A communication management system for a data center flywheel energy storage system

Abstract

The application relates to the technical field of remote control, in particular to a communication management system for a flywheel energy storage system of a data center, which comprises a state synchronization module, an interval determination module, a trend mapping module and an instruction feedback module. In the application, the time sequence deviation caused by path difference is eliminated through unified clock interval acquisition combined with a sliding window average method, the multi-source synchronization accuracy is improved, a time delay fluctuation rate dynamic detection model is constructed based on speed interval transition, the adaptive deviation threshold is set through historical benchmark comparison, the abnormal recognition accuracy is enhanced, a speed range fluctuation curve and a packet loss rate Spearman rank correlation model are established, multi-dimensional coupling analysis of mechanical state and communication quality is realized, the traditional single-dimensional monitoring mode is broken through, a dynamic switching strategy is generated through a double-instruction triggering mechanism combined with a link topology table, a time delay parameter real-time feedback closed-loop control system is constructed, and the real-time performance and reliability of communication scheduling under complex working conditions are improved.

Description

Technical Field

[0001] This invention relates to the field of remote control technology, and in particular to a communication management system for a flywheel energy storage system in a data center. Background Technology

[0002] The field of remote control technology encompasses functions such as command control and status monitoring of remote devices, systems, or networks through information transmission. The core of this technology lies in sending control commands from local devices to remotely controlled objects via electrical or electronic communication, thereby dynamically adjusting or operating their working status and parameters. Remote control technology typically covers communication link establishment, command parsing, status feedback, and real-time monitoring, and is widely used in various scenarios including power systems, industrial automation, smart grids, and remote equipment management. The development of this field has deepened alongside the evolution of communication technology, microprocessors, and embedded systems, demanding high levels of security, real-time performance, and accuracy in the transmission of control commands.

[0003] The communication management system refers to the system structure and implementation mechanism used to coordinate and control communication between remote power equipment. This patent addresses the technical aspects of access management for various terminal devices in a remote power system, dynamic allocation of communication resources, and real-time monitoring and switching control of communication status. Specifically, it achieves this by constructing software control logic that includes functions such as communication parameter identification, channel scheduling control, and data routing configuration, combined with physical link relay and transmission protocol settings. The communication management system relies on information interaction between nodes in the power communication network structure, executing connection switching and path optimization between nodes according to predetermined control strategies to achieve comprehensive coordination and control of the communication process.

[0004] Existing technologies employ a fixed-period polling mechanism for data acquisition. When flywheel speed changes abruptly, the mismatch between the acquisition period and the rate of change in equipment status leads to data timeliness discrepancies. Static threshold detection methods cannot capture the nonlinear fluctuations in communication latency, making them prone to misjudging abnormal states in dynamic operating scenarios. Traditional monitoring systems rely on single-parameter threshold alarm mechanisms, lacking a modeling relationship between mechanical vibration characteristics and communication quality parameters, making it difficult to identify systemic risks. Communication link switching depends on manually configuring preset paths, and response speed is limited by manual operation delays in sudden failure scenarios. For example, when flywheel speed jumps, existing technologies, lacking a correlation model between the rate of speed change and latency fluctuations, cannot predict communication quality degradation trends, resulting in a time lag between control commands and the actual equipment state. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and propose a communication management system for flywheel energy storage systems in data centers.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a communication management system for a data center flywheel energy storage system includes:

[0007] The status synchronization module is used to collect the real-time rotation speed of the flywheel energy storage unit through a unified clock interval, synchronously obtain the response delay and packet loss rate of the communication link, input the parameters into the sliding window averaging method to complete the timestamp alignment, generate the aligned dataset, and transmit it to the interval determination module.

[0008] The interval determination module is used to call the preset speed interval threshold table, perform interval transition judgment on the speed in the alignment dataset, calculate the time delay volatility when a transition is detected, and generate a time delay deviation value by subtracting the current volatility from the historical benchmark value. When the deviation exceeds the limit, a risk trigger instruction is generated and the risk trigger instruction and the alignment dataset are transmitted to the trend mapping module.

[0009] The trend mapping module is used to parse the rotational speed sequence in the aligned dataset, extract the rotational speed range, input the range sequence into the sliding window averaging method to generate an oscillation curve, call the packet loss rate in the dataset, calculate the curve correlation using the Spearman rank correlation coefficient, and generate a trend warning instruction when the coefficient exceeds the limit, which is then transmitted to the instruction feedback module.

[0010] As a further embodiment of the present invention, the aligned dataset includes timestamp-aligned rotation speed data, communication link response delay sequence, and packet loss rate parameter set. The risk triggering instruction specifically includes a delay volatility exceeding limit event identifier and a transition interval code. The delay deviation value includes absolute value deviation and relative change rate. The trend warning instruction includes oscillation curve correlation coefficient and range fluctuation mode. The target link number specifically refers to the link topology priority sequence and switching fault tolerance threshold parameter.

[0011] As a further aspect of the present invention, the historical benchmark value is the sliding average of the latency volatility over the previous N detection periods, where N is an integer greater than or equal to 5;

[0012] The warning coefficient threshold for the Spearman rank correlation coefficient is an absolute value greater than 0.7.

[0013] As a further aspect of the present invention, the state synchronization module includes:

[0014] The data acquisition and link monitoring submodule acquires the three-phase current frequency of the flywheel energy storage unit through a unified clock interval, monitors the signal transmission time difference and data packet reception interval of the communication link, counts the timestamp difference of adjacent data packets and calculates the number of packet losses per unit time, measures the average time difference between the start and reception of signal transmission, and generates a real-time speed sequence and link delay rate.

[0015] The timestamp alignment processing submodule extracts the standard deviation of the timestamp interval of the real-time rotation speed sequence, uses the sliding window averaging method to perform linear interpolation compensation on three adjacent sampling points within the window, dynamically adjusts the window coverage to twice the delay period according to the link delay rate, performs secondary smoothing filtering on the compensated timestamps, and generates an aligned timestamp sequence.

[0016] The alignment dataset generation submodule matches the aligned timestamp sequence with the original rotation speed values ​​using Euclidean distance, removes data points whose distance values ​​exceed twice the standard deviation, and integrates the three parameters of rotation speed value, latency rate, and number of packet loss items in the order of timestamps to generate the aligned dataset.

[0017] As a further aspect of the present invention, the double delay period and the secondary smoothing filter are used in conjunction to reduce the interference of delay jitter on timestamp alignment.

[0018] As a further aspect of the present invention, the interval determination module includes:

[0019] The transition detection submodule calls the preset speed range threshold table, scans adjacent speed sampling points in the aligned dataset point by point, compares the current speed value with the previous and next interval boundary values, and when the speed is detected to cross the threshold boundary, records the transition timestamp and marks the interval number to generate an interval transition event sequence.

[0020] The fluctuation calculation submodule extracts the time difference between adjacent transitions to form a sequence based on the interval transition event sequence, using the formula:

[0021] ;

[0022] Calculate the time delay dispersion coefficient, and subtract the dispersion coefficient from the historical baseline value to generate the time delay deviation value;

[0023] in, This represents the time delay deviation value. This represents the time difference between the k-th adjacent transitions. This represents the average time difference within the current detection cycle. , These represent the speed range, For the number of effective jumps, The historical baseline standard deviation To avoid the extremely small constant of division by zero;

[0024] The deviation assessment submodule compares the time delay deviation value with a preset deviation threshold. When the deviation value exceeds the upper limit of the threshold, a risk trigger command is generated, and the risk trigger command is output after being timestamped with the alignment dataset.

[0025] As a further aspect of the present invention, the trend mapping module includes:

[0026] The range sequence extraction submodule parses the speed sampling values ​​in the aligned dataset, calculates the absolute difference between the peak and valley values ​​of speed in each sampling period, arranges the differences in time order to form a continuous sequence, and generates the speed range sequence.

[0027] The oscillation curve generation submodule, based on the fluctuation period characteristics of the speed range sequence, uses the sliding window averaging method to perform quadratic polynomial fitting on seven consecutive range values ​​within the window, calculates the standard deviation of the fitted curve and the original sequence, adjusts the window width until the standard deviation is lower than the preset fluctuation tolerance threshold, and generates the range oscillation curve.

[0028] The trend warning determination submodule calls the packet loss rate parameter in the aligned dataset, extracts the amplitude change rate of the range oscillation curve, and uses the Spearman rank correlation coefficient to calculate the rank correlation between the amplitude change rate and the packet loss rate. When the absolute value of the correlation coefficient exceeds the set warning coefficient threshold, a trend warning instruction is generated.

[0029] As a further aspect of the present invention, the system further includes:

[0030] The instruction feedback module is used to receive the risk trigger instruction and the trend warning instruction. When both instructions exist, it generates a target link number based on the link topology table, performs a switching operation, and then feeds back the current latency parameter to the status synchronization module.

[0031] As a further aspect of the present invention, the current latency parameter includes a latency baseline value after handover and a set of link health indicators.

[0032] As a further aspect of the present invention, the instruction feedback module includes:

[0033] The dual-instruction verification submodule receives the risk trigger instruction and the trend warning instruction, detects the overlap interval of the timestamps of the two instructions, and generates a dual-instruction valid identifier when the duration of the overlap interval exceeds the preset instruction synchronization threshold.

[0034] The link switching execution submodule queries the link topology table based on the valid identifier of the dual instructions, filters the three link numbers with the highest latency parameters among the currently active links, calculates the average latency growth rate of the three links over the last five periods, selects the link number with the lowest average growth rate, and generates the target link number.

[0035] The delay parameter feedback submodule executes the communication link switching operation corresponding to the target link number, records the first data transmission delay value of the link after the switch, calculates the difference between the delay value and the delay value before the switch, and generates the current delay parameter.

[0036] The average delay growth rate is the delay change rate calculated using the weighted moving average method, with the weighting coefficients in reverse chronological order being 0.5, 0.3, 0.15, 0.04, and 0.01.

[0037] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0038] This invention eliminates timing deviations caused by transmission path differences by using a unified clock interval acquisition combined with a sliding window averaging method, thereby improving the synchronization accuracy of multi-source parameters. A dynamic detection model for delay fluctuation rate is generated based on speed range transition judgment, and an adaptive deviation threshold is established by comparing historical benchmark values, enhancing the accuracy of abnormal state identification. A Spearman rank correlation analysis model of speed range oscillation curve and packet loss rate is constructed to achieve multi-dimensional coupled analysis of equipment mechanical state and communication quality parameters, breaking through the traditional single-dimensional monitoring mode. A dual-instruction triggering mechanism combined with a link topology table generates a dynamic switching strategy, forming a closed-loop control system for real-time feedback of delay parameters, improving the real-time performance and reliability of communication resource scheduling under complex operating conditions. Attached Figure Description

[0039] Figure 1 This is a system flowchart of the present invention;

[0040] Figure 2 This is a flowchart of the state synchronization module of the present invention;

[0041] Figure 3 This is a flowchart of the interval determination module of the present invention;

[0042] Figure 4 This is a flowchart of the trend mapping module of the present invention;

[0043] Figure 5 This is a flowchart of the instruction feedback module of the present invention. Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0045] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0046] Example 1

[0047] Please see Figure 1 A communication management system for a data center flywheel energy storage system includes:

[0048] The status synchronization module is used to collect the real-time rotation speed of the flywheel energy storage unit through a unified clock interval, synchronously obtain the response delay and packet loss rate of the communication link, input the parameters into the sliding window averaging method to complete the timestamp alignment, generate the aligned dataset, and transmit it to the interval determination module.

[0049] The interval determination module is used to call the preset speed interval threshold table, perform interval transition judgment on the speed in the alignment dataset, calculate the time delay volatility when a transition is detected, and generate a time delay deviation value by subtracting the current volatility from the historical benchmark value. When the deviation exceeds the limit, a risk trigger instruction is generated and the risk trigger instruction and the alignment dataset are passed to the trend mapping module.

[0050] The trend mapping module is used to parse the speed sequence in the aligned dataset, extract the speed range value, input the range value sequence into the sliding window averaging method to generate an oscillation curve, call the packet loss rate in the dataset, use the Spearman rank correlation coefficient to calculate the curve correlation, and generate a trend warning instruction when the coefficient exceeds the limit, which is then passed to the instruction feedback module.

[0051] The instruction feedback module is used to receive risk trigger instructions and trend warning instructions. When both instructions are present, it generates the target link number based on the link topology table, performs the switching operation, and feeds back the current latency parameters to the status synchronization module.

[0052] The alignment dataset includes timestamp-aligned rotation speed data, communication link response delay sequence, and packet loss rate parameter set. The risk triggering instruction specifically includes delay volatility exceeding limit event identifier and transition interval code. The delay deviation value includes absolute value deviation and relative change rate. The trend warning instruction includes oscillation curve correlation coefficient and range fluctuation mode. The target link number specifically refers to the link topology priority sequence and handover fault tolerance threshold parameter. The current delay parameter includes the delay baseline value after handover and the link health indicator set.

[0053] The historical benchmark value is the moving average of the latency volatility over the previous N detection periods, where N is an integer greater than or equal to 5;

[0054] The warning threshold for the Spearman rank correlation coefficient is an absolute value greater than 0.7.

[0055] Please see Figure 2 The state synchronization module includes:

[0056] The data acquisition and link monitoring submodule acquires the three-phase current frequency of the flywheel energy storage unit through a unified clock interval, monitors the signal transmission time difference and data packet reception interval of the communication link, counts the timestamp difference of adjacent data packets and calculates the number of packet losses per unit time, measures the average time difference between the start and reception of signal transmission, and generates a real-time speed sequence and link delay rate.

[0057] The data acquisition and link monitoring submodule starts operating, synchronously sampling the real-time parameters of the three-phase AC power of the flywheel energy storage unit at a precisely set uniform clock interval, specifically 10 milliseconds. At a certain sampling moment, for example... At 10:00 Hz, the sensor system acquired the current frequencies of phase A (49.98 Hz), phase B (50.01 Hz), and phase C (49.99 Hz). These frequency data, along with the precise sampling time, were then used to obtain the current frequencies. The data is recorded and archived by the system. This data acquisition process continues at a fixed interval of 10 milliseconds, thereby constructing a time-series three-phase current frequency database.

[0058] Simultaneously, the communication link monitoring function is activated, focusing on assessing the real-time status of the industrial Ethernet link upon which data transmission depends. The system actively sends probe data packets with a specific structure and precisely records the complete time from the sending of these data packets to the remote device's confirmation of receipt and return. This is used to determine the one-way transmission time of the signal between the two ends of the link. For example, if a probe data packet is sent at 10:00:00.000, the remote end confirms receipt at 10:00:00.015 and immediately responds, and the response packet arrives at the source at 10:00:00.030, then one observation of the one-way transmission time difference is 15 milliseconds. The system also rigorously examines the data packet reception sequence and meticulously analyzes the precise time differences between adjacent data packets arriving at the receiving end. If data packet P1 arrives at time point... Arrival, followed by data packet P2 at time point The arrival time, the receiving interval between the two is... This interval will be compared and analyzed with the expected transmission interval of 10 milliseconds.

[0059] Furthermore, the system performs timestamp verification on all successfully received data packets within a preset statistical period (here, 1 second), extracts the original sending timestamp contained in each data packet, and calculates the difference between the timestamps of any two consecutive data packets in the sequence. If the... The timestamp of each data packet is , No. The timestamp of each data packet is The difference between the two timestamps is then... Based on this analysis, the system accurately counts the exact number of data packets that fail to arrive within a unit statistical period (1 second). If the system expects to receive 100 data packets within that 1-second period, but only 98 data packets actually arrive successfully, the number of packet losses within that unit of time is counted as 2. Furthermore, by performing multiple independent measurements of the complete transmission time from signal transmission to final acknowledgment of reception, the system calculates the arithmetic mean of these time differences. Specifically, this involves performing 50 independent transmission time difference measurements within 1 second to obtain... The average time difference of such a set of time difference data in milliseconds is The result is 16.2 milliseconds.

[0060] Based on the collected three-phase current frequency data, the system utilizes the inherent frequency-to-speed conversion relationship of the flywheel energy storage system (assuming the flywheel motor has 2 pole pairs, the speed calculation formula is: Speed ​​(RPM) = Frequency (Hz) × 60 / 2 = Frequency (Hz) × 30). The system converts the average three-phase frequency at each sampling moment (e.g., (49.98 + 50.01 + 49.99) / 3 = 49.993 Hz) into the corresponding real-time speed (49.993 × 30 = 1499.79 RPM). This process continues, forming a data pair sequence consisting of timestamps and corresponding speed values, i.e., the real-time speed sequence. Simultaneously, based on the calculated number of packet losses per unit time (2 times / second) and the average signal transmission time difference (16.2 milliseconds), a comprehensive evaluation and generation of the link latency index is performed. Finally, this submodule outputs the real-time speed sequence and link latency, providing crucial raw data support for subsequent accurate system status analysis and decision-making.

[0061] The timestamp alignment processing submodule extracts the standard deviation of the timestamp interval of the real-time rotation speed sequence, uses the sliding window averaging method to perform linear interpolation compensation on three adjacent sampling points within the window, dynamically adjusts the window coverage to twice the delay period according to the link delay rate, and performs secondary smoothing filtering on the compensated timestamps to generate an aligned timestamp sequence.

[0062] The timestamp alignment processing submodule receives the real-time rotational speed sequence generated by the aforementioned data acquisition and link monitoring submodule. Due to inherent network transmission delays and jitter, the timestamps of the data points in this sequence are not strictly evenly distributed. The first step is to extract all timestamp data from this real-time rotational speed sequence, calculate the interval between each pair of timestamps, and further calculate the standard deviation of these intervals. Assume there exists a timestamp subsequence. The given timestamp intervals are 0.000 seconds, 0.012 seconds, 0.019 seconds, 0.031 seconds, and 0.040 seconds, respectively. The calculated timestamp intervals are 0.012 seconds, 0.007 seconds, 0.012 seconds, and 0.009 seconds. The average of these intervals is (0.012 + 0.007 + 0.012 + 0.009) / 4 = 0.010 seconds. The sum of the squares of the differences between each interval and the average is... Second Therefore, the standard deviation is The standard deviation, measured in seconds, objectively reflects the unevenness of the timestamp sequence.

[0063] Next, the system uses the sliding window averaging method to perform linear interpolation compensation on the timestamps of three adjacent speed sampling points within the window. The initial sliding window size is set to 3 data points. For three consecutive sampling points within the window (denoted as their timestamps), If the midpoint Compared to based on Theoretical timestamps derived from the ideal sampling interval (10 milliseconds) If there is a deviation in seconds, then according to and right The timestamps are linearly adjusted, and the adjusted timestamps are... The system will further adjust the coverage of this sliding window based on the link status reflected by the link latency rate calculated by the previous submodule. Specifically, the window size will be expanded to twice the average latency period. The average latency period here is determined based on the average time difference between the data acquisition and link monitoring submodules (16.2 milliseconds), hence twice the latency period is... In milliseconds, considering the original sampling interval is 10 milliseconds, the number of data points covered by the adjusted window is approximately 10 ... The window size is dynamically adjusted by rounding up to 4 data points. The reason for this is that longer network latency often causes a systematic shift in the timestamps of multiple consecutive data points. Using a larger window can more accurately capture and compensate for this collective timestamp drift caused by latency jitter.

[0064] After interpolation compensation, a preliminary aligned timestamp sequence is obtained. This sequence is then subjected to a second smoothing filter, specifically a simple moving average filter with a window size of 3. For any timestamp in the sequence... (Non-boundary points), their smoothed value is This step aims to further eliminate residual minute, high-frequency jitter in the timestamp sequence. After the above processing, a more uniformly distributed and more accurate aligned timestamp sequence is finally generated.

[0065] The alignment dataset generation submodule matches the alignment timestamp sequence with the original rotation speed values ​​using Euclidean distance, removes data points whose distance values ​​exceed twice the standard deviation, and integrates the three parameters of rotation speed value, latency rate, and number of packet loss items in the order of timestamp to generate the alignment dataset.

[0066] The alignment dataset generation submodule operates on the aligned timestamp sequence output from the previous stage and its associated original rotational speed measurements. Its primary task is to match and associate each optimized timestamp in the aligned timestamp sequence with the potentially biased timestamps recorded during the acquisition of the original rotational speed data. This matching operation is not a simple replacement; rather, it assigns the precisely aligned timestamp to its closest temporally adjacent original rotational speed measurement, ensuring the time reference for the rotational speed data points is corrected. For each formed (aligned timestamp, original rotational speed value) data pair, the system calculates the degree of difference in rotational speed values ​​between it and its neighboring data points. One specific approach is to compare the rotational speed values ​​of the current data point within a specific small window. The local average of the rotational speed values ​​of several adjacent data points before and after it. absolute difference The system calculates the standard deviation of the sequence based on these differences among all data points. For example, if the calculated series of differences is {2,3,1,5,2,4,30,3,2} RPM, its average is 6.22 RPM and its standard deviation is... The calculated value is 8.95 RPM.

[0067] The system sets a data removal threshold to twice the standard deviation, i.e. RPM, this threshold is set based on statistical principles. Generally, sample points exceeding twice the standard deviation of the mean are considered potential outliers. Then, the system iterates through all data points; if the RPM of a data point differs from its local mean... If the RPM exceeds 17.9 (such as 30 RPM in the above sequence), the data point is identified as an outlier and removed from the current sequence. After all outlier data points are identified and removed, the core data points with timestamps aligned and valid RPM values ​​are retained.

[0068] Finally, the submodule integrates these filtered and valid (aligned timestamp, rotational speed value) data pairs with three key parameters generated by the data acquisition and link monitoring submodule at the same timestamp (or the closest aligned timestamp obtained through interpolation): link latency (specifically 16.2 milliseconds) and packet loss count per unit time (specifically 2 times / second). All integrated data entries are arranged strictly according to the chronological order of the aligned timestamps, forming structured record units. Thus, a complete aligned dataset containing time-synchronized rotational speed, link latency, and packet loss count is constructed.

[0069] Double the delay period and double smoothing filter work together to reduce the interference of delay jitter on timestamp alignment.

[0070] Please see Figure 3 The interval determination module includes:

[0071] The transition detection submodule calls the preset speed range threshold table, scans adjacent speed sampling points in the aligned dataset point by point, compares the current speed value with the previous and next interval boundary values, and when the speed is detected to cross the threshold boundary, records the transition timestamp and marks the interval number to generate an interval transition event sequence.

[0072] The primary component of the interval determination module is the transition detection submodule, whose core function is to accurately identify the transition of the flywheel energy storage system speed between different preset operating state intervals. This process begins by retrieving an internally stored, detailed speed interval threshold table, which clarifies the upper and lower limits of the speed value corresponding to each operating state.

[0073] Table 1 Threshold Table for Flywheel Energy Storage Speed ​​Range:

[0074] .

[0075] Referring to Table 1, this table specifically lists the speed ranges and their precise boundary values ​​for the flywheel energy storage system under different operating conditions. These thresholds are determined based on the design specifications, safety operation requirements, and typical operating condition characteristics summarized from long-term operating experience of the flywheel energy storage system. Each row defines a specific operating range. For example, range number 3 is defined as the "rated operating range," and its corresponding speed range is 1500 RPM to 3000 RPM.

[0076] The transition detection submodule strictly follows the chronological order, scanning the centrally recorded rotational speed samples in the aligned dataset point by point. For the currently processed rotational speed sample point, let its rotational speed value be 2980 RPM, and the corresponding timestamp be... The system retrieves the rotational speed value of the immediately preceding sampling point, which is 2950 RPM in this case. Simultaneously, it compares the current rotational speed value of 2980 RPM with the boundary values ​​of all intervals in the threshold table. Since 2980 RPM is still within the boundary of interval 3 (1500 RPM-3000 RPM), the system determines that no cross-interval state transition has occurred. If the next sampling point in the sequence shows a rotational speed value of 3020 RPM, its timestamp is... At this point, the system detects that the rotational speed has increased from 2980 RPM (still within interval 3) to 3020 RPM. This value clearly exceeds the upper limit of 3000 RPM in interval 3 and enters the range of interval 4 (3000 RPM-3500 RPM). This phenomenon constitutes a speed transition event that crosses the preset threshold boundary. The system will accurately record the timestamp of this transition at this time. This timestamp can be taken as... or and The average value is calculated, and a new interval number of 4 is simultaneously assigned. If the previous state of the sampling point was interval 3, the system generates a transition event record from interval 3 to interval 4. This record contains the exact time of the transition and the target interval number 4. Through continuous scanning and judgment of the entire aligned dataset, a sequence of interval transition events arranged in chronological order is finally generated. Each event in the sequence is recorded in detail with the time point of the transition and the new interval number entered.

[0077] The fluctuation calculation submodule extracts the time difference between adjacent transitions from the interval transition event sequence to form a sequence, using the formula:

[0078] ;

[0079] Calculate the time delay dispersion coefficient, and subtract the dispersion coefficient from the historical baseline value to generate the time delay deviation value;

[0080] in, This represents the time delay deviation value. This represents the time difference between the k-th adjacent transitions. This represents the average time difference within the current detection cycle. , These represent the speed range, For the number of effective jumps, The historical baseline standard deviation To avoid the extremely small constant of division by zero;

[0081] The fluctuation calculation submodule operates immediately following the transition detection submodule. Its processing object is the interval transition event sequence generated in the previous stage. First, the submodule extracts the time intervals between all consecutive transition events from this event sequence, thus forming a new time difference data sequence. If the timestamps of each transition recorded in the transition event sequence are sequentially... The extracted sequence of adjacent transition time differences is then... .

[0082] Within a specific detection period, suppose the system observes This effective interval transition correspondingly yielded... The time difference between each transition is denoted as... Specifically, a total of 5 effective transitions occurred within the current detection cycle, namely... The five time differences measured are as follows: Second, Second, Second, Second, Seconds. Next, the submodule applies the following defined formula to calculate the time delay discrepancy coefficients:

[0083] ;

[0084] The parameters in this formula and how they are obtained are explained below: Representing the The observed time differences between adjacent transitions are directly derived from the analysis of transition event sequences. Represents the current testing cycle Time difference The arithmetic mean, through The calculation yielded the result. and These represent the highest and lowest rotational speeds recorded during the operation of the flywheel energy storage unit within the current evaluation period. These two extreme speed values ​​were extracted from the aligned dataset. After querying, the data within this period... RPM, RPM. This refers to the number of valid transitions within the current detection period. Based on actual observations, this period... . This represents a historical baseline standard deviation, which is calculated by analyzing the transition time differences recorded over a large number of historical periods under long-term stable operation of the flywheel energy storage system. The standard deviation is determined through statistical analysis, and its setup process is as follows: collect continuous... A stable operating cycle Standard deviation data The average of these standard deviations is calculated as the historical baseline standard deviation. , After calculation, we obtained Second. It is a minimal normal quantity introduced to prevent the denominator from being zero in calculations; its value is much smaller than 0. Typical values ​​are set to ensure numerical stability of the calculations. Second.

[0085] To ensure the correctness of the calculations of each physical quantity in the formula, all time-related parameters ( All parameters related to rotational speed use "seconds" as the unified unit. All of them use "RPM" (revolutions per minute) as a unified unit, and the formula structure itself is designed to correctly combine these different types of physical quantities.

[0086] Perform a calculation example: First calculate :

[0087] Second.

[0088] Next calculation :

[0089] ;

[0090] ;

[0091] ;

[0092] ;

[0093] ;

[0094] so Then the square root term Seconds. Speed ​​range term RPM. Sum in the denominator. Seconds. Substitute into the formula to calculate. : RPM. Calculated time delay dispersion coefficients. RPM.

[0095] Subsequently, the system compares the current delay dispersion coefficient with a preset historical benchmark value and calculates the difference. This historical benchmark value is calculated based on the system's long-term stable operation period. The values ​​are obtained through statistical analysis (taking the median or average), and this historical baseline value is set at 750 RPM. The process for determining this baseline value is as follows: the data collection system continuously... Within a confirmed stable operating cycle value, After sorting them, we take the median or calculate their arithmetic mean, resulting in 750 RPM. The generated delay deviation value is: RPM. The advantage of this formula is that it not only measures the discreteness of the transition time interval but also combines it with the amplitude of rotational speed changes during operation and normalizes it relative to historical fluctuation levels and the observed sample size, thus providing a more comprehensive quantitative indicator of system instability. This time delay deviation value of 155.18 RPM will be passed to subsequent evaluation stages.

[0096] The deviation assessment submodule compares the time delay deviation value with the preset deviation threshold. When the deviation value exceeds the upper limit of the threshold, a risk trigger command is generated, and the risk trigger command is output after being timestamped with the alignment dataset.

[0097] The deviation assessment submodule receives the time delay deviation value, specifically 155.18 RPM, precisely calculated and transmitted by the fluctuation calculation submodule. It then rigorously compares this value with an internally set deviation threshold. The specific value of this threshold is set with full consideration of the stringent operational stability requirements of flywheel energy storage systems, as well as the results of in-depth statistical analysis of the distribution characteristics of time delay deviation values ​​preceding significant operational anomalies in a large amount of historical operational data. The typical setting process involves analyzing the time delay deviation values ​​under massive amounts of normal operating state data and critical abnormal state data using statistical methods, selecting a specific value that effectively distinguishes between these two states while ensuring sufficient advance warning. In this embodiment, the deviation threshold is set as follows: The upper limit of the value is 120 RPM. This threshold was determined through rigorous experimental verification: In a controlled laboratory environment, different levels of communication link interference and system load fluctuations were actively simulated, and the time delay deviation value generated by the system was recorded synchronously. It was observed that when the flywheel energy storage system began to show initial signs of instability, such as when its output control accuracy decreased by 5%, the corresponding time delay deviation value consistently appeared in the range of 110 RPM to 130 RPM in multiple repeated experiments. Therefore, considering both the sensitivity and reliability of the early warning, 120 RPM was selected as the critical threshold for judging risk. At the same time, the experimental data also showed that when the time delay deviation value was below 80 RPM, the system exhibited a highly stable operating state.

[0098] The core comparison operation of the submodule is to check whether the currently acquired latency deviation value of 155.18 RPM is greater than the preset upper limit of the deviation threshold of 120 RPM. In this specific example, since 155.18 RPM > 120 RPM, this judgment condition is met, indicating that the latency deviation of the current system has exceeded the normal fluctuation range. When the deviation value is accurately determined to exceed the preset upper limit, the submodule will immediately generate a risk trigger instruction. This instruction contains a precise timestamp, which strictly corresponds to the timestamp of the latest data point in the aligned dataset used to calculate the excess deviation value. Specifically, if the data sequence used for deviation value calculation ends at the timestamp... Then the timestamp within the risk trigger instruction is assigned a value. Or in A tiny fixed time offset is added to the existing data. The generated risk trigger instruction (containing a risk indication and its corresponding timestamp) is then strictly timestamped with the complete alignment dataset (which also contains precise timestamp information). Both are then passed as output to subsequent processing units. This synchronization process ensures a precise time match between the risk instruction and its corresponding data state. One approach is to insert the risk trigger instruction as a special marker event into the closest timestamp position in the alignment dataset, or to output it as a parallel metadata stream synchronized with the main data stream. This output clearly indicates that the current system's latency deviation exceeds the acceptable range, and the system operation may face potential risks.

[0099] Please see Figure 4 The trend mapping module includes:

[0100] The range sequence extraction submodule parses the speed sampling values ​​in the aligned dataset, calculates the absolute difference between the peak and valley values ​​of speed in each sampling period, arranges the differences in time order to form a continuous sequence, and generates the speed range sequence.

[0101] The primary task of the range sequence extraction submodule is to extract deeper dynamic behavioral characteristics from the speed data of the flywheel energy storage system. Its initial processing is handled by this submodule, which first analyzes the speed sampling value sequence contained in the aligned dataset. This sequence consists of a series of precise timestamps and their corresponding speed values, represented as... The submodule uses a pre-defined sampling period as the basic unit to segment the speed data. The duration of this sampling period is fixed at 1 second. Within each 1-second sampling period, the submodule accurately identifies the highest (peak) and lowest (valley) speed values ​​that have occurred within that time window. Let's say a 1-second time window... arrive Within this 1-second period, the system observed a sequence of rotational speed values ​​of {1495, 1500, 1510, 1505, 1490, 1515, 1508, 1498, 1502, 1512} RPM (assuming one point every 0.1 seconds, for a total of 10 points). Therefore, within this 1-second period, the recorded peak rotational speed is 1515 RPM, and the valley rotational speed is 1490 RPM.

[0102] Next, the submodule calculates the absolute difference between the identified peak and trough values, i.e. RPM, the calculated difference (25 RPM), is defined as the speed range within that specific 1-second sampling period. The submodule divides the entire aligned dataset into several 1-second sampling periods sequentially, and repeats the peak-valley value lookup and range calculation process for each period, thus obtaining a series of speed range values ​​arranged chronologically. If the speed ranges calculated for three consecutive 1-second sampling periods are 25 RPM, 30 RPM, and 22 RPM, these values ​​constitute a continuous time series of speed ranges. Finally, the submodule generates a speed range sequence that dynamically reflects the magnitude of speed fluctuations over different consecutive time periods. This sequence is represented as... ,in It is the first The speed range value calculated within a 1-second sampling period.

[0103] The oscillation curve generation submodule is based on the fluctuation period characteristics of the speed range sequence. It uses the sliding window averaging method to perform a quadratic polynomial fitting on seven consecutive range values ​​within the window, calculates the standard deviation of the fitted curve and the original sequence, and adjusts the window width until the standard deviation is lower than the preset fluctuation tolerance threshold to generate the range oscillation curve.

[0104] The oscillation curve generation submodule receives the rotational speed range sequence output by the range sequence extraction submodule; a specific sequence segment is... RPM, and based on the inherent oscillation periodicity of this sequence, constructs an oscillation curve that smoothly reflects its trend. The core method of the sub-module is to use sliding window processing technology, performing a quadratic polynomial fitting operation on seven consecutive speed range values ​​within the window. Seven data points are selected as the initial window width, based on the balance between signal smoothing and trend tracking effects in engineering practice. This number can reflect the main fluctuation trend in a short period of time, and effectively smooth out some random noise. For a specific seven-point window, such as from... arrive Data point sequence The submodule is dedicated to solving a problem of the form . The goal is to find a quadratic polynomial whose curve minimizes the mean square error between the polynomial curve and the seven actual data points. The coefficients are then solved. The standard procedure is to apply the least squares method to these seven data points to construct and solve the corresponding normal equation system.

[0105] After obtaining the fitted quadratic polynomial (which has been determined) After obtaining the values, use this polynomial to calculate the fitted values ​​corresponding to the x-coordinates (1 to 7) of the original seven data points. Then, the standard deviation between the fitted curve and the original data sequence segment is calculated precisely using the following formula: A specific standard deviation is obtained and set to 2.5 RPM. The submodule compares this calculated standard deviation with a pre-set fluctuation tolerance threshold, which is the fluctuation tolerance threshold mentioned above. The specific value is set based on the statistical characteristics analysis of the inherent natural fluctuation range of the speed range sequence of the flywheel energy storage system under long-term stable operation conditions, combined with engineering experience and experimental data on the vibration intensity that the system can safely withstand. The specific setting process is as follows: through a series of experiments, the vibration response of the system under different disturbances is monitored. When the local fitting standard deviation of the speed range sequence is lower than 3.0 RPM, all operating indicators of the system show stability, and the control system does not need to make additional intervention. Based on this, the fluctuation tolerance threshold is prudently set to 3.0 RPM.

[0106] If the standard deviation calculated by the current window (2.5 RPM) is lower than the fluctuation tolerance threshold (3.0 RPM), the system determines that the current window width of 7 is appropriate. If the calculated standard deviation is higher than 3.0 RPM, the submodule will automatically adjust the window width, typically by increasing the number of data points within the window, for example, from 7 to 9 points. Then, it will re-perform quadratic polynomial fitting and subsequent standard deviation calculations using the new window data. This iterative adjustment process will continue until the newly calculated standard deviation successfully falls below the fluctuation tolerance threshold (3.0 RPM), or the window width reaches a preset maximum allowable upper limit. This dynamic adjustment of the window width... The mechanism aims to ensure that the final fitted curve accurately captures and reflects the main fluctuation trend of the speed range without overfitting local noise due to an excessively narrow window. Once a suitable window width is determined (or if an initial width of 7 is sufficient), the submodule moves this sliding window forward by one data point in the time series and repeats the complete fitting and verification process on the new window's data set. This is achieved by smoothly connecting the fitted values ​​of each window's center point (or other representative points selected according to a specific strategy), or by performing a smooth transition after segmented fitting of the entire speed range sequence. Ultimately, this submodule successfully generates a range oscillation curve that clearly and smoothly represents the main fluctuation trend of the speed range.

[0107] The trend warning determination submodule calls the packet loss rate parameter in the aligned dataset, extracts the amplitude change rate of the range oscillation curve, and uses the Spearman rank correlation coefficient to calculate the rank correlation between the amplitude change rate and the packet loss rate. When the absolute value of the correlation coefficient exceeds the set warning coefficient threshold, a trend warning instruction is generated.

[0108] The core responsibility of the trend warning and judgment submodule is to deeply analyze potential trend characteristics in the system that indicate a possible deterioration in the operating status. This submodule first precisely retrieves the packet loss rate parameter corresponding to the current analysis time window from the aligned dataset. Here, the packet loss rate is defined as the number of data packets actually lost per unit time (consistent with the sampling period of the range sequence, which is 1 second), divided by the total number of data packets expected to be received within that time period, multiplied by 100%. For example, if the system expects to receive 100 data packets within a certain 1-second period, but actually detects that 2 data packets failed to be received, then the packet loss rate for that period is calculated as follows: Simultaneously, the submodule retrieves the range oscillation curve generated by the previous stage (oscillation curve generation submodule) and accurately calculates the rate of change of the curve's amplitude over time. Here, the amplitude can be defined as half the difference between the peak and trough values ​​of the range oscillation curve within a local complete cycle. Alternatively, more directly, the numerical sequence of the oscillation curve itself (smoothed range values) can be used as the analysis object, and its rate of change is obtained by calculating the numerical difference between adjacent time points on the curve (corresponding to the sampling period of the range sequence). If the range oscillation curve at time point... The value is At the next adjacent time point The value is A simplified representation of the rate of change of amplitude (or curve value) at that moment is: ,because Since the sampling period is fixed (1 second), the rate of change is... By performing this calculation on the entire range oscillation curve, a time series of amplitude change rate is formed.

[0109] Next, the submodule uses the Spearman rank correlation coefficient to quantitatively evaluate the rank (order of magnitude) correlation strength between the obtained amplitude change rate sequence and the packet loss rate sequence observed in the same time period. The calculation process of the Spearman rank correlation coefficient is as follows: First, each data point in the amplitude change rate sequence and the packet loss rate sequence is independently sorted according to its numerical value, and each data point is assigned a rank (i.e., ranking number) in the sequence. If there are data points with the same value in the sequence, the rank of these data points with the same value is the average of the ranks they should occupy. Suppose there are two segments of length... Synchronous sequence segment: Amplitude change rate sequence (unit: RPM / second) is Its corresponding rank sequence is Packet loss rate sequence (unit: %) is Its corresponding rank sequence is Then, calculate the difference between the ranks of each pair of data points that correspond in time. The rank difference sequence is obtained. And calculate the sum of squares of these rank differences. Spearman rank correlation coefficient The calculation formula is: Substitute the above values ​​into: .

[0110] After obtaining a correlation coefficient value of 0.9, the submodule compares the absolute value of this correlation coefficient with a pre-set warning coefficient threshold. This threshold is based on statistical analysis of Spearman's rank correlation coefficient values ​​from a large amount of historical operational data, specifically when a flywheel energy storage system experiences a significant decline in communication link quality (manifested as a high packet loss rate), leading to increased system instability (manifested as drastic changes in speed range amplitude). Specifically, by analyzing historical failure case data, it observes the relationship between the amplitude change rate and the packet loss rate within a specific time window before a serious system failure. The absolute value of the Spearman rank correlation coefficient typically exceeds 0.7. Therefore, to ensure the effectiveness and timeliness of the early warning, the early warning coefficient threshold is set to 0.7. This threshold selection needs to be verified through simulation experiments. In the experiments, different levels of communication interference are artificially introduced to change the packet loss rate, while simultaneously monitoring the changes in the amplitude of the speed range difference and calculating the correlation. It is confirmed that when the absolute value of the correlation coefficient reaches 0.7, the system is already in a potentially high-risk state, and this threshold can provide effective early warning without excessive false alarms. If the absolute value of the Spearman rank correlation coefficient calculated in this example is... It exceeded the preset warning coefficient threshold of 0.7, that is... If the conditions are met, the submodule will generate a trend warning instruction. This trend warning instruction clearly indicates that the current system has detected a significant synchronous correlation between the deteriorating trend of a key operating parameter (speed range amplitude) and the degradation of communication link quality.

[0111] Please see Figure 5 The instruction feedback module includes:

[0112] The dual-instruction verification submodule receives risk trigger instructions and trend warning instructions, detects the overlap interval of the timestamps of the two instructions, and generates a dual-instruction valid identifier when the duration of the overlap interval exceeds the preset instruction synchronization threshold.

[0113] The initial processing unit of the instruction feedback module is a dual-instruction verification submodule. This submodule is designed to simultaneously receive and process risk-triggered instructions generated by the deviation assessment submodule and trend warning instructions generated by the trend warning determination submodule. Each received instruction precisely carries the timestamp information of its generation time. Specifically, the risk-triggered instruction... The included timestamp is (Hour:Minute:Second.Millisecond), Trend Warning Command The included timestamp is The core function of the submodule is to carefully detect the inherent correlation between these two independently generated instructions in the time dimension. Specifically, it is implemented by accurately calculating the absolute difference between the timestamps carried by the two instructions. This difference is then compared with a pre-set instruction synchronization threshold.

[0114] The setting of the instruction synchronization threshold comprehensively considers the inherent computational processing latency of the two independent analysis paths that generate these two instructions (i.e., the risk assessment path and the trend warning path), as well as the system's performance requirements for immediate response to emergency events. The instruction synchronization threshold is set at 200 milliseconds. This threshold is established based on experimental data: in tests of various simulated fault scenarios, it was observed that when the same potential fault source triggers two alarm paths to generate responses, the timestamp difference of their output instructions is usually within 150 milliseconds. Therefore, 200 milliseconds is selected as the upper limit benchmark for judging whether two instructions are synchronized. This setting aims to ensure that these two instructions triggered by the same or closely related system events can be effectively captured.

[0115] Submodule performs calculations: Seconds, or 50 milliseconds. Next, the submodule determines whether this calculated timestamp difference of 50 milliseconds is less than or equal to the preset instruction synchronization threshold of 200 milliseconds. In this specific instance, since 50 milliseconds... The condition of 200 milliseconds is met, clearly indicating that the two instructions are highly synchronized in time. Therefore, they can be reasonably considered as a composite response from the system to system state anomalies of similar or common root causes. When the system confirms that the timestamp difference between the two instructions falls precisely within the preset synchronization threshold, indicating that their "timestamp overlap interval duration" (simplified here as the closeness of the timestamps of the two instructions) meets the predetermined conditions, the submodule will generate a dual-instruction valid identifier. This identifier can be a boolean value (TRUE in this scenario) or a specific status code to indicate that the two critical instructions have successfully passed the synchronization check. If the timestamps of the two instructions meet the set synchronization conditions, the generation of the dual-instruction valid identifier indicates that the system has simultaneously detected potential operational problems from two different analytical dimensions (risk of state change and deterioration trend).

[0116] The link switching execution submodule queries the link topology table based on the dual-instruction valid identifier, filters the three link numbers with the highest latency parameters among the currently active links, calculates the average latency growth rate of the three links over the last five periods, selects the link number with the lowest average growth rate, and generates the target link number.

[0117] After successfully receiving a valid dual instruction flag (with a value of TRUE) from the dual instruction verification submodule, the link switching execution submodule immediately initiates its preset link switching decision process. Its primary core task is to quickly query and parse the internally stored link topology table, which records in detail all available communication links in the current flywheel energy storage system and their latest operating status parameters.

[0118] Table 2 Communication Link Topology Table:

[0119] .

[0120] Referring to Table 2, the table clearly lists the unique number of each communication link in the system, whether it is currently active and available, the current transmission latency monitored in real time, and a key performance evaluation indicator—the weighted average latency growth rate of the last five monitoring periods. The setting and updating of these parameters are continuously performed by the system's link monitoring module. The submodule first filters all links marked as "active" in the topology table. In this table, the active links are LNK001, LNK002, LNK004, and LNK005. Then, these active links are sorted in descending order according to their "current latency" parameter value, and the three links with the highest latency parameter (i.e., the largest latency value) are selected as candidate analysis objects. The current latency of the active links are LNK001 (15ms), LNK002 (25ms), LNK004 (30ms), and LNK005 (12ms). After sorting in descending order, they are LNK004 (30ms), LNK002 (25ms), LNK001 (15ms), and LNK005 (12ms). Therefore, the three links with the highest latency are LNK004, LNK002, and LNK001.

[0121] Next, for these three selected candidate links (LNK004, LNK002, LNK001), the submodule needs to accurately calculate and compare the weighted average of their respective latency growth rates over the most recent five consecutive monitoring periods. Here, the "latency growth rate" is calculated based on the link latency values ​​measured periodically (e.g., per second), and the calculation method is as follows: The "mean" here is not a simple arithmetic average, but a specific weighted moving average method. The weighting coefficients are strictly set in reverse chronological order (i.e., the most recent monitoring period has the highest weight, and the earliest period has the lowest weight), at 0.5, 0.3, 0.15, 0.04, and 0.01 respectively. These weighting coefficients are set to give higher influence to recent data, as they better reflect the current real-time trend of the link. Furthermore, the sum of the weights (0.5 + 0.3 + 0.15 + 0.04 + 0.01) equals 1.0, ensuring the rationality of the weighting. Taking link LNK004 as an example, if its latency growth rate for the most recent five consecutive monitoring periods (from the most recent period to before the fifth period) is as follows... The calculation process for its weighted average delay growth rate is as follows: The same weighted average calculation process was performed on LNK002 and LNK001, and their respective weighted average latency growth rates were 2.5% and 0.8% (as shown in Table 2).

[0122] Finally, the submodule precisely compares the weighted average latency growth rates calculated for the three candidate links (LNK004: 3.578%, LNK002: 2.5%, LNK001: 0.8%), and selects the link with the lowest average growth rate as the target communication link to which the system will switch next. In this specific example, the weighted average latency growth rate of link LNK001 (0.8%) is the lowest among the three. Therefore, the submodule ultimately generates and outputs the target link number LNK001. This number explicitly instructs the communication control unit of the flywheel energy storage system to immediately switch the data transmission task to this communication link, which is expected to be more stable and has the least obvious latency deterioration trend.

[0123] The delay parameter feedback submodule performs the communication link switching operation corresponding to the target link number, records the first data transmission delay value of the link after the switch, calculates the difference between the delay value and the delay value before the switch, and generates the current delay parameter.

[0124] After accurately receiving the target link number (LNK001 in this example) from the link switching execution submodule, the latency parameter feedback submodule immediately initiates and executes the actual communication link switching operation. This operation involves sending explicit control commands to the underlying network management system or communication control unit. The commands require that the current system's data communication traffic be forcibly switched from the active link that may have performance problems (assuming that the system was using the problematic link LNK004 before the switch, and its "current latency" record before the switch was 30 milliseconds) to the newly selected target link LNK001, which is expected to have better performance.

[0125] After the link switching command is successfully executed and data communication is confirmed to be established and running stably on the new link LNK001, the primary task of this submodule is to immediately begin to closely monitor the various performance indicators of the newly activated link LNK001, with particular focus on and accurate recording of the actual transmission latency value exhibited during the first successful data transmission cycle after the switchover is completed. Suppose that after switching to link LNK001, the system accurately measures the round-trip latency of its first complete data interaction to be 12 milliseconds by sending test data packets or analyzing the transmission time of the first batch of service data packets. This 12 milliseconds is regarded as the "first data transmission latency value" of the new link LNK001 after the switchover.

[0126] Subsequently, the submodule performs a precise difference calculation between the newly measured initial data transmission delay value of 12 milliseconds on the new link LNK001 and the "current delay" value recorded by the original active link LNK004 in its last working cycle before the link switching operation (this value is 30 milliseconds, and the information comes from the status data of LNK004 obtained from the link topology table by the link switching execution submodule before the decision). The calculation process is as follows: New link delay - Old link delay = The calculated difference of -18 milliseconds intuitively quantifies the immediate improvement in transmission latency brought about by this link switching operation. Finally, this submodule generates and outputs the "current latency parameter." This parameter can specifically refer to the initial data transmission latency of the new link LNK001 after the switch, which is 12 milliseconds, or it can be the calculated difference of -18 milliseconds reflecting the latency change. The core purpose of this parameter is to evaluate the immediate effectiveness of the link switching decision in real time and to update the system's internal link status records and performance monitoring database, providing a basis for subsequent system optimization and decision-making.

[0127] The average delay growth rate is the delay change rate calculated using the weighted moving average method, with the weighting coefficients in reverse chronological order being 0.5, 0.3, 0.15, 0.04, and 0.01.

[0128] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A communication management system for a data center flywheel energy storage system, characterized by, The system includes: The status synchronization module is used to collect the real-time rotation speed of the flywheel energy storage unit through a unified clock interval, synchronously obtain the response delay and packet loss rate of the communication link, input the parameters into the sliding window averaging method to complete the timestamp alignment, generate the aligned dataset, and transmit it to the interval determination module. The interval determination module is used to call the preset speed interval threshold table, perform interval transition judgment on the speed in the alignment dataset, calculate the time delay volatility when a transition is detected, and generate a time delay deviation value by subtracting the current volatility from the historical benchmark value. When the deviation exceeds the limit, a risk trigger instruction is generated and the risk trigger instruction and the alignment dataset are transmitted to the trend mapping module. The trend mapping module is used to parse the rotation speed sequence in the aligned dataset, extract the rotation speed range, input the range sequence into the sliding window averaging method to generate an oscillation curve, call the packet loss rate in the dataset, calculate the curve correlation using the Spearman rank correlation coefficient, and generate a trend warning instruction when the coefficient exceeds the limit, which is then transmitted to the instruction feedback module. The trend mapping module includes: The range sequence extraction submodule parses the speed sampling values ​​in the aligned dataset, calculates the absolute difference between the peak and valley values ​​of speed in each sampling period, arranges the differences in time order to form a continuous sequence, and generates the speed range sequence. The oscillation curve generation submodule, based on the fluctuation period characteristics of the speed range sequence, uses the sliding window averaging method to perform quadratic polynomial fitting on seven consecutive range values ​​within the window, calculates the standard deviation of the fitted curve and the original sequence, adjusts the window width until the standard deviation is lower than the preset fluctuation tolerance threshold, and generates the range oscillation curve. The trend warning determination submodule calls the packet loss rate parameter in the aligned dataset, extracts the amplitude change rate of the range oscillation curve, and uses the Spearman rank correlation coefficient to calculate the rank correlation between the amplitude change rate and the packet loss rate. When the absolute value of the correlation coefficient exceeds the set warning coefficient threshold, a trend warning instruction is generated.

2. The communication management system for a data center flywheel energy storage system of claim 1, wherein, The aligned dataset includes timestamp-aligned rotation speed data, communication link response delay sequence, and packet loss rate parameter set. The risk triggering instruction specifically includes a delay volatility exceeding limit event identifier and a transition interval code. The delay deviation value includes absolute value deviation and relative change rate. The trend warning instruction includes oscillation curve correlation coefficient and range fluctuation mode.

3. The communication management system for a data center flywheel energy storage system according to claim 2, characterized in that, The historical benchmark value is the moving average of the latency volatility over the previous N detection periods, where N is an integer greater than or equal to 5; The warning coefficient threshold for the Spearman rank correlation coefficient is an absolute value greater than 0.

7.

4. The communication management system for a data center flywheel energy storage system of claim 3, wherein, The status synchronization module includes: The data acquisition and link monitoring submodule acquires the three-phase current frequency of the flywheel energy storage unit through a unified clock interval, monitors the signal transmission time difference and data packet reception interval of the communication link, counts the timestamp difference of adjacent data packets and calculates the number of packet losses per unit time, measures the average time difference between the start and reception of signal transmission, and generates a real-time speed sequence and link delay rate. The timestamp alignment processing submodule extracts the standard deviation of the timestamp interval of the real-time rotation speed sequence, uses the sliding window averaging method to perform linear interpolation compensation on three adjacent sampling points within the window, dynamically adjusts the window coverage to twice the delay period according to the link delay rate, performs secondary smoothing filtering on the compensated timestamps, and generates an aligned timestamp sequence. The alignment dataset generation submodule matches the aligned timestamp sequence with the original rotation speed values ​​using Euclidean distance, removes data points whose distance values ​​exceed twice the standard deviation, and integrates the three parameters of rotation speed value, latency rate, and number of packet loss items in the order of timestamps to generate the aligned dataset.

5. The communication management system for a data center flywheel energy storage system of claim 4, wherein, The double delay period, together with the secondary smoothing filter, is used to reduce the interference of delay jitter on timestamp alignment.

6. The communication management system for a data center flywheel energy storage system of claim 5, wherein, The interval determination module includes: The transition detection submodule calls the preset speed range threshold table, scans adjacent speed sampling points in the aligned dataset point by point, compares the current speed value with the previous and next interval boundary values, and when the speed is detected to cross the threshold boundary, records the transition timestamp and marks the interval number to generate an interval transition event sequence. The fluctuation calculation submodule extracts the time difference between adjacent transitions to form a sequence based on the interval transition event sequence, using the formula: ； Calculate the time delay dispersion coefficient, and subtract the dispersion coefficient from the historical baseline value to generate the time delay deviation value; in, This represents the time delay deviation value. This represents the time difference between the k-th adjacent transitions. This represents the average time difference within the current detection cycle. , These represent the speed range, For the number of effective jumps, The historical baseline standard deviation To avoid the extremely small constant of division by zero; The deviation assessment submodule compares the time delay deviation value with a preset deviation threshold. When the deviation value exceeds the upper limit of the threshold, a risk trigger command is generated, and the risk trigger command is output after being timestamped with the alignment dataset.

7. The communication management system for a data center flywheel energy storage system according to claim 1, characterized in that, The system also includes: The instruction feedback module is used to receive the risk trigger instruction and the trend warning instruction. When both instructions exist, it generates a target link number based on the link topology table, performs a switching operation, and feeds back the current latency parameter to the status synchronization module. The target link number specifically refers to the link topology priority sequence and the switching fault tolerance threshold parameter.

8. The communication management system for a data center flywheel energy storage system of claim 7, wherein, The current latency parameters include the latency baseline value after handover and a set of link health indicators.

9. The communication management system for a data center flywheel energy storage system of claim 8, wherein, The instruction feedback module includes: The dual-instruction verification submodule receives the risk trigger instruction and the trend warning instruction, detects the overlap interval of the timestamps of the two instructions, and generates a dual-instruction valid identifier when the duration of the overlap interval exceeds the preset instruction synchronization threshold. The link switching execution submodule queries the link topology table based on the valid identifier of the dual instructions, filters the three link numbers with the highest latency parameters among the currently active links, calculates the average latency growth rate of the three links over the last five periods, selects the link number with the lowest average growth rate, and generates the target link number. The delay parameter feedback submodule executes the communication link switching operation corresponding to the target link number, records the first data transmission delay value of the link after the switch, calculates the difference between the delay value and the delay value before the switch, and generates the current delay parameter. The average delay growth rate is the delay change rate calculated using the weighted moving average method, with the weighting coefficients in reverse chronological order being 0.5, 0.3, 0.15, 0.04, and 0.01.

Images