Iot data hierarchical processing system based on edge computing

Abstract

The application discloses an Internet of Things data hierarchical processing system based on edge computing and relates to the technical field of edge computing, which comprises a wind speed change energy period identification module for collecting and constructing a wind speed change sequence and determining an initial high-priority processing period proportion; a high change segment time redistribution module for adjusting the high change energy segment coverage density to generate a plurality of second proportions; a processing cycle division and queue configuration module for forming a hierarchical processing configuration and applying the hierarchical processing configuration in an edge node hierarchical processing execution module to realize hierarchical processing of wind speed data; and a proportion incentive reinforcement learning evaluation module for obtaining a proportion incentive value, and finally, an optimal high-priority processing period proportion application module is used to screen an optimal proportion, so that the problems of incomplete processing of edge node Internet of Things data in a critical wind speed change section and waste of computing resources are solved.

Description

Technical Field

[0001] This invention relates to the field of edge computing technology, and more specifically, to an IoT data hierarchical processing system based on edge computing. Background Technology

[0002] Hierarchical processing of IoT data based on edge computing has become an important means to improve data processing efficiency and decision-making response capabilities in wind turbines and similar scenarios. By deploying high-performance computing resources at edge nodes, wind speed, pitch angle, and generator monitoring data can be processed locally and quickly, thereby reducing data transmission latency and enabling priority response to critical control events. Existing methods typically hierarchically process wind speed data and control events according to a fixed sampling frequency, or prioritize important events to ensure that edge nodes can complete real-time decisions as efficiently as possible with limited computing power.

[0003] The proportion of high-priority wind speed processing time is one of the core factors affecting the effectiveness of tiered processing. The proportion of high-priority wind speed processing time refers to the percentage of data time segments included in the high-priority processing link within the entire continuous operating cycle of an edge node. This proportion directly determines the duration of the high-priority processing link, thus affecting the continuity and completeness of wind speed data and control events in the processing link. A properly designed proportion can ensure that critical changes in wind speed within the critical wind speed zone are fully analyzed, improving the accuracy of start / stop judgments and pitch control.

[0004] However, both excessively high and excessively low proportions of high-priority wind speed processing periods present technical flaws, negatively impacting the hierarchical processing of IoT data based on edge computing. An excessively high proportion leads to a large amount of non-critical wind speed data entering the high-priority processing chain, consuming edge node computing power and reducing attention paid to processing critical start-stop judgment data. Conversely, an excessively low proportion results in the fragmentation of critical continuous change segments, leading to fragmented start-stop judgment criteria and making it difficult for the processing chain to fully perceive wind speed changes within a single context.

[0005] When the proportion is low, the high-priority processing link is only activated for a very short period of time, and continuous wind speed changes before and after entering the critical wind speed zone cannot be fully incorporated into the same processing context. This fragmentation causes start-up and shutdown decisions to be frequently affected by discontinuous data, reducing the rapid response capability of edge nodes to critical states. It may also cause delays or misjudgments in pitch control decisions, reducing the overall control accuracy of the wind turbine.

[0006] When the proportion of wind speeds is high, especially during prolonged periods of low wind speed, a large number of invalid wind speed periods are sent to the high-priority processing chain, causing high-priority processing threads at edge nodes to continuously consume computing power. At this time, the high-priority processing chain's focus on critical control events is diluted, resulting in a decrease in attention being paid to the key wind speed changes that actually trigger pitch adjustments or start / stop decisions, thus affecting the efficiency of tiered processing and the accuracy of decision-making.

[0007] Existing technologies lack sophisticated strategies for controlling the proportion of high-priority wind speed processing periods, typically employing fixed time intervals or simple thresholds to activate high-priority processing links. This results in critical changes not being fully perceived in graded processing scenarios involving low wind speeds and the critical wind speed cutoff zone. High-priority processing links may consume computing power for extended periods, or critical data may be fragmented, affecting the accurate response of edge nodes to start / stop judgments and pitch control events. This limits the performance of IoT data graded processing in actual wind turbine control.

[0008] To address the above problems, this invention proposes a solution. Summary of the Invention

[0009] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide an IoT data hierarchical processing system based on edge computing. This system uses adaptive hierarchical processing based on the high priority processing ratio of wind speed change energy and reinforcement learning optimization to solve the problems of incomplete processing and wasted computing resources of IoT data from edge nodes in critical wind speed change zones.

[0010] To achieve the above objectives, the present invention provides the following technical solution:

[0011] An edge computing-based IoT data hierarchical processing system includes a wind speed change energy period identification module, a high-change segment time redistribution module, a processing cycle time segmentation and queue configuration module, an edge node hierarchical processing execution module, a proportion-based incentive reinforcement learning evaluation module, and an optimal high-priority processing period proportion application module. The wind speed change energy period identification module collects wind speed time series and constructs wind speed change sequences. Through time-frequency analysis, it determines the concentrated intervals of change energy on the time axis, using the proportion of these concentrated intervals to the overall time as the initial wind speed high-priority processing period proportion. The high-change segment time redistribution module redistributes the initial wind speed high-priority processing period proportion based on the temporal clustering characteristics of the wind speed change sequence. By adjusting the coverage density of high-change energy segments on the time axis, it generates several second wind speed high-priority processing period proportions. The processing cycle time... The module for segmentation and queue configuration incorporates the proportion of high-priority wind speed processing periods into the edge node processing cycle, segments the processing cycle over time, and forms a processing configuration that includes processing time allocation and queue hierarchy based on the temporal density distribution of wind speed changes. The edge node hierarchical processing execution module applies the processing configurations corresponding to each proportion of the second high-priority wind speed processing period to the edge computing nodes to perform hierarchical processing of wind speed data. The proportion-incentivized reinforcement learning evaluation module extracts the processing characteristics under each proportion of the high-priority wind speed processing period, constructs a reinforcement learning state space, and trains it using the corresponding processing configuration as actions to obtain the proportion incentive value for each proportion of the second high-priority wind speed processing period. The optimal high-priority processing period proportion application module selects the optimal proportion of the high-priority wind speed processing period based on the proportion incentive value and applies it to the hierarchical processing of IoT data.

[0012] The technical effects and advantages of the IoT data hierarchical processing system based on edge computing of this invention are as follows:

[0013] 1. This invention, through the collaborative work of multiple modules, achieves refined hierarchical processing of wind speed data and key control events, exhibiting significant technical effects and advantages. Technically, this invention utilizes a wind speed change energy period identification module and a high-change segment time redistribution module to precisely characterize and adjust the distribution of high-change energy segments in wind speed data along the time axis, thereby generating several proportions of second-level high-priority processing periods. Combined with a processing cycle time segmentation and queue configuration module, the system can accurately match high-priority processing time periods under different proportion conditions with the processing cycle, and establish high-priority processing queues and regular processing queues, achieving continuous, complete, and orderly hierarchical processing of wind speed data and control events at edge nodes. This technical solution effectively ensures that key change segments entering the critical wind speed zone are fully perceived, improving the accuracy of start / stop judgments and pitch control, enhancing the real-time response capability of edge nodes to critical states, and simultaneously improving the overall efficiency and reliability of IoT data hierarchical processing.

[0014] 2. This invention uses reinforcement learning to model and train processing features under different high-priority processing ratios to obtain ratio incentive values ​​for selecting the optimal ratio. This design can avoid long-term computational resource occupation by high-priority processing links, reduce interference from non-critical data on processing resources, and balance the integrity of continuous sensing with fluctuations in start / stop judgments, while ensuring the continuity and integrity of critical data and processing. Finally, the optimal high-priority processing time period ratio is configured to edge nodes to achieve dynamic, intelligent, and controllable hierarchical data processing. This method significantly improves the processing efficiency, control decision accuracy, and system resource utilization of IoT data in the edge computing environment of wind turbine units, while also having good adaptability to cope with different wind speed variation scenarios, improving the overall stability and reliability of the system. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the structure of the IoT data hierarchical processing system based on edge computing of the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0017] Example 1, Figure 1 This invention presents an IoT data hierarchical processing system based on edge computing, comprising a wind speed change energy period identification module, a high-variability segment time reallocation module, a processing cycle time segmentation and queue configuration module, an edge node hierarchical processing execution module, a proportion-based incentive reinforcement learning evaluation module, and an optimal high-priority processing period proportion application module.

[0018] The wind speed change energy period identification module is used to collect wind speed time series and construct wind speed change series. It determines the concentrated interval of change energy on the time axis through time-frequency analysis, and uses the proportion of the concentrated interval to the total time as the proportion of the initial high-priority processing period of wind speed.

[0019] In this embodiment, the process of collecting wind speed time series and constructing wind speed change series, determining the concentrated intervals of energy change on the time axis through time-frequency analysis, and using the proportion of the concentrated intervals to the total time as the proportion of the initial high-priority processing period for wind speed, specifically involves:

[0020] The wind speed data output by the wind speed sensor node at a fixed sampling interval during a continuous operating cycle is collected and formed into a wind speed time series in chronological order.

[0021] The wind speed time series is subjected to first-order difference processing to obtain the wind speed change between adjacent sampling times, and a wind speed change sequence is formed in chronological order to characterize the instantaneous change behavior of wind speed.

[0022] The wind speed change sequence is expanded at multiple time scales based on a continuous wavelet transform process to obtain the time-frequency distribution of wind speed changes on the time axis and scale axis.

[0023] Based on the time-frequency distribution results, the energy distribution characteristics corresponding to each time position are extracted along the time axis and limited to a high-frequency scale range adaptively determined by the overall energy spectrum.

[0024] Within the high-frequency scale range, the changing energy corresponding to each sampling moment is arranged along the time axis to form a time series of changing energy.

[0025] The time series of energy changes is analyzed for local continuity. The energy change trend of adjacent sampling times is used as the basis for continuous determination. When the energy change direction of adjacent times is consistent, it is determined to be the same change segment.

[0026] Based on the above continuity determination results, the change energy time series is divided into several energy change segments, and the duration of each segment is recorded.

[0027] From the energy change segments, segments whose duration is longer than the average segment length of the overall time axis are selected as the concentrated intervals of wind speed change energy on the time axis.

[0028] The coverage of the concentrated wind speed energy intervals over the entire sampling time axis is statistically analyzed, and the time length of the concentrated intervals is proportional to the overall sampling time length to obtain the proportion of the initial wind speed high-priority processing period.

[0029] It should be noted that the following is a feasible example of calculating the time-frequency expansion of continuous wavelet transform:

[0030] Perform continuous wavelet transform on the wind speed variation sequence:

[0031] ;

[0032] In the formula, These are the wavelet coefficients at scale a and time position b. For the mother wavelet function; It is the conjugate form of the mother wavelet; The wind speed change occurs within the i-th sampling interval; N is the total number of sampling points. This represents the wind speed value at the i-th sampling time. The time interval between adjacent sampling points is a fixed value.

[0033] Time-frequency distribution matrix representation:

[0034] ;

[0035] In the formula, the direction of the row is: ~ ,column direction: ~ .

[0036] In this embodiment, the wind speed time series refers to the raw wind speed data continuously output by the wind speed sensing nodes of the wind turbine during continuous operation, at fixed sampling intervals. The wind speed data is arranged in chronological order of its generation, forming a continuous and ordered numerical sequence on the time axis. This wind speed time series does not perform smoothing, filtering, or thresholding processing on the wind speed values; its purpose is to completely preserve the original change process of wind speed under natural operating conditions, providing basic data support for subsequent analysis of change behavior.

[0037] In this embodiment, the wind speed change sequence refers to a numerical sequence obtained by differential processing of wind speed values ​​at adjacent sampling times based on the wind speed time series. This sequence reflects the magnitude and direction of wind speed changes within adjacent time intervals. By highlighting the relationship between adjacent moments, this change sequence separates the "change behavior" of wind speed from the original absolute value, facilitating subsequent analysis of the concentration and suddenness of wind speed changes along the time axis.

[0038] In this embodiment, the time-frequency distribution result refers to the two-dimensional distribution structure that reflects the intensity of wind speed changes with time and scale after the wind speed change sequence is unfolded at multiple time scales. This distribution structure retains the specific time of the change in the time axis direction and reflects the rate of change in the scale axis direction, so that the energy distribution characteristics of wind speed changes at different time scales can be observed and compared simultaneously.

[0039] In this embodiment, the change energy distribution characteristic refers to the energy characterization value obtained by comprehensively describing the wind speed change intensity corresponding to each time position in the time-frequency distribution results. This value is used to characterize the severity of wind speed change at that time position. The change energy is not a single instantaneous change, but rather a comprehensive consideration of the contribution of that time position to changes at multiple scales, thereby reflecting the importance of that moment in the overall change process.

[0040] In this embodiment, the high-frequency scale range refers to a set of scale intervals corresponding to rapid wind speed changes across the entire scale distribution. This scale range is not manually fixed but is adaptively determined based on the overall distribution of changing energy along the scale axis, primarily carrying rapidly changing energy segments to eliminate interference from slow changes in critical state identification.

[0041] In this embodiment, the change energy time series refers to a one-dimensional sequence formed by arranging the change energy corresponding to each sampling moment in chronological order within a defined high-frequency scale range. This sequence continuously expresses the fluctuation process of wind speed change energy on the time axis, providing a direct basis for subsequent analysis of the aggregation and dispersion of change energy over time.

[0042] In this embodiment, the local continuity analysis refers to the process of analyzing the continuity of energy change along the time axis by using the changing trend of adjacent sampling moments in the energy change time series as the basis for judgment. When the energy change at adjacent moments shows a consistent upward or downward trend, it is regarded as a continuous extension of the same change process, which is used to avoid mistakenly splitting the essentially continuous change behavior into multiple discrete segments.

[0043] In this embodiment, the energy change segment refers to a time interval that exhibits a continuous trend in the energy change time series, formed as a whole based on local continuity analysis. Each energy change segment corresponds to a relatively complete wind speed change process, and its duration is used to measure the stability and impact range of this change process on the time axis.

[0044] In this embodiment, the energy concentration interval refers to a set of time intervals obtained by selecting segments from all energy change segments whose duration is significantly higher than the overall average level. This concentration interval is used to characterize the region where wind speed change energy continuously accumulates on the time axis, reflecting that wind speed change is not occasional noise, but a key change stage with a certain degree of continuity.

[0045] In this embodiment, the initial high-priority processing period percentage for wind speed refers to the ratio between the time length covered by the energy variation concentration interval across the entire sampling time axis and the overall sampling time length. This ratio describes the proportion of time under natural operating conditions when wind speed changes are in a high-energy variation state, providing an initial constraint benchmark for the subsequent high-priority processing time configuration.

[0046] In this embodiment, adaptive determination does not refer to online learning or model prediction, but rather to a process of dynamically limiting the analysis scope based on the overall statistical distribution characteristics of the current wind speed change sequence itself, without introducing external thresholds. This method ensures that the determined analysis scale matches the current wind speed operating state, avoiding the problem of insufficient adaptability to different operating conditions due to fixed parameter settings.

[0047] It should be noted that the above-described process for identifying wind speed change energy periods is not intended to detect single instantaneous wind speed anomalies, but rather to identify stages of wind speed change that exhibit a continuous, high-density distribution over time. By separating the changing behavior from the original wind speed values ​​and further analyzing the aggregation state of the changing energy over time, the configuration of subsequent high-priority processing periods no longer relies on fixed thresholds or human experience, but is based on the temporal structural characteristics of wind speed changes themselves. This provides a continuous and engineering-feasible temporal basis for the hierarchical processing of subsequent edge nodes.

[0048] The high-variability segment time redistribution module is used to redistribute the proportion of the initial high-priority processing period based on the temporal aggregation characteristics of the wind speed variation sequence. By adjusting the coverage density of the high-variability energy segment on the time axis, it generates several second high-priority processing period proportions for wind speed.

[0049] In this embodiment, based on the temporal aggregation characteristics of the wind speed change sequence, the proportion of the initial high-priority processing period for wind speed is time-redistributed. By adjusting the coverage density of high-variable energy segments on the time axis, several proportions of the second high-priority processing period for wind speed are generated, specifically as follows:

[0050] Based on the time series of energy changes and their corresponding energy change segments obtained in the previous step, the duration of each energy change segment on the time axis, the time interval between segments, and the distribution order of the segments on the overall time axis are statistically analyzed to form a time clustering structure description that reflects the clustering pattern of high energy changes on the time axis.

[0051] The time aggregation structure is described as a constraint condition, the time sequence of energy change segments is fixed, and the start and end positions of each segment are abstracted as segment units that can be continuously translated along the time axis, forming a segment constraint sequence for subsequent redistribution.

[0052] While keeping the energy distribution pattern within each energy change segment unchanged, the spacing of the segment constraint sequence on the time axis is continuously compressed or expanded, so that the mutual distance between high energy change segments on the time axis changes, thereby changing their degree of aggregation within the local time range.

[0053] By performing the time axis redistribution operation multiple times, several time distribution results with different concentration intensities in high energy variation segments are obtained, and each time distribution result corresponds to a concentration state of high energy variation on the time axis.

[0054] Based on the time distribution results, the continuous coverage interval of the high-change energy segment on the time axis is redefined, forming multiple new sets of wind speed change energy concentration intervals;

[0055] The coverage length of each new wind speed change energy concentration interval on the overall sampling time axis is statistically analyzed and proportionally related to the overall sampling time length to obtain the proportion of several second wind speed high priority processing periods.

[0056] In this embodiment, the high-variable-energy segment refers to a continuous time segment with consistently high variable energy and a certain length on the time axis, obtained after local continuity analysis of the wind speed variable energy time series in the previous module. This segment reflects the rapid change behavior of wind speed within a continuous time range and serves as the basic unit for the redistribution of high-priority processing periods.

[0057] In this embodiment, the temporal clustering structure description refers to an abstract description formed by statistically analyzing the position, duration, and inter-interval relationships of all high-energy-varying segments on the time axis. This description characterizes the distribution pattern and clustering degree of high-energy-varying segments on the overall time axis. It reflects the arrangement order, density, and possible local clustering characteristics of the segments on the time axis, providing a reference for subsequent redistribution operations.

[0058] In this embodiment, the segment constraint sequence refers to a sequence formed by abstracting the start and end positions of high-energy-varying segments into units that can be continuously translated along the time axis, while maintaining the energy distribution pattern within each segment. The segment constraint sequence is used to define the relative order and adjustable temporal position range of each segment, so that, without changing the energy characteristics within the segments, the local concentration intensity of high-energy-varying regions on the overall time axis can be altered by adjusting the segment spacing.

[0059] In this embodiment, the time axis redistribution operation refers to the continuous compression or expansion of the spacing between high-energy-varying segments in the segment constraint sequence on the time axis, thereby changing the clustering density and distribution pattern of the segments within a local time range. This operation ensures that the wind speed variation characteristics within each segment remain unchanged, while changing the relative positions of the segments to make some segments more compact or more dispersed in a local time, thus forming different temporal clustering states.

[0060] In this embodiment, the high-variable energy concentration state refers to the specific clustering pattern of high-variable energy segments on the time axis after time axis reallocation. Each state corresponds to a different local density between segments. This state is used to generate multiple sets of new wind speed variation energy concentration intervals, providing a variety of possible high-priority processing time ratio schemes for edge node hierarchical processing.

[0061] In this embodiment, the proportion of the second high-priority processing period for wind speed refers to the proportion obtained by recalculating the continuous coverage time of the high-variable energy segment on the overall sampling time axis under each high-variable energy concentration state and proportionally relating it to the total length of the entire sampling period. This proportion reflects the relative amount of time occupied by the high-priority processing period in the total sampling period under different time axis redistribution states, and is used for the configuration selection of subsequent edge node hierarchical processing.

[0062] It should be noted that the temporal clustering structure description not only reflects the individual length of the segment, but also includes the relative intervals between segments on the time axis. This interval information is crucial for adjusting the local clustering intensity, and can avoid problems such as loss of key change segments or over-concentration due to excessive compression or over-dispersion.

[0063] It should be noted that the multiple executions of the time axis redistribution operation are not random adjustments, but rather a systematic operation based on maintaining the segment order. By continuously fine-tuning the segment spacing, several highly variable energy concentration states are obtained, ensuring that each state can be practically applied through the edge node scheduling cycle in engineering implementation, thereby forming multiple feasible schemes for the proportion of high priority processing time periods for the second wind speed.

[0064] It should be noted that, to ensure the continuity and integrity of each high-energy-change segment during the redistribution process, the energy change patterns and fluctuation characteristics within each segment remain unchanged during the adjustment process; only the start and end positions of the segment on the time axis change. This ensures that when generating the proportion of the second high-priority wind speed processing period, the actual wind speed change information of each segment is still fully preserved, facilitating the continuous perception of key change processes by edge nodes during actual processing.

[0065] The processing cycle time segmentation and queue configuration module is used to introduce the proportion of high-priority wind speed processing periods into the edge node processing cycle, segment the processing cycle by time, and form a processing configuration that includes processing time allocation and queue hierarchy based on the time density distribution of wind speed changes.

[0066] In this embodiment, the proportion of high-priority wind speed processing periods is introduced into the edge node processing cycle, the processing cycle is divided into time segments, and a processing configuration including processing time allocation and queue hierarchy is formed based on the time-density distribution of wind speed changes. Specifically:

[0067] Based on the fixed scheduling cycle of edge computing nodes, the continuous operation process of nodes is divided into several processing cycles;

[0068] The proportion of each high-priority processing period for the second wind speed obtained in the previous step is introduced into a single processing cycle as a constraint on the proportion of high-priority processing time in the entire cycle within that processing cycle.

[0069] Within a single processing cycle, based on the proportion of the corresponding high-priority processing period for wind speed, a continuous high-priority processing sub-period is divided along the time axis, and the remaining time is used as a regular processing sub-period, thereby decomposing the processing cycle into a high-priority processing segment and a regular processing segment.

[0070] Based on the distribution of the concentrated wind speed change energy range on the time axis obtained in the previous stage, the concentrated range is time-aligned with the current processing cycle to determine the time location where wind speed changes occur with high density within the processing cycle.

[0071] The wind speed data streams entering the edge nodes are sorted according to their timestamps and the correspondence between the high-priority processing sub-periods and the range of concentrated energy changes, so that the data that falls into the high-priority processing sub-periods and the corresponding range of concentrated energy changes enters the processing and scheduling process before other data.

[0072] The sorted wind speed data stream is divided into a high-priority processing queue and a regular processing queue. The high-priority processing queue corresponds to the data within the high-priority processing sub-period, while the regular processing queue corresponds to the data within the other time periods.

[0073] The time segmentation results within the processing cycle are combined with the queue hierarchy to form a processing configuration that corresponds one-to-one with the proportion of the second high-priority wind speed processing period.

[0074] It should be noted that the following is a calculation example of the feasible introduction of the proportion of high-priority wind speed processing periods:

[0075] Let the set of the proportion of the second high-priority processing time periods obtained from the previous stage be:

[0076] ;

[0077] in, The proportion of time allocated to high-priority processing within a processing cycle;

[0078] High-priority sub-period length:

[0079] For any processing cycle and percentage Define the length of the high-priority processing sub-period as:

[0080] ;

[0081] In the formula, The length of the high-priority sub-period is used for processing. The duration of a single processing cycle.

[0082] It should be noted that the following is a feasible example of calculating the time mapping and sorting of wind speed data streams:

[0083] Suppose the wind speed data stream entering the edge node is:

[0084] ;

[0085] In the formula, For wind speed data stream, This is the wind speed value. This corresponds to the timestamp.

[0086] High-priority decision function:

[0087] ;

[0088] In the formula, It is a high-priority decision function.

[0089] ; This refers to the time locations where wind speed changes occur most frequently within the current processing cycle; This is the set of concentrated intervals of changing energy output from the previous module. Sub-times are processed with high priority; This is the kth processing cycle.

[0090] In this embodiment, the processing cycle refers to the basic time unit into which the continuous operation process of the edge computing node is divided according to a fixed scheduling rhythm. Each processing cycle contains several consecutive time periods, used to perform high-priority processing and regular processing of wind speed data on the node. The length of the processing cycle is related to the node scheduling rhythm to ensure that the node can perform hierarchical processing according to a preset time strategy during continuous operation.

[0091] In this embodiment, the high-priority processing sub-period refers to a continuous time period defined along the time axis within a single processing cycle based on the proportion of the second wind speed high-priority processing sub-period. This time period is specifically used to process data within the concentrated energy range of wind speed changes. The high-priority processing sub-period ensures that critical wind speed change segments receive priority computing resources at edge nodes, thereby enabling rapid response and detailed analysis.

[0092] In this embodiment, the regular processing sub-period refers to the remaining time period in the processing cycle excluding the high-priority processing sub-period. During the regular processing sub-period, wind speed data is processed according to the basic processing chain and does not have high-priority computing power allocation, but it still maintains the complete acquisition and basic analysis functions of wind speed data to supplement the data processing needs outside the high-priority processing sub-period.

[0093] In this embodiment, wind speed data stream sorting refers to rearranging continuous wind speed data entering the edge node according to the correspondence between its timestamp and the high-priority processing sub-periods and the concentrated energy change intervals defined within the processing cycle. This sorting ensures that during scheduling, data falling into high-priority processing sub-periods and corresponding to high-energy change intervals are preferentially allocated to the node's high-priority processing links, thereby achieving priority perception and processing of critical changes.

[0094] In this embodiment, the high-priority processing queue refers to the queue consisting of all data falling into the high-priority processing sub-time period after sorting. This queue is scheduled for execution in the node corresponding to the high-priority processing link. The high-priority processing queue ensures that critical wind speed change data can be continuously and completely analyzed and processed by the edge nodes in each processing cycle, without being disturbed by the conventional processing flow.

[0095] In this embodiment, the regular processing queue refers to the queue of data that falls into the regular processing sub-period within the processing cycle. This part of the data is executed sequentially along the basic processing link in the node. It does not occupy computing resources first, but still participates in the overall wind speed analysis of the edge node to ensure the integrity and continuity of wind speed data throughout the entire time period.

[0096] It should be noted that the correspondence between high-priority processing sub-periods and concentrated wind speed change energy intervals is not fixed, but dynamically determined based on the actual distribution of change energy intervals within each processing cycle. This ensures that high-priority processing sub-periods can cover continuous data segments of local key changes, thereby avoiding the situation where key change data is segmented or delayed in edge node scheduling.

[0097] It should be noted that the overall design of processing cycle time segmentation and queue hierarchy configuration ensures that the proportion of high-priority processing periods can be flexibly mapped to different time positions within a continuous running cycle, and data priority scheduling is achieved through sorting and queue partitioning. This design can also be extended to multi-node collaborative processing scenarios, enabling multiple nodes in the edge computing network to consistently and efficiently perceive and process key wind speed change data through unified processing cycle segmentation rules and queue hierarchy strategies.

[0098] The edge node hierarchical processing execution module is used to apply the processing configuration corresponding to the proportion of each second wind speed high priority processing period to the edge computing nodes to perform hierarchical processing of wind speed data.

[0099] In this embodiment, the processing configuration corresponding to the proportion of each second high-priority processing period for wind speed is applied to the edge computing nodes to perform hierarchical processing of wind speed data, specifically as follows:

[0100] In the edge nodes of the wind turbine, the basic processing link for wind speed data is kept running continuously, and a high-priority processing link for start / stop judgment and critical state analysis is configured. The high-priority processing link is only activated within a limited time range.

[0101] For each high-priority processing period of the second wind speed, during the operation of the wind turbine, the available time length of the high-priority processing link is determined according to this proportion within the continuous operation time, so that the high-priority processing link remains open within the time length and is closed at other times.

[0102] When the high-priority processing link is enabled, the wind speed data that continuously flows in within the corresponding time period will be connected to the high-priority processing link as a whole, so that the wind speed change process within that time period can be fully analyzed in the same processing context.

[0103] During the high-priority processing link is active, when wind speed data triggers control events related to pitch angle adjustment or start / stop judgment, the control events and the corresponding continuous wind speed data segments are sent to the high-priority processing link for joint processing.

[0104] Once the high-priority processing link reaches the specified activation duration, new wind speed data will stop being input into the high-priority processing link, and subsequent wind speed data will be restored to the basic processing link for further processing.

[0105] The wind speed analysis results and control judgment results generated during the high-priority processing link will be marked and output according to the proportion of the corresponding second wind speed high-priority processing period, for subsequent comparative analysis of processing effects under different proportion conditions.

[0106] In this embodiment, the basic processing link refers to the continuously running wind speed data processing channel in the edge node of the wind turbine, which is used to complete the routine calculation, recording, and basic analysis of wind speed data. The basic processing link ensures that even if the high-priority processing link is not activated, all collected wind speed data can still be completely accessed to the node for basic processing, thereby maintaining a continuous data flow and node status monitoring.

[0107] In this embodiment, the high-priority processing link refers to an independent processing channel within a node specifically used for critical wind speed data analysis, start / stop determination, and critical state identification. This link has a priority computing power allocation function and is activated only within a pre-defined time range to ensure that data in high-energy-varying areas receive a continuous and complete analysis environment during processing.

[0108] In this embodiment, the high-priority processing period length refers to the available time of the high-priority processing link within the continuous operation cycle of the node, based on the proportion of each second-wind-speed high-priority processing period. This time length determines the continuous operation range of the high-priority processing link, ensuring that key wind speed change segments can be fully analyzed within this time period without being cut off or delayed.

[0109] In this embodiment, a continuous wind speed data segment refers to a set of wind speed data that flows into the node in chronological order during the high-priority processing link's operation. Continuous wind speed data segments are accessed through the high-priority processing link within the same processing context to ensure that the dynamic characteristics of key changes and event triggering conditions remain complete and consistent during analysis.

[0110] In this embodiment, control events refer to pitch angle adjustments, start / stop judgments, or other operational signals related to the operational safety and performance of wind turbines triggered during the high-priority processing link analysis of wind speed data. Control events, along with corresponding continuous wind speed data segments for the same time period, are sent to the high-priority processing link, enabling the analysis results to simultaneously reflect data change characteristics and the triggering conditions for operational decisions.

[0111] In this embodiment, the analysis result identification refers to clearly marking the wind speed data analysis results and control judgment results completed during the high-priority processing link's operation period according to the corresponding proportion of the second wind speed high-priority processing period. The purpose of the identification is to support subsequent comparison of processing effects and performance evaluation under different proportion conditions, so that the actual effect of each high-priority processing strategy can be independently quantified and tracked.

[0112] It should be noted that the opening and closing of the high-priority processing link is not a fixed switch, but is strictly executed according to the time length defined by the proportion of the high-priority processing period of the second wind speed, to ensure that the proportion is completely consistent with the actual processing time of the node, and to avoid discontinuity or insufficient coverage of high-priority data analysis caused by human factors or node scheduling fluctuations.

[0113] It should be noted that the design of the edge node hierarchical processing execution module not only ensures the priority processing of critical wind speed changes, but can also be extended to multi-node collaborative scenarios. Through unified high-priority processing link management and proportion time allocation, multiple nodes can synchronously and efficiently perform continuous analysis of critical wind speed events, and provide multi-scale data support for system-level wind turbine operation optimization.

[0114] The percentage incentive reinforcement learning evaluation module is used to extract the processing characteristics of each wind speed high priority processing period, construct a reinforcement learning state space, and train it with the corresponding processing configuration as the action to obtain the percentage incentive value of each second wind speed high priority processing period.

[0115] In this embodiment, the process features of each high-priority processing period for each wind speed are extracted, a reinforcement learning state space is constructed, and the corresponding processing configuration is used as an action for training to obtain the proportion incentive value of each second high-priority processing period for wind speed. Specifically:

[0116] For each applied second wind speed high priority processing period, during the continuous operation of the wind turbine edge node, record the opening and closing time periods of the high priority processing link under the corresponding percentage condition, and take a single complete opening-closing process as a hierarchical processing sample unit.

[0117] Within each graded processing sample unit, the duration of high-priority processing links being in the open state is statistically analyzed, along with the average utilization rate of high-priority processing threads during that time period, forming a continuous computing power utilization characteristic. This is used to characterize the situation where high-priority links occupy computing power for a long time when the proportion of high-priority processing periods for wind speed is relatively high.

[0118] During the high-priority processing link is activated, based on the continuous distribution of the wind speed change sequence within the critical wind speed cut-in zone, the length of continuous wind speed change segments that are completely included in the same high-priority processing link is calculated, and the coverage of the segments within the relevant time range of start-stop judgment is recorded to form a continuous perception integrity feature, which is used to characterize the problem of key change processes being segmented when the proportion is low.

[0119] Within each graded processing sample unit, the sequence of start and stop judgment results output by the high-priority processing link is recorded, the number of times adjacent judgment results are reversed is counted, and combined with the consistency of wind speed change direction within the corresponding time period, start and stop judgment fluctuation characteristics are formed to reflect the impact of incomplete processing context on decision stability.

[0120] The continuous computing power occupancy feature, continuous perception integrity feature, and start / stop judgment fluctuation feature are combined in a fixed order to form a processing state vector corresponding to the proportion of each second wind speed high priority processing period, which is used as the state input for reinforcement learning.

[0121] Using the processing configuration corresponding to the proportion of each second wind speed high priority processing period as the input of reinforcement learning actions, during the training process, the improvement of continuous perception integrity feature and the reduction of start-stop judgment fluctuation feature are used as positive reward factors, and the excessively high computing power continuous occupation feature is used as negative reward factors. The expected reward value corresponding to the proportion of each second wind speed high priority processing period is output as the proportion incentive value of that proportion.

[0122] In this embodiment, the hierarchical processing sample unit refers to the complete time period from the start to the stop of a single high-priority processing link, recorded in the edge node for each applied second wind speed high-priority processing period percentage. Each sample unit includes the running time of the high-priority processing link under that percentage condition, the processed data segment, and the triggered control events, which are used for subsequent feature extraction and reinforcement learning state construction.

[0123] In this embodiment, the sustained computing power occupancy characteristic refers to the proportion of computing resources consumed during the period when the high-priority processing link is active within the hierarchical processing sample unit, and the proportion of its duration. This characteristic depicts the situation where a high proportion of the high-priority processing link may lead to long-term concentrated use of node computing power, providing feedback on resource occupancy for reinforcement learning.

[0124] In this embodiment, the continuous sensing integrity feature refers to the length and coverage of continuous wind speed change segments that are fully incorporated into the same link processing during the high-priority processing link's operation. This feature reflects that when the proportion of high-priority processing periods is low, key wind speed change processes may be segmented or omitted, thereby affecting the link's ability to sense continuous changes.

[0125] In this embodiment, the start / stop judgment fluctuation characteristic refers to the comprehensive characteristic formed by combining the number of times adjacent judgment results reverse in the start / stop judgment result sequence output by the high-priority processing link within each hierarchical processing sample unit with the consistency of wind speed change direction. This characteristic is used to reflect the sensitivity of the high-priority processing link to the continuity of wind speed changes and control stability. When the proportion is unreasonable, the judgment fluctuation will increase, which may lead to instability in the control strategy.

[0126] In this embodiment, the processing state vector refers to a vectorized representation formed by combining the continuous computing power occupancy feature, the continuous perception integrity feature, and the start / stop judgment fluctuation feature in a fixed order, which is used as the state input for reinforcement learning. This vector systematizes the processing characteristics of nodes under different proportion conditions, facilitating the establishment of state-action correspondences during training.

[0127] In this embodiment, the reinforcement learning action input refers to the processing configuration adopted for each hierarchical processing sample unit, including the duration of the high-priority processing link, the queue hierarchy, and the data scheduling strategy. The action input enables reinforcement learning to explore the impact of processing strategies at different proportions on the state vector feedback, thereby evaluating the effectiveness of proportion adjustments.

[0128] It should be noted that the design of the expected reward value takes into account both the improvement of continuous perception integrity features and the reduction of start-stop judgment fluctuation features as positive rewards, and also considers the negative penalty of excessive continuous computing power occupation features, so as to balance the processing effect and resource consumption in reinforcement learning training, and make the percentage incentive value truly reflect the comprehensive performance of different high priority processing periods in node execution.

[0129] It should be noted that the edge node proportion incentive reinforcement learning evaluation module not only evaluates the processing effect of a single node, but can also be extended to multi-node collaborative scenarios. By unifying state vectors and action design, it can achieve cross-node proportion strategy optimization, thereby obtaining the optimal scheduling effect in system-level wind speed data hierarchical processing and control decision-making.

[0130] The optimal high-priority processing period ratio application module is used to select the optimal wind speed high-priority processing period ratio based on the ratio incentive value and apply it to the hierarchical processing of IoT data.

[0131] In this embodiment, the step of selecting the optimal wind speed high-priority processing period based on the proportion incentive value and applying it to the hierarchical processing of IoT data specifically involves:

[0132] The incentive values ​​corresponding to the proportion of each high-priority treatment period of the second wind speed are sorted in order of proportion size, and a sequence of proportion-incentive value correspondence is constructed.

[0133] In the ratio-excitation value correspondence sequence, identify the segment where the ratio excitation value increases monotonically and then tends to level off as the ratio increases, and determine the position where the excitation value first reaches a local maximum and the subsequent change is less than a preset threshold.

[0134] The proportion of the second wind speed high-priority processing period corresponding to the location is determined as the candidate optimal proportion, which is used to characterize the avoidance of excessive opening of high-priority processing links under the premise of ensuring that the processing benefits no longer significantly increase.

[0135] When there are multiple proportions of second wind speed high priority processing periods that meet the aforementioned change conditions, the one with the smallest proportion is selected as the final proportion of wind speed high priority processing periods to reduce the long-term occupation time of the high priority processing link.

[0136] The proportion of the final high-priority processing period is configured to the edge computing nodes of the wind turbine. During operation, the opening duration of the high-priority processing link is limited according to this proportion, and the wind speed data is processed in a hierarchical manner according to the corresponding processing configuration.

[0137] In this embodiment, the proportion-incentive value correspondence sequence refers to the sequence formed by arranging the proportion of each second wind speed high-priority processing period and its corresponding proportion incentive value in ascending order of proportion. This sequence is used to visually demonstrate the impact of proportion changes on the processing effect, providing a data foundation for subsequently finding the optimal proportion.

[0138] In this embodiment, the monotonically increasing and then leveling-off transition period refers to the situation where, in the proportion-incentive value correspondence sequence, as the proportion increases, the upward trend of the proportion incentive value reaches a certain point, after which the increase significantly decreases and approaches stability. This period reflects the stage where continuing to increase the proportion of high-priority processing periods can no longer significantly improve processing revenue, and is an important criterion for selecting the optimal proportion.

[0139] In this embodiment, a local maximum position refers to the position in the percentage-incentive value correspondence sequence where the percentage incentive value first reaches a relatively high level, and in subsequent consecutive percentages, the change in the incentive value is less than a preset threshold. This position serves as a candidate optimal percentage, indicating that, while ensuring high processing efficiency, no further significant benefits are gained by increasing the percentage, thereby avoiding the over-activation of high-priority processing links.

[0140] In this embodiment, the candidate optimal proportion refers to the proportion of the second high-priority processing period corresponding to the identified local maximum location. This proportion serves as a reference standard for the operation of the high-priority processing link, ensuring that the high-priority processing link remains active during critical time periods while avoiding unnecessary long-term resource occupation.

[0141] In this embodiment, the final high-priority processing time percentage for wind speed refers to selecting the smallest percentage value as the final configuration value when multiple candidate optimal percentages exist. This selection strategy balances processing efficiency and resource utilization, reducing the continuous operation time of high-priority processing links while ensuring processing benefits and optimizing the use of computing power at edge nodes.

[0142] In this embodiment, the hierarchical processing configuration application refers to allocating the proportion of the final high-priority processing time period to the edge computing nodes of the wind turbine. During operation, the activation duration of the high-priority processing link is limited according to this proportion, and the wind speed data stream is processed hierarchically in conjunction with the corresponding processing configuration. This operation ensures that the actual data processing behavior is consistent with the optimized proportion result, achieving refined edge processing guided by the proportion incentive value.

[0143] It should be noted that this module not only determines the optimal proportion of a single node, but also combines multi-node edge computing scenarios to weight or collaboratively analyze the proportion incentive values ​​of different nodes to form a system-level optimal proportion strategy, thereby achieving the optimal allocation of high-priority processing resources in the entire wind farm or IoT data network.

[0144] It should be noted that the selection of the final high-priority processing period ratio for wind speed depends not only on the trend analysis of the ratio incentive value, but also on a comprehensive consideration of node computing power, data traffic, and historical wind speed change patterns. This aims to further improve the processing efficiency of edge nodes and the stability of control decisions, while avoiding conflicts or computing power bottlenecks in high-priority processing links at different time periods. In the embodiments provided by this invention, it should be understood that the disclosed system can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may exist in actual implementation.

[0145] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0146] Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be embraced within the invention. No appended diagram markings in the claims should be construed as limiting the scope of the claims.

[0147] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0148] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0149] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the system can be divided into different functional units or modules to complete all or part of the functions described above.

[0150] In the embodiments provided in this disclosure, it should be understood that the disclosed systems and methods can also be implemented in other ways. The system embodiments described above are merely illustrative; for example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0151] It should be noted that, in this disclosure, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element limited by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0152] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. An IoT data hierarchical processing system based on edge computing, characterized in that, This includes modules for identifying wind speed variation energy periods, reallocating time segments with high variation, dividing processing cycles into time segments and configuring queues, executing hierarchical processing at edge nodes, evaluating performance through reinforcement learning based on proportion incentives, and applying the proportion of optimal high-priority processing periods. The wind speed change energy period identification module is used to collect wind speed time series and construct wind speed change series. It determines the concentrated interval of change energy on the time axis through time-frequency analysis, and uses the proportion of the concentrated interval to the total time as the proportion of the initial high-priority processing period of wind speed. The high-variability segment time redistribution module is used to redistribute the proportion of the initial high-priority processing period based on the temporal aggregation characteristics of the wind speed variation sequence. By adjusting the coverage density of the high-variability energy segment on the time axis, it generates several second high-priority processing period proportions for wind speed. The processing cycle time segmentation and queue configuration module is used to introduce the proportion of high-priority wind speed processing periods into the edge node processing cycle, segment the processing cycle by time, and form a processing configuration that includes processing time allocation and queue hierarchy based on the time density distribution of wind speed changes. The edge node hierarchical processing execution module is used to apply the processing configuration corresponding to the proportion of each second wind speed high priority processing period to the edge computing nodes to perform hierarchical processing of wind speed data. The percentage incentive reinforcement learning evaluation module is used to extract the processing characteristics of each wind speed high priority processing period, construct a reinforcement learning state space, and train it with the corresponding processing configuration as the action to obtain the percentage incentive value of each second wind speed high priority processing period. The optimal high-priority processing period ratio application module is used to select the optimal wind speed high-priority processing period ratio based on the ratio incentive value and apply it to the hierarchical processing of IoT data.

2. The IoT data hierarchical processing system based on edge computing according to claim 1, characterized in that, The process involves collecting wind speed time series data and constructing wind speed change sequences. Time-frequency analysis is used to determine the concentrated intervals of energy changes on the time axis. The proportion of these concentrated intervals to the total time is used as the initial high-priority processing period for wind speed. Specifically: The wind speed data output by the wind speed sensor node at a fixed sampling interval during a continuous operating cycle is collected and formed into a wind speed time series in chronological order. The wind speed time series is subjected to first-order difference processing to obtain the wind speed change between adjacent sampling times, and a wind speed change sequence is formed in chronological order to characterize the instantaneous change behavior of wind speed. The wind speed change sequence is expanded at multiple time scales based on a continuous wavelet transform process to obtain the time-frequency distribution of wind speed changes on the time axis and scale axis. Based on the time-frequency distribution results, the energy distribution characteristics corresponding to each time position are extracted along the time axis and limited to a high-frequency scale range adaptively determined by the overall energy spectrum. Within the high-frequency scale range, the changing energy corresponding to each sampling moment is arranged along the time axis to form a time series of changing energy. The time series of energy changes is analyzed for local continuity. The energy change trend of adjacent sampling times is used as the basis for continuous determination. When the energy change direction of adjacent times is consistent, it is determined to be the same change segment. Based on the above continuity determination results, the change energy time series is divided into several energy change segments, and the duration of each segment is recorded. From the energy change segments, segments whose duration is longer than the average segment length of the overall time axis are selected as the concentrated intervals of wind speed change energy on the time axis. The coverage of the concentrated wind speed energy intervals over the entire sampling time axis is statistically analyzed, and the time length of the concentrated intervals is proportional to the overall sampling time length to obtain the proportion of the initial wind speed high-priority processing period.

3. The IoT data hierarchical processing system based on edge computing according to claim 2, characterized in that, Based on the temporal aggregation characteristics of the wind speed variation sequence, the proportion of the initial high-priority processing period for wind speed is time-redistributed. By adjusting the coverage density of high-variable energy segments on the time axis, several second high-priority processing period proportions for wind speed are generated, specifically: Based on the time series of energy changes and their corresponding energy change segments obtained in the previous step, the duration of each energy change segment on the time axis, the time interval between segments, and the distribution order of the segments on the overall time axis are statistically analyzed to form a time clustering structure description that reflects the clustering pattern of high energy changes on the time axis. The time aggregation structure is described as a constraint condition, the time sequence of energy change segments is fixed, and the start and end positions of each segment are abstracted as segment units that can be continuously translated along the time axis, forming a segment constraint sequence for subsequent redistribution. While keeping the energy distribution pattern within each energy change segment unchanged, the spacing of the segment constraint sequence on the time axis is continuously compressed or expanded, so that the mutual distance between high energy change segments on the time axis changes, thereby changing their degree of aggregation within the local time range. By performing the time axis redistribution operation multiple times, several time distribution results with different concentration intensities in high energy variation segments are obtained, and each time distribution result corresponds to a concentration state of high energy variation on the time axis. Based on the time distribution results, the continuous coverage interval of the high-change energy segment on the time axis is redefined, forming multiple new sets of wind speed change energy concentration intervals; The coverage length of each new wind speed change energy concentration interval on the overall sampling time axis is statistically analyzed and proportionally related to the overall sampling time length to obtain the proportion of several second wind speed high priority processing periods.

4. The IoT data hierarchical processing system based on edge computing according to claim 3, characterized in that, The process of incorporating the proportion of high-priority wind speed processing periods into the edge node processing cycle, dividing the processing cycle into time segments, and forming a processing configuration that includes processing time allocation and queue hierarchy based on the temporal density distribution of wind speed changes, specifically: Based on the fixed scheduling cycle of edge computing nodes, the continuous operation process of nodes is divided into several processing cycles; The proportion of each high-priority processing period for the second wind speed obtained in the previous step is introduced into a single processing cycle as a constraint on the proportion of high-priority processing time in the entire cycle within that processing cycle. Within a single processing cycle, based on the proportion of the corresponding high-priority processing period for wind speed, a continuous high-priority processing sub-period is divided along the time axis, and the remaining time is used as a regular processing sub-period, thereby decomposing the processing cycle into a high-priority processing segment and a regular processing segment. Based on the distribution of the concentrated wind speed change energy range on the time axis obtained in the previous stage, the concentrated range is time-aligned with the current processing cycle to determine the time location where wind speed changes occur with high density within the processing cycle. The wind speed data streams entering the edge nodes are sorted according to their timestamps and the correspondence between the high-priority processing sub-periods and the range of concentrated energy changes, so that the data that falls into the high-priority processing sub-periods and the corresponding range of concentrated energy changes enters the processing and scheduling process before other data. The sorted wind speed data stream is divided into a high-priority processing queue and a regular processing queue. The high-priority processing queue corresponds to the data within the high-priority processing sub-period, while the regular processing queue corresponds to the data within the other time periods. The time segmentation results within the processing cycle are combined with the queue hierarchy to form a processing configuration that corresponds one-to-one with the proportion of the second high-priority wind speed processing period.

5. The IoT data hierarchical processing system based on edge computing according to claim 4, characterized in that, The process configuration corresponding to the proportion of each second wind speed high-priority processing period is applied to the edge computing nodes to perform hierarchical processing of wind speed data, specifically as follows: In the edge nodes of the wind turbine, the basic processing link for wind speed data is kept running continuously, and a high-priority processing link for start / stop judgment and critical state analysis is configured. The high-priority processing link is only activated within a limited time range. For each high-priority processing period of the second wind speed, during the operation of the wind turbine, the available time length of the high-priority processing link is determined according to this proportion within the continuous operation time, so that the high-priority processing link remains open within the time length and is closed at other times. When the high-priority processing link is enabled, the wind speed data that continuously flows in within the corresponding time period will be connected to the high-priority processing link as a whole, so that the wind speed change process within that time period can be fully analyzed in the same processing context. During the high-priority processing link is active, when wind speed data triggers control events related to pitch angle adjustment or start / stop judgment, the control events and the corresponding continuous wind speed data segments are sent to the high-priority processing link for joint processing. Once the high-priority processing link reaches the specified activation duration, new wind speed data will stop being input into the high-priority processing link, and subsequent wind speed data will be restored to the basic processing link for further processing. The wind speed analysis results and control judgment results generated during the high-priority processing link will be marked and output according to the proportion of the corresponding second wind speed high-priority processing period, for subsequent comparative analysis of processing effects under different proportion conditions.

6. The IoT data hierarchical processing system based on edge computing according to claim 5, characterized in that, The process features of the processing under the proportion of each high-priority processing period for wind speed are extracted, a reinforcement learning state space is constructed, and the corresponding processing configuration is used as the action for training to obtain the proportion incentive value of each second high-priority processing period for wind speed. Specifically: For each applied second wind speed high priority processing period, during the continuous operation of the wind turbine edge node, record the opening and closing time periods of the high priority processing link under the corresponding percentage condition, and take a single complete opening-closing process as a hierarchical processing sample unit. Within each graded processing sample unit, the duration of high-priority processing links being in the open state is statistically analyzed, along with the average utilization rate of high-priority processing threads during that time period, forming a continuous computing power utilization characteristic. This is used to characterize the situation where high-priority links occupy computing power for a long time when the proportion of high-priority processing periods for wind speed is relatively high. During the high-priority processing link is activated, based on the continuous distribution of the wind speed change sequence within the critical wind speed cut-in zone, the length of continuous wind speed change segments that are completely included in the same high-priority processing link is calculated, and the coverage of the segments within the relevant time range of start-stop judgment is recorded to form a continuous perception integrity feature, which is used to characterize the problem of key change processes being segmented when the proportion is low. Within each graded processing sample unit, the sequence of start and stop judgment results output by the high-priority processing link is recorded, the number of times adjacent judgment results are reversed is counted, and combined with the consistency of wind speed change direction within the corresponding time period, start and stop judgment fluctuation characteristics are formed to reflect the impact of incomplete processing context on decision stability. The continuous computing power occupancy feature, continuous perception integrity feature, and start / stop judgment fluctuation feature are combined in a fixed order to form a processing state vector corresponding to the proportion of each second wind speed high priority processing period, which is used as the state input for reinforcement learning. Using the processing configuration corresponding to the proportion of each second wind speed high priority processing period as the input of reinforcement learning actions, during the training process, the improvement of continuous perception integrity feature and the reduction of start-stop judgment fluctuation feature are used as positive reward factors, and the excessively high computing power continuous occupation feature is used as negative reward factors. The expected reward value corresponding to the proportion of each second wind speed high priority processing period is output as the proportion incentive value of that proportion.

7. The IoT data hierarchical processing system based on edge computing according to claim 6, characterized in that, The method of selecting the optimal wind speed high-priority processing period based on the proportion incentive value and applying it to the hierarchical processing of IoT data is as follows: The incentive values ​​corresponding to the proportion of each high-priority treatment period of the second wind speed are sorted in order of proportion size, and a sequence of proportion-incentive value correspondence is constructed. In the ratio-excitation value correspondence sequence, identify the segment where the ratio excitation value increases monotonically and then tends to level off as the ratio increases, and determine the position where the excitation value first reaches a local maximum and the subsequent change is less than a preset threshold. The proportion of the second wind speed high-priority processing period corresponding to the location is determined as the candidate optimal proportion, which is used to characterize the avoidance of excessive opening of high-priority processing links under the premise of ensuring that the processing benefits no longer significantly increase. When there are multiple proportions of second wind speed high priority processing periods that meet the aforementioned change conditions, the one with the smallest proportion is selected as the final proportion of wind speed high priority processing periods to reduce the long-term occupation time of the high priority processing link. The proportion of the final high-priority processing period is configured to the edge computing nodes of the wind turbine. During operation, the opening duration of the high-priority processing link is limited according to this proportion, and the wind speed data is processed in a hierarchical manner according to the corresponding processing configuration.

Images