Server power supply switching method and device supporting multi-node power supply

Abstract

The invention relates to the technical field of power supply switching, in particular to a server power supply switching method and device supporting multi-node power supply. The method comprises the following steps: acquiring power supply efficiency and module internal resistance data of each power supply module in a power supply system, and load power of each power supply module for external power supply; obtaining a performance degradation coefficient of each power supply module according to the power supply efficiency and the module internal resistance data of each power supply module; according to the performance degradation coefficient and the load power of each power supply module, obtaining the switching necessity of each power supply module for implementing power switching under the condition that the current load is supplied; and according to the switching necessity and the load power of each power supply module and the power fluctuation degree of each load, obtaining the switching feasibility degree of the corresponding load switching from the current power supply module to other power supply modules, and when the switching feasibility degree is greater than a preset feasibility degree threshold value, executing a power supply switching task on the current load. According to the technical scheme, reliable switching of the power supply can be implemented based on the difference of different loads.

Description

Technical Field

[0001] This invention relates to the technical field of power supply switching, and specifically to a server power supply switching method and apparatus that supports multi-node power supply. Background Technology

[0002] Servers supporting multi-node power supply are used in conjunction with automatic power switching to redundant power supply paths, achieving zero-interruption switching in the event of a single node or line failure, ensuring business continuity. Simultaneously, multi-node power supply can share current, improving system power margin and reliability, and allowing online power maintenance and upgrades, greatly enhancing the resilience of the overall infrastructure. However, traditional methods that switch power supplies based solely on their current load size may lead to abnormal data fluctuations or even power outages after the switch, resulting in insufficient reliability of power switching for multi-node powered servers. Summary of the Invention

[0003] To address the technical problem of insufficient reliability in power switching for servers with multi-node power supply, the present invention aims to provide a method and apparatus for power switching of servers with multi-node power supply. The specific technical solution adopted is as follows: In a first aspect, embodiments of the present invention provide a server power switching method supporting multi-node power supply, the method comprising: Acquire the power supply efficiency and internal resistance data of each power supply module in the power system, as well as the load power supplied by each power supply module to the external system; Based on the power supply efficiency and internal resistance data of each power supply module, the performance degradation coefficient of each power supply module is obtained; Based on the performance degradation coefficient and load power of each power supply module, the switching necessity of each power supply module to switch power supply under the current load is obtained; Based on the switching necessity of each power supply module, the load power, and the power fluctuation of each load, the switching feasibility of switching the corresponding load from the current power supply module to other power supply modules is obtained; When the switching feasibility is greater than the preset feasibility threshold, a power supply switching task is performed on the current load so that the target power supply module supplies power to the current load.

[0004] In one optional embodiment, the performance degradation coefficient of each power supply module is obtained based on the power supply efficiency and internal resistance data of each power supply module, including: The efficiency decay characteristics of each power supply module are analyzed to obtain the first performance index characterizing the power supply efficiency decay of each power supply module. Based on the current internal resistance value of each power supply module's internal resistance data, a second performance index characterizing the internal resistance variation of each power supply module is obtained. The performance degradation coefficient of each power supply module is obtained based on the first and second performance indicators of each power supply module.

[0005] In one optional embodiment, the efficiency degradation characteristics of each power supply module are analyzed to obtain a first performance index characterizing the power supply efficiency degradation of each power supply module, including: Based on the efficiency difference of each power supply module at adjacent times in the historical period, the efficiency difference set of each power supply module is obtained. Each set of efficiency differences is compared with a preset difference threshold to obtain the number of efficiency differences in each set that are greater than the difference threshold. Based on the number of efficiency differences for each power supply module and the number of power supply efficiencies in historical periods, the first characteristic coefficient representing the power supply efficiency decay of each power supply module is obtained. Based on the difference between the current power supply efficiency of each power supply module and the efficiency extreme value of the historical period, the second characteristic coefficient representing the power supply efficiency decay of each power supply module is obtained. The first performance index of each power supply module is obtained based on the first characteristic coefficient and the second characteristic coefficient of each power supply module.

[0006] In one optional embodiment, the switching necessity of each power supply module for power switching under the current load is obtained based on the performance degradation coefficient and load power of each power supply module, including: The attenuation degradation index of each power supply module is obtained based on the difference between the maximum degradation coefficient and the performance degradation coefficient of each power supply module in historical power supply tasks. A correlation characteristic analysis is performed on the load power of the current power supply module and other power supply modules to obtain load fluctuation indicators that characterize the load stability performance of each power supply module. Based on the attenuation and degradation index and load fluctuation index of each power supply module, the switching necessity of each power supply module to switch power supply under the current load is obtained.

[0007] In one optional embodiment, a correlation feature analysis is performed on the load power of the current power supply module and other power supply modules to obtain load fluctuation indicators characterizing the load stability performance of each power supply module, including: Based on the difference between the first load power of the current power supply module and the second load power of each other power supply module, obtain the power difference set of the current power supply module; Based on the cumulative summation of all data in the current power difference set of the power supply module and the stability index of the current power supply module, the load fluctuation index of each power supply module is obtained.

[0008] In an optional embodiment, before obtaining the switching feasibility of switching a corresponding load from the current power supply module to other power supply modules based on the switching necessity of each power supply module, the load power, and the power fluctuation of each load, the method further includes: The number of fluctuations of each load in multiple historical operation tasks is statistically analyzed to obtain the fluctuation frequency of each load's operating power fluctuation exceeding the power threshold. The peak power coefficient of each load is obtained by statistically analyzing the maximum operating power of each load in multiple historical operation tasks. The power fluctuation of each load is obtained based on its fluctuation frequency and peak power factor.

[0009] In one optional embodiment, the switching feasibility of switching a corresponding load from the current power supply module to other power supply modules is obtained based on the switching necessity of each power supply module, the load power, and the power fluctuation of each load, including: Based on the operating power and power fluctuation of each load, obtain the fluctuation impact coefficient of each load under its fluctuating operating conditions; Based on the switching necessity and load power of each power supply module, the acceptance capacity index of each power supply module is obtained; Based on the fluctuation impact coefficient of each load and the acceptance capacity index of power supply through different power supply modules, the switching feasibility of the corresponding load from the current power supply module to other power supply modules is obtained.

[0010] In one optional embodiment, based on the operating power and power fluctuation of each load, the fluctuation impact coefficient characterized by each load under its fluctuating operating conditions is obtained, including: The fluctuating power of the current load is estimated by multiplying the current load power and the power fluctuation. The power fluctuation estimate of the current load is normalized to obtain the fluctuation impact coefficient of the current load under its fluctuating operating conditions.

[0011] In one optional embodiment, the acceptance capability index of each power supply module is obtained based on the switching necessity and load power of each power supply module, including: The necessity difference of the current power supply module is obtained by comparing the maximum value of all switching necessities in historical power supply tasks with the switching necessity of the current power supply module. Based on the current power supply module's necessity difference and load power, the acceptance capacity index of the current power supply module is obtained.

[0012] Secondly, embodiments of the present invention also provide a server power switching device supporting multi-node power supply. The device is the same as any power switching method in the first aspect, and includes: The acquisition module is used to acquire the power supply efficiency and internal resistance data of each power supply module in the power system, as well as the load power supplied by each power supply module to the outside. The first acquisition module is used to obtain the performance degradation coefficient of each power supply module based on the power supply efficiency and internal resistance data of each power supply module. The second acquisition module is used to obtain the switching necessity of each power supply module in supplying the current load based on the performance degradation coefficient and load power of each power supply module. The third acquisition module is used to obtain the switching feasibility of the corresponding load from the current power supply module to other power supply modules based on the switching necessity of each power supply module, the load power, and the power fluctuation of each load. The power supply switching module is used to perform a power supply switching task on the current load when the switching feasibility is greater than a preset feasibility threshold, so that the target power supply module can supply power to the current load.

[0013] The present invention has the following beneficial effects: The technical solution of this invention obtains the power supply efficiency and internal resistance data of each power supply module in the power system, as well as the load power supplied by each power supply module. To measure the health of the power supply modules, a performance degradation coefficient is obtained for each power supply module based on its power supply efficiency and internal resistance data. To identify the power supply modules that need to be switched, a switching necessity for each power supply module to switch power to the current load is obtained based on its performance degradation coefficient and load power. Furthermore, based on the switching necessity of each power supply module, the load power, and the power fluctuation of each load, a switching feasibility for switching the corresponding load from the current power supply module to another power supply module is obtained, giving different loads corresponding migration priorities. When the switching feasibility is greater than a preset feasibility threshold, a power supply switching task is performed on the current load so that the target power supply module supplies power to the current load. This technical solution, when switching power supplies for servers with multi-node power supply, schedules different loads of power supply modules according to the load characteristics and stability of different power supply modules in the power supply system. This makes the switching of power supply modules more reasonable and more conducive to ensuring the stable operation of the power supply system. It can reliably switch power supplies based on the differences in different loads. Attached Figure Description

[0014] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1A flowchart illustrating a server power switching method supporting multi-node power supply, provided as an embodiment of the present invention; Figure 2 A flowchart illustrating the calculation of the performance degradation coefficient provided in one embodiment of the present invention; Figure 3 A flowchart illustrating the calculation of switching necessity provided in one embodiment of the present invention; Figure 4 A flowchart illustrating the calculation of switching feasibility according to an embodiment of the present invention; Figure 5 This is a schematic diagram of a server power switching device that supports multi-node power supply, provided as an embodiment of the present invention. Detailed Implementation

[0016] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a server power switching method and apparatus supporting multi-node power supply according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0017] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0018] The technical solution provided in this invention can be applied to power switching control of servers employing multi-node redundant power supply. Specific server types include dual-power server power supply systems, N+1 redundant power supply architecture power supply systems, rack-level A / B-path power supply systems, and multi-UPS (Uninterruptible Power System) parallel power supply systems. Because traditional methods only consider the current load size when implementing power switching, if the size and fluctuation characteristics of different loads are not suitable for the corresponding power supply module, it will cause abnormal data fluctuations when changing power supplies, thus failing to guarantee the stable operation of the power system.

[0019] This invention achieves intelligent switching and optimized load scheduling among multiple power nodes through comprehensive analysis of the health status and load fluctuation characteristics of the power supply module, thereby ensuring continuous and stable power supply to the server under different load conditions. The specific solution of a server power switching method and device supporting multi-node power supply provided by this invention will be described in detail below with reference to the accompanying drawings.

[0020] Please see Figure 1 , Figure 1This is a flowchart illustrating a server power switching method supporting multi-node power supply, provided as an embodiment of the present invention. This power switching method can be applied to the operation of a power system controller. The controller can be a control terminal composed of a PLC (Programmable Logic Controller) or a microcontroller, as long as the controller can run this switching method; no specific limitation is made to the type of controller here. The power switching method includes: S11. Obtain the power supply efficiency and internal resistance data of each power supply module in the power supply system, as well as the load power supplied by each power supply module to the outside.

[0021] Specifically, when the power system is in operation, the power supply efficiency and internal resistance data of each power supply module can be obtained based on a preset frequency. For example, through I... 2 The system uses a C or PMBus interface to continuously collect key parameters of the server power supply, including input / output voltage, current, power, and output ripple. Through an independent out-of-band management network, relying on standard protocols such as Redfish or IPMI, the key parameter data is reliably transmitted to a central time-series database, providing a foundation for subsequent in-depth trend analysis. Based on the real-time collected input and output voltage and current data, the power supply's power value is calculated using the formula: Calculate the power supply efficiency of power supply module a. ,in, The output power of power supply module a, Input power to power supply module a, and then obtain the efficiency of each power supply module at different times. It should be noted that a certain input power value is required for power supply efficiency calculations, i.e. Greater than 0, in When the value is 0, it indicates that there is no power input, and the power supply efficiency can be fixed at 0.

[0022] It should be noted that, for ease of calculation, all indicator data involved in the calculation in this embodiment of the invention have undergone data preprocessing to eliminate the influence of dimensions. The specific methods for eliminating the influence of dimensions are well known to those skilled in the art and are not limited here.

[0023] It is understandable that the module internal resistance data represents the power supply internal resistance of the corresponding power supply module. The module internal resistance data can be acquired based on preset sensors. Since the ripple voltage is positively correlated with the load current and the equivalent series resistance, the ripple and load data can also be read in real time through the intelligent power management bus to indirectly deduce the dynamic change of the power supply internal resistance, i.e., the module internal resistance data.

[0024] A positive correlation indicates that the independent variable and the dependent variable change in the same direction, and the larger the independent variable is, the larger the dependent variable is. A negative correlation indicates that the independent variable and the dependent variable change in opposite directions, and the smaller the independent variable is, the larger the dependent variable is. The specific manifestation of positive and negative correlation is determined by practical application, and this application does not impose any special restrictions.

[0025] It should be noted that a power supply module can be a single battery pack, while the power system consists of multiple power supply modules. Some power supply modules are configured to operate, supplying power to the load; others charge the load; and the remaining modules, after charging, are configured as redundant backup power. Each power supply module collects data during charging and discharging to obtain corresponding power supply efficiency and module internal resistance data. Both power supply efficiency and module internal resistance data are stored as data columns based on different collection times. Taking power supply efficiency as an example, the current time... The power supply efficiency is denoted as Each power supply module may supply power to the same load or different loads. The power supply objects can be communication equipment, computer equipment, server equipment, etc. The corresponding load power is derived based on the output power of the power supply module, and the load power is also a data column based on time changes.

[0026] At this point, the power supply efficiency, module internal resistance data, and load power have been obtained, and we proceed to step S12.

[0027] S12. Based on the power supply efficiency and internal resistance data of each power supply module, obtain the performance degradation coefficient of each power supply module.

[0028] Specifically, as the power supply module's usage time increases, its power supply efficiency gradually decreases, and the module's internal resistance gradually increases. The continuous decline in the power supply module's efficiency means that more input electrical energy is converted into heat energy, which directly leads to an increase in the operating temperature of internal components (such as MOSFETs, magnetic core components, and electrolytic capacitors). Taking electrolytic capacitors as an example, high temperatures drastically accelerate the drying of the electrolyte and the thermal fatigue of the semiconductor material, forming a feedback loop of decreased efficiency → increased temperature → accelerated aging → further decreased efficiency, thus significantly shortening the power supply module's lifespan. Increased internal resistance is a direct indicator of component performance degradation. Therefore, by analyzing the changes in the power supply module's efficiency and internal resistance, a performance degradation coefficient can be derived. Based on this coefficient, the health of the power supply is assessed, and the performance degradation coefficient characterizes the remaining lifespan of the corresponding power supply module.

[0029] For example, please refer to Figure 2 Step S12 includes sub-steps S12-1 to S12-3, which are described in detail below: S12-1. Analyze the efficiency degradation characteristics of each power supply module to obtain a primary performance index characterizing the power supply efficiency degradation of each power supply module. This can be done by analyzing the data variation characteristics of power supply efficiency over time, then statistically calculating the change in power supply efficiency for each power supply module, and quantifying it using the primary performance index.

[0030] The calculation steps for the first performance indicator will be explained in detail below, including: The first step is to obtain the efficiency difference set for each power supply module based on the efficiency difference between adjacent moments in the historical time period. Then, by extrapolating N moments backward from the current moment t, the historical time period is obtained. N is a natural number greater than 1, for example, set to 1000. Calculate historical time periods. The difference in internal power supply efficiency between adjacent times i-1 and i That is, the difference between the power supply efficiency at the previous moment and the current moment, in historical time periods. All efficiency differences can form an efficiency difference set.

[0031] It is understandable that during a cold start, i.e. when the i-th time is the first time, the efficiency difference can be the power supply efficiency value at the i-th time itself.

[0032] The second step involves comparing each set of efficiency differences with a preset difference threshold to obtain the number of efficiency differences in each set that exceed the threshold. The difference threshold can be set based on the actual performance of the power supply module. When an efficiency difference exceeds the threshold (set to be greater than the tolerance for normal minor fluctuations, adjusted according to the actual scenario; preferably, it can be set to 1% of the rated efficiency of the power supply module, which is determined by the actual performance of the power supply module; if the rated efficiency is 95%, then the difference threshold is 0.95%), it indicates that the power supply module has power efficiency degradation. This is then statistically analyzed to determine the historical time periods. The amount of data is denoted as the number of efficiency differences. .

[0033] The third step involves obtaining the first characteristic coefficient representing the power supply efficiency degradation of each power supply module based on the number of efficiency differences for each module and the number of power supply efficiencies over historical periods. There are N time periods in the historical timeframe, each with a power supply efficiency, and the number of power supply efficiencies is also N. Therefore, the coefficient can be determined based on the number of efficiency differences. The ratio of the power supply efficiency number N to the power supply efficiency number N is used to derive the first characteristic coefficient of each power supply module. .

[0034] The fourth step involves obtaining the second characteristic coefficient representing the power supply efficiency degradation of each power supply module based on the difference between the current power supply efficiency and the historical efficiency extreme values ​​for each time period. The current power supply efficiency at time t is denoted as... The efficiency extremum can be the minimum value. Minimize the efficiency extreme value. For example, the second characteristic coefficient can be obtained through The calculations show that, understandably, during the calculation process, it is necessary to... and Dimensionless processing is performed, and the specific dimensionless processing methods are commonly used in this field and are not limited thereto.

[0035] The fifth step is to obtain the first performance index of each power supply module based on its first and second characteristic coefficients. The first performance index can be derived from the product of the first and second characteristic coefficients. After the product calculation, the values ​​can be normalized based on the actual situation to eliminate dimensions and standardize them for easier subsequent data analysis.

[0036] In one embodiment of the present invention, the normalization process can be specifically, for example, maximum and minimum value normalization. Furthermore, the normalization in subsequent steps can all adopt maximum and minimum value normalization. In other embodiments of the present invention, other normalization methods can be selected according to the specific range of the numerical values, which will not be elaborated further.

[0037] S12-2. Based on the current internal resistance value of each power supply module, obtain a second performance index characterizing the internal resistance variation of each power supply module. The current internal resistance value at time t is... Normalization can be performed to obtain As a second performance indicator.

[0038] S12-3. Based on the first and second performance indicators of each power supply module, obtain the performance degradation coefficient of each power supply module. This can be achieved using the formula: The performance degradation coefficient of power supply module a at the current time t is calculated. When the efficiency difference is large The ratio of the power supply efficiency number N The larger the difference The smaller, and the second performance index The larger the value, the greater the performance degradation of power supply module a, and the worse its health condition; conversely, a smaller value indicates that the power supply module has good power supply performance. Based on the above method, the degradation status of each power supply module at different times can be evaluated in real time, serving as a reference for subsequent power supply switching. (This is from an embodiment of the invention.) The normalization method can be either maximum or minimum value normalization, which achieves dimensionless and standardized processing. The processing of exp(-) can also perform data standardization, thereby obtaining the performance degradation coefficient.

[0039] in, , and These are preset weighting coefficients, summing to 1, representing the contributions of efficiency decay frequency, efficiency decay magnitude, and increased internal resistance to performance degradation, respectively. These weights can be calibrated based on the design characteristics of different server power supplies and aging experience data from actual operation; for example, they can be set to w1=0.4, w2=0.4, w3=0.2.

[0040] At this point, the performance degradation coefficient of each power supply module has been obtained, and we proceed to step S13.

[0041] S13. Based on the performance degradation coefficient and load power of each power supply module, obtain the switching necessity of each power supply module in implementing power switching under the current load.

[0042] Specifically, when a power supply module exhibits significant instability, such as severe voltage spikes / drops, voltage surges / peaks, voltage fluctuations, frequency drift, or harmonic distortion, and its degradation is substantial, even if there is no immediate power outage, it is necessary to switch power to a portion of the load in this module to prevent further exacerbation of the instability and potential power outage due to continued supply of a large load. Therefore, it is crucial to analyze the performance degradation coefficient and load power of each power supply module to determine whether switching is necessary for the current load. The switching necessity rating indicates the degree to which a power supply module needs to switch power to the current load. A higher rating means the current power supply module cannot sustain the current load for an extended period, necessitating timely power switching; conversely, a lower rating indicates that monitoring can continue.

[0043] For example, please refer to Figure 3 Step S13 includes sub-steps S13-1 to S13-3, which are described in detail below: S13-1. Based on the difference between the maximum degradation coefficient and the performance degradation coefficient of each power supply module in historical power supply tasks, obtain the attenuation degradation index of each power supply module. Each power supply module has a maximum performance degradation coefficient in historical power supply tasks, which is recorded as the maximum degradation coefficient. The performance degradation coefficient of power supply module a at the current time t is The degradation index can be based on The calculation yielded the result.

[0044] Maximum degradation coefficient This is the design performance degradation threshold for the power supply module. It is a fixed value preset based on module specifications or experimental data, representing the upper limit at which the module performance reaches a critical state and requires close monitoring or maintenance.

[0045] S13-2. Perform a correlation characteristic analysis on the load power of the current power supply module and other power supply modules to obtain a load fluctuation index characterizing the load stability performance of each power supply module. Since the load of each power supply module supplying power to the outside world may be unbalanced, and since both the current power supply module and other power supply modules supply power to the outside world, a correlation characteristic analysis of the load power of the current power supply module and other power supply modules can determine whether the current power supply module is under excessive power supply pressure, which can be quantified through the load fluctuation index.

[0046] By summing the load differences between the current power supply module and all other power supply modules, the absolute load of the power supply can be converted into a quantitative indicator of overall relative pressure. This allows the system to identify power supply nodes bearing excessive pressure in the entire multi-node power supply system, thus providing an accurate load distribution basis for determining switching necessity. Specifically, this includes: The first step is to obtain the power difference set of the current power supply module based on the difference between the first load power of the current power supply module and the second load power of each other power supply module. The first load power is denoted as... The second load power is denoted as The number of other power supply modules is The difference between the power of the first load and the power of the second load is 1. .

[0047] The second step is to obtain the load fluctuation index of each power supply module based on the cumulative sum of all data in the current power difference set and the stability index of the current power supply module. The cumulative sum of all data is as follows: Since the load power of the current power supply module is known, it can be normalized based on the standard deviation of all load power over historical periods, using the negative of the standard deviation as a stability index. The stability index of power supply module a at the current time t is: The load fluctuation index is The addition of 0.1 to the denominator is a safety parameter set to prevent the denominator from being zero, and the error caused by its value setting is within an acceptable range.

[0048] S13-3. Based on the attenuation degradation index and load fluctuation index of each power supply module, obtain the switching necessity of each power supply module in supplying the current load. The attenuation degradation index and load fluctuation index are quantitative results obtained from the analysis of the power supply module itself and the load fluctuation characteristics, respectively. The switching necessity can be derived from the product of the two. For example, according to the formula: Calculate the switching necessity of power supply module a at the current time t. The max-min normalization method can also be used to... After normalization, we get Its range is [0,1].

[0049] It is understandable that the stability of power supply module a at the current time t is... The smaller the summation result The larger the value, the greater the performance degradation coefficient of power supply module a at the current time t. , with maximum degradation The difference The smaller the time, the greater the risk of power outage in power supply module a at the current time t, and its load is also larger. In order to avoid single-circuit overload and mitigate or avoid the occurrence of power outage, the power supply of the load in power supply module a should be switched.

[0050] In some embodiments of the present invention, the need for power switching can be determined directly by the switching necessity of each power supply module. Taking a switching necessity threshold of 0.7 as an example, if the switching necessity is greater than the necessity threshold of 0.7, it means that the load corresponding to the current power supply module needs to be switched; otherwise, the switching necessity is calculated and monitored.

[0051] It should be noted that the necessity threshold of 0.7 is an empirical value obtained under typical hardware configurations and test scenarios, intended to facilitate understanding of the present invention. In practical applications, those skilled in the art can adjust, calibrate, or optimize these parameters according to specific hardware performance, scenario complexity, and data characteristics, which does not constitute a limitation of the present invention.

[0052] At this point, the switching necessity of each power supply module in supplying the current load has been obtained, and the process proceeds to step S14.

[0053] S14. Based on the switching necessity of each power supply module, the load power, and the power fluctuation of each load, obtain the switching feasibility of the corresponding load switching from the current power supply module to other power supply modules.

[0054] Specifically, in multi-node power supply servers, relying solely on the switching necessity of power supply modules as a switching reference is insufficient. Power switching depends not only on whether a switch is necessary but also on whether a safe switch can be made at the current moment. For example, some loads experience significant power fluctuations, such as those performing AI training tasks. To ensure stable operation and avoid triggering protection due to instantaneous peak power, these loads need to be placed on power supplies with sufficient power margins. The magnitude of load power fluctuations can be assessed by referencing historical power fluctuations of similar loads. Power fluctuation characterizes the degree of power fluctuation during the operation of a corresponding load. Based on the switching necessity of each power supply module, load power, and power fluctuation of each load, a more comprehensive analysis can be performed to determine the feasibility of switching the corresponding load from the current power supply module to another. A calculation model is developed based on the relationship between switching feasibility and other parameters, and the switching feasibility is solved using this model. Alternatively, a neural network model can be used to predict the feasibility based on switching necessity, load power, and power fluctuation; no specific limitations are imposed here.

[0055] It should be noted that power fluctuation characterizes the magnitude of power fluctuations during the operation of a corresponding load. It can be calculated by taking the standard deviation of the operating power of each load and then further derived. Standard deviation can measure the dispersion and volatility of data. However, in multi-node server load management scenarios, the characteristic is that the power is relatively stable most of the time, with occasional extremely high peaks. These peaks are short in duration but have a huge impact on the power supply. When calculating power fluctuation based on standard deviation, the standard deviation will be diluted by a large amount of stable data, resulting in insufficient accuracy of the power fluctuation calculation. The calculation method of power fluctuation will be explained in detail below.

[0056] The first step is to statistically analyze the frequency of power fluctuations for each load across multiple historical runs to determine the frequency at which power fluctuations exceed a power threshold. During load data collection, the loads can be connected to a smart socket. The smart socket's built-in sensors and chips can then be used to read real-time power data via a network interface (such as SNMP or HTTP API) to obtain the operating power of each load. The difference between the maximum and minimum power values ​​at different times during the v-th historical run of the same type of load k is then calculated. The difference in runtime values ​​for all historical instances of the same type of load (k) is obtained through comparison. maximum value ; Analyze the differences obtained during historical runs of similar load k. The number of runs with a normalized value greater than 0.75 (this value is a preset value, obtained through analysis of historical data in actual test scenarios). The total number of historical runs for all loads of the same type k is The fluctuation frequency is Understandably, This represents the total number of historical runs. When performing specific analysis, its value should be greater than 0. This applies if the load is a newly added load, or if it represents the total number of historical runs. If the value is 0, then its fluctuation frequency is 0.

[0057] Among them, the same type of load k refers to server loads that run the same type of business (such as database, web service, AI training) or are grouped into one category after being clustered according to the characteristics of historical power curves (such as mean, peak, variance).

[0058] The second step involves statistically analyzing the maximum operating power of each load across multiple historical operation tasks to obtain the peak power coefficient representing the maximum power variation characteristics of each load. This is achieved by comparing the maximum operating power of the same load (k) across all historical operation runs. And the maximum power of each load operating independently. The peak power factor is It should be noted that, To prevent The safety item set to 0 can have a specific value of 0.1, with dimensions similar to... To maintain consistency, the error caused by setting its value is within an acceptable range.

[0059] The third step is to obtain the power fluctuation of each load based on its fluctuation frequency and peak power factor. This can be done using the formula: Calculate the power fluctuation of load k. The greater the power fluctuation, the greater the power fluctuation of the same type of load k during operation, and the higher the maximum power it may have during operation. Therefore, it is more necessary to switch to a safer power supply with a smaller load. Based on this, accurate power fluctuation calculations can be performed on loads requiring power switching. The power of each load at the current time t is analyzed sequentially from highest to lowest, serving as a reference for subsequent load power switching.

[0060] For example, please refer to Figure 4 Step S14 includes sub-steps S14-1 to S14-3, which are described in detail below: S14-1. Based on the operating power and power fluctuation of each load, obtain the fluctuation impact coefficient of each load under its fluctuating operating conditions. When the power of the load itself is large and the possibility of fluctuation is high, in order to prevent power outages caused by current surges and voltage dips when the load increases sharply, the corresponding load should be switched using a power supply module with a smaller current load and better stability.

[0061] We can first obtain the estimated power fluctuation of the current load by multiplying its current operating power by its power fluctuation. The operating power of load k at the current time t is... Power fluctuation is The estimated power fluctuation of the current load is then... The power fluctuation estimate of the current load is further normalized to obtain the fluctuation impact coefficient of the current load under its fluctuating operating conditions, i.e., the fluctuation impact coefficient is... .

[0062] S14-2. Based on the switching necessity and load power of each power supply module, obtain the acceptance capacity index of each power supply module. The current actual total load power of power supply module a is: It can obtain the actual total load power of each power supply module at the current time, and obtain the maximum value by comparing the switching necessity of all power supply modules in the system at the current moment. This dynamic benchmark value reflects the situation of the module under the greatest pressure in the current system, making the calculation of the necessity difference more meaningful. The necessity difference of the current power supply module is obtained by comparing the maximum value of all switching necessities in historical power supply tasks with the switching necessity of the current power supply module. The necessity difference is... .

[0063] Based on the current necessity difference of the power supply module and the load power, the acceptance capacity index of the current power supply module is obtained. The acceptance capacity index of power supply module a is obtained through... Calculations show that To avoid setting parameter parameters with a denominator of 0, their value can be, for example, 0.1, with dimensions consistent with the actual total load power. The dimensions are consistent, and the error caused by setting its value is within an acceptable range.

[0064] It should be noted that the lower the necessity of switching the current power supply module, the smaller the difference in necessity. The larger the value, the less necessary the corresponding load switching is, and the lower the requirements for the power supply module. In this case, the actual total load power of power supply module a is... The smaller the value, the higher the load redundancy of power supply module a, the more efficient it is at handling loads, the stronger its capacity to accept loads, and the more feasible it is to switch loads in the future.

[0065] S14-3. Based on the fluctuation impact coefficient of each load and the acceptance capacity index of power supply through different power supply modules, obtain the switching feasibility of the corresponding load from the current power supply module to other power supply modules. The feasibility of switching load k to power supply module a at the current time t is: , Using the max-min normalization method to... After normalization, we get Its range is [0,1].

[0066] S15. When the switching feasibility is greater than the preset feasibility threshold, perform a power supply switching task on the current load so that the target power supply module supplies power to the current load.

[0067] Specifically, the feasibility threshold can be set based on the experience of technical personnel or based on calibration experiments. For example, when the range of switching feasibility is [0,1], the feasibility threshold can be set to 0.7, and the switching feasibility is greater than... At time t, it is determined that switching load k on power supply module a is feasible. To conserve resources and facilitate the conversion of other loads, among all candidate power supply modules with a switching feasibility greater than the feasibility threshold, the target power supply module with the highest switching feasibility value can be selected for switching. This ensures that the selected solution is the globally optimal solution considering both load risk and module capacity. After the switch, the original power supply module supplies power to the outside under the new load, and the recalculated switching feasibility is less than or equal to... When this happens, the conversion of the original power supply module can be stopped.

[0068] The feasibility threshold (0.7) given in the embodiments of this invention are empirical values ​​obtained under typical hardware configurations and test scenarios, intended to facilitate understanding of this invention. In practical applications, those skilled in the art can adjust, calibrate, or optimize the feasibility threshold according to specific hardware performance, scenario complexity, and data characteristics, which does not constitute a limitation of this invention.

[0069] It should be noted that the above methods are used to analyze the abnormal conditions of different power supply modules and perform corresponding switching analyses on their loads. The switching requirements are stored in a database along with the monitored power, voltage, and current data. The switching command for the power supply module is transmitted to the ATS (Automatic Transfer Switching Equipment). After a delay confirmation, the ATS immediately sends an action command to its own switching mechanism to execute the power switch. During this critical switching process, the downstream UPS plays a crucial role. The UPS needs to use its internal battery to instantly discharge, providing an uninterrupted power buffer for the load and ensuring power continuity during the ATS switching process.

[0070] Based on the same technical concept as the switching method, this embodiment of the invention also provides a server power switching device that supports multi-node power supply. This device corresponds to any of the power switching methods described above. Please refer to... Figure 5 , Figure 5 This is a schematic diagram of the switching device, which includes an acquisition module 501, a first acquisition module 502, a second acquisition module 503, a third acquisition module 504, and a power supply switching module 505.

[0071] The acquisition module 501 is used to acquire the power supply efficiency and internal resistance data of each power supply module in the power supply system, as well as the load power supplied by each power supply module to the outside.

[0072] The first obtaining module 502 is used to obtain the performance degradation coefficient of each power supply module based on the power supply efficiency and internal resistance data of each power supply module.

[0073] The second obtaining module 503 is used to obtain the switching necessity of each power supply module in supplying the current load based on the performance degradation coefficient and load power of each power supply module.

[0074] The third obtaining module 504 is used to obtain the switching feasibility of the corresponding load from the current power supply module to other power supply modules based on the switching necessity of each power supply module, the load power, and the power fluctuation of each load.

[0075] The power supply switching module 505 is used to perform a power supply switching task on the current load when the switching feasibility is greater than a preset feasibility threshold, so that the target power supply module can supply power to the current load.

[0076] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0077] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A server power switching method supporting multi-node power supply, characterized in that, The method includes: Acquire the power supply efficiency and internal resistance data of each power supply module in the power system, as well as the load power supplied by each power supply module to the external system; Based on the power supply efficiency and internal resistance data of each power supply module, the performance degradation coefficient of each power supply module is obtained; Based on the performance degradation coefficient of each power supply module and the load power, the switching necessity of each power supply module to switch power supply under the current load is obtained; Based on the switching necessity of each power supply module and the load power and the power fluctuation of each load, the switching feasibility of switching the corresponding load from the current power supply module to other power supply modules is obtained; When the switching feasibility is greater than a preset feasibility threshold, a power supply switching task is performed on the current load so that the target power supply module supplies power to the current load.

2. The server power switching method supporting multi-node power supply according to claim 1, characterized in that, The process of obtaining the performance degradation coefficient of each power supply module based on its power supply efficiency and internal resistance data includes: The efficiency decay characteristics of each power supply module are analyzed to obtain the first performance index characterizing the power supply efficiency decay of each power supply module. Based on the current internal resistance value of each power supply module's internal resistance data, a second performance index characterizing the internal resistance variation of each power supply module is obtained. The performance degradation coefficient of each power supply module is obtained based on the first and second performance indicators of each power supply module.

3. The server power switching method supporting multi-node power supply according to claim 2, characterized in that, The analysis of the efficiency degradation characteristics of each power supply module to obtain a first performance index characterizing the power supply efficiency degradation of each power supply module includes: Based on the efficiency difference of each power supply module at adjacent times in the historical period, the efficiency difference set of each power supply module is obtained. Each set of efficiency differences is compared with a preset difference threshold to obtain the number of efficiency differences in each set that are greater than the difference threshold. Based on the number of efficiency differences for each power supply module and the number of power supply efficiencies during the historical period, a first characteristic coefficient representing the power supply efficiency decay of each power supply module is obtained. Based on the difference between the current power supply efficiency of each power supply module and the efficiency extreme value of the historical period, a second characteristic coefficient representing the power supply efficiency decay of each power supply module is obtained. The first performance index of each power supply module is obtained based on the first characteristic coefficient and the second characteristic coefficient of each power supply module.

4. The server power switching method supporting multi-node power supply according to claim 1, characterized in that, Based on the performance degradation coefficient of each power supply module and the load power, the switching necessity of each power supply module in supplying the current load is obtained, including: The attenuation degradation index of each power supply module is obtained based on the difference between the maximum degradation coefficient and the performance degradation coefficient of each power supply module in historical power supply tasks. A correlation characteristic analysis is performed on the load power of the current power supply module and other power supply modules to obtain load fluctuation indicators that characterize the load stability performance of each power supply module. Based on the attenuation and degradation index and load fluctuation index of each power supply module, the switching necessity of each power supply module to switch power supply under the current load is obtained.

5. The server power switching method supporting multi-node power supply according to claim 4, characterized in that, The correlation feature analysis of the load power of the current power supply module with other power supply modules to obtain load fluctuation indicators characterizing the load stability performance of each power supply module includes: Based on the difference between the first load power of the current power supply module and the second load power of each other power supply module, obtain the power difference set of the current power supply module; Based on the cumulative summation of all data in the current power difference set of the power supply module and the stability index of the current power supply module, the load fluctuation index of each power supply module is obtained.

6. The server power switching method supporting multi-node power supply according to claim 1, characterized in that, Before obtaining the switching feasibility of switching a corresponding load from the current power supply module to other power supply modules based on the switching necessity of each power supply module, the load power, and the power fluctuation of each load, the method further includes: The number of fluctuations of each load in multiple historical operation tasks is statistically analyzed to obtain the fluctuation frequency of each load's operating power fluctuation exceeding the power threshold. The peak power coefficient of each load is obtained by statistically analyzing the maximum operating power of each load in multiple historical operation tasks. The power fluctuation of each load is obtained based on its fluctuation frequency and peak power factor.

7. The server power switching method supporting multi-node power supply according to claim 1, characterized in that, The step of obtaining the switching feasibility of switching a corresponding load from the current power supply module to other power supply modules based on the switching necessity of each power supply module, the load power, and the power fluctuation of each load includes: Based on the operating power and power fluctuation of each load, obtain the fluctuation impact coefficient of each load under its fluctuating operating conditions; Based on the switching necessity and load power of each power supply module, the acceptance capacity index of each power supply module is obtained; Based on the fluctuation impact coefficient of each load and the acceptance capacity index of power supply through different power supply modules, the switching feasibility of the corresponding load from the current power supply module to other power supply modules is obtained.

8. The server power switching method supporting multi-node power supply according to claim 7, characterized in that, The process of obtaining the fluctuation impact coefficient of each load under fluctuating operating conditions based on the operating power and power fluctuation of each load includes: The fluctuating power of the current load is estimated by multiplying the current load power and the power fluctuation. The power fluctuation estimate of the current load is normalized to obtain the fluctuation impact coefficient of the current load under its fluctuating operating conditions.

9. The server power switching method supporting multi-node power supply according to claim 7, characterized in that, The process of obtaining the acceptance capacity index of each power supply module based on the switching necessity and load power of each power supply module includes: The necessity difference of the current power supply module is obtained by comparing the maximum value of all switching necessities in historical power supply tasks with the switching necessity of the current power supply module. Based on the current power supply module's necessity difference and load power, the acceptance capacity index of the current power supply module is obtained.

10. A server power switching device supporting multi-node power supply, characterized in that, The device is the device corresponding to any one of the power switching methods according to claims 1-9, and the device includes: The acquisition module is used to acquire the power supply efficiency and internal resistance data of each power supply module in the power system, as well as the load power supplied by each power supply module to the outside. The first acquisition module is used to obtain the performance degradation coefficient of each power supply module based on the power supply efficiency and internal resistance data of each power supply module. The second obtaining module is used to obtain the switching necessity of each power supply module to perform power switching under the current load based on the performance degradation coefficient of each power supply module and the load power. The third obtaining module is used to obtain the switching feasibility of the corresponding load from the current power supply module to other power supply modules based on the switching necessity of each power supply module, the load power, and the power fluctuation of each load. The power supply switching module is used to perform a power supply switching task on the current load when the switching feasibility is greater than a preset feasibility threshold, so that the target power supply module supplies power to the current load.

Images