A power distribution cabinet online monitoring method and system based on 5G Internet of Things

Abstract

The present application relates to load switching technical field, specifically to a kind of power distribution cabinet online monitoring method and system based on 5G internet of things, comprising the following steps: through 5G acquisition branch current signal input sliding window extraction dynamic baseline parameter, kalman filter trend prediction generates offset characteristic parameter, if prediction difference is over first threshold and negative, then output adjustment instruction, difference is between two and three threshold then call time slice reconstruction strategy to reduce current group time slice, send switching instruction to activate new link and set initial time slice.In the present application, by constructing dynamic baseline and trend prediction composite mechanism, sliding window extracts current characteristics to form benchmark, kalman filter modeling generates offset parameter, interval division is combined with time slice reconstruction to adjust load, double condition locks abnormal attenuation, multi-level threshold realizes gradual control, fusion prediction characteristic and adaptive control improve coordination ability, optimize load configuration and stabilize power quality.

Description

Technical Field

[0001] The present invention relates to the field of load switching technology, and in particular to a method and system for online monitoring of a distribution cabinet based on 5G Internet of Things. Background Art

[0002] The field of load switching technology encompasses various methods used to regulate electrical energy distribution in power systems to achieve voltage stability, current balance, and safe system operation. The core of this technology is to dynamically adjust the power consumption of each branch or device by switching or distributing different loads during grid operation, thereby optimizing the operating efficiency and load balancing level of the entire system. Its technical system primarily encompasses AC voltage regulation, automatic load transfer, load priority management, fault response mechanisms, and supporting power monitoring methods. It is widely used in power distribution systems, backup power switching systems, and industrial automation power supply scenarios to ensure the continuity of power supply and the stability of power quality.

[0003] Among them, the online monitoring method for distribution cabinets refers to a technical solution for real-time acquisition and analysis of various operating parameters within the distribution cabinet and remote data exchange. The technical matters covered include real-time collection of parameters such as voltage, current, temperature, and humidity, remote transmission of signals, and analysis and processing of background data. Specifically, by integrating power collection devices and sensor units in the distribution cabinet, uploading the collected parameters to the monitoring center using a 5G communication module, and combining remote control instructions to perform load switching operations, comprehensive online monitoring and control adjustment of the distribution system's operating status are achieved. The core basis of this method is the use of a load status identification and distribution system parameter monitoring mechanism based on 5G Internet of Things communication.

[0004] Traditional monitoring methods use preset fixed parameter thresholds for load management, which struggles to adapt to the dynamic and nonlinear fluctuations of grid loads. Static benchmark parameter systems cannot track the continuous evolution of equipment operating conditions. They are prone to parameter drift in scenarios with periodic or intermittent load fluctuations, causing threshold determinations to deviate from actual operating requirements. Most systems lack the ability to predict and analyze load trends, responding only passively based on current data. This can lead to decision lags in scenarios like equipment startup and sudden load increases, making it difficult to proactively mitigate potential risks. Single threshold trigger mechanisms lack multi-level coordinated control strategies, resulting in frequent switching when responding to gradual load changes and insufficient response to sudden load fluctuations, potentially causing equipment protection systems to malfunction. Time-slicing management models typically use fixed time configurations and lack dynamic optimization based on real-time load characteristics. This can easily disrupt the power supply sequence for critical equipment or cause secondary loads to occupy excessive time slices. The coordinated control of data acquisition and command execution lacks precision, and multi-level delays in parameter processing can cause phase deviations between control signals and actual conditions under fast transient conditions. For example, the power distribution system of a manufacturing workshop adopts a fixed time slice allocation strategy, which leads to frequent power supply timing conflicts during the alternating start and stop of production line equipment, resulting in periodic operation disorders of the automation equipment group. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and propose an online monitoring method and system for distribution cabinets based on 5G Internet of Things.

[0006] In order to achieve the above objectives, the present invention adopts the following technical solution: a method for online monitoring of distribution cabinets based on 5G Internet of Things, comprising the following steps:

[0007] S1: The real-time current signals of multiple branches of the power distribution cabinet are collected through the 5G communication unit, and the real-time current signals are input into the sliding window algorithm to extract the minimum value, mean value and range of the current within 5 seconds to generate the dynamic baseline parameters of the current;

[0008] S2: Based on the current dynamic baseline parameters, the Kalman filter algorithm is called to predict the current trend, the difference between the current and next moment predictions is calculated, and the current offset characteristic parameters are generated by combining the absolute value and the direction index;

[0009] S3: comparing the absolute value of the predicted difference in the current offset characteristic parameter with a first threshold value, and if the difference exceeds the threshold and the direction is negative, triggering a dynamic load threshold adjustment instruction, and synchronously outputting the dynamic load threshold adjustment instruction and the current offset characteristic parameter;

[0010] S4: Based on the dynamic load threshold adjustment instruction, the predicted difference is classified using an interval partitioning function. When the difference is between the second and third threshold values, the time slice reconstruction strategy is called to reduce the current group time slice, generate a group switching time parameter, and synchronously update the target group time slice.

[0011] As a further solution of the present invention, the current dynamic baseline parameters include a minimum baseline, a mean baseline, and a range baseline; the current offset characteristic parameters include the absolute value of the predicted difference, a directional index, and a trend confidence; the dynamic load threshold adjustment instruction includes a negative offset flag, a threshold trigger level, and an adjustment amplitude ratio; the group switching time parameters include the current group reduction duration, the target group preset duration, and the switching priority identifier.

[0012] As a further solution of the present invention, the steps of obtaining the current dynamic baseline parameters are specifically as follows:

[0013] S101: Current signals of multiple branches of the power distribution cabinet are collected through the 5G communication unit, the current strength of the multiple branches is recorded every 10 milliseconds, the current signals within the collection period are integrated, and a multi-branch current sequence is generated;

[0014] S102: Based on the multi-branch current sequence, intercept a continuous 5-second data window in chronological order, take the minimum value of the current data of each branch in the window, calculate the arithmetic mean, and count the difference between the maximum and minimum values ​​to generate the window extreme difference, window mean and window minimum value of the multi-branch;

[0015] S103: Call the window extreme value, window mean and window minimum of the multiple branches, arrange the three parameters of the same branch into triplets in order, encapsulate the triplets of all branches of the distribution cabinet in a unified format, and generate current dynamic baseline parameters.

[0016] As a further solution of the present invention, the step of obtaining the current offset characteristic parameter is specifically as follows:

[0017] S201: Calling the Kalman filter algorithm, based on the current dynamic baseline parameters, passing the estimated value of the previous moment through the state equation, combining the observation noise covariance matrix to update the current moment prediction value, calculating the current moment prediction value and the next moment prediction current value, and generating a predicted current trend;

[0018] S202: extracting the algebraic difference between the predicted current value at the current moment and the predicted current value at the next moment according to the predicted current trend, identifying the direction attribute of the difference through a sign determiner, and obtaining the predicted difference;

[0019] S203: Based on the predicted difference, extract the absolute value and superimpose the direction index, using the formula:

[0020] ;

[0021] Obtaining offset characteristic components through calculation to generate current offset characteristic parameters;

[0022] in, represents the current offset characteristic parameter, Represents the current prediction difference between the current moment and the next moment, Represents the direction indicator, represents the square mean of the dynamic baseline parameter, A sliding mean representing recent current changes.

[0023] As a further solution of the present invention, the step of obtaining the dynamic load threshold adjustment instruction is specifically as follows:

[0024] S301: Based on the absolute value of the predicted difference in the current offset characteristic parameter, calling the first threshold value, performing numerical comparison item by item and determining whether the threshold is exceeded, and generating an over-limit direction determination value;

[0025] S302: judging whether the predicted difference direction is negative based on the limit-crossing direction determination value, combining the sign information and the offset data to jointly verify the direction validity, and outputting a direction and limit-crossing joint confirmation value;

[0026] S303: For the direction and limit-crossing joint confirmation value, if the state is established, call the load offset value, current change rate and load fluctuation frequency, and use the formula:

[0027] ;

[0028] Calculating and obtaining a dynamic load adjustment value to obtain a dynamic load threshold adjustment signal;

[0029] in, Represents the dynamic load adjustment value, Represents the current offset value caused by load change behavior at the current time node, Representatives in the The current change rate value collected at each time node, Represents the instantaneous value of the frequency component in the currently monitored load power change. Represents the total number of time nodes for sampling the current change rate, Represents the sequence number of a single time node traversed during the summation process, is the dimension of the current change rate.

[0030] As a further solution of the present invention, the step of obtaining the group switching time parameter is specifically as follows:

[0031] S401: Based on the dynamic load threshold adjustment instruction, calling an interval partitioning function to compare the predicted difference value with the second threshold value using a greater than or equal to operator, and with the third threshold value using a less than or equal to operator, and determining whether the predicted difference value is within a left-closed and right-open interval formed by the second threshold value and the third threshold value based on the double comparison results, thereby generating a classification difference interval;

[0032] S402: Extracting the absolute difference between the predicted difference and the second threshold value according to the classification difference interval, combining the dynamic load fluctuation factor and the compression weight factor, and using the formula:

[0033] ;

[0034] By calculating the time slice compression coefficient and mapping it to the time slice reconstruction strategy, the reduced time slice length is generated;

[0035] in, Represents the length of the time slice after reconstruction, Represents the initial length of the group time slice, represents the absolute value of the forecast difference, represents the second threshold value, represents the third threshold value, represents the compression weight factor, represents the dynamic load correction factor;

[0036] S403: Calling the reduced time slice length to calculate the absolute difference between the initial length and the reduced length, converting the difference into a standard time unit, adding the thread switching delay compensation value preset in the multi-core scheduling cycle parameter, and generating a group switching time parameter.

[0037] As a further embodiment of the present invention, the method further comprises:

[0038] S5: Sending a switching execution instruction according to the group switching time parameter, marking the current group as to be switched, activating the target group power supply link, and setting the link initial time slice to the value defined by the group switching time parameter;

[0039] The switching execution instruction includes a target group identifier, a time slice validity flag, and a link switching trigger signal.

[0040] As a further solution of the present invention, the step of setting the link initial time slice as the group switching time parameter definition value is specifically:

[0041] S501: Based on the group switching time parameter, extract the group identifier, match the current group running state with the active state identifier, and if there is no fault alarm in the state, write the running identifier into the waiting switching queue and generate a waiting switching mark value;

[0042] S502: Calling the power supply link database, traversing the node address according to the to-be-switched tag value, sending an activation instruction and receiving the response signal strength, comparing it with a preset activation threshold, and if the signal strength meets the threshold, updating the link status register with the activation status code, and generating a link activation status code;

[0043] S503: Based on the initial time slice definition value, the single node allocation value is calculated in combination with the number of nodes with the link activation status code, and the value is written into the timing configuration register after comparison with the maximum load capacity to generate the link initial time slice value.

[0044] A 5G Internet of Things-based distribution cabinet online monitoring system, the 5G Internet of Things-based distribution cabinet online monitoring system is used to execute the above-mentioned 5G Internet of Things-based distribution cabinet online monitoring method, the system comprising:

[0045] A real-time acquisition module is used to collect current signals from each branch of the power distribution cabinet through a 5G communication unit, input the current signals for 5 consecutive seconds into a sliding window algorithm, extract the minimum value, mean value, and range, generate current dynamic baseline parameters, and transmit the current dynamic baseline parameters to a trend prediction module;

[0046] A trend prediction module is used to call the Kalman filter algorithm based on the current dynamic baseline parameter to predict the current value at the next moment, calculate the difference between the current moment and the predicted value, generate a current offset characteristic parameter in combination with the direction indicator, and pass the current offset characteristic parameter to the threshold judgment module;

[0047] a threshold judgment module, configured to compare a difference in the current offset characteristic parameter with a preset first threshold value, and if the absolute value of the difference exceeds the first threshold value and is negative, generate a dynamic load threshold adjustment instruction, and synchronously transmit the dynamic load threshold adjustment instruction and the current offset characteristic parameter to the time slice reconstruction module;

[0048] a time slice reconstruction module, configured to input a difference in the current offset characteristic parameter into an interval partitioning function through the dynamic load threshold adjustment instruction, and when the difference is between a second threshold value and a third threshold value, reduce the current group time slice, generate a group switching time parameter, and transmit the group switching time parameter to the execution control module;

[0049] The execution control module is used to generate a power supply link switching instruction according to the group switching time parameter, mark the current group state as waiting for switching, activate the target group power supply link, and set the link initial time slice to the defined value of the group switching time parameter.

[0050] Compared with the prior art, the advantages and positive effects of the present invention are:

[0051] This invention significantly enhances the environmental adaptability and predictive capabilities of load state identification by constructing a composite analysis mechanism that combines real-time dynamic baseline parameters with trend prediction. A sliding window algorithm continuously extracts current characteristic parameters to form a dynamic baseline, breaking through the traditional fixed threshold model's reliance on steady-state conditions and enabling the monitoring system to automatically track the natural fluctuations of the load. A Kalman filter algorithm models current trends in multiple dimensions, combining absolute value and directional indicators to form offset characteristic parameters, effectively capturing potential abnormalities in equipment operating conditions. An interval partitioning strategy and a dynamic time slice reconstruction mechanism work together to implement differentiated load distribution strategies based on the magnitude of predicted differences, maintaining power supply stability for critical equipment while enhancing overall system response flexibility. A dual-condition trigger mechanism, through combined determination of absolute value violations and directional characteristics, accurately identifies the critical state of abnormal current attenuation, avoiding misoperation caused by normal fluctuations. A multi-level threshold system establishes a progressive control framework, enabling rapid intervention in unexpected conditions while maintaining normal operating stability, resulting in a load distribution solution that balances safety and economy. The deep integration of predictive feature extraction and adaptive control strategies significantly improves the dynamic coordination capability of the distribution system in dealing with complex working conditions, achieves more accurate load optimization configuration in current mutation scenarios, and reduces the impact of power quality fluctuations on sensitive equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 It is a schematic diagram of the workflow of the present invention;

[0053] Figure 2 Flowchart of the steps for obtaining the current dynamic baseline parameters of the present invention;

[0054] Figure 3 Flowchart of the steps for obtaining the current offset characteristic parameters of the present invention;

[0055] Figure 4 A flow chart of the steps for obtaining a dynamic load threshold adjustment instruction according to the present invention;

[0056] Figure 5 Flowchart of the steps for obtaining the group switching time parameters of the present invention;

[0057] Figure 6 The present invention is a flowchart of the steps of setting the initial time slice of the link as the defined value of the group switching time parameter. DETAILED DESCRIPTION

[0058] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present 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 only used to explain the present invention and are not intended to limit the present invention.

[0059] In the description of the present invention, it should be understood that the terms "length," "width," "up," "down," "front," "back," "left," "right," "vertical," "horizontal," "top," "bottom," "inside," "outside," and the like, indicating positions or relationships, are based on the positions or relationships shown in the accompanying drawings and are intended only to facilitate the description of the present invention and simplify the description. They do not indicate or imply that the devices or elements referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limiting the present invention. Furthermore, in the description of the present invention, "plurality" means two or more, unless otherwise expressly and specifically defined.

[0060] Example 1

[0061] See also Figure 1 The present invention provides a technical solution: a method for online monitoring of a distribution cabinet based on 5G Internet of Things, comprising the following steps:

[0062] S1: The 5G communication unit collects real-time current signals from multiple branches of the power distribution cabinet. The real-time current signals are input into the sliding window algorithm to extract the minimum, mean, and range of the current within 5 seconds to generate the dynamic baseline parameters of the current.

[0063] S2: Based on the current dynamic baseline parameters, the Kalman filter algorithm is called to predict the current trend, the difference between the current and next moment predictions is calculated, and the current offset characteristic parameters are generated by combining the absolute value and direction indicators;

[0064] S3: Compare the absolute value of the predicted difference in the current offset characteristic parameter with the first threshold value. If the difference exceeds the limit and the direction is negative, trigger the dynamic load threshold adjustment instruction and synchronously output the dynamic load threshold adjustment instruction and the current offset characteristic parameter;

[0065] S4: Based on the dynamic load threshold adjustment instruction, the predicted difference is classified using the interval partitioning function. When the difference is between the second and third threshold values, the time slice reconstruction strategy is invoked to reduce the current group time slice, generate the group switching time parameters, and synchronously update the target group time slice;

[0066] S5: Send a switching execution instruction according to the group switching time parameter, mark the current group as to be switched, activate the target group power supply link, and set the link initial time slice to the value defined by the group switching time parameter.

[0067] The current dynamic baseline parameters include the minimum baseline, mean baseline, and range baseline. The current offset characteristic parameters include the absolute value of the predicted difference, directional index, and trend confidence. The dynamic load threshold adjustment instructions include the negative offset flag, threshold trigger level, and adjustment amplitude ratio. The group switching time parameters include the current group reduction duration, the target group preset duration, and the switching priority flag. The switching execution instructions include the target group flag, the time slice effectiveness flag, and the link switching trigger signal.

[0068] See also Figure 2 , the specific steps for obtaining the current dynamic baseline parameters are:

[0069] S101: Current signals of multiple branches of the power distribution cabinet are collected through the 5G communication unit, the current strength of the multiple branches is recorded every 10 milliseconds, the current signals within the collection period are integrated, and a multi-branch current sequence is generated;

[0070] A 5G communication unit (standard industrial-grade 5GDTU: GM800 series) deployed on the third-floor power distribution cabinet PDU-A3 in Building A of a specific data center continuously collects current signals from the eight server rack branches (internally labeled ZL01 to ZL08) that it supplies power to. The communication unit integrates a high-precision Hall effect current sensor (accuracy level 0.5), capable of recording a snapshot of the effective AC current value of each branch at a fixed interval of every 10 milliseconds. A portion of the collected data within the initial continuous 100 milliseconds is shown below:

[0071] Table 1 Initial current data collection table of some branches of the power distribution cabinet PDU-A3

[0072]

[0073] Table 1 shows the current readings for branches ZL01, ZL02, and ZL08 within 100 milliseconds of the start of acquisition. In practice, all eight branches are continuously recorded. The current intensity data collected from all eight branches within a complete acquisition cycle (set to 5 seconds, corresponding to 500 sampling time points) is integrated and structured in chronological order and branch identification, forming a data matrix of 500 rows (time points) × 8 columns (branches). This matrix represents the multi-branch current sequence for that cycle.

[0074] S102: Based on the multi-branch current sequence, intercept a continuous 5-second data window in chronological order, take the minimum value of the current data of each branch in the window, calculate the arithmetic mean, and count the difference between the maximum and minimum values ​​to generate the window range, window mean, and window minimum value of the multi-branch;

[0075] The width of the data window is set to 5 seconds, that is, it contains 500 data points continuously. The window slides according to the time step (the step length is 1 sampling interval, 10ms), and the data is intercepted for analysis. Considering the data of the first complete window (time point 1 to 500), the data of each branch in the window is processed separately. Taking the ZL01 branch as an example, its 500 current data points are traversed and the minimum value is identified. After searching, At the same time, the arithmetic mean of these 500 data points is calculated. The calculation process is to convert the 500 current values ​​( ) and then divided by 500, we get Next, find the maximum value among these 500 data points and get , calculate the range value of the ZL01 branch within the window, that is, the difference between the maximum and minimum values: ;

[0076] The same minimum value extraction, arithmetic mean calculation, and range calculation are performed on the remaining 7 branches (ZL02 to ZL08) in the window to obtain the window minimum value set, window mean value set, and window range value set of all 8 branches in the time window.

[0077] S103: Call the window range value, window mean value and window minimum value of multiple branches, arrange the three parameters of the same branch into triplets in order, encapsulate the triplets of all branches of the distribution cabinet in a unified format, and generate current dynamic baseline parameters.

[0078] The three statistical parameters of the same branch are organized into a triplet data structure in the given order of (window range, window mean, window minimum). For branch ZL01, the triplet of the first window is (1.98A, 15.48A, 14.53A). Corresponding triplet are generated for all 8 branches under the power distribution cabinet PDU-A3 to form a list structure. This list contains the statistical information of all branches in the current time window, for example: [(1.98, 15.48 ,14.53),(1.82,15.05,14.14),(2.15,15.88,14.81),(1.90,15.55,14.60),(2.01,15.70,14.75),(1.77,14.95,14.06),(2.20,16.10,15.00),(2.08,15.96,14.92)], this encapsulated list structure is the current dynamic baseline parameters of the current analysis window.

[0079] See also Figure 3 , the specific steps for obtaining the current offset characteristic parameters are:

[0080] S201: Calling the Kalman filter algorithm, based on the current dynamic baseline parameters, passing the estimated value of the previous moment through the state equation, combining the observation noise covariance matrix to update the current moment prediction value, calculating the current moment prediction value and the next moment prediction current value, and generating a predicted current trend;

[0081] Using the window mean of each branch As the observation input of the filter, combined with the previous time step ( ) state estimation results (including the predicted current value at the previous moment and its error covariance ) and a predefined system state transition model, which assumes that the current changes follow a random walk pattern: ,in represents the process noise, and its covariance Reflecting the inherent uncertainty of the system, The value of is determined based on the long-term analysis of the volatility of historical current data and is set to , the filtering process first performs state prediction: calculate the current moment ( ) prior state estimate and the prior error covariance Then, the observation value at the current moment is introduced, that is, the branch mean of the window calculated by S102 , and the observation noise covariance , The value is calibrated based on the accuracy level (0.5 level) and range of the current sensor used and is set to , calculate the Kalman gain , use the gain to update the state estimate: the posterior state estimate at the current moment (i.e. the predicted current value at the current moment) and the posterior error covariance , consider the ZL01 branch, if at the last moment The predicted value is And the error covariance , current window The observed mean is , then the prior estimate , prior error covariance , Kalman gain (unitless);

[0082] The predicted value at the current moment:

[0083] ;

[0084] Next, using the updated state Predict the next moment ( ) current value , we can conclude Repeat this filtering and prediction calculation for all branches to obtain the current moment of each branch and the next moment The sequence of predicted current values ​​constitutes the predicted current trend.

[0085] S202: extracting the algebraic difference between the predicted current value at the current moment and the predicted current value at the next moment based on the predicted current trend, identifying the direction attribute of the difference through a sign determiner, and obtaining the predicted difference;

[0086] Extract the predicted current value of each branch at the current moment and the predicted current value at the next moment , calculate the algebraic difference between the two , for the ZL01 branch, based on the calculation results of S201 (in the specific case of using the random walk model), To demonstrate the non-zero case, consider another branch ZL05, whose predicted value after filtering is and , then the algebraic difference

[0087] ;

[0088] Next, use a symbolic judgment logic to process this algebraic difference: If , determine the difference direction attribute is +1; if ,determination is -1; if ,determination is 0, for ZL01, ,so , for ZL05, ,therefore , the calculated algebraic difference and its corresponding direction attributes It is output as the prediction difference information of the branch.

[0089] S203: Based on the predicted difference, extract the absolute value and superimpose the direction indicator, using the formula:

[0090] ;

[0091] Obtaining offset characteristic components through calculation to generate current offset characteristic parameters;

[0092] in, represents the current offset characteristic parameter, Represents the current prediction difference between the current moment and the next moment, Represents the direction indicator, represents the square mean of the dynamic baseline parameter, A sliding mean representing recent current changes.

[0093] The calculation formula is:

[0094] ;

[0095] In this formula, is the predicted algebraic difference (unit: A), is a unitless direction indicator (+1, -1, 0), Represents the square mean of the dynamic baseline parameter (unit: ), Represents the sliding mean of recent current changes (unit: A). The acquisition method is: call the current dynamic baseline parameter list generated by S103 and extract the window mean of the latest N time windows (set N=10) of the specific branch , calculate the square of these means, and then find their average. For the ZL05 branch, assuming that the means of the past 10 windows are [15.1, 15.0, 15.2, 14.9, 15.1, 15.0, 14.8, 15.2, 14.9, 15.0]A, then ;

[0096] parameter The method of obtaining is to calculate the difference between the current values ​​of the branch at the latest M original sampling points (set M=20, that is, the past 200ms) , take the absolute values ​​of these differential values, and then calculate the average value of these absolute differences. For the ZL05 branch, the sum of the absolute current differences in the last 20 sampling intervals (10ms) is 1.28A, then . It reflects the severity of the recent fluctuation of the current. When the value is low (less than 0.05A), it means that the current changes are stable in the near future; A high value (above 0.1A) indicates that the current has fluctuated significantly recently.

[0097] Example (ZL05 branch): Given , , calculated , . Substitute into the formula:

[0098] ;

[0099] ;

[0100] ;

[0101] ;

[0102] Note: The unit under square root is , which is consistent with physical meaning.

[0103] , ;

[0104] The calculated The value (59.333A for ZL05) is output as the current offset characteristic parameter of the branch.

[0105] The formula is useful because it combines the forecast difference itself (which reflects the direction and instantaneous change) ) and a baseline containing historical levels ( ) and the recent volatility rate ( ) generates a comprehensive feature parameter ,This parameter not only reflects the predicted offset direction and size, but also incorporates historical background information, making it a more stable and comprehensive assessment of the degree of deviation of the current state.

[0106] This result shows that the calculated It is a comprehensive indicator. Its positive value indicates that the overall assessment tends to increase the current (although the instantaneous prediction is a decrease, the historical baseline and volatility dominate). The numerical value represents the characteristic strength of the offset. This value will be used in subsequent steps to determine whether the dynamic load threshold needs to be adjusted.

[0107] See also Figure 4 ,The steps for obtaining the dynamic load threshold adjustment instruction are as follows:

[0108] S301: Based on the absolute value of the predicted difference in the current offset characteristic parameter, call the first threshold value, compare the values ​​item by item, determine whether the threshold is exceeded, and generate an over-limit direction determination value;

[0109] Based on the absolute value of the forecast difference and the preset first threshold , to make an over-limit judgment, threshold value The setting is based on the normal operation of each branch of the distribution cabinet Statistical analysis of historical data, select the 95th percentile of historical data distribution as the benchmark value, and add a 10% safety margin. The specific calculation process is: collect the data of all branches during normal operation for one month The data is calculated to have a 95th percentile of 0.45A, then ;

[0110] Set it to 0.5A, and set the current calculated branches One by one with Perform numerical comparison and execute judgment: If , it is judged as out of limit, and the out of limit direction judgment value is generated and assigned a value of 1; if , it is determined that the limit is not exceeded, and the limit direction judgment value is generated and assigned to 0. For the ZL01 branch, ,because , its over-limit direction judgment value is 0. For ZL05 branch, ;

[0111] because , its over-limit direction judgment value is also 0. Consider another branch ZL09. If its calculation result is ,but ,because , so ZL09 is judged as out of limit, and its out of limit direction judgment value is 1.

[0112] S302: Based on the limit-crossing direction determination value, determine whether the predicted difference direction is negative, combine the sign information and the offset data to jointly verify the direction validity, and output the direction and limit-crossing joint confirmation value;

[0113] According to the over-limit direction judgment value, check whether its value is 1. Only when the judgment value is 1, continue to judge whether the predicted difference direction of the branch is negative. This judgment is made by checking the direction index obtained in S202. Is it equal to -1 to complete, combining these two conditions (the limit judgment value is 1 and ), conduct joint verification to confirm whether there is a significant negative current offset. For the ZL09 branch, its over-limit direction judgment value is 1, and its Corresponding , both conditions are met, at this time the output direction and the limit-crossing joint confirmation value are set to 1 (confirmed state), for ZL01 and ZL05, their limit-crossing direction judgment values ​​are both 0, which does not meet the first condition, so their joint confirmation values ​​are set to 0 (unconfirmed state), if one branch exceeds the limit (judgment value is 1) but its or 0, its joint confirmation value is also 0.

[0114] S303: For the direction and limit-crossing joint confirmation value, if the state is established, call the load offset value, current change rate and load fluctuation frequency, and use the formula:

[0115] ;

[0116] Calculating and obtaining a dynamic load adjustment value to obtain a dynamic load threshold adjustment signal;

[0117] in, Represents the dynamic load adjustment value, Represents the current offset value caused by load change behavior at the current time node, Representatives in the The current change rate value collected at each time node, Represents the instantaneous value of the frequency component in the currently monitored load power change. Represents the total number of time nodes for sampling the current change rate, Represents the sequence number of a single time node traversed during the summation process, is the dimension of the current change rate.

[0118] Check the direction and limit-crossing joint confirmation value. If the confirmation value is 1 (confirmation state, such as ZL09), the dynamic load adjustment value needs to be calculated. To do this, the following parameters need to be obtained: the current offset value caused by the specific load change behavior at the current time node , the current change rate sequence collected recently , and the currently monitored load power change frequency , then apply the formula:

[0119] ;

[0120] parameter (Unit: A) represents the current step change caused by a clear load event (server offline, task completion), which is obtained through the event monitoring system. When the server shutdown event occurs on the ZL09 branch, the difference between the stable current before and after the event is recorded, and the measured value is ,parameter (Unit: A / s) represents the Sampling time points (intervals ) The collected branches (i.e. the current change rate of ZL09) is calculated as follows: , collect recent Time nodes (covering The current change rate values ​​in the time window (time window) are calculated and their square sums are shown in Table 2.

[0121] Table 2 Example data of the current change rate of the ZL09 branch at the last five time nodes

[0122]

[0123] Table 2 shows the calculation process of the last 5 points. The actual calculation needs to include all points, calculated .parameter (Unit: Hz) represents the main frequency component of the currently monitored load current fluctuation, obtained by performing Fast Fourier Transform (FFT) analysis on the ZL09 current data of the last 1 second (100 sampling points). If the spectrum analysis shows a peak at 0.5Hz, and the peak amplitude exceeds the preset significance threshold (the threshold is set according to the historical noise level, and the amplitude is greater than 3 times the average background noise to be considered significant), then , if there is no significant low-frequency component, The value is 0.

[0124] Unit processing and calculation example (ZL09 branch): The numerator of the formula contains (A) and ( ), the direct addition units are inconsistent. In this specific application scenario, the formula structure is intended to combine the event-driven offset and the energy of random fluctuations, implicitly The steps of converting to equivalent energy contribution, or considering the formula as an operation based on normalized / scaled values, to complete the calculation, consider it as a combination of numerical levels, and convert The value of The unit of the result is determined by the final application scenario (adjusting the ampere threshold). The unit is ampere (A). (Hz) appears in the denominator In order to ensure consistency, the "1" here is unitless. The value of is directly involved in the calculation. , , . Substitute into the formula:

[0125] ;

[0126] ;

[0127] ;

[0128] ;

[0129] ;

[0130] Calculated This is the dynamic load adjustment value. This value will generate a specific dynamic load threshold adjustment signal, which will instruct the current load lower limit threshold of the ZL09 branch to be reduced by 5.812A. If the original lower limit threshold is 8.0A, the adjusted value will be , output this adjustment instruction (update the ZL09 lower limit threshold to 2.188A). If the S302 joint confirmation value is 0, the calculation of S303 is not performed and no adjustment signal is generated.

[0131] The formula is beneficial in that it combines specific event offsets 、Recent volatility energy and fluctuation frequency , can dynamically calculate the threshold adjustment amount based on the specific characteristics of the current change (whether it is a sudden event or a continuous high-frequency fluctuation) , which makes threshold adjustment more accurate and targeted, avoiding the drawbacks of using a single adjustment strategy for all types of changes.

[0132] This result shows that the calculated The threshold adjustment amplitude is calculated by the system based on the detected significant negative current excursion and its characteristics (including event-driven and random fluctuation components). This value directly quantifies how many amperes the threshold needs to be adjusted downward to adapt to the current load state change and form a specific adjustment instruction.

[0133] See also Figure 5 , the specific steps for obtaining the group switching time parameters are as follows:

[0134] S401: Based on the dynamic load threshold adjustment instruction, the interval partitioning function is called to compare the predicted difference value with the second threshold value using a greater than or equal to operator, and with the third threshold value using a less than or equal to operator. Based on the double comparison results, it is determined whether the predicted difference value is within the left-closed and right-open interval formed by the second threshold value and the third threshold value, and a classification difference interval is generated.

[0135] Use the absolute value of the forecast difference , execute the interval division logic, which depends on two preset thresholds: the second threshold and the third threshold ,The setting of these threshold values ​​needs to distinguish the different degrees of urgency of current changes. Used to identify changes that are more significant than normal fluctuations, the value should be set higher than , reference history The data distribution is set at the 98th percentile, which is calculated to be 0.98A, and the rounding is set , Used to define very dramatic changes that may indicate serious problems. Set at the 99.9th percentile of historical data, calculated as 2.85A, rounded to the nearest integer. , the division logic is as follows: judge Whether to satisfy both (Right now )and (Right now ), if both conditions are met, that is, , then the predicted difference is determined to be within the target interval, and the classification difference interval identifier is generated and assigned a value of 1. Otherwise, the identifier is assigned a value of 0. For ZL01( ) and ZL05( ),because , their interval identifiers are all 0, for ZL09 ( ),because , whose interval identifier is also 0, now introduce another branch ZL10, and calculate ,but ,because , the conditions are met, so the classification difference interval of ZL10 is marked as 1.

[0136] S402: Extract the absolute difference between the predicted difference and the second threshold value according to the classification difference interval, combine the dynamic load fluctuation factor and the compression weight factor, and use the formula:

[0137] ;

[0138] By calculating the time slice compression coefficient and mapping it to the time slice reconstruction strategy, the reduced time slice length is generated;

[0139] in, Represents the length of the time slice after reconstruction, Represents the initial length of the group time slice, represents the absolute value of the forecast difference, represents the second threshold value, represents the third threshold value, represents the compression weight factor, represents the dynamic load correction factor;

[0140] Check the classification difference interval flag. If it is 1 (such as in the case of ZL10), perform time slice reconstruction calculation and extract the absolute value of the predicted difference. (Unit: A) and the second threshold (Unit: A) Absolute difference , and combined with the dynamic load fluctuation factor and compression weight factor , apply the formula to calculate the length of the reconstructed time slice (Unit: ms):

[0141] ;

[0142] in, It is the initial basic time slice length of the group to which the branch belongs, which is preset by the system scheduling strategy. , compression weight factor For adjustment The nonlinear effect of relative position on the degree of compression within the interval is determined by simulation tests. The goal is to near Provides stronger compression effect, setting (unitless), dynamic load correction factor Used according to The absolute size of provides linear compression correction, and its value is also set by the trade-off test between system stability and response speed. The value is smaller to avoid excessive compression. (unitless).

[0143] Example (ZL10 branch): It is known that the interval identifier of ZL10 is 1. , , , , , . Substitute into the formula:

[0144] ;

[0145] Note: The unit A in fractions and ratios cancels out, resulting in a unitless ratio.

[0146] ;

[0147] ;

[0148] ;

[0149] ;

[0150] ;

[0151] ;

[0152] The calculated Output, as the reduced time slice length of the ZL10 branch after dynamic adjustment based on the current prediction difference. If the classification difference interval is marked as 0, this calculation is not performed and the subsequent steps will use the basic time slice length .

[0153] The formula is beneficial in that it implements dynamic, nonlinear adjustment of the time slice length, and only adjusts the time slice length when the current forecast change falls into the predefined "medium risk" interval. Compression is triggered when the compression level is not only related to the change value The relative position within the interval is related (by factors), and also their absolute size (via Factor correction) enables the system to shorten the response time proportionally according to the urgency of the change, thereby improving processing efficiency and adaptability.

[0154] This result shows that the calculated The system determines the shorter time slice length to be used based on the predicted current change of 1.5A (falling between 1.0A and 3.0A) detected and the set compression strategy applied. This shortened time slice reflects the system's decision to respond to this change more quickly (such as checking or switching more frequently).

[0155] S403: Calling the reduced time slice length to calculate the absolute difference between the initial length and the reduced length, converting the difference into a standard time unit, adding the thread switching delay compensation value preset in the multi-core scheduling cycle parameter, and generating a group switching time parameter.

[0156] Determine the time slice length applicable to the current branch , if the interval identifier is 1, then (For ZL10 it is 61.7ms), if the flag is 0, then (For ZL01, ZL05, and ZL09, it is 100ms). Next, get the thread switching delay compensation value preset in the system configuration. This value is to compensate for the fixed time overhead incurred by the operating system or hardware when performing task switching. It is determined based on the performance benchmark test of the target computing platform (Intel Xeon server, running Linux Kernel 5.x). , the final group switching time parameter It is composed of the currently applicable time slice length and compensation value, and is defined as , for ZL10, its , for branches marked as 0, such as ZL01, ZL05, and ZL09, , the calculated The value (unit: ms) is output as the group switching time parameter.

[0157] See also Figure 6 The steps for setting the initial link time slice to the group switching time parameter definition value are as follows:

[0158] S501: Based on the group switching time parameter, extract the group identifier, match the current group running state with the active state identifier, and if there is no fault alarm in the state, write the running identifier into the waiting switching queue and generate the waiting switching mark value;

[0159] Based on group switching time parameters First, identify the group corresponding to the parameter. By querying the branch-group mapping relationship, it is determined that the branch (ZL10) that triggers the calculation belongs to group G3. Next, query the current operating status monitoring information of group G3 to confirm that its status is "ACTIVE" (active status identifier) ​​and there are no active serious fault alarms (alarm level lower than Critical). If these two conditions are met, the group identifier "G3" is added to a pending switching task queue named pending_switch_queue. This is a first-in-first-out queue used to cache groups that need to update time slices. At the same time, the status of group G3 is updated to "PENDING_SWITCH", indicating that its time slice parameters are about to be updated. This tag value is the pending switching tag value.

[0160] S502: Call the power link database, traverse the node address according to the value of the tag to be switched, send an activation command and receive the response signal strength, compare it with the preset activation threshold, and if the signal strength meets the standard, update the link status register to the activation status code and generate a link activation status code;

[0161] An independent scheduling controller process periodically checks the pending_switch_queue queue. When it detects that the queue is not empty, it takes out the group identifier at the head of the queue, which is currently "G3". Based on the information in the system configuration library, it traverses all power supply nodes belonging to group G3 (server power management interface IPMI address or intelligent PDU socket network address). Group G3 includes nodes N3-1 and N3-2. The scheduler sends a specific activation or heartbeat detection command to the addresses of these two nodes and listens for their response signals. It evaluates the received response signals and checks the received signal strength indicator (RSSI) of the response signals. The preset activation thresholds are: RSSI not less than -75dBm and RTT not more than 50ms. The RSSI of the response signal of N3-1 is -70dBm and the RTT is 30ms, which meet the threshold requirements; the RSSI of the response signal of N3-2 is -72dBm and the RTT is 40ms, which also meet the requirements. Since all nodes in the group have successfully responded and the signal quality meets the standard, the system updates the status fields of these two nodes in the link state database to the activation status code "0x01" and generates the link activation status code "GROUP_ACTIVE" for the entire group G3.

[0162] S503: Based on the initial time slice definition value, the single node allocation value is calculated in combination with the number of nodes in the link activation status code, and the value is written into the timing configuration register after comparison with the maximum load capacity to generate the link initial time slice value.

[0163] Use the group switching time parameters calculated for group G3 This value is used as the initial time slice definition value for the next scheduling cycle of the group. Combined with the number of activated nodes in group G3 confirmed in S502 (a total of 2 nodes: N3-1, N3-2), the average time slice value allocated to each activated node is calculated, that is, the single node allocation value Before writing this value to the node's configuration, perform a safety check: compare The minimum safe time slice allowed by the node , this value is set according to the processing capacity of the node and the minimum operation response time, Set to 10ms, because , the verification is passed, and the calculated single node allocation value is The initial time slice of each node in group G3 is set by writing it into the timing configuration registers of nodes N3-1 and N3-2 through the management protocol (SNMP or IPMI). These nodes will run according to this time slice in subsequent scheduling.

[0164] A 5G Internet of Things-based distribution cabinet online monitoring system, which is used to execute the above-mentioned 5G Internet of Things-based distribution cabinet online monitoring method, includes:

[0165] The real-time acquisition module is used to collect the current signals of each branch of the distribution cabinet through the 5G communication unit, input the current signals for 5 consecutive seconds into the sliding window algorithm, extract the minimum value, mean value and range, generate the dynamic baseline parameters of the current, and pass the dynamic baseline parameters of the current to the trend prediction module;

[0166] The trend prediction module is used to call the Kalman filter algorithm based on the current dynamic baseline parameters to predict the current value at the next moment, calculate the difference between the current moment and the predicted value, generate the current offset characteristic parameters based on the direction indicator, and pass the current offset characteristic parameters to the threshold judgment module;

[0167] a threshold judgment module, configured to compare a difference in the current offset characteristic parameter with a preset first threshold value, and if the absolute value of the difference exceeds the first threshold value and is negative, generate a dynamic load threshold adjustment instruction, and synchronously transmit the dynamic load threshold adjustment instruction and the current offset characteristic parameter to the time slice reconstruction module;

[0168] a time slice reconstruction module, configured to input the difference in the current offset characteristic parameter into an interval partitioning function through a dynamic load threshold adjustment instruction, and when the difference is between a second threshold value and a third threshold value, reduce the current group time slice, generate a group switching time parameter, and transmit the group switching time parameter to the execution control module;

[0169] The execution control module is used to generate a power supply link switching instruction according to the group switching time parameter, mark the current group state as pending switching, activate the target group power supply link, and set the link initial time slice to the group switching time parameter definition value.

[0170] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A method for online monitoring of distribution cabinets based on 5G Internet of Things, characterized in that: The following steps are involved: S1: The real-time current signals of multiple branches of the power distribution cabinet are collected through the 5G communication unit, and the real-time current signals are input into the sliding window algorithm to extract the minimum value, mean value and range of the current within 5 seconds to generate the dynamic baseline parameters of the current; S2: Based on the current dynamic baseline parameters, the Kalman filter algorithm is called to predict the current trend, the difference between the current and next moment predictions is calculated, and the current offset characteristic parameters are generated by combining the absolute value and the direction index; The step of combining the absolute value and the direction index to generate the current offset characteristic parameter is specifically as follows: S203: Based on the predicted difference, extract the absolute value and superimpose the direction index, using the formula: ； Obtaining offset characteristic components through calculation to generate current offset characteristic parameters; in, represents the current offset characteristic parameter, Represents the current prediction difference between the current moment and the next moment, Represents the direction indicator, represents the square mean of the dynamic baseline parameter, A sliding mean representing recent current changes; S3: comparing the absolute value of the predicted difference in the current offset characteristic parameter with a first threshold value, and if the difference exceeds the threshold and the direction is negative, triggering a dynamic load threshold adjustment instruction, and synchronously outputting the dynamic load threshold adjustment instruction and the current offset characteristic parameter; S4: Based on the dynamic load threshold adjustment instruction, the predicted difference is classified using an interval partitioning function. When the difference is between the second and third thresholds, a time slice reconstruction strategy is invoked to reduce the current group time slice, generate a group switching time parameter, and synchronously update the target group time slice. The steps for obtaining the group switching time parameter are specifically as follows: S401: Based on the dynamic load threshold adjustment instruction, calling an interval partitioning function to compare the predicted difference value with the second threshold value using a greater than or equal to operator, and with the third threshold value using a less than or equal to operator, and determining whether the predicted difference value is within a left-closed and right-open interval formed by the second threshold value and the third threshold value based on the double comparison results, thereby generating a classification difference interval; S402: Extracting the absolute difference between the predicted difference and the second threshold value according to the classification difference interval, combining the dynamic load fluctuation factor and the compression weight factor, and using the formula: ； By calculating the time slice compression coefficient and mapping it to the time slice reconstruction strategy, the reduced time slice length is generated; in, Represents the length of the time slice after reconstruction, Represents the initial length of the group time slice, represents the absolute value of the forecast difference, represents the second threshold value, represents the third threshold value, represents the compression weight factor, represents the dynamic load correction factor; S403: Calling the reduced time slice length to calculate the absolute difference between the initial length and the reduced length, converting the difference into a standard time unit, adding the thread switching delay compensation value preset in the multi-core scheduling cycle parameter, and generating a group switching time parameter.

2. The online monitoring method for distribution cabinets based on 5G Internet of Things according to claim 1 is characterized in that: The current dynamic baseline parameters include a minimum baseline, a mean baseline, and a range baseline; the current offset characteristic parameters include the absolute value of the predicted difference, a directional index, and a trend confidence; the dynamic load threshold adjustment instruction includes a negative offset flag, a threshold trigger level, and an adjustment amplitude ratio; the group switching time parameters include the current group reduction duration, the target group preset duration, and the switching priority identifier.

3. The online monitoring method for distribution cabinets based on 5G Internet of Things according to claim 2 is characterized in that: The steps for obtaining the current dynamic baseline parameters are specifically as follows: S101: Current signals of multiple branches of the power distribution cabinet are collected through the 5G communication unit, the current strength of the multiple branches is recorded every 10 milliseconds, the current signals within the collection period are integrated, and a multi-branch current sequence is generated; S102: Based on the multi-branch current sequence, intercept a continuous 5-second data window in chronological order, take the minimum value of the current data of each branch in the window, calculate the arithmetic mean, and count the difference between the maximum and minimum values ​​to generate the window extreme difference, window mean and window minimum value of the multi-branch; S103: Call the window range value, window mean value and window minimum value of the multiple branches, arrange the three parameters of the same branch into triplets in order, encapsulate the triplets of all branches of the distribution cabinet in a unified format, and generate current dynamic baseline parameters.

4. The online monitoring method for distribution cabinets based on 5G Internet of Things according to claim 1 is characterized in that: Based on the current dynamic baseline parameters, the Kalman filter algorithm is called to predict the current trend, and the steps of calculating the difference between the current and next moment predictions are as follows: S201: Calling the Kalman filter algorithm, based on the current dynamic baseline parameters, passing the estimated value of the previous moment through the state equation, combining the observation noise covariance matrix to update the current moment prediction value, calculating the current moment prediction value and the next moment prediction current value, and generating a predicted current trend; S202: extracting the algebraic difference between the predicted current value at the current moment and the predicted current value at the next moment according to the predicted current trend, identifying the difference direction attribute through a sign determiner, and obtaining the predicted difference.

5. The online monitoring method for distribution cabinets based on 5G Internet of Things according to claim 4 is characterized in that: The steps for obtaining the dynamic load threshold adjustment instruction are specifically as follows: S301: Based on the absolute value of the predicted difference in the current offset characteristic parameter, calling the first threshold value, performing numerical comparison item by item and determining whether the threshold is exceeded, and generating an over-limit direction determination value; S302: judging whether the predicted difference direction is negative based on the limit-crossing direction determination value, combining the sign information and the offset data to jointly verify the direction validity, and outputting a direction and limit-crossing joint confirmation value; S303: For the direction and limit-crossing joint confirmation value, if the state is established, call the load offset value, current change rate and load fluctuation frequency, and use the formula: ； Calculating and obtaining a dynamic load adjustment value to obtain a dynamic load threshold adjustment signal; in, Represents the dynamic load adjustment value, Represents the current offset value caused by load change behavior at the current time node, Representatives in the The current change rate value collected at each time node, Represents the instantaneous value of the frequency component in the currently monitored load power change. Represents the total number of time nodes for sampling the current change rate, Represents the sequence number of a single time node traversed during the summation process, is the dimension of the current change rate.

6. The online monitoring method for distribution cabinets based on 5G Internet of Things according to claim 1 is characterized in that: The method further comprises: S5: Sending a switching execution instruction according to the group switching time parameter, marking the current group as to be switched, activating the target group power supply link, and setting the link initial time slice to the value defined by the group switching time parameter; The switching execution instruction includes a target group identifier, a time slice validity flag, and a link switching trigger signal.

7. The online monitoring method for distribution cabinets based on 5G Internet of Things according to claim 6 is characterized in that: The step of setting the link initial time slice as the group switching time parameter definition value is specifically as follows: S501: Based on the group switching time parameter, extract the group identifier, match the current group running state with the active state identifier, and if there is no fault alarm in the state, write the running identifier into the waiting switching queue and generate a waiting switching mark value; S502: Calling the power supply link database, traversing the node address according to the to-be-switched tag value, sending an activation instruction and receiving the response signal strength, comparing it with a preset activation threshold, and if the signal strength meets the threshold, updating the link status register with the activation status code, and generating a link activation status code; S503: Based on the initial time slice definition value, the single node allocation value is calculated in combination with the number of nodes with the link activation status code, and the value is written into the timing configuration register after comparison with the maximum load capacity to generate the link initial time slice value.

8. A 5G Internet of Things-based distribution cabinet online monitoring system, characterized in that: The system is used to implement the online monitoring method for a distribution cabinet based on 5G Internet of Things according to any one of claims 1 to 7, and the system includes: A real-time acquisition module is used to collect current signals from each branch of the power distribution cabinet through a 5G communication unit, input the current signals for 5 consecutive seconds into a sliding window algorithm, extract the minimum value, mean value, and range, generate current dynamic baseline parameters, and transmit the current dynamic baseline parameters to a trend prediction module; A trend prediction module is used to call the Kalman filter algorithm based on the current dynamic baseline parameter to predict the current value at the next moment, calculate the difference between the current moment and the predicted value, generate a current offset characteristic parameter in combination with the direction indicator, and pass the current offset characteristic parameter to the threshold judgment module; a threshold judgment module, configured to compare a difference in the current offset characteristic parameter with a preset first threshold value, and if the absolute value of the difference exceeds the first threshold value and is negative, generate a dynamic load threshold adjustment instruction, and synchronously transmit the dynamic load threshold adjustment instruction and the current offset characteristic parameter to the time slice reconstruction module; a time slice reconstruction module, configured to input a difference in the current offset characteristic parameter into an interval partitioning function through the dynamic load threshold adjustment instruction, and when the difference is between a second threshold value and a third threshold value, reduce the current group time slice, generate a group switching time parameter, and transmit the group switching time parameter to the execution control module; The execution control module is used to generate a power supply link switching instruction according to the group switching time parameter, mark the current group state as waiting for switching, activate the target group power supply link, and set the link initial time slice to the defined value of the group switching time parameter.

Images