CN-121858306-B - Internet of things data hierarchical processing system based on edge calculation

CN121858306BCN 121858306 BCN121858306 BCN 121858306BCN-121858306-B

Abstract

The invention discloses an Internet of things data grading processing system based on edge calculation, which relates to the technical field of edge calculation and comprises the steps of acquiring and constructing a wind speed change sequence through a wind speed change energy period identification module, determining an initial high-priority processing period proportion, utilizing a high change segment time redistribution module to adjust high change energy segment coverage density to generate a plurality of second proportion, forming grading processing configuration through a processing period segmentation and queue configuration module, applying the grading processing configuration in an edge node grading processing execution module to realize grading processing of wind speed data, acquiring a duty ratio excitation value through a duty ratio excitation reinforcement learning evaluation module, and finally screening an optimal duty ratio through an optimal high-priority processing period duty ratio application module, thereby solving the problems of incomplete processing and calculation resource waste of the edge node Internet of things data in a key change segment of a wind speed critical region.

Inventors

ZHANG LIANG
ZHANG SHENGXUAN
XING HAILONG

Assignees

上海鼎为物联技术有限公司

Dates

Publication Date: 20260508
Application Date: 20260317

Claims (7)

1. The internet of things data hierarchical processing system based on edge calculation is characterized by comprising a wind speed change energy period identification module, a high change fragment time redistribution module, a processing period time segmentation and queue configuration module, an edge node hierarchical processing execution module, a duty ratio excitation reinforcement learning evaluation module and an optimal high-priority processing period duty ratio application module: The wind speed change energy period recognition module is used for acquiring a wind speed time sequence and constructing a wind speed change sequence, determining a concentrated interval of change energy on a time axis through time-frequency analysis, and taking the proportion of the concentrated interval to the whole time as the initial wind speed high-priority processing period proportion; The high-change-segment time redistribution module is used for time redistribution of the initial wind speed high-priority processing time period duty ratio based on the time aggregation characteristic of the wind speed change sequence, and a plurality of second wind speed high-priority processing time period duty ratios are generated by adjusting the coverage density of the high-change energy segments on a time axis; the processing period time segmentation and queue configuration module is used for introducing the wind speed high-priority processing period duty ratio into the edge node processing period, performing time segmentation on the processing period, and forming processing configuration comprising processing time distribution and queue classification according to the time dense distribution of wind speed change; The edge node grading processing execution module is used for respectively applying processing configurations corresponding to the duty ratio of each second wind speed high-priority processing period to the edge computing nodes and executing grading processing of wind speed data; The duty ratio excitation reinforcement learning evaluation module is used for extracting the processing characteristics of each wind speed high-priority processing time period under the duty ratio, constructing reinforcement learning state space, and training by taking corresponding processing configuration as action to obtain the duty ratio excitation value of each second wind speed high-priority processing time period; And the optimal high-priority processing time period duty ratio application module is used for screening the duty ratio of the optimal wind speed high-priority processing time period based on the duty ratio excitation value and applying the optimal wind speed high-priority processing time period duty ratio to the data classification processing of the Internet of things.
2. The internet of things data hierarchical processing system based on edge calculation according to claim 1, wherein the collecting wind speed time sequence and constructing a wind speed change sequence, determining a concentrated interval of change energy on a time axis through time-frequency analysis, taking a proportion of the concentrated interval to the whole time as an initial wind speed high priority processing time period duty ratio, specifically: Collecting wind speed data output by a wind speed sensing node in a continuous operation period according to fixed sampling intervals, and forming a wind speed time sequence according to time sequence; performing first-order differential processing on the wind speed time sequence to obtain wind speed variation between adjacent sampling moments, and forming a wind speed variation sequence according to time sequence for representing transient variation behavior of wind speed; Expanding the wind speed change sequence under multiple time scales based on a continuous wavelet transformation process to obtain a time-frequency distribution result of wind speed change on a time axis and a scale axis; based on the time-frequency distribution result, extracting the change energy distribution characteristics corresponding to each time position along a time axis, and limiting the change energy distribution characteristics in a high-frequency scale range determined by the whole energy spectrum in a self-adaption mode; in the high-frequency scale range, the variable energy corresponding to each sampling moment is arranged moment by moment along a time axis to form a variable energy time sequence; Carrying out local continuity analysis on the change energy time sequence, taking the energy change trend at the adjacent sampling moment as a continuous judgment basis, and judging the change trend as the same change section when the energy change directions at the adjacent moments are consistent; dividing the time sequence of the change energy into a plurality of energy change sections according to the continuity judging result, and recording the duration time of each section; screening a section with duration time longer than the average section length of the whole time axis from the energy change sections as a concentration section of wind speed change energy on the time axis; And counting the coverage range of the wind speed change energy concentration interval on the whole sampling time axis, and forming a proportional relation between the time length of the concentration interval and the whole sampling time length to obtain the initial wind speed high-priority processing time period duty ratio.
3. The internet of things data hierarchical processing system based on edge computing according to claim 2, wherein the time-redistribution is performed on the initial wind speed high-priority processing period duty ratio based on the time aggregation characteristic of the wind speed change sequence, and the coverage density of the high-change energy segments on the time axis is adjusted to generate a plurality of second wind speed high-priority processing period duty ratios, specifically: Based on the time sequence of the change energy obtained in the last step and the corresponding energy change sections, counting the duration of each energy change section on a time axis, the time interval between the sections and the distribution sequence of the sections on the whole time axis, and forming a time aggregation structure description reflecting the aggregation form of high change energy on the time axis; the time aggregation structure is used as a constraint condition, the time sequence of energy change sections is fixed, the start and stop positions of the sections are abstracted into section units which can be continuously translated along a time axis, and a section constraint sequence for subsequent reassignment is formed; continuously compressing or expanding the interval of the section constraint sequence on the time axis on the premise of keeping the internal change energy distribution form of each energy change section unchanged, so that the mutual distance of the high change energy section on the time axis is changed, and the aggregation degree of the high change energy section in a local time range is changed; The time axis redistribution operation is carried out for a plurality of times, so that a plurality of time distribution results which are different in the aggregation intensity of the high-change energy section are obtained, and each time distribution result corresponds to a concentrated state of the high-change energy on the time axis; re-determining continuous coverage areas of the high-variation energy sections on a time axis based on each time distribution result to form a plurality of groups of new wind speed variation energy concentration sections; And counting the coverage length of each group of new wind speed change energy concentration intervals on the whole sampling time axis, and forming a proportional relation with the whole sampling time length to obtain a plurality of second wind speed high-priority processing time period occupation ratios.
4. The internet of things data hierarchical processing system based on edge computing according to claim 3, wherein the introducing the wind speed high priority processing period duty ratio into the edge node processing period, performing time slicing on the processing period, and forming a processing configuration including processing time allocation and queue classification according to the time-intensive distribution of wind speed variation, specifically comprises: dividing a node continuous operation process into a plurality of processing periods based on fixed scheduling beats of the edge calculation nodes; Introducing the duty ratio of each second wind speed high-priority treatment period obtained in the previous step into a single treatment period as the proportion constraint of the high-priority treatment time in the treatment period in the whole period; In a single processing period, dividing a continuous high-priority processing subperiod along a time axis according to the corresponding wind speed high-priority processing period duty ratio, and taking the rest time as a conventional processing subperiod, so that the processing period is decomposed into a high-priority processing section and a conventional processing section; Based on the distribution position of the wind speed change energy concentration interval on the time axis, which is obtained in the previous stage, time alignment is carried out on the concentration interval and the current processing period, and the time position of high-density occurrence of wind speed change in the processing period is determined; the wind speed data stream entering the edge node is ordered according to the corresponding relation between the time stamp and the high-priority processing subinterval and the change energy concentration interval, so that the data which falls into the high-priority processing subinterval and corresponds to the change energy concentration interval enter a processing scheduling flow before other data; dividing the sequenced wind speed data stream into a high-priority processing queue and a conventional processing queue, wherein the high-priority processing queue corresponds to data in a high-priority processing subperiod, and the conventional processing queue corresponds to data in other time periods; And combining the time slicing result in the processing period with the queue hierarchical structure to form processing configuration corresponding to the second wind speed high-priority processing time period duty ratio one by one.
5. The internet of things data classification processing system based on edge computing according to claim 4, wherein the processing configuration corresponding to the duty ratio of each second wind speed high priority processing period is applied to the edge computing node respectively, and the classification processing of wind speed data is performed specifically as follows: In the edge node of the wind turbine generator, a basic processing link for keeping wind speed data continuously runs, and a high-priority processing link for start-stop judgment and critical state analysis is configured, wherein the high-priority processing link is activated only in a limited time range; for each second wind speed high-priority processing time period duty ratio, in the running process of the wind turbine generator, defining the available time length of the high-priority processing link in the continuous running time according to the duty ratio, so that the high-priority processing link is kept on in the time length, and the rest time is in a closed state; when the high-priority processing link is in an open state, the wind speed data continuously flowing in the corresponding time period is integrally connected to the high-priority processing link, so that the wind speed change process in the time period is completely analyzed in the same processing context; During the opening period of the high-priority processing link, when wind speed data trigger a control event related to pitch angle adjustment or start-stop judgment, the control event and a corresponding wind speed continuous data segment are sent into the high-priority processing link together for joint processing; When the high-priority processing link reaches the opening time limited by the duty ratio, stopping inputting new wind speed data to the high-priority processing link, and recovering the subsequent wind speed data to the basic processing link for continuous processing; and outputting the identification according to the corresponding second wind speed high-priority processing time period duty ratio by using the wind speed analysis result and the control judgment result formed during the high-priority processing link opening period for the comparison analysis of the processing effect under different subsequent duty ratio conditions.
6. The internet of things data hierarchical processing system based on edge computing according to claim 5, wherein the extracting the processing process features under the duty ratio of each wind speed high priority processing period, constructing a reinforcement learning state space, and training with the corresponding processing configuration as an action, to obtain the duty ratio excitation value of each second wind speed high priority processing period, specifically: Recording the opening and closing time periods of the high-priority processing links under the corresponding proportion conditions in the continuous operation process of the edge nodes of the wind turbine generator, and taking a single complete opening-closing process as a grading processing sample unit; In each grading processing sample unit, counting the duration ratio of the high-priority processing link in an open state and the average occupancy rate of the high-priority processing thread in the time period to form a calculation force continuous occupancy feature for describing the situation that the high-priority link occupies calculation force for a long time when the wind speed high-priority processing time period is higher in occupancy rate; During the starting period of the high-priority processing link, according to the continuous distribution condition of the wind speed change sequence in the cut-in wind speed critical area, counting the length of continuous wind speed change fragments which are completely incorporated into the same high-priority processing link, and recording the coverage degree of the fragments in the relevant time range of start-stop judgment, so as to form continuous perception integrity characteristics for describing the problem that the key change process is segmented when the duty ratio is lower; In each hierarchical processing sample unit, recording a start-stop judging result sequence output by a high-priority processing link, counting the times of inversion of adjacent judging results, and forming start-stop judging fluctuation characteristics by combining the consistency of wind speed change directions in corresponding time, wherein the start-stop judging fluctuation characteristics are used for reflecting the influence of incomplete processing context on decision stability; Combining the calculated force continuous occupation feature, the continuous perception integrity feature and the start-stop judgment fluctuation feature according to a fixed sequence to form a processing state vector corresponding to the duty ratio of each second wind speed high-priority processing period, and using the processing state vector as the state input of reinforcement learning; And taking processing configuration corresponding to the duty ratio of each second wind speed high priority processing period as reinforcement learning action input, taking continuous perception integrity feature improvement and start-stop judgment fluctuation feature reduction as positive return factors in the training process, taking excessive calculation power continuous occupation feature as negative return factors, and outputting expected return values corresponding to the duty ratio of each second wind speed high priority processing period as duty ratio excitation values of the duty ratio.
7. The internet of things data classification processing system based on edge computing according to claim 6, wherein the screening of the optimal wind speed high priority processing period duty ratio based on the duty ratio excitation value is applied to internet of things data classification processing, specifically comprises: sequencing the duty ratio excitation values corresponding to the duty ratio of each second wind speed high-priority treatment period according to the order of the duty ratio, and constructing a duty ratio-excitation value corresponding relation sequence; In the corresponding relation sequence of the duty ratio and the excitation value, a change section which tends to be gentle after the duty ratio excitation value increases monotonically along with the increase of the duty ratio is identified, and the position where the excitation value reaches the local maximum for the first time and the subsequent change amplitude is smaller than a preset threshold value is determined; Determining the duty ratio of the second wind speed high-priority processing period corresponding to the position as a candidate optimal duty ratio for representing that the high-priority processing link is prevented from being excessively opened on the premise of ensuring that the processing income is not obviously improved; when a plurality of second wind speed high priority processing time period duty ratios meeting the changing conditions exist, selecting one with the smallest duty ratio value as the final wind speed high priority processing time period duty ratio so as to reduce the long-term occupied time of a high priority processing link; And configuring the duty ratio of the final wind speed high-priority processing period to an edge computing node of the wind turbine, limiting the opening time of the high-priority processing link according to the duty ratio in the operation process, and executing grading processing on wind speed data according to the corresponding processing configuration.

Description

Internet of things data hierarchical processing system based on edge calculation Technical Field The invention relates to the technical field of edge computing, in particular to an Internet of things data hierarchical processing system based on edge computing. Background The internet of things data hierarchical processing based on edge calculation has become an important means for improving data processing efficiency and decision response capability in wind turbines and similar scenes. By deploying high-performance computing resources at the edge nodes, wind speed, pitch angle and generator monitoring data can be processed locally and rapidly, so that data transmission delay is reduced, and preferential response to key control events is realized. The existing method generally carries out grading treatment on wind speed data and control events according to fixed sampling frequency or carries out priority scheduling on important events so as to ensure that edge nodes can complete real-time decisions as efficiently as possible under limited computing power. The wind speed high priority processing period duty ratio becomes one of the core factors affecting the gradation processing effect. By wind speed high priority processing period duty cycle is meant the proportion of the data period incorporated into the high priority processing link to the whole period during the edge node continuous operation period. This duty cycle directly determines the length of time that the high priority processing link is on, thereby affecting the continuity and integrity of the wind speed data and control events in the processing link. If the duty ratio is reasonable in design, the key change data in the critical area of the cut-in wind speed can be ensured to be completely analyzed, and the accuracy of start-stop judgment and pitch control is improved. However, the technical defect exists in that the wind speed high-priority processing time period is too high or too low in ratio, and the data classification processing of the internet of things based on edge calculation is negatively affected. The higher duty ratio can lead a large amount of non-critical wind speed data to enter a high-priority processing link, occupy the calculation force of edge nodes and reduce the processing attention to critical start-stop judgment data, and the lower duty ratio can lead a critical continuous change section to be divided, so that the start-stop judgment is based on fragmentation, and the processing link is difficult to completely sense the wind speed change in a single context. In the case of a lower duty cycle, the high priority processing link is only opened in a very short period of time, and the continuous wind speed variation before and after the cut-in wind speed critical section cannot be completely incorporated into the same processing context. The fragmentation causes that the start-stop judgment is frequently influenced by discontinuous data, reduces the quick response capability of the edge node to the critical state, and simultaneously can cause delay or misjudgment of pitch control decision and reduce the overall control precision of the wind turbine generator. In the case of a higher duty cycle, in particular in the long-term low wind speed phase, a large number of dead wind speed time periods are fed into the high priority processing links, resulting in a continuous occupation of the edge node high priority processing threads. At this time, the attention of the high-priority processing link to the key control event is diluted, so that the attention of the key wind speed change truly triggering the pitch adjustment or the start-stop judgment is reduced in the processing link, thereby affecting the efficiency of the grading processing and the accuracy of the decision. The prior art lacks a fine strategy for controlling the high priority processing period duty ratio of wind speed, and a fixed time or a simple threshold is generally adopted for high priority processing link opening. In the grading processing scene of the critical area of low wind speed-cut-in wind speed, the key change process cannot be completely perceived, the high-priority processing link may occupy calculation power for a long time or key data is fragmented, accurate response of the edge node to start-stop judgment and pitch control events is affected, and performance of the grading processing of the data of the Internet of things in actual wind turbine generator set control is limited. The present invention proposes a solution to the above-mentioned problems. Disclosure of Invention In order to overcome the defects in the prior art, the embodiment of the invention provides an Internet of things data grading processing system based on edge calculation, which solves the problems of incomplete processing and resource waste of calculation force of the edge node Internet of things data in a critical change section of a wind speed critical area through high-priority