ​​can be scaled proportionally according to the total number of terminals in the area.

[0047] Backoff base value The selection method is as follows: ; in, Indicates a closed interval An integer value is randomly selected from the uniformly distributed inner range. The upper limit of the first backoff interval. Less than the lower limit of the second retreat interval The upper limit of the second interval Less than the lower limit of the third interval This ensures that the longest backoff time of a high-confidence terminal is shorter than the shortest backoff time of a medium-confidence terminal, and the longest backoff time of a medium-confidence terminal is shorter than the shortest backoff time of a low-confidence terminal. This allows high-confidence data to occupy the first time slot of the regular reception window first, reducing the probability of high-value data colliding due to channel contention.

[0048] In some embodiments, after the backoff base value is selected, the micro-jitter overlay unit independently generates a random micro-jitter value. And it follows a uniform distribution: ; in, It is a positive integer, and This ensures that the amplitude of the micro-jitter does not exceed one-tenth of the minimum retreat interval length. Total retreat time for: ; The introduction of this micro-jitter value means that even if multiple terminals of the same level randomly select the same backoff base value, there will still be a sub-slot-level random offset in the actual transmission time, thereby transforming synchronous competition into asynchronous competition and further reducing the probability of physical layer frame collision.

[0049] In some embodiments, the upload triggering unit continuously listens for the start trigger beacon broadcast by the edge node. The start trigger beacon is a broadcast frame from the Media Access Control layer, whose frame payload includes a round number R (represented by a 16-bit unsigned integer) and a timestamp. The timestamp indicates the precise time when the edge node sent the beacon, serving as a unified reference time base.

[0050] After the terminal receives the beacon, it uses Synchronize and calibrate the local clock, and The backoff timer is started as the timing start point. If the terminal fails to receive the start trigger beacon in the current round, it continues to run based on the beacon time of the previous round maintained by the local clock and reports the beacon loss count to the upper layer; if more than 3 rounds of consecutive losses occur, the terminal will actively withdraw from the upload attempt in the current round and wait for the beacon to be restored in the next round.

[0051] The timer is initially set to the total backoff duration. (Unit: number of time slots) When the timer reaches zero, the uplink trigger unit sends a transmit trigger pulse to the RF front-end, and the control terminal sends uplink data frames at this moment. The frame header of the uplink data frame carries the current round number R, so that the edge nodes can assign the data frames of each terminal to the correct aggregation period after receiving them.

[0052] In some embodiments, edge nodes are at a fixed period The broadcast start trigger beacon, and its lower limit of duration is determined by the following formula: ; In terms of timing, the following are the conventional receiving windows. Compensation micro-time slots and protection interval Arranged in order.

[0053] in, The normal reception window duration is not less than the maximum backoff time of all terminals in this round. The sum of the single-frame air transmission duration and the inter-frame interval, To compensate for the micro-slot duration and accommodate retransmission attempts in compensated upload mode, the single-frame air transmission duration and inter-frame interval both adopt the original duration values ​​specified in the adopted short-range communication protocol standard, without further scaling; in participating During the summation operation, the sum of the single-frame air transmission duration and the inter-frame interval is calculated using the basic time slot. The multiples thereof are rounded up and included as a whole to ensure that the timing of the standard protocol is not disrupted. To protect the interval duration, it is used to offset the clock drift accumulated by the crystal oscillator frequency deviation of each terminal and the difference in spatial propagation delay of the wireless signal, and to prevent the data frames of the previous round from infiltrating into the receiving window of the next round and causing cross-round interference.

[0054] The unified reference time base point is obtained by the edge nodes through GPS / BeiDou time synchronization, with a timing accuracy of no less than ±1 millisecond.

[0055] In some embodiments, the duration of the compensation micro-slot is dynamically adjusted by the edge node based on the collision rate; the edge node calculates the collision rate CR (the ratio of the number of unsuccessfully received frames to the total number of transmitted frames) for the current round, and if the CR exceeds a preset probability threshold η for three consecutive rounds, the terminal is instructed to... Extend the duration to 1.5 to 2 times the current duration, and remain in effect until the CR falls to a certain level for three consecutive rounds. The following will revert to the base duration; The value ranges from 30% to 50%, determined by the edge node based on the number of currently online terminals in the area. The following adaptive calculation method is used: ; in, The maximum terminal capacity designed for the distribution area (pre-configured by edge nodes, typically 50) will be calculated as follows: The threshold is truncated to the 30%~50% range; the more terminals there are, the more intense the channel competition, so the threshold is lowered accordingly to trigger the compensation micro-timeslot extension earlier, and vice versa to avoid over-adjustment.

[0056] The preset extension duration is 1.5 to 2 times the current compensation micro-slot duration. The above value range is determined based on the collision rate tolerance recommended by the IEEE 802.15.4 standard and the retransmission slot requirements under the typical number of terminals in a distribution area (20 to 50 terminals). The new value is carried in the frame payload of the next round's start trigger beacon. The terminal will take effect in the next round after receiving the beacon, and the currently ongoing compensation upload process will not be affected. The initial value of the terminal-side contention backoff counter and the maximum contention window remain unchanged during the change.

[0057] The collision rate is the ratio of the number of data frames that were not successfully received in this round to the total number of data frames transmitted.

[0058] In some embodiments, the ready flag is a one-bit atomic operation register in the terminal's local memory, where logic 1 indicates a valid state (unconfirmed data awaiting upload) and logic 0 indicates an invalid state. After the terminal-side mutual authentication module completes the comprehensive trust level calculation and writes the collected data packet into the transmission buffer, this unit immediately sets the ready flag to valid.

[0059] After the upload triggering unit confirms the data frame has been sent at the physical layer, it starts an acknowledgment waiting timer, the timeout period of which is specified in the original text. Set as: ; in, This refers to the one-way airborne propagation delay of the signal at its maximum communication distance. The maximum processing latency required for edge node reception, processing, and response frame generation. The duration of over-the-air transmission for the link layer acknowledgment frame. If in If an ACK frame is received from the edge node, the upload is considered successful, the ready flag is cleared, and the send buffer is released.

[0060] If no ACK is received within the timeout period, the ready flag remains valid, and the terminal automatically switches to compensated upload mode. In this mode, the terminal opens compensated micro-slots at the edge nodes. Within the process, a carrier sense multiple access and collision avoidance mechanism is used to retry uploading. Specifically, the terminal continuously monitors the channel status. When it detects that the channel idle time is greater than the distributed inter-frame interval, it starts a contention backoff counter. The counter decrements during idle time slots, and a retransmission is initiated when it reaches zero. If the channel is detected to be busy during the backoff process, the counter is suspended until the channel becomes idle again.

[0061] Among them, the initial value of the contention backoff counter 15 time slots, maximum competition window There are 1023 time slots; after each collision, the contention window is updated according to a binary exponential backoff strategy: ; The length of a single time slot adopts the time slot unit specified in the IEEE 802.15.4 standard; the distributed inter-frame interval is 2 time slots in length.

[0062] In some embodiments, if no ACK is received after multiple retransmissions within the compensation micro-slot, and the compensation micro-slot is nearing its end, the unit forcibly clears the ready flag (sets it to logic 0) at the boundary moment of the end of the compensation micro-slot and discards the data packets in the current transmission buffer. This forced clearing mechanism prevents the terminal from permanently locking the ready flag due to repeated retransmission failures, ensuring that the terminal can discard expired data and start a new round of collection in subsequent rounds, thus avoiding data timeliness exceeding the limit.

[0063] In some embodiments, the edge-side classification and aggregation module is deployed in the application layer processing unit of the edge node and interfaces with the physical layer receive buffer through the data plane interface. This module is triggered to execute after the regular receive window closes in each round. First, it performs round-based attribution binding and time window alignment on the raw data frames accumulated in the buffer. Then, according to multidimensional classification rules, it divides the data into three data sets with different confidence levels, performs baseline aggregation calculation and boundary floating interval calculation for each, and finally outputs the aggregation result with uncertainty boundaries.

[0064] The edge-side classification and aggregation module includes a window alignment unit, a data classification unit, a baseline aggregation unit, a floating interval calculation unit, and an untrusted data processing unit.

[0065] In some embodiments, the window alignment unit extracts the frame header information of each uplink data frame from the physical layer receive buffer and reads the round number field R carried therein. This field is the round number entered by the terminal during upload from the latest start trigger beacon it received. The window alignment unit then aligns all frames that meet the specified conditions... Data frames are incorporated into the current aggregation cycle. And the terminal ID, overall trust level L, communication quality flag Q, and compensation channel identifier in each data frame will be included. The net load P measured was extracted into a structured record.

[0066] For terminals that have not reported any data frames in the current aggregation cycle, meaning that the terminal's data is completely missing in this round (including cases where the ready flag is forcibly cleared and data packets are discarded due to failure to upload successfully before the compensation micro-slot ends), the window alignment unit generates a filling record for the terminal. The terminal ID of the filling record is taken from the missing terminal ID, the overall reliability level L is set to low, the communication quality flag Q is set to poor, and the compensation channel identifier... The measurement payload P is set to 0, and is calculated as the measurement value of the terminal in the previous valid aggregation cycle multiplied by (1 + the average change rate of trusted terminals in this round). The average change rate of trusted terminals in this round is defined as (the average measurement value of trusted terminals reported in this round - the average measurement value of the same trusted terminal set in the previous round) / the average measurement value of the same trusted terminal set in the previous round. If the average value of the previous round is 0, the change rate is 0. If the terminal has no historical valid data, the measurement payload P is set to 50% of the terminal's rated capacity. All filled records carry a placeholder marker field of 1. In subsequent data classification units, they are directly classified into the untrusted data category and do not participate in the determination of physical conservation conditions or the calculation of alternative values ​​for untrusted data. For untrusted filled records carrying placeholder markers, the edge node encapsulates its terminal ID, rated capacity, and placeholder marker into structured supplementary information and includes it in the missing terminal list subfield of the uncovered quantity field, which is reported with the aggregation results to retain the record trace of missing data for the terminal, but it is not included in the physical quantity statistics of the uncovered quantity.

[0067] The compensation channel identifier is carried in the protocol reserved field defined by the short-range communication protocol in the physical layer frame header of the uplink data frame. The terminal sets this reserved field to a valid value in the compensation upload mode. The edge node parses the field after receiving the data to identify the compensation channel identifier. In the normal upload mode, the field remains at the default value.

[0068] In some embodiments, the upload times of data from different terminals within the same aggregation period may differ, as some terminals may complete the upload at the beginning of the normal receiving window, while others may only complete the upload near the closing of the window. The window alignment unit ignores these upload time discrepancies and treats all data under the same round number as valid samples collected within the same aggregation period. This operation avoids cross-round data misalignment caused by differential delays in backoff scheduling, ensuring that data collected by the same group of terminals under the same round number can be fused and calculated at the same time cross-section, conforming to the engineering specifications for synchronous processing of power system measurement data.

[0069] In some embodiments, the data classification unit divides the window-aligned terminal data into three categories: reliable data, questionable data, and unreliable data according to the following multi-dimensional judgment rules. The classification criteria include three dimensions: comprehensive reliability level. Communication quality flags Compensation channel signage .

[0070] Specifically, the classification rules are as follows: First, perform the highest priority judgment: if a record carries... If the data packet was uploaded via compensated micro-timeslot retransmission, it is directly classified as suspicious data and will not be subject to further processing based on... and The determination is based on the following principle: Uploading through the compensation channel means that the data packet has undergone at least one failed retransmission within the normal receiving window, and its transmission delay has exceeded the normal range. Although it may contain valid measurement values, its timeliness and transmission integrity do not have the same level of reliability as the data uploaded in normal mode, so it is not suitable to participate in the benchmark aggregation.

[0071] for Data, classification units based on and The following judgment is made based on the combination of: like =High and =Good, then it is reliable data; if =in and =Good, then it is questionable data; if =low and =Good, then it is unreliable data; if and If the difference is significant, the data is considered unreliable.

[0072] Among them, the communication quality flag bit The physical layer of the edge node generates the following in real time after each data frame is received: If the physical layer cyclic redundancy check (CRC) of the data frame passes, and the received signal strength indicator (RSSI) is higher than the demodulation sensitivity threshold and the signal-to-noise ratio (SNR) is higher than the minimum demodulation threshold corresponding to the current modulation and coding scheme, then Q = good; otherwise, Q = poor. Data frames that fail the CRC check are not included in the data classification process.

[0073] In some embodiments, the benchmark aggregation unit only includes records whose classification result is reliable data in the benchmark aggregation calculation. Assume there are a total of [number missing] records in the current aggregation period. The first trusted data record, The power measurement value of the record The corresponding overall credibility level is high. (Baseline aggregation result) The calculation method is a weighted arithmetic mean: ; Among them, the weighting coefficient Based on the mapping of the terminal's overall credibility level, since the data participating in the benchmark aggregation are all of high credibility level, each... It can be uniformly set to 1, i.e., equal-weighted average. Alternatively, it can be fine-tuned based on the specific signal-to-noise ratio values ​​in the communication quality flags of each terminal. ; in, The preset proportionality coefficient is preferred. The value is 0.01. The baseline signal-to-noise ratio threshold is set to the lowest demodulation threshold corresponding to the current modulation and coding scheme, typically 10 dB. This weighting method ensures that reliable data with better communication quality contributes more to the baseline aggregation, further improving the accuracy of the aggregation results.

[0074] In some embodiments, the criteria for determining low-confidence aggregation results include any one of the following: (a) the number of credible data in the current aggregation period is 0; (b) the number of credible data in the current aggregation period is greater than 0 but less than 30% of the total number of online terminals in the area, and the number of questionable data entries accounts for more than 50% of the total number of reported data entries in the period; wherein, the total number of online terminals refers to the number of terminals that the edge node actually receives data frames in the current aggregation period, excluding unreported terminals; (c) the total incoming line measurement value of the area on which the baseline aggregation result is based in the current aggregation period is invalid or missing; (d) after the untrusted data substitution value is calculated, the number of untrusted data entries for which the substitution value is successfully calculated still exceeds 20% of the total number of terminals in the area in the current aggregation period, and this count does not include filler records and untrusted data that does not meet the conservation condition and is directly included in the uncovered quantity field.

[0075] If any of the above conditions are met, the aggregation result for that period is marked as a low-confidence aggregation result. Low-confidence aggregation results are not included in the calculation of the ramp difference of the baseline aggregation value within the window; if any source period within the window is marked as low-confidence, the interval propagation unit performs extended compensation; if both source periods are low-confidence, the overlay interval is set to the full-range interval.

[0076] Specifically, in this embodiment, for power measurement, the lower limit of the full-range interval is taken as 0kW, and the upper limit is taken as the maximum measurable power value corresponding to the transformation ratio of the total incoming current transformer of the transformer area. This maximum measurable power value is calculated by the edge node during the initialization phase based on the transformation ratio of the total incoming current transformer and the rated voltage of the transformer area. The calculation formula is as follows: ,in This is the rated voltage of the transformer substation. The upper limit is the rated current of the primary side of the CT; if the CT ratio of the main incoming line of the transformer substation is not configured, the upper limit is 120% of the rated capacity of the transformer in that substation. For voltage measurement, the lower limit of the full range is 0V, and the upper limit is 120% of the rated voltage of that voltage level. For current measurement, the lower limit of the full range is 0A, and the upper limit is the rated current value of the circuit breaker connected to the terminal; if the rated current value of the circuit breaker cannot be obtained, the full-scale value nominal on the nameplate of the terminal's measuring module is used. For temperature measurement, the lower limit of the full range is -40°C, and the upper limit is 125°C. When none of the above physical range upper limits can be obtained, the default range upper limit of this type of terminal issued by the cloud is used as the upper limit of the full range.

[0077] The full-range interval is stored in floating-point form and written to the local cache of the edge node along with the aggregation result of the period, for use by the interval propagation unit.

[0078] In some embodiments, the floating interval calculation unit calculates the current reading of each record for which the classification result is questionable data. Based on this, retrieve the historical fluctuation statistics of the terminal. and Calculate the potential deviation boundary of the questionable data from the benchmark aggregation result.

[0079] Historical fluctuation statistics and This is the statistical upper limit of the maximum absolute relative deviation of the terminal's valid data from the trusted data aggregation benchmark within a preset number of historical samples (a fixed 20 most recent valid aggregation periods). The valid aggregation periods only include periods where the terminal's data is classified as trusted or questionable, excluding periods classified as untrustworthy. For terminals with no questionable data in the historical valid periods, and The absolute value of the maximum relative deviation between the terminal's historical trusted data and the benchmark aggregate value is used as the initial equivalent value; if the terminal also has no historical trusted data, it is processed in the same way as the initialization of a newly accessed terminal.

[0080] If a terminal is classified as untrustworthy for five consecutive aggregation cycles, its δ⁺ and δ⁻ values ​​are forcibly reset to the median of the aforementioned statistical values ​​for all terminals within that area at the end of the fifth cycle. The consecutive untrustworthy count is then reset to zero, and the statistics restart. If the area median is unavailable, it is reset to the factory default value. This default value is 5% for power measurements, and 1% and 2% for voltage and current measurements, respectively, to ensure that terminals classified as untrustworthy for extended periods have usable initial statistical values ​​after recovery. For aggregation cycles marked as low confidence, their baseline aggregation value is not included. and Statistical calculations are performed, but questionable data readings for that period are still included in the historical samples of the sliding window to maintain the continuity of fluctuation statistics. After each aggregation period, the edge node updates the terminal's [data / data] according to the sliding window mechanism. and value.

[0081] The statistical upper limit of the absolute value of the maximum relative deviation of its valid data relative to the trusted data aggregation benchmark within that period. Specifically: ; in, For the set of historical effective cycles; For this terminal in the first Readings from historical periods that are considered questionable data; This is the baseline aggregate value corresponding to this period. If it is 0, then this period will not participate in the above statistical calculations to avoid division by zero anomalies.

[0082] The above statistical values ​​are pre-stored in the terminal historical feature database of the edge nodes and retrieved by the floating interval calculation unit according to the terminal ID index.

[0083] Current readings of questionable data As the baseline, the maximum positive deviation of the aggregation result from the baseline. and maximum negative deviation They are respectively: ; The resulting boundary floating range is represented as follows: ; This interval represents the range within which the "true value" of the questionable data point is most likely to fall, inferred from historical fluctuation patterns. This boundary fluctuation range will be transmitted as uncertainty information along with the aggregation results to the ramp value calculation and interval propagation units during the subsequent feature extraction stage, enabling the cloud to perceive the confidence boundary of each questionable data point.

[0084] In some embodiments, the untrusted data processing unit performs an independent processing procedure for records whose classification result is untrusted data. This unit first determines whether the physical conservation condition is satisfied in the current aggregation cycle.

[0085] Specifically, the physical conservation condition is as follows: within the same busbar or the same transformer area, whether the absolute value of the deviation between the algebraic sum of the measured values ​​of reliable data and doubtful data and the total incoming line measured value is less than a preset conservation tolerance threshold. : ; Among them, in the calculation of terminal-side mutual verification and edge-side substitution value... Both use three times the combined standard uncertainty as the benchmark term. However, in the terminal-side mutual verification, the larger of this benchmark term and 2% of the rated capacity of the transformer in the distribution area is taken to strictly filter unreliable data in the neighboring verification. In the edge-side substitution value calculation, 5% of the current total load of the distribution area is further introduced as a dynamic adjustment term to take into account the impact of relative error amplification under light load scenarios. The difference in the value rules of the two is for different optimization objectives at different processing stages and does not constitute a system logic conflict.

[0086] This refers to the total number of terminals within the station area that participated in this round of reporting. This is the measured value of total incoming power. For a reliable dataset, the threshold for determining physical conservation conditions. Take it in the following way: ,in, Three times the combined standard uncertainty of all terminal measuring devices. The value is the smaller of 2% of the rated capacity of the transformer in the distribution area and 5% of the current total load of the distribution area. When the current total load of the distribution area is less than 5% of the rated capacity of the transformer, Take directly Furthermore, it does not perform substitution value calculations for unreliable data.

[0087] If the aforementioned physical conservation conditions are satisfied, then this unit, according to Kirchhoff's current law, uses all reliable data within the same busbar / station area to infer the replacement value of unreliable data. Let the set of reliable data within this station area be... The total incoming line measurement value is If the number of untrusted data in the current aggregation period is 1, then the replacement value for the m-th untrusted data is... The calculation method is as follows: ; If the number of untrusted data points in the current aggregation cycle is greater than 1, then all untrusted data points are treated as a whole, and the sum replacement value is calculated.

[0088] ;

[0089] Then, based on the proportion of the rated capacity corresponding to each untrusted terminal (obtained and stored by the edge node when the terminal is registered) to the total rated capacity of untrusted terminals, it is allocated to each record and then classified into the questionable data category.

[0090] If the sum of the rated capacities of untrusted terminals is zero or cannot be obtained, the allocation is made proportionally to the ratio of the historical average measurement value of each untrusted terminal (the average of the last 5 valid aggregation periods) to the sum of the historical average measurement values ​​of all untrusted terminals. If the historical average measurement value is also unavailable, the allocation is made proportionally to the total number of untrusted terminals in the current period and distributed to each record, with a double fluctuation compensation coefficient added to the calculation of the fluctuation range of that record. and Multiply each by 2) to characterize the additional uncertainty brought about by the equal distribution.

[0091] in, For questionable data sets, For the first The current reading of the questionable data. If the set of reliable data is empty, the alternative value of the unreliable data cannot be reliably calculated using the above method. In this case, the unreliable data is directly included in the uncovered quantity field, and no alternative value is calculated.

[0092] The above calculation, assuming the total incoming power is known, deducts all reliable data and other questionable data, and the remaining power is the alternative estimated value of the unreliable data.

[0093] In some embodiments, alternative values Data categorized as questionable will participate in subsequent floating range calculation units but will not be included in the baseline aggregation unit calculation. This approach transforms some unreliable data into quantifiable uncertainties through physical conservation constraints, expanding the data coverage for boundary range calculations while preventing low-quality data from contaminating the baseline aggregation results.

[0094] If the current aggregation cycle does not meet the physical conservation conditions, for example, due to unknown line losses or a large-scale failure of the measuring device causing the conservation deviation to exceed the tolerance threshold, then this unit will not perform substitution value calculations for unreliable data, but will instead use the terminal ID and original readings of each unreliable data point. The unreliability reasons (such as low overall credibility, poor communication quality, and excessive physical conservation deviation) are recorded one by one in the uncovered quantity field according to the data format. The uncovered quantity field is appended to the baseline aggregation result in a structured manner and transmitted to the feature extraction module along with the aggregation result. It is also retained in the final cloud report for offline review or manual correction by cloud operation and maintenance personnel.

[0095] In some embodiments, the edge-side feature extraction module is deployed in the data processing pipeline of the edge node, located after the classification and aggregation module and before the upload module. This module takes the aggregation result sequence of multiple consecutive aggregation cycles (time windows) as input, and through three processing steps—anchor point extraction, extreme value ramp-up detection, and uncertainty interval propagation—compresses the high-dimensional time series data into compact feature tuples containing start / end anchor points, maximum positive / negative rates of change, and their confidence boundaries, for the upload module to send to the cloud. The technical effect of this dimensionality reduction extraction strategy is that, while preserving the main characteristics of power load fluctuations, it compresses the original measurement data of all terminals in each round into a fixed-length feature vector, significantly reducing uplink bandwidth usage and cloud storage pressure.

[0096] The edge-side feature extraction module includes an anchor point extraction unit, a ramp calculation unit, an interval propagation unit, a feature encapsulation unit, and an interval merging unit.

[0097] In some embodiments, the anchor point extraction unit continuously The input consists of a sequence of baseline aggregation results from each aggregation cycle, where... This is the length of the time window, set by parameters configured and distributed from the cloud. The default value is 6, corresponding to 6 consecutive rounds, covering the main load change trends within the most recent scheduling cycle.

[0098] The starting period index of the time window is... The termination period index is ,satisfy The anchor point extraction unit extracts the initial periodic aggregation value from the sequence. and the aggregate value of the termination period And bind the corresponding boundary floating ranges respectively. and .

[0099] Furthermore, the boundary floating range is output by the edge-side classification aggregation module in each cycle along with the baseline aggregation result, in the following form: ; in, and The first The absolute value of the maximum negative deviation and the maximum positive deviation of all questionable data in the period from the benchmark aggregation result are calculated using the method described in the floating range calculation unit above.

[0100] In some embodiments, if a benchmark aggregation result for a certain period within a time window is marked as missing due to insufficient reliable data, then that period will not participate in anchor point extraction and ramp-up calculation. The window length is... Slide sequentially forward until a sufficient number of valid aggregation periods are covered. Specifically, for each missing period removed, a newest aggregation period is added forward, until the number of valid periods within the window reaches a certain threshold. If the number of consecutive missing periods exceeds If the value is 2, then the sliding will stop and the time window will be marked as an invalid window, and the feature tuple will not be uploaded.

[0101] In some embodiments, the slope calculation unit receives the output from the anchor point extraction unit. Given a baseline aggregation value sequence for each effective aggregation cycle, calculate the difference between aggregation values ​​in adjacent cycles. : ; like If this difference is a ramp value, it represents the rate of increase in load (or power) between adjacent cycles; if This is a descending ramp value, representing the magnitude of the load decrease (recorded as a negative value).

[0102] The climbing calculation unit extracts the maximum value from all ascending climbing values ​​and the maximum absolute value from all descending climbing values. Let the set of ascending climbing values ​​be... The set of descending and climbing values ​​is Then the maximum climbing value and maximum descent ramp value They are respectively: ; Equivalently, if the decrease is expressed in absolute value, then the maximum decrease slope value is... However, its negative sign is retained when recording to indicate the direction of change.

[0103] In some embodiments, if or If the set is empty, meaning there are no upward or downward changes within the window, then the corresponding maximum ramp value is set to 0, and the floating range is set to the zero range.

[0104] The ramp calculation unit also records the termination cycle number corresponding to each maximum value. Specifically, the termination cycle index corresponding to the maximum ramp value is: That is, the difference The corresponding next period number The index of the termination period corresponding to the maximum descent ramp value is The termination cycle number is used by the cloud to reconstruct the specific time and location of the ramp-up after receiving the feature tuple.

[0105] If a period marked as missing baseline exists within the window, that period will not participate in the difference calculation, and the entire window will be extended one period forward and retried until the proportion of valid periods within the window recovers to above 50%. If a period marked as low confidence exists within the window, the baseline aggregate value of that period will only be used as the source value for interval propagation in the expansion compensation of the floating interval, and will not participate in the difference calculation of adjacent periods. The numerical calculation (i.e., this period does not participate in the extraction of the ramp value, but the window length is not reduced, and the floating range of this period is still passed to the interval propagation unit for expansion to adjacent periods); if all within the window If both are 0, then the maximum upward and downward ramp values ​​are both set to 0, and the floating range is set to the zero range.

[0106] In some embodiments, the interval propagation unit calculates the corresponding floating interval for each ramp value. Let the maximum ramp value be... ,in, The two corresponding source periods are the initial period and the... and termination period The corresponding boundary floating ranges are respectively and .in, , , , All are non-negative values. and This indicates a negative deviation. and This indicates a positive deviation.

[0107] The interval propagation unit superimposes the boundary floating intervals of two source periods, resulting in an overlaid interval. for: ; Aggregation results marked as low confidence will have their boundary floating intervals processed according to the following rules before participating in interval propagation: If the low confidence is due to the number of credible data being 0 (i.e., the aforementioned case (a)), the floating interval will be set to 3 times the median of all historical valid period floating intervals within the station area; if the low confidence is due to the number of credible data being less than 30%, invalid incoming line measurements, or the number of substitute values ​​exceeding the limit (i.e., the aforementioned cases (b), (c), and (d)), the existing floating interval will be expanded outward by 100%; if multiple of the above conditions are triggered simultaneously, the strictest expansion rule will be executed, i.e., the maximum value of the expansion multiple corresponding to each rule will be taken.

[0108] The theoretical basis for the above superposition rules is: the deviation error of two independent questionable data points in the difference operation In the linear accumulation, the negative error accumulates to form the negative boundary of the superimposed interval, and the positive error accumulates to form the positive boundary of the superimposed interval. Specifically, the maximum positive deviation occurs... Take the maximum negative deviation and Taking the case of the maximum positive deviation, that is: ; The maximum negative deviation occurred Take the maximum positive deviation and The case of taking the maximum negative deviation is as follows: ; Therefore, the slope value The floating range is: ; If we take the absolute value of the deviation and If we express this as an expression, then the above interval can be rewritten as: The interval representation format is consistent with that of the anchor point. The maximum upward and downward ramp values ​​are each calculated independently using the interval propagation method described above, yielding their respective floating intervals. and .

[0109] It should be noted that, through the above-mentioned interval propagation mechanism, the uncertainty boundary of the ramp value can be accurately transmitted from the uncertainty of the two source cycles to the feature tuple, avoiding the loss of confidence boundary information due to only transmitting the absolute value of the ramp. This enables the cloud to effectively assess the confidence range of each rate of change when performing load forecasting or anomaly detection based on the feature tuple.

[0110] In some embodiments, the feature encapsulation unit combines the above extraction results into a structured feature tuple. The data structure of the feature tuple is defined as a fixed-length field sequence, which includes at least the following fields: the initial period aggregation value, the boundary floating range of the initial period, the final period aggregation value, the boundary floating range of the final period, the maximum upward ramp value, the floating range of the maximum upward ramp value, the final period number corresponding to the maximum upward ramp value, the maximum downward ramp value, the floating range of the maximum downward ramp value, and the final period number corresponding to the maximum downward ramp value.

[0111] In some embodiments, the feature tuple is serialized into a binary byte stream in Protocol Buffers or JSON format, and appended with the start absolute time and window ID fields of the current time window, and then delivered to the edge-side upload module for transmission to the cloud. This fixed-length feature tuple design ensures that the amount of data uploaded in each round does not change with the number of terminals, making uplink bandwidth usage predictable and facilitating resource planning for continuous 24 / 7 operation.

[0112] In some embodiments, the edge-side feature extraction module further includes an interval merging unit, used to perform lateral comparison and compression merging of the floating intervals of different aggregation groups before feature tuple generation. Here, different aggregation groups refer to data sets of different bus data, different feeders, or different equipment types within the same distribution area, and their respective boundary floating intervals vary in magnitude due to differences in terminal measurement characteristics and communication quality.

[0113] The interval merging unit receives floating intervals from multiple aggregation groups. The aggregation group is pre-divided by the edge nodes according to the physical topology of each terminal in the transformer area (including the bus, feeder or equipment type to which it belongs). The terminal data in the same aggregation group are independently aggregated on the baseline and calculated on the floating range in the classification aggregation module.

[0114] in, ,in, and These are the lower and upper bounds of the m-th interval, respectively. The interval merging unit compares the floating intervals pairwise, with the following criteria: and ; in, This is a preset percentage threshold, with a default value of 15%. If the upper and lower bound deviations of both floating intervals do not exceed the larger of the two interval values... If the two intervals are determined to be of the same magnitude, a merge operation is performed.

[0115] Furthermore, the smaller of the lower bounds of the two intervals is taken as the lower bound of the merged interval, and the larger of the upper bounds of the two intervals is taken as the upper bound of the merged interval, which is called the outward expansion and merging of the set. After the merge is completed, the interval merging unit generates a general floating interval field, along with a list of aggregate group indexes that share the interval.

[0116] This universal floating range replaces the original floating ranges of each independent aggregation group and is uploaded with the feature tuples as a shared confidence boundary field for multiple aggregation groups.

[0117] In some embodiments, if the merged general floating range is shared by three or more aggregation groups, the general floating range is promoted to the baseline floating range and remains unchanged across multiple consecutive feature tuples. When a new aggregation group's floating range deviates from the baseline floating range by more than a preset percentage threshold, the interval merging unit triggers a recalculation and rebroadcast process.

[0118] It can be seen that when multiple aggregation groups have the same order of magnitude of measurement uncertainty, merging repeated floating intervals into a single general field with a shared index list effectively reduces the amount of redundant data in feature tuples and improves uplink transmission efficiency. The compression effect is particularly significant in scenarios with a large number of terminals and a fixed station topology.

[0119] In some embodiments, the edge-side upload module is deployed in the uplink communication protocol stack of the edge node, establishing a bidirectional data channel with the cloud data center through a wide area network interface. This module receives the feature tuples output by the feature encapsulation unit, encapsulates and encrypts them before sending them to the cloud, and is responsible for ensuring transmission reliability.

[0120] In some embodiments, the upload module will include the feature tuple along with the edge node identifier, the transformer area topology number, the time window start stamp, and the window length. The data is concatenated into an upload data frame, appended with a 32-bit CRC checksum, and then losslessly compressed (LZ4 or Zstandard) before being sent to the cloud. If the interval merging unit has already been merged, the general floating interval field and its corresponding aggregation group index list are uploaded along with the feature tuple.

[0121] In some embodiments, the upload module employs a hybrid strategy of primarily triggering per window and secondarily triggering based on periodic timeout: Upload is triggered immediately after feature extraction for each time window is completed; simultaneously, a periodic timer (default 5 minutes) is started, and if the timeout occurs, the currently extracted feature tuples are forcibly uploaded to prevent prolonged periods without data updates in the cloud. When the maximum ramp value exceeds a preset alarm threshold, the periodic timer is interrupted, and an immediate upload is initiated with an emergency flag.

[0122] In some embodiments, the upload module communicates with the cloud using the MQTT over TLS 1.3 protocol. After sending, an acknowledgment waiting timer is started; if no ACK is received within the timeout period or a NACK is received, a retransmission is initiated.

[0123] In some embodiments, the upload module has a built-in non-volatile circular queue buffer (capacity 200 feature tuples) to temporarily store data to be uploaded during network interruptions. When the buffer is full, the oldest frame is discarded. After the network is restored, failed frames are retransmitted first, with the retransmission rate limited to 50% of the normal rate. Random backoff intervals of 0-500ms are inserted between frames to smooth the load. This mechanism ensures that transient network failures do not lead to permanent loss of feature data, while also preventing traffic surges after prolonged network outages.

[0124] The edge data aggregation and real-time upload scheduling system for virtual power plant monitoring terminals provided in this embodiment of the invention can execute the edge data aggregation and real-time upload scheduling system for virtual power plant monitoring terminals provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0125] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0126] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. An edge data aggregation and real-time upload scheduling system for virtual power plant monitoring terminals, characterized in that: include: The terminal-side mutual verification module is used to exchange and verify data with adjacent terminals through short-range communication, obtain mutual verification marks, and generate a comprehensive credibility level based on the number and level of the mutual verification marks. The terminal-side backoff scheduling module is used to calculate the backoff base value based on the comprehensive credibility level, and then upload the collection data packet to the edge node after superimposing the backoff base value on the unified reference time base point indicated by the start trigger beacon broadcast by the edge node. The start trigger beacon contains the round number. The edge-side classification and aggregation module is used to receive the collected data packets and bind each data packet to its respective round number. Data belonging to the same round number are assigned to the same aggregation period. The data is divided into trusted data, questionable data and untrusted data according to the comprehensive credibility level and communication quality flag. Data carrying compensation channel identifiers is not restricted by the comprehensive credibility level and is directly classified into the questionable data category. The trusted data is aggregated to generate a benchmark aggregation result, and the boundary floating range is calculated for the questionable data. The edge-side feature extraction module is used to extract features from the aggregation results of multiple consecutive aggregation cycles along the time window sequence, retain the aggregation values ​​of the start cycle and the end cycle, extract the maximum rising and falling ramp values ​​between adjacent cycles, and superimpose the boundary floating ranges of the two cycles corresponding to each ramp value as the floating range of the ramp value. The edge-side upload module is used to upload the extracted feature tuples to the cloud.

2. The edge data aggregation and real-time upload scheduling system for virtual power plant monitoring terminals according to claim 1, characterized in that, The terminal-side mutual authentication module includes: The tag acquisition unit is used to exchange collected data with neighboring terminals through short-range communication, verify the data of neighboring terminals according to preset physical constraint rules, and obtain the mutual verification tags returned by each neighboring terminal. The levels of the mutual verification tags include trustworthy, doubtful, and untrustworthy. The rating determination unit is used to determine the overall trustworthiness level of the terminal itself based on the number of mutual verification marks and the overall rating. When the number of mutual verification tags is three, the level determination unit determines the overall credibility level according to the majority principle; when the number of mutual verification tags is two, if both mutual verification tags are credible, it is determined to be a high level; if one credible tag is paired with one questionable tag, it is determined to be a medium level; otherwise, it is determined to be a low level; when the number of mutual verification tags is less than two, it is forcibly determined to be a low level and an isolated status identifier is reported to the edge node.

3. The edge data aggregation and real-time upload scheduling system for virtual power plant monitoring terminals according to claim 1, characterized in that, The terminal-side backoff scheduling module includes: The backoff baseline calculation unit is used to calculate the backoff baseline value based on the comprehensive credibility level. When the comprehensive credibility level is high, it corresponds to the first backoff interval; when it is medium, it corresponds to the second backoff interval; and when it is low, it corresponds to the third backoff interval. The upper limit of the first backoff interval is less than the lower limit of the second backoff interval, and the upper limit of the second backoff interval is less than the lower limit of the third backoff interval. A micro-jitter overlay unit is used to add random micro-jitter values ​​to the backoff base value; The upload triggering unit is used to upload the acquisition data packet to the edge node after receiving the start triggering beacon broadcast by the edge node, taking the unified reference time base point indicated by the start triggering beacon as the timing start point, and waiting for the total duration of the backoff base value and the micro-jitter value superimposed. The edge node broadcasts the start trigger beacon at a fixed period, the lower limit of which is determined by the sum of the regular receive window duration, the compensation microslot duration, and the guard interval duration.

4. The edge data aggregation and real-time upload scheduling system for virtual power plant monitoring terminals according to claim 3, characterized in that, The terminal-side backoff scheduling module also includes a ready flag management unit, which is used to set the ready flag to an effective state after the comprehensive trust level calculation is completed, and to clear the ready flag after receiving the link layer confirmation response returned by the edge node. If no confirmation response is received from the link layer within the preset timeout period, the ready flag remains valid, and the process switches to compensated upload mode. The compensated upload mode is as follows: within the compensated micro-slot opened by the edge node, the upload is retried using the carrier sense multiple access and collision avoidance mechanism. If no link layer acknowledgment response is received from the edge node, the ready flag is forcibly cleared at the end of the compensated micro-slot.

5. The edge data aggregation and real-time upload scheduling system for virtual power plant monitoring terminals according to claim 1, characterized in that, The edge-side classification aggregation module includes: The window alignment unit is used to uniformly assign all terminal data belonging to the same round number to the same aggregation period, ignoring the upload time deviation of each terminal data within the period. The data classification unit is used to divide window-aligned data into three categories: reliable data, questionable data, and unreliable data based on the overall reliability level, communication quality flag, and compensation channel identifier. Data carrying the compensation channel identifier is not restricted by the overall reliability level and is directly classified into the questionable data category. The benchmark aggregation unit is used to directly incorporate all reliable data into the calculation and generate benchmark aggregation results. The floating range calculation unit is used to retrieve the historical fluctuation statistics of the terminal, and calculate the maximum positive deviation and the maximum negative deviation of the aggregated result of the benchmark based on the current reading of the questionable data, thus forming the boundary floating range; The untrusted data processing unit is used to calculate the energy balance difference using reliable data from the same bus or transformer area according to Kirchhoff's current law when the physical conservation conditions are met, to obtain a substitute value for the untrusted data, and to classify the substitute value into the questionable data category for boundary floating range calculation; when the physical conservation conditions are not met, the untrusted data is included in the uncovered quantity field, and the uncovered quantity field is appended to the baseline aggregation result.

6. The edge data aggregation and real-time upload scheduling system for virtual power plant monitoring terminals according to claim 5, characterized in that, The historical fluctuation statistics retrieved by the floating range calculation unit are the statistical values ​​of the upper limit of the historical valid data fluctuation of the terminal. The historical valid data includes historical readings that are classified as reliable data and questionable data, but excludes historical readings that are classified as unreliable data.

7. The edge data aggregation and real-time upload scheduling system for virtual power plant monitoring terminals according to claim 1, characterized in that, The edge-side feature extraction module includes: Anchor point extraction unit is used to extract the start period aggregation value and the end period aggregation value from a sequence of multiple consecutive aggregation periods, and bind the corresponding boundary floating intervals respectively. The ramp calculation unit is used to calculate the difference between the aggregate values ​​of adjacent periods, extract the maximum upward ramp value and the maximum downward ramp value according to the absolute value of the difference, and record the termination period number corresponding to each maximum value. The interval propagation unit is used to superimpose the boundary floating intervals of the two source cycles corresponding to each ramp value, and use the superimposed interval as the floating interval of the corresponding ramp value; when there is a low confidence aggregation result in the two source cycles, the superimposed interval is expanded outward to compensate; when both source cycles are low confidence, the superimposed interval is set to the full range interval. The feature encapsulation unit is used to combine the starting aggregation value and floating range, the ending aggregation value and floating range, the maximum upward ramp value and floating range and the ending cycle number, and the maximum downward ramp value and floating range and the ending cycle number to form a feature tuple.

8. The edge data aggregation and real-time upload scheduling system for virtual power plant monitoring terminals according to claim 7, characterized in that, The interval propagation unit adds the negative absolute values ​​of the boundary floating intervals of the two source periods as the negative boundary of the superimposed interval, and adds the positive absolute values ​​of the boundary floating intervals of the two source periods as the positive boundary of the superimposed interval; the edge-side feature extraction module also includes an interval merging unit, which is used to perform horizontal comparison of the floating intervals of different aggregation groups. When the upper and lower bound deviations of two floating intervals do not exceed a preset percentage threshold of the larger interval value, they are determined to be of the same magnitude and are merged. The merged interval is used as a general floating interval field and is accompanied by an aggregated group index list that shares the interval.

9. The edge data aggregation and real-time upload scheduling system for virtual power plant monitoring terminals according to claim 4, characterized in that, The duration of the compensation micro-slot is dynamically adjusted by the edge node based on the collision rate; When the collision rate continuously exceeds a preset probability threshold, the terminal is instructed in the next round of broadcast beacons to extend the compensation micro-slot duration to a preset extended duration.

10. The edge data aggregation and real-time upload scheduling system for virtual power plant monitoring terminals according to claim 1, characterized in that, If the ready flag is invalid when the next round of start trigger beacon arrives, the terminal-side backoff scheduling module will trigger the terminal-side mutual authentication module to perform a new round of data collection and mutual authentication. If the ready flag is detected as valid, the terminal-side mutual authentication module will be retried after a forced reset.

Images