A self-adaptive load-regulated switching power supply control system and method

Abstract

The application belongs to the technical field of switching power supply, and provides a switching power supply control system and method for adaptive load adjustment, which comprises the following steps: determining a load limit value by screening historical load data of a switching power supply in a historical period, combining real-time load data collected in a preset collection period, and evaluating whether the load of the switching power supply in the collection period is normal, wherein the historical load data comprises historical load; analyzing the load change trend of the switching power supply, and judging whether the load of the switching power supply has an overload trend; presetting a warning monitoring period, obtaining warning load data collected in the warning monitoring period and analyzing the warning load data, and identifying whether the switching power supply reaches an overload warning state in the warning monitoring period; and the application maximally utilizes an adjustment time window, avoids lag adjustment, and determines the load adjustment amount, which is beneficial to avoiding overload caused by insufficient adjustment or affecting system stability caused by excessive adjustment.

Description

Technical Field

[0001] This invention belongs to the field of switching power supply technology, specifically an adaptive load regulation switching power supply control system and method. Background Technology

[0002] In modern electronic equipment and power systems, switching power supplies are widely used due to their advantages such as high efficiency and small size. However, dynamic changes in the load during operation can significantly affect the stability and reliability of the power supply. When the load fluctuates abnormally or shows an overload trend, it may lead to a decline in power supply performance, equipment failure, or even system shutdown, resulting in economic losses.

[0003] Currently, existing switching power supply load regulation technologies have several shortcomings. Firstly, in terms of load anomaly assessment, the assessment dimensions are limited, relying solely on real-time absolute load values, which fails to comprehensively reflect abnormal load conditions. Secondly, in overload trend prediction, existing technologies lack multi-time-dimensional trend analysis methods, failing to effectively capture short-term fluctuations and long-term trends in load. Early warning mechanisms are lagging, often only responding when approaching dangerous loads, lacking sufficient buffer time. Thirdly, at the load adjustment strategy level, existing solutions lack accurate predictive models, making it difficult to scientifically set the timing and amount of load adjustments, easily leading to untimely, insufficient, or excessive adjustments, affecting system stability. Furthermore, existing technologies are mostly reactive fault handling, lacking proactive, end-to-end load management mechanisms, making it difficult to effectively prevent switching power supply failures and ensure reliable system operation.

[0004] Therefore, the present invention provides an adaptive load regulation switching power supply control system and method. Summary of the Invention

[0005] In order to overcome the shortcomings of the prior art, at least one technical problem raised in the background art is solved.

[0006] In a first aspect, the present invention provides an adaptive load-adjustable switching power supply control method, comprising:

[0007] After filtering the historical load data of the switching power supply within the historical period, the load limit value is determined. Combined with the real-time load data collected within the preset collection period, the load of the switching power supply within the collection period is evaluated to determine whether the load is normal. The historical load data includes historical load.

[0008] If the load of the switching power supply is normal during the data collection period, analyze the load change trend of the switching power supply to determine whether there is an overload trend in the load of the switching power supply.

[0009] If the switching power supply shows an overload trend, a pre-set warning monitoring period is set, the warning load data collected during the warning monitoring period is acquired and analyzed, the load status of the switching power supply is output, and it is identified whether the switching power supply has reached the overload warning state during the warning monitoring period.

[0010] If the load status of the switching power supply is in an overload warning state during the early warning monitoring period, the load after the overload warning is predicted based on the LSTM model. The load adjustment time and the load adjustment amount of the switching power supply before it reaches the overload state are determined. The load adjustment program is started to adjust the load of the switching power supply before it reaches the overload state, so as to prevent the switching power supply from entering the overload state.

[0011] As a further aspect of the present invention: the load limit value is obtained in the following way:

[0012] Obtain the historical load within the historical period of the switching power supply, sum all historical loads and take the average to obtain the historical average load; if the historical load is greater than or equal to the historical average load, then the corresponding historical load is retained.

[0013] Extract all retained historical loads, sum and average all retained historical loads to obtain the average retained load, introduce a compensation coefficient, and calculate the load limit value.

[0014] As a further aspect of the present invention, the specific process for evaluating whether the load of the switching power supply is normal during the data acquisition period is as follows:

[0015] The data collection period is preset and divided into several data collection time points. The real-time load of the switching power supply is collected at each data collection time point and analyzed to obtain the percentage of abnormal loads and the degree of load abnormality.

[0016] The load anomaly characterization value is obtained by weighting and summing the percentage of abnormal loads and the load anomaly severity value.

[0017] If the load anomaly indicator value is less than the load anomaly indicator threshold, it indicates that the load of the switching power supply is normal during the data collection period.

[0018] As a further aspect of the present invention: the method for obtaining the percentage of abnormal loads and the load abnormality characterization value is as follows:

[0019] If the real-time load is greater than or equal to the load limit, the corresponding collection time point will be recorded as an abnormal load point.

[0020] Calculate the percentage of abnormal load points across all data collection time points;

[0021] The difference between the real-time load corresponding to the abnormal load point and the load limit value is calculated to obtain the abnormal load deviation corresponding to the abnormal load point. The abnormal load deviations corresponding to all abnormal load points are summed and averaged, and then the ratio is calculated with the load limit value to obtain the load abnormality level value.

[0022] As a further aspect of the present invention: the specific process for determining whether the load of the switching power supply has an overload tendency is as follows:

[0023] Extract the load at each collection time point within the collection period, and analyze it in conjunction with the set short-term and long-term sliding window sizes to determine the load growth window;

[0024] Calculate the percentage of load growth windows within a long-term sliding window;

[0025] If the proportion of load growth windows in a long-term sliding window is greater than or equal to the threshold for the proportion of load growth windows, it indicates that the switching power supply has an overload trend.

[0026] As a further aspect of the present invention: the method for obtaining the load growth window is as follows:

[0027] For the load at each sampling time point within the sampling period, a short-time moving average is first performed to remove high-frequency noise, and then a long-time moving average is performed on the short-time average to obtain the long-time average load at the sampling time points of the long-time moving window.

[0028] Based on the long-term load average value at each sampling time point of the long-term sliding window, the load change rate of the long-term sliding window at the current sampling time point is calculated based on any sampling time point.

[0029] If the load change rate of the long-term sliding window at the current data collection time point is greater than the standard value of the load change rate, then the corresponding long-term sliding window is recorded as the load growth window.

[0030] As a further aspect of the present invention: the specific process of identifying whether the switching power supply reaches an overload warning state during the warning monitoring period is as follows:

[0031] The early warning monitoring period is preset and divided into several early warning monitoring time points with equal time intervals. Monitoring load data is collected at the early warning monitoring time points. The monitoring load data includes the monitoring load. Deviation analysis is performed on the monitoring load data collected at the early warning monitoring time points to obtain the proportion of abnormal monitoring points and the load early warning level value.

[0032] The load status value is obtained by weighted summation of the proportion of abnormal monitoring points and the load warning level value.

[0033] If the load status value is greater than or equal to the load status threshold, it indicates that the switching power supply load status is in an overload warning state during the early warning monitoring period.

[0034] As a further aspect of the present invention: the method for obtaining the proportion of abnormal monitoring points and the load warning level value is as follows:

[0035] If the monitored load is greater than or equal to the load warning value, the corresponding warning monitoring time point will be recorded as an abnormal monitoring point.

[0036] Calculate the percentage of abnormal monitoring points across all early warning monitoring time points;

[0037] The load warning value corresponding to the abnormal monitoring point is subtracted from the load warning value to obtain the load warning deviation corresponding to the abnormal monitoring point. The load warning deviations corresponding to all abnormal monitoring points are summed and averaged, and then the ratio is calculated with the load warning value to obtain the load warning level value.

[0038] As a further aspect of the present invention: the process for obtaining the load adjustment time and the load adjustment amount of the switching power supply is as follows:

[0039] The prediction period is preset and divided into several prediction time points with equal time intervals. The monitoring load and corresponding time series within the early warning monitoring period are used as sample points. The load after the overload early warning is predicted based on the LSTM model, and the predicted load at each prediction time point is output.

[0040] Extract all predicted loads that are greater than or equal to the load limit value and integrate them into a predicted load sequence in chronological order. Record the predicted time point corresponding to the first predicted load in the predicted load sequence as the latest control time point.

[0041] Extract the time point corresponding to the maximum monitored load when the switching power supply enters the switch warning state, and record it as the earliest control time point. Subtract the latest control time point from the earliest control time point to obtain the load adjustment time.

[0042] The load adjustment amount of the switching power supply is obtained by subtracting the first predicted load in the predicted load sequence from the maximum monitored load when the switching power supply enters the switching warning state.

[0043] In a second aspect, the present invention also provides an adaptive load regulation switching power supply control system, the system comprising:

[0044] Load anomaly assessment module: After filtering the historical load data of the switching power supply within the historical period, the load limit value is determined. Combined with the real-time load data collected within the preset collection period, the module assesses whether the load of the switching power supply is normal within the collection period. The historical load data includes historical loads.

[0045] Overload trend analysis module: If the load of the switching power supply is normal during the data collection period, analyze the load change trend of the switching power supply to determine whether there is an overload trend in the load of the switching power supply.

[0046] Load warning analysis module: If the switching power supply has an overload trend, a preset warning monitoring period is set, the warning load data collected during the warning monitoring period is acquired and analyzed, the load status of the switching power supply is output, and it is identified whether the switching power supply has reached the overload warning state during the warning monitoring period.

[0047] Control and processing module: If the load status of the switching power supply is in an overload warning state during the early warning monitoring period, the load after the overload warning is predicted based on the LSTM model, the load adjustment time and the load adjustment amount of the switching power supply before it reaches the overload state are determined, and the load adjustment program is started to adjust the load of the switching power supply before it reaches the overload state to prevent the switching power supply from entering the overload state.

[0048] The beneficial effects of this invention are as follows:

[0049] 1. This invention determines load limits by filtering historical load data of the switching power supply within a historical period. Combined with real-time load data collected within a preset collection period, it assesses whether the load of the switching power supply is normal during the collection period. Historical load data includes historical load values. If the load of the switching power supply is normal during the collection period, the load change trend is analyzed to determine if there is an overload trend. This invention filters historical load data of the switching power supply within a historical period, removing data with loads lower than the historical average, and determines load limits, providing a reliable basis for load anomaly assessment. It comprehensively assesses whether the load of the switching power supply is abnormal during the collection period from two aspects: the proportion of abnormal load points and the degree of load anomaly. When the load is normal, short-term and long-term sliding windows are used to process the load data, analyze the load change trend, and determine if there is an overload trend. This provides a basis for taking measures to address overload in advance, promptly detecting load anomalies and carrying out maintenance, and predicting overload trends in advance to prepare for responses. This can reduce the occurrence of switching power supply failures, reduce downtime and other losses caused by failures, and improve the reliability and stability of the system.

[0050] 2. When the switching power supply shows an overload trend, this invention presets a warning monitoring period, acquires and analyzes the warning load data collected during the monitoring period, outputs the load status of the switching power supply, and identifies whether the switching power supply has reached an overload warning state during the monitoring period. If the switching power supply load status is in an overload warning state during the monitoring period, the load after the overload warning is predicted based on an LSTM model, determining the load adjustment time and the load adjustment amount before the switching power supply reaches an overload state, and initiating a load adjustment program to adjust the switching power supply load before it reaches an overload state, preventing the switching power supply from entering an overload state. This invention sets a load warning value lower than the load limit value to prevent the switching power supply from entering an overload state when it approaches the limit. Early warnings are triggered before dangerous loads occur to prevent direct entry into overload conditions and to provide a buffer time for load adjustment. By combining the proportion of abnormal monitoring points with the load warning level, the load risk during the warning period is assessed, which facilitates the system to initiate subsequent prediction and adjustment processes in a targeted manner. The prediction period is set according to requirements, and the LSTM model outputs load prediction values ​​at multiple time points, providing sufficient time dimension reference for load adjustment. The time point when the predicted load first exceeds the load limit is used as the bottom line to ensure that the adjusted load does not reach the load limit. The time point of the maximum load when entering the overload warning state is used as the starting point to maximize the use of the adjustment time window and avoid delayed adjustment. Determining the amount of load adjustment helps to avoid under-adjustment leading to overload or over-adjustment affecting system stability. Attached Figure Description

[0051] The invention will now be further described with reference to the accompanying drawings.

[0052] Figure 1 This is a flowchart of the acquisition steps of an adaptive load regulation switching power supply control method according to an embodiment of the present invention;

[0053] Figure 2 This is a system block diagram of an adaptive load regulation switching power supply control system according to an embodiment of the present invention. Detailed Implementation

[0054] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0055] Example 1

[0056] Please see Figure 1 As shown in the figure, an adaptive load regulation switching power supply control method according to an embodiment of the present invention includes the following steps:

[0057] Step 1: After filtering the historical load data of the switching power supply within the historical period, determine the load limit value. Combined with the real-time load data collected within the preset collection period, evaluate whether the load of the switching power supply is normal within the collection period. The historical load data includes historical load.

[0058] It is understandable that historical periods include, but are not limited to, 30 days, 60 days, and 90 days;

[0059] The historical load within a historical cycle of the switching power supply is obtained by using the non-volatile memory inside the switching power supply. The historical average load is obtained by summing all the historical loads within the historical cycle and taking the average value.

[0060] Compare historical load within a historical period with historical average load:

[0061] If the historical load is greater than or equal to the historical average load, the corresponding historical load will be retained.

[0062] If the historical load is less than the historical average load, the corresponding historical load will be removed.

[0063] Extract all retained historical loads, sum and average them to obtain the average retained load BF. Introduce a compensation coefficient λ to calculate the load limit value FX. The specific calculation formula is as follows:

[0064] FX=λ·BF

[0065] It should be noted that the compensation coefficient should be set according to the rated power margin of the power supply.

[0066] The preset data collection period is divided into several data collection time points with equal time intervals, and then processed via I... 2 The C-bus connects to the power supply, and standard PMBus commands are used to collect the real-time load of the switching power supply at the specified acquisition time. Data preprocessing is then performed on the collected real-time load data, specifically including:

[0067] Kalman filtering is used to remove high-frequency noise, convert the raw values ​​acquired by the ADC into actual physical quantities, and normalize them for easier unified analysis.

[0068] Based on any real-time load;

[0069] If the real-time load is greater than or equal to the load limit, the corresponding collection time point will be recorded as an abnormal load point.

[0070] If the real-time load is less than the load limit, the corresponding collection time point will be recorded as the normal load point.

[0071] Count the number of abnormal load points at all data collection time points, and calculate the percentage of abnormal load points at all data collection time points;

[0072] The difference between the real-time load corresponding to the abnormal load point and the load limit value is calculated to obtain the abnormal load deviation corresponding to the abnormal load point. The abnormal load deviations corresponding to all abnormal load points are summed and averaged. Then, the ratio is calculated with the load limit value to obtain the load abnormality level value.

[0073] The load anomaly characterization value is obtained by weighting and summing the percentage of abnormal loads and the load anomaly severity value.

[0074] In some embodiments, the load anomaly representation value is compared with the load anomaly representation threshold. The specific comparison process is as follows:

[0075] If the load anomaly indicator value is greater than or equal to the load anomaly indicator threshold, it indicates that the power supply has an abnormal load during the data collection period, and the power supply should be inspected.

[0076] If the load anomaly indicator value is less than the load anomaly indicator threshold, it indicates that the load of the switching power supply is normal during the data collection period.

[0077] Step 2: If the load of the switching power supply is normal during the data collection period, analyze the load change trend of the switching power supply to determine whether there is an overload trend.

[0078] Extract the load at each acquisition time point within the acquisition period, and set the short-time sliding window size to N. s The long-term sliding window size is N. l First, a short-time moving average is performed to remove high-frequency noise, resulting in a short-time load average value for the short-time sliding window. Then, a long-time moving average is performed on the short-time average value to obtain a long-time load average value for the long-time sliding window, thus converting the switching power supply load into a smooth trend point.

[0079] The short-time load average at the t-th sampling time point of the short-time sliding window The calculation method is as follows:

[0080]

[0081] Among them, FZ k This represents the load at the k-th sampling time point, where the value of k ranges from tN. s +1 to t, that is, tracing back N from the current sampling time point t. s -1 sampling time point;

[0082] Long-term load average at the t-th sampling time point of the long-term sliding window The calculation method is as follows:

[0083]

[0084] in, This represents the short-time load average at the m-th sampling time point, where m ranges from tN. l +1 to t, that is, tracing back N from the current sampling time point t. l -1 sampling time point is used to calculate the long-term trend;

[0085] Based on the long-term load average value at the t-th sampling time point of the long-term sliding window, the long-term sliding window load change rate FB at the current sampling time point is calculated based on any sampling time point. t The specific calculation formula is as follows:

[0086]

[0087] Among them, FZ t-Δt (long) represents the long-term load average value Δt minutes ago, where Δt is the time interval corresponding to the sliding step size of the long-term sliding window.

[0088] Compare the load change rate of the long-term sliding window at the current data collection time point with the standard value of the load change rate:

[0089] If the load change rate of the long-term sliding window at the current data collection time point is greater than the standard value of the load change rate, then the corresponding long-term sliding window is recorded as the load growth window.

[0090] If the load change rate of the long-term sliding window at the current data collection time point is less than or equal to the standard value of the load change rate, then the corresponding long-term sliding window is recorded as a non-growing load window.

[0091] It should be noted that the standard value for the load change rate was set by those skilled in the art based on historical experience;

[0092] Count the number of load-increasing windows in a long-term sliding window and calculate the percentage of load-increasing windows in the long-term sliding window.

[0093] If the proportion of load growth windows in a long-term sliding window is greater than or equal to the threshold for the proportion of load growth windows, it indicates that the switching power supply has an overload trend.

[0094] If the proportion of load growth windows in a long-term sliding window is less than the threshold for the proportion of load growth windows, it indicates that the switching power supply does not have an overload trend.

[0095] The technical solution of this embodiment is as follows: After filtering the historical load data of the switching power supply within a historical period, the load limit value is determined. Combined with the real-time load data collected within a preset collection period, the load of the switching power supply within the collection period is evaluated to determine whether the load of the switching power supply within the collection period is normal. The historical load data includes historical load. If the load of the switching power supply is normal within the collection period, the load change trend of the switching power supply is analyzed to determine whether there is an overload trend. This invention filters the historical load data of the switching power supply within a historical period, removes data with loads less than the historical average load, and determines the load limit value, providing a reliable basis for load anomaly assessment. The load of the switching power supply within the collection period is comprehensively evaluated from two aspects: the proportion of abnormal load points and the degree of load anomaly. When the load is normal, the load data is processed through short-term and long-term sliding windows to analyze the load change trend of the switching power supply and determine whether there is an overload trend. This provides a basis for taking measures to deal with overload in advance, timely detection of load anomalies and maintenance, and early judgment of overload trends and preparation for response. This can reduce the occurrence of switching power supply failures, reduce losses such as downtime caused by failures, and improve the reliability and stability of the system.

[0096] Example 2

[0097] Please see Figure 1 As shown in the embodiment of the present invention, an adaptive load regulation switching power supply control method further includes the following steps:

[0098] Step 3: If the switching power supply shows an overload trend, preset the early warning monitoring period, acquire and analyze the early warning load data collected during the early warning monitoring period, output the load status of the switching power supply, and identify whether the switching power supply has reached the overload warning state during the early warning monitoring period.

[0099] The pre-defined early warning monitoring period is divided into several equally spaced early warning monitoring time points, also via I... 2 The C bus is connected to the power supply, and standard PMBus commands are used to collect monitoring load data at the early warning monitoring time point. The monitoring load data includes the monitoring load.

[0100] If the monitored load is greater than or equal to the load warning value, the corresponding warning monitoring time point will be recorded as an abnormal monitoring point.

[0101] If the monitored load is less than the load warning value, the corresponding warning monitoring time point will be recorded as the normal monitoring point.

[0102] It should be noted that the load warning value is a load value set before the load limit value is reached. It is set by those skilled in the art based on historical experience. The load warning value is less than the load limit value. Its purpose is to provide early warning and control of the load of the switching power supply after the load warning value is reached but before the load limit value is reached.

[0103] Count the number of abnormal monitoring points at all early warning monitoring time points, and calculate the percentage of abnormal monitoring points at all early warning monitoring time points;

[0104] The load warning value corresponding to the abnormal monitoring point is subtracted from the load warning value to obtain the load warning deviation corresponding to the abnormal monitoring point. The load warning deviations corresponding to all abnormal monitoring points are summed and averaged, and then the ratio is calculated with the load warning value to obtain the load warning level value.

[0105] The load status value is obtained by weighted summation of the proportion of abnormal monitoring points and the load warning level value.

[0106] In some embodiments, the load status value is compared with a load status value threshold. The specific comparison process is as follows:

[0107] If the load status value is greater than or equal to the load status threshold, it means that the load status of the switching power supply is in an overload warning state during the early warning monitoring period. At this time, the switching power supply is about to enter an overload state and the overload time prediction mechanism needs to be activated.

[0108] If the load status value is less than the load status threshold, it means that the load status of the switching power supply is not in an overload warning state during the early warning monitoring period.

[0109] Step 4: If the load status of the switching power supply is in an overload warning state during the early warning monitoring period, predict the load after the overload warning based on the LSTM model, determine the load adjustment time and the load adjustment amount of the switching power supply before it reaches the overload state, start the load adjustment program, and adjust the load of the switching power supply before it reaches the overload state to prevent the switching power supply from entering the overload state.

[0110] Based on the fact that the switching power supply load is in an overload warning state during the early warning monitoring period, a prediction period is preset and divided into several prediction time points with equal time intervals. The monitoring load and the corresponding time series during the early warning monitoring period are used as sample points. The load after the overload warning is predicted based on the LSTM model, and the predicted load at each prediction time point is output.

[0111] All collected switching power supply load data are input into the trained LSTM model. The LSTM model predicts the load data at the predicted time point based on the learned load data features. For example, if the load data at the next 10 predicted time points is set, the LSTM model will output the predicted load at each predicted time point in sequence.

[0112] To more accurately capture the load change patterns after overload warning, the LSTM model structure was optimized by setting three LSTM network layers, each containing 128 neurons. The multi-layer network structure enhances the ability to extract data features. At the same time, a Dropout layer was introduced to randomly discard some neuron connections, preventing the LSTM model from overfitting and improving its generalization ability. For example, the Dropout probability was set to 0.2, meaning that during training, each neuron has a 20% probability of not participating in the calculation.

[0113] The output layer of the LSTM model uses a fully connected layer to output the load forecast value for one or more future time points according to the forecast requirements, and the loss function is the mean squared error.

[0114] Extract all predicted loads that are greater than or equal to the load limit value and integrate them into a predicted load sequence in chronological order. Record the predicted time point corresponding to the first predicted load in the predicted load sequence as the latest control time point.

[0115] Extract the time point corresponding to the maximum monitored load when the switching power supply enters the switch warning state, and record it as the earliest control time point. Subtract the latest control time point from the earliest control time point to obtain the load adjustment time.

[0116] To ensure that the load of the switching power supply does not exceed the load limit at the overload time point, the load adjustment amount of the switching power supply is obtained by subtracting the first predicted load in the predicted load sequence from the maximum monitored load when the switching power supply enters the switching warning state.

[0117] Initiate the load adjustment program to adjust the load on the switching power supply before it reaches an overload state, thus preventing the switching power supply from entering an overload state.

[0118] The technical solution of this embodiment is as follows: If the switching power supply shows an overload trend, a pre-set warning monitoring period is established. Warning load data collected during the warning monitoring period is acquired and analyzed. The load status of the switching power supply is output, and it is determined whether the switching power supply reaches an overload warning state during the warning monitoring period. If the load status of the switching power supply during the warning monitoring period is in an overload warning state, the load after the overload warning is predicted based on an LSTM model. The load adjustment time and the load adjustment amount before the switching power supply reaches an overload state are determined. The load adjustment program is started to adjust the load of the switching power supply before it reaches an overload state, preventing the switching power supply from entering an overload state. This invention sets a load warning value lower than the load limit value to prevent the switching power supply from entering an overload state. The system triggers an early warning before the source approaches a dangerous load, preventing direct entry into an overload state and providing a buffer time for load adjustment. By combining the proportion of abnormal monitoring points with the load warning level, the system assesses the load risk during the warning period, facilitating the targeted initiation of subsequent prediction and adjustment processes. The prediction period is set according to requirements, and the LSTM model outputs load prediction values ​​at multiple time points, providing sufficient time dimension reference for load adjustment. The time point when the predicted load first exceeds the load limit is used as the baseline to ensure that the adjusted load does not reach the load limit. The time point of the maximum load when entering the overload warning state is used as the starting point to maximize the use of the adjustment time window and avoid delayed adjustments. Determining the amount of load adjustment helps to avoid under-adjustment leading to overload or over-adjustment affecting system stability.

[0119] Example 3

[0120] Please see Figure 2 As shown in the embodiment of the present invention, an adaptive load regulation switching power supply control system includes the following modules:

[0121] Load anomaly assessment module: After filtering the historical load data of the switching power supply within the historical period, the load limit value is determined. Combined with the real-time load data collected within the preset collection period, the module assesses whether the load of the switching power supply is normal within the collection period. The historical load data includes historical loads.

[0122] Overload trend analysis module: If the load of the switching power supply is normal during the data collection period, analyze the load change trend of the switching power supply to determine whether there is an overload trend in the load of the switching power supply.

[0123] Load warning analysis module: If the switching power supply has an overload trend, a preset warning monitoring period is set, the warning load data collected during the warning monitoring period is acquired and analyzed, the load status of the switching power supply is output, and it is identified whether the switching power supply has reached the overload warning state during the warning monitoring period.

[0124] Control and processing module: If the load status of the switching power supply is in an overload warning state during the early warning monitoring period, the load after the overload warning is predicted based on the LSTM model, the load adjustment time and the load adjustment amount of the switching power supply before it reaches the overload state are determined, and the load adjustment program is started to adjust the load of the switching power supply before it reaches the overload state to prevent the switching power supply from entering the overload state.

[0125] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A switching power supply control method for adaptive load regulation, characterized in that: include: After filtering the historical load data of the switching power supply within the historical period, the load limit value is determined. Combined with the real-time load data collected within the preset collection period, the load of the switching power supply within the collection period is evaluated to determine whether the load is normal. The historical load data includes historical load. The load limit value is obtained as follows: Obtain the historical load within the historical period of the switching power supply, sum all historical loads and take the average to obtain the historical average load; if the historical load is greater than or equal to the historical average load, then the corresponding historical load is retained. Extract all retained historical loads, sum and average all retained historical loads to obtain the average retained load, introduce a compensation coefficient, and calculate the load limit value; If the load of the switching power supply is normal during the data collection period, analyze the load change trend of the switching power supply to determine whether there is an overload trend in the load of the switching power supply. The specific process for determining whether the load of the switching power supply has an overload tendency is as follows: Extract the load at each collection time point within the collection period, and analyze it in conjunction with the set short-term and long-term sliding window sizes to determine the load growth window; Calculate the percentage of load growth windows within a long-term sliding window; If the proportion of load growth windows in a long-term sliding window is greater than or equal to the threshold for the proportion of load growth windows, it indicates that the switching power supply has an overload trend. If the switching power supply shows an overload trend, a pre-set warning monitoring period is set, the warning load data collected during the warning monitoring period is acquired and analyzed, the load status of the switching power supply is output, and it is identified whether the switching power supply has reached the overload warning state during the warning monitoring period. If the load status of the switching power supply is in an overload warning state during the early warning monitoring period, the load after the overload warning is predicted based on the LSTM model, the load adjustment time and the load adjustment amount of the switching power supply before it reaches the overload state are determined, and the load adjustment program is started to adjust the load of the switching power supply before it reaches the overload state to prevent the switching power supply from entering the overload state. The process for obtaining the load adjustment time and the load adjustment amount of the switching power supply is as follows: The prediction period is preset and divided into several prediction time points with equal time intervals. The monitoring load and corresponding time series within the early warning monitoring period are used as sample points. The load after the overload early warning is predicted based on the LSTM model, and the predicted load at each prediction time point is output. Extract all predicted loads that are greater than or equal to the load limit value and integrate them into a predicted load sequence in chronological order. Record the predicted time point corresponding to the first predicted load in the predicted load sequence as the latest control time point. Extract the time point corresponding to the maximum monitored load when the switching power supply enters the switch warning state, and record it as the earliest control time point. Subtract the latest control time point from the earliest control time point to obtain the load adjustment time. The load adjustment amount of the switching power supply is obtained by subtracting the first predicted load in the predicted load sequence from the maximum monitored load when the switching power supply enters the switching warning state.

2. The adaptive load regulation switching power supply control method according to claim 1, characterized in that: The specific process for assessing whether the load of the switching power supply is normal during the data collection period is as follows: The data collection period is preset and divided into several data collection time points. The real-time load of the switching power supply is collected at each data collection time point and analyzed to obtain the percentage of abnormal loads and the degree of load abnormality. The load anomaly characterization value is obtained by weighting and summing the percentage of abnormal loads with the load anomaly severity value. If the load anomaly indicator value is less than the load anomaly indicator threshold, it indicates that the load of the switching power supply is normal during the data collection period.

3. The adaptive load regulation switching power supply control method according to claim 2, characterized in that: The method for obtaining the percentage of abnormal loads and the load anomaly characterization value is as follows: If the real-time load is greater than or equal to the load limit, the corresponding collection time point will be recorded as an abnormal load point. Calculate the percentage of abnormal load points across all data collection time points; The difference between the real-time load corresponding to the abnormal load point and the load limit value is calculated to obtain the abnormal load deviation corresponding to the abnormal load point. The abnormal load deviations corresponding to all abnormal load points are summed and averaged, and then the ratio is calculated with the load limit value to obtain the load abnormality level value.

4. The adaptive load regulation switching power supply control method according to claim 1, characterized in that: The load growth window is obtained as follows: For the load at each sampling time point within the sampling period, a short-time moving average is first performed to remove high-frequency noise, and then a long-time moving average is performed on the short-time average to obtain the long-time average load at the sampling time points of the long-time moving window. Based on the long-term load average value at each sampling time point of the long-term sliding window, the load change rate of the long-term sliding window at the current sampling time point is calculated based on any sampling time point. If the load change rate of the long-term sliding window at the current data collection time point is greater than the standard value of the load change rate, then the corresponding long-term sliding window is recorded as the load growth window.

5. The adaptive load regulation switching power supply control method according to claim 1, characterized in that: The specific process for identifying whether the switching power supply reaches an overload warning state during the early warning monitoring period is as follows: The early warning monitoring period is preset and divided into several early warning monitoring time points with equal time intervals. Monitoring load data is collected at the early warning monitoring time points. The monitoring load data includes the monitoring load. Deviation analysis is performed on the monitoring load data collected at the early warning monitoring time points to obtain the proportion of abnormal monitoring points and the load early warning level value. The load status value is obtained by weighted summation of the proportion of abnormal monitoring points and the load warning level value. If the load status value is greater than or equal to the load status threshold, it indicates that the switching power supply load status is in an overload warning state during the early warning monitoring period.

6. The adaptive load regulation switching power supply control method according to claim 5, characterized in that: The method for obtaining the percentage of abnormal monitoring points and the load warning level value is as follows: If the monitored load is greater than or equal to the load warning value, the corresponding warning monitoring time point will be recorded as an abnormal monitoring point. Calculate the percentage of abnormal monitoring points across all early warning monitoring time points; The load warning value corresponding to the abnormal monitoring point is subtracted from the load warning value to obtain the load warning deviation corresponding to the abnormal monitoring point. The load warning deviations corresponding to all abnormal monitoring points are summed and averaged, and then the ratio is calculated with the load warning value to obtain the load warning level value.

7. An adaptive load regulation switching power supply control system, characterized in that, The system is used to perform the method according to any one of claims 1-6, the system comprising: Load anomaly assessment module: After filtering the historical load data of the switching power supply within the historical period, the load limit value is determined. Combined with the real-time load data collected within the preset collection period, the module assesses whether the load of the switching power supply is normal within the collection period. The historical load data includes historical loads. Overload trend analysis module: If the load of the switching power supply is normal during the data collection period, analyze the load change trend of the switching power supply to determine whether there is an overload trend in the load of the switching power supply. Load warning analysis module: If the switching power supply has an overload trend, a preset warning monitoring period is set, the warning load data collected during the warning monitoring period is acquired and analyzed, the load status of the switching power supply is output, and it is identified whether the switching power supply has reached the overload warning state during the warning monitoring period. Control and processing module: If the load status of the switching power supply is in an overload warning state during the early warning monitoring period, the load after the overload warning is predicted based on the LSTM model, the load adjustment time and the load adjustment amount of the switching power supply before it reaches the overload state are determined, and the load adjustment program is started to adjust the load of the switching power supply before it reaches the overload state to prevent the switching power supply from entering the overload state.

Images