Intelligent management method and system for server modular power supply

Abstract

The application discloses an intelligent management method and system of a server modular power supply, relates to the technical field of computers, and discloses the intelligent management method and system of the server modular power supply, which collects server load and power module running state data in real time, performs load trend prediction in combination with historical data, dynamically schedules power module output power based on a prediction result, makes the modules always run in an optimal efficiency interval, simultaneously realizes real-time monitoring of fault parameters and seamless switching of redundant modules, and can improve the fault tolerance capability and energy efficiency of the server modular power supply.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to intelligent management methods and systems for modular power supplies of servers. Background Technology

[0002] Against the backdrop of rapid evolution in information technology and widespread penetration of internet services, data centers and server systems face ever-increasing computing and storage pressures, making the demand for high reliability, high availability, and flexible scalability increasingly urgent. As a core component ensuring the continuous and stable operation of servers, the performance of the server power system directly affects the overall system's operational quality. However, current mainstream server power systems generally adopt a static, fixed configuration model, where the capacity and number of power modules are locked during the equipment manufacturing stage, preventing users from dynamically adjusting them based on load fluctuations during actual operation. When server load increases, service must be interrupted and the entire power system replaced, which not only drives up hardware procurement and maintenance costs but also causes unnecessary business downtime, affecting service continuity. In terms of system reliability, existing power architectures lack scientific redundancy design; a single power module failure can easily trigger a global power outage, leading to the loss of critical data and prolonged business paralysis. Simultaneously, fault monitoring mechanisms are simplistic and crude, relying on basic threshold alarms and lacking in-depth analysis capabilities, failing to achieve real-time fault identification and rapid response, resulting in significant delays in the recovery process. In terms of energy efficiency, the power supply system cannot dynamically adjust its output power according to real-time load, resulting in a sharp deterioration in conversion efficiency under low-load conditions. Furthermore, the lack of intelligent coordination mechanisms between power modules leads to some modules operating under prolonged light or overload conditions, running in inefficient ranges and wasting energy while incurring additional heat dissipation burdens. Regarding maintenance, the existing system does not support hot-swapping, requiring system shutdown for faulty module replacement, further reducing system availability and operational flexibility.

[0003] The above content is only used to help understand the technical solution of this application and does not represent an admission that the above content is prior art. Summary of the Invention

[0004] The main purpose of this application is to provide an intelligent management method and system for server modular power supplies, which aims to improve the fault tolerance and energy efficiency of server modular power supplies.

[0005] To achieve the above objectives, this application proposes an intelligent management method for modular power supplies in servers, the method comprising: The system collects server load data, operating status data of each power module, and specification parameter data of each power module. It then preprocesses the server load data to obtain preprocessed load data. The server load data includes processor utilization data, memory utilization data, and network traffic data. The operating status data includes voltage data, current data, temperature data, current output power data, and cumulative runtime data for each power module. The specification parameter data includes the rated power value of each power module. The preprocessed load data is matched with the historical load data for the corresponding time period in the historical load data. Based on the matching result, the direction and magnitude of load change in the future preset time period are determined, and load prediction data is generated. Based on the load forecast data, the operating status data of each power module, and the specification parameter data, combined with the preset power scheduling strategy and redundancy configuration strategy, a power module scheduling scheme is determined; the power module scheduling scheme includes the identification information of the power module to be adjusted, the adjustment type, and the target output power of each power module after adjustment. According to the power module scheduling scheme, control commands are sent to the corresponding power modules to adjust the output power and operating status of each power module, so that each power module operates within the preset optimal efficiency range. The operating status data of each power module is continuously monitored, and each parameter in the operating status data is compared with a preset safety threshold. When any parameter exceeds the corresponding preset safety threshold, the power module is determined to be a faulty power module, triggering the protection mechanism and generating a fault alarm message. At the same time, the automatic switching of the backup power module is performed according to the redundancy configuration strategy.

[0006] In one embodiment, the step of performing trend matching between the preprocessed load data and historical load data for the corresponding time period in the historical load data, and determining the direction and magnitude of load change within a future preset time period based on the matching result, to generate load prediction data includes: Load peak features, load average features, and load fluctuation features are extracted from the preprocessed load data, and the load peak features, load average features, and load fluctuation features are combined into load feature data. Historical load peak features, historical load average features, and historical load fluctuation features for the corresponding time period are extracted from the historical load data and combined into historical load features. The load feature data is then trend-matched with the historical load features, and the similarity between the load feature data and the historical load features is calculated. The subsequent change trend associated with the historical load feature corresponding to the highest similarity is selected as a reference trend. Based on the degree of deviation between the load feature data and the reference trend, the direction and magnitude of load change within a preset time period are determined, and the load prediction data is generated.

[0007] In one embodiment, based on the load forecast data, the operating status data of each power module, and the specification parameter data, combined with a preset power scheduling strategy and redundancy configuration strategy, a power module scheduling scheme is determined; the power module scheduling scheme includes the steps of identifying the power modules to be adjusted, the adjustment type, and the target output power of each power module after adjustment, including: The total power demand value for a future preset period is calculated based on the load forecast data, and the total power demand value is compared with the rated power value of each power module to determine the minimum number of working power modules required to meet the total power demand value. The number of redundant power modules is determined based on the redundancy configuration strategy, and the minimum number is added to the number of redundant power modules to obtain the target number of working modules; wherein, the redundancy configuration strategy is an N+1 redundancy strategy or an N+N redundancy strategy. The current output power data, temperature data, and cumulative running time data of each power module in the operation status data are comprehensively analyzed to determine the priority ranking of each power module, and the power modules are selected as working power modules in descending order of priority. Based on the total power demand and the target number of working modules, the total power is allocated to the selected working power modules, and the target output power of each working power module is determined. The power module scheduling scheme is generated by summarizing the information. The power module scheduling scheme includes: the identifier of the power module to be started, the identifier of the power module to be shut down, and the target output power value of each working power module.

[0008] In one embodiment, the step of allocating the total power to selected working power modules based on the total power demand and the target number of working modules, and determining the target output power of each working power module, includes: Obtain the current output power data and rated power value of each working power module, and calculate the current load rate of each working power module; The total power demand is initially allocated according to the ratio of the rated power value of each working power module to obtain the initial allocated power of each working power module. For a power supply module whose initial power allocation exceeds the upper limit of the preset optimal efficiency range, the excess power is transferred to other power supply modules whose current load rate is lower than the lower limit of the preset optimal efficiency range. After iterative adjustment, the actual load rate of each power supply module is within the preset optimal efficiency range. Alternatively, when the number of iterative adjustments reaches the preset maximum iteration threshold, the adjustment is stopped and the current power allocation is used as the target output power to obtain the target output power of each power supply module.

[0009] In one embodiment, when the total power demand is lower than a preset low load threshold, the power module scheduling scheme further includes: At least one power module currently in operation is marked as a module to be hibernated, and a hibernation command is sent to the module to be hibernated to make it enter a low-power hibernation state; wherein, the power module in hibernation state maintains communication connectivity and is able to receive a wake-up command and return to the operating state; When the subsequently generated load prediction data indicates that the total power demand value rises above the preset low load threshold, a wake-up command is sent to the power module in sleep mode to restore it to working state.

[0010] In one embodiment, after the step of sending control commands to the corresponding power modules according to the power module scheduling scheme to adjust the output power and operating status of each power module so that each power module operates within a preset optimal efficiency range, the method further includes: Obtain the temperature data of each power module and the ambient temperature data inside the server; The temperature data of each power module is compared with a preset graded temperature threshold to determine the current temperature level of each power module; the preset graded temperature threshold includes at least a first temperature threshold and a second temperature threshold, and the first temperature threshold is lower than the second temperature threshold. Based on the current temperature level of each power module, a corresponding cooling fan speed adjustment strategy is determined; wherein, when the current temperature level exceeds the first temperature threshold but does not exceed the second temperature threshold, the cooling fan speed is increased at a first rate; when the current temperature level exceeds the second temperature threshold, the cooling fan speed is increased at a second rate, wherein the second rate is greater than the first rate. According to the cooling fan speed adjustment strategy, a speed adjustment command is sent to the cooling fan control module.

[0011] In one embodiment, the operating status data of each power module is continuously monitored, and each parameter in the operating status data is compared with a preset safety threshold one by one. When any parameter exceeds the corresponding preset safety threshold, the power module is determined to be a faulty power module, triggering a protection mechanism and generating a fault alarm message. Simultaneously, the automatic switching of the backup power module is performed according to the redundancy configuration strategy, including: The voltage data of each power module is compared with the preset overvoltage threshold and the preset undervoltage threshold, the current data is compared with the preset overcurrent threshold, and the temperature data is compared with the preset overtemperature threshold to obtain the comparison results. When the comparison result indicates that any parameter of any power module exceeds the corresponding preset safety threshold, the power module is determined to be a faulty power module, and fault identification data containing the faulty power module identifier, fault parameter type and fault parameter value is generated. Based on the fault identification data, an output cut-off operation is performed on the faulty power module, and the output power of the faulty power module before the cut-off is recorded as the power value to be switched. At the same time, a target module for switching is selected from the power modules in standby state. A start command is sent to the target switching module to adjust the output power of the target switching module from the initial value to the power value to be switched, thereby completing the automatic switching of the backup power module.

[0012] In one embodiment, after the automatic switching of the backup power module is completed, the method further includes: After detecting that the faulty power module has been physically removed from the server, the fault alarm information corresponding to the faulty power module is cleared. After a new power module is detected being inserted into the server, an initialization test is performed on the newly inserted power module. The initialization test includes voltage calibration test, current calibration test, and communication connectivity test. Once the initialization test passes, the newly inserted power module is configured as a standby module and brought under the jurisdiction of the redundancy configuration strategy.

[0013] In one embodiment, before the steps of collecting server load data, operating status data of each power module, and specification parameter data of each power module, and preprocessing the server load data to obtain preprocessed load data, the method further includes: Perform a power-on self-test on each power module to confirm that the communication connectivity and functions of each power module are normal. Obtain the specification parameter data of each power module, which also includes the rated voltage range and rated current range; Based on the specification parameters and redundancy configuration strategy of each power module, the initial role assignment of each power module is determined, and the initial role assignment divides each power module into working modules and standby modules. Each power module is configured with a corresponding preset safety threshold, which includes an overvoltage threshold, an undervoltage threshold, an overcurrent threshold, and an overtemperature threshold.

[0014] Furthermore, to achieve the above objectives, this application also proposes an intelligent management system for a server modular power supply. The intelligent management system for the server modular power supply includes: a memory, a processor, and an intelligent management program for the server modular power supply stored in the memory and executable on the processor. The intelligent management program for the server modular power supply is configured to implement the steps of the intelligent management method for the server modular power supply.

[0015] The intelligent management method and system for modular power supplies for servers proposed in this application collect server load and power module operating status data in real time, combine historical data to predict load trends, and dynamically schedule the output power of power modules based on the prediction results, so that the modules always operate in the optimal efficiency range. At the same time, it realizes real-time monitoring of fault parameters and seamless switching of redundant modules, which can improve the fault tolerance and energy efficiency of modular power supplies for servers. Attached Figure Description

[0016] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart illustrating an embodiment of the intelligent management method for modular power supplies in the server according to this application; Figure 2 This is a schematic diagram of a structural design for an embodiment of the intelligent management system for modular power supplies of the server in this application.

[0019] Explanation of icon numbers: 10. Memory; 20. Processor.

[0020] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0021] The technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of this application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0022] It should be understood that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0023] In existing technologies, server power systems suffer from high replacement costs and system downtime when the power capacity and quantity are fixed and cannot be flexibly adjusted. When a single power module fails, the entire server system may lose power, and there is a lack of real-time monitoring and automatic switching mechanisms. Under low load conditions, energy efficiency is poor, and there is a lack of load balancing mechanisms, resulting in energy waste. Furthermore, maintenance operations usually require downtime, affecting system availability and business continuity.

[0024] Based on this, embodiments of this application provide an intelligent management method for modular power supplies of servers, referring to... Figure 1 The intelligent management method for the modular power supply of the server includes steps S100 to S500, wherein: Step S100: Collect server load data, operating status data of each power module, and specification parameter data of each power module, and preprocess the server load data to obtain preprocessed load data; wherein, the server load data includes processor utilization data, memory utilization data, and network traffic data; the operating status data includes voltage data, current data, temperature data, current output power data, and cumulative runtime data of each power module; the specification parameter data includes the rated power value of each power module; Step S200: Perform trend matching between the preprocessed load data and the historical load data for the corresponding time period in the historical load data, determine the direction and magnitude of load change within the future preset time period based on the matching result, and generate load prediction data. Step S300: Based on the load prediction data, the operating status data of each power module, and the specification parameter data, and in conjunction with the preset power scheduling strategy and redundancy configuration strategy, a power module scheduling scheme is determined; the power module scheduling scheme includes the identification information of the power module to be adjusted, the adjustment type, and the target output power of each power module after adjustment. Step S400: According to the power module scheduling scheme, send control commands to the corresponding power modules to adjust the output power and working status of each power module so that each power module operates within the preset optimal efficiency range. Step S500: Continuously monitor the operating status data of each power module, compare each parameter in the operating status data with the preset safety threshold one by one, and when any parameter exceeds the corresponding preset safety threshold, determine that the power module is a faulty power module, trigger the protection mechanism and generate fault alarm information, and at the same time execute the automatic switching of the backup power module according to the redundancy configuration strategy.

[0025] In this embodiment, the power scheduling strategy refers to a set of rules guiding power modules to start / stop, allocate power, and balance load according to load demand. This strategy aims to optimize the overall energy efficiency and response speed of the power system, ensuring that power modules can operate efficiently and stably under different load conditions. The redundancy configuration strategy refers to a power module backup mechanism set up to ensure the continuity and reliability of power supply to the server system. This strategy defines the additional power capacity or number of backup modules that the system should maintain during normal operation or in the event of a failure, to prevent system interruption due to a single point of failure. The power module scheduling scheme refers to a set of specific operation instructions calculated based on load forecasting, power module operating status, and preset strategies. This scheme specifies in detail which power modules need to be adjusted, the type of adjustment, and the target output power value of each working power module after adjustment.

[0026] In this embodiment, the optimal efficiency range refers to the range within which the power module's energy conversion efficiency reaches or approaches its peak when operating within a specific output power range. Operating the power module within this range helps reduce energy loss and improve the overall energy efficiency of the power system. Preset safety thresholds refer to the upper or lower limits set for various operating parameters to ensure the safe and stable operation of the power module. When parameters such as voltage, current, and temperature of the power module exceed these thresholds, it indicates that the module has an abnormality or failure risk. The protection mechanism refers to a series of automatic countermeasures taken by the system when an abnormality or failure of the power module is detected. This mechanism aims to prevent the fault from escalating, protect other system components, and ensure the continuity of power supply.

[0027] In this embodiment, the intelligent management method for modular server power supplies first collects server load data, operating status data of each power module, and specification parameter data of each power module. Specifically, server load data can be obtained by the system through software interfaces, such as periodically reading processor utilization data, memory utilization data, and network traffic data through the API interface of the operating system or virtualization platform. This data can be stored directly in its raw form. The operating status data of each power module can be obtained through the power management unit (PMU) or intelligent sensors, such as reading voltage data, current data, temperature data, current output power data, and cumulative runtime data of each power module through I2C or PMBus bus. This data can be used directly without calibration or filtering. The specification parameter data of each power module, such as rated power values, can be pre-stored in the system configuration table or manually entered when the power module is first installed. After data collection, the server load data is preprocessed, for example, the raw load data can be simply converted to ensure data consistency, thereby obtaining preprocessed load data.

[0028] Furthermore, the preprocessed load data is compared with historical load data for the corresponding time period to determine trends. For example, a simple numerical comparison can be made between the current preprocessed load data and historical load data for the same time period in the past to determine whether the current load is higher, lower, or in line with the historical average. Based on the results of this simple numerical comparison, the direction and approximate magnitude of load changes within a preset future time period can be preliminarily determined. This generates load forecast data.

[0029] Based on this, and using the load forecast data, the operating status data of each power module, and the specification parameter data, combined with preset power scheduling and redundancy configuration strategies, a power module scheduling scheme is determined. For example, the power scheduling strategy can be set as follows: when the load forecast data indicates that the total power demand exceeds 80% of the current working module's total capacity, a standby module is activated; when the load is below 40% of the current working module's total capacity, a working module is shut down. The redundancy configuration strategy can simply stipulate that a standby module is always maintained. According to these rules, the identification information of the power modules to be adjusted, the adjustment type, and the target output power of each working power module after adjustment can be determined; for example, the total power demand can be evenly distributed among all working modules.

[0030] In this embodiment, control commands are sent to the corresponding power modules according to the power module scheduling scheme. For example, a simple on / off command can be sent to the power modules via the PMBus interface, or a fixed output power percentage command can be set. This command aims to adjust the output power and operating state of each power module to ensure it operates within a preset optimal efficiency range. For example, if the scheduling scheme requires a module to output 50% of its rated power, this command is sent directly without fine-tuning iterative adjustments to ensure it precisely falls within the optimal efficiency range.

[0031] Furthermore, in this embodiment, the operating status data of each power module is continuously monitored. For example, the system can poll the voltage, current, and temperature data of each power module every few seconds. Each parameter in the operating status data is compared with a preset safety threshold. For example, a fixed overvoltage threshold, a fixed undervoltage threshold, a fixed overcurrent threshold, and a fixed overtemperature threshold can be set. When any parameter exceeds the corresponding preset safety threshold, the power module is determined to be a faulty power module. At this time, a protection mechanism is triggered, for example, issuing an audible alarm and displaying fault information, while simultaneously generating a fault alarm message. As a simple redundancy configuration strategy, the first available backup power module can be automatically switched to the working state in a preset order to replace the faulty module.

[0032] In this embodiment, intelligent scheduling of power modules is achieved through comprehensive collection and preprocessing of server load data, power module operating status data, and specification parameter data, combined with load prediction and preset strategies. This method dynamically adjusts the output power and operating status of the power modules, ensuring they operate within their optimal efficiency range, thereby effectively reducing energy consumption. Simultaneously, by continuously monitoring the operating status and responding promptly to faults, automatically switching to backup power modules improves the reliability and power supply continuity of the server power system, preventing system downtime and data loss due to power failures, and solving the problems of low energy efficiency, poor reliability, and inconvenient maintenance inherent in traditional power systems.

[0033] In some of the embodiments described above in this application, although it is proposed to perform trend matching between preprocessed load data and historical load data to generate load prediction data, in practical applications, if only simple trend matching is performed, it may be difficult to accurately capture the complex dynamic changes of server load, such as instantaneous peaks, continuous average levels, and short-term fluctuations, thereby affecting the accuracy of load prediction. This may lead to an underoptimized power module scheduling scheme, which may fail to fully leverage the efficiency advantages of modular power supplies.

[0034] To address this, this application further proposes a step of trend matching between preprocessed load data and historical load data for the corresponding time period, determining the direction and magnitude of load changes within a future preset time period based on the matching results, and generating load prediction data. Specifically, this step includes: In this embodiment, load peak characteristics, load average characteristics, and load fluctuation characteristics are extracted from the preprocessed load data, and these characteristics are combined into load feature data. The load peak characteristic aims to capture the maximum load demand of the preprocessed load data within a specific time window. This can be obtained by identifying the highest data point or applying a peak detection algorithm, which is crucial for evaluating the instantaneous load capacity of the power system. The load average characteristic reflects the average load level of the preprocessed load data over a period of time and can be obtained by calculating the arithmetic mean, used to characterize the typical operating load of the server. The load fluctuation characteristic quantifies the degree of change or instability of the preprocessed load data over time. For example, it can be measured using standard deviation, variance, or the difference between the maximum and minimum values, which helps to understand the dynamic characteristics of load changes. By combining these independent features, a multi-dimensional and comprehensive description of the current load pattern is formed, i.e., load feature data, providing a rich information foundation for subsequent trend matching.

[0035] In this embodiment, historical load peak features, historical load average features, and historical load fluctuation features for the corresponding time period are extracted from the historical load data and combined into historical load features. The load feature data is then trend-matched with the historical load features to calculate the similarity between the load feature data and the historical load features. In this step, the system extracts the load peak features, load average features, and load fluctuation features from the stored historical load data for the historical time period corresponding to the current preprocessed load data (e.g., the same day of the same week, the same hour of the same day, etc.), and combines them into historical load features. Subsequently, the currently generated load feature data is compared with these historical load features to perform trend matching. Trend matching can be implemented using various algorithms, such as Euclidean distance, cosine similarity, correlation coefficient, or dynamic time warping (DTW) to quantify the similarity between the current load feature data and each historical load feature. The calculated similarity value will serve as an indicator of the degree of fit between the current load pattern and the historical pattern.

[0036] In this embodiment, the subsequent change trend associated with the historical load feature corresponding to the highest similarity is selected as a reference trend. Combining the deviation between the load feature data and the reference trend, the direction and magnitude of load change within a preset future time period are determined, generating the load prediction data. After identifying the historical load feature most similar to the current load pattern, the system extracts the subsequent load change trend that accompanied the actual occurrence of that historical load feature and uses it as a reference trend for predicting future load. To further improve prediction accuracy, the system also considers subtle deviations between the current load feature data and the selected historical load feature. For example, if the peak value of the current load is slightly higher than the peak value of a similar historical pattern, the predicted value of the future load is adjusted upwards based on the reference trend. In this way, by comprehensively considering the evolution of historical patterns and the actual situation of the current load, the expected direction of server load change (e.g., rising, falling, or remaining stable) and the specific magnitude of change within a preset future time period can be accurately determined, ultimately generating accurate load prediction data.

[0037] In this embodiment, through the above technical solution, this application extracts load peak characteristics, load average characteristics, and load fluctuation characteristics from preprocessed load data and historical load data, and combines them into multi-dimensional load characteristic data, which can more comprehensively and finely depict the dynamic characteristics of server load. Therefore, by performing trend matching and similarity calculation between current load characteristic data and historical load characteristics, the historical context most similar to the current load pattern can be accurately identified. Based on this, the subsequent historical change trend corresponding to the maximum similarity is selected as a reference, and corrections are made based on the deviation between the current load characteristics and this reference trend, thereby improving the prediction accuracy of the load change direction and magnitude within a preset future time period. This refined load prediction data enables the power module scheduling scheme to respond more accurately to actual load demands, avoiding frequent start-stop or inefficient operation of power modules due to inaccurate predictions, thus ensuring stable operation of each power module within a preset optimal efficiency range, effectively improving the overall energy efficiency and reliability of the power system.

[0038] To address this, this application further proposes a method for determining a power module scheduling scheme. This method can dynamically and intelligently determine the power module scheduling scheme based on real-time server load forecasts and the operating status of the power modules. In practical applications, determining the power module scheduling scheme solely based on load forecast data and preset strategies may not fully consider the actual operating conditions and lifespan balancing issues of the power modules. For example, simply allocating the load according to rated power may result in some modules operating at high loads or high temperatures for extended periods, while others remain idle, thus affecting the overall lifespan of the power modules and system energy efficiency. Furthermore, ensuring that necessary system redundancy is maintained while meeting load demands is also a problem requiring refined management.

[0039] The method proposed in this application includes: calculating the total power demand value within a preset future time period based on load forecast data, comparing the total power demand value with the rated power value of each power module, and determining the minimum number of working power modules required to meet the total power demand value; determining the number of redundant power modules based on a redundancy configuration strategy, and adding the minimum number to the number of redundant power modules to obtain the target number of working power modules; wherein, the redundancy configuration strategy is an N+1 redundancy strategy or an N+N redundancy strategy; comprehensively analyzing the current output power data, temperature data, and cumulative running time data of each power module in the operating status data, determining the priority ranking of each power module, and selecting power modules as working power modules in descending order of priority; allocating the total power to the selected working power modules according to the total power demand value and the target number of working power modules, and determining the target output power of each working power module; and generating a power module scheduling scheme, which includes: the identifier of the power module to be started, the identifier of the power module to be shut down, and the target output power value of each working power module.

[0040] In this embodiment, when calculating the total power demand for a future preset time period based on load forecast data and determining the minimum number of working power modules required to meet the total power demand, the system can first input the load forecast data (e.g., processor utilization data, memory utilization data, and network traffic data) into a pre-established power consumption model to obtain a comprehensive and quantified total power demand value. This power consumption model can be trained and optimized based on historical operating data to improve prediction accuracy. Once the total power demand value is obtained, the system can compare it with the rated power value of each power module. For example, by dividing the total power demand value by the rated power value of a single power module and rounding up, the minimum number of working power modules required to meet the current load demand can be preliminarily determined. This step lays the foundation for subsequent power module selection and power allocation.

[0041] Building upon this, to ensure high availability and reliability of the system, this application further determines the number of redundant power modules based on a redundancy configuration strategy, and adds the minimum number to the number of redundant power modules to obtain the target number of working modules. Redundancy configuration strategy is a key technology in server power management, designed to cope with sudden failures and prevent system interruption due to single-point failures by configuring additional backup power modules. In this application, the redundancy configuration strategy can adopt an N+1 redundancy strategy or an N+N redundancy strategy. An N+1 redundancy strategy means configuring one additional backup module in addition to the N working modules required to meet the total power demand, providing fault tolerance for single-module failures. An N+N redundancy strategy means configuring N additional backup modules in addition to the N working modules required to meet the total power demand, i.e., the number of working modules is equal to the number of backup modules, providing higher fault tolerance. System administrators can flexibly choose an appropriate redundancy strategy based on the server's criticality, budget constraints, and specific requirements for system reliability.

[0042] In this embodiment, to optimize the operating efficiency of the power modules and extend their service life, this application comprehensively analyzes the current output power data, temperature data, and cumulative runtime data of each power module in the operating status data to determine the priority ranking of each power module, and selects power modules as working power modules in descending order of priority. For example, the system can prioritize power modules with lower current output power, lower temperature, or shorter cumulative runtime as working modules. By setting different weights for these operating parameters and using weighted summation or other multi-index decision-making algorithms (such as fuzzy comprehensive evaluation, analytic hierarchy process, etc.), a comprehensive score for each power module can be calculated, thereby forming a dynamic priority ranking. This intelligent selection mechanism helps to achieve load balancing and rotation of power modules, avoiding premature failure of some modules due to long-term high load or high temperature operation, thereby extending the overall service life of the power system.

[0043] In this embodiment, the total power is allocated to the selected power modules based on the total power requirement and the target number of operating modules, and the target output power of each power module is determined. This step is the core of achieving refined power management. The system can employ various allocation strategies. For example, initial allocation can be performed based on the rated power ratio of each power module, or dynamic allocation can be performed based on the real-time efficiency curve of each module to ensure that each module operates as close as possible to its optimal efficiency point. During the allocation process, it is crucial to strictly ensure that the power allocated to each module does not exceed its rated power, and that the sum of the output power of all operating modules is precisely equal to the total power requirement.

[0044] In this embodiment, the system aggregates all the above decision results to generate a power module scheduling scheme. This scheduling scheme is a structured instruction set that explicitly includes the identifiers of power modules to be started, power modules to be shut down, and the target output power value for each working power module. For example, if the current number of working modules is insufficient to meet the target number, power modules that need to switch from standby to working status will be identified; if the current number of working modules exceeds the target number, power modules that need to switch from working status to standby or shutdown status will be identified. This information provides clear and executable guidance for subsequent power module control.

[0045] In this embodiment, by accurately calculating the total power demand and combining it with a redundancy configuration strategy, the system can ensure that it maintains necessary system reliability while meeting load requirements. Simultaneously, by comprehensively analyzing the operating status data of the power modules, such as current output power, temperature, and cumulative runtime, intelligent switching and load balancing of power modules can be achieved, avoiding overload or excessive wear of a single module, thereby extending the overall lifespan of the power modules. Furthermore, rationally allocating the total power demand to the selected working power modules allows each module to operate within its optimal efficiency range as much as possible, improving the overall energy efficiency of the power system and reducing server operating costs. The final generated scheduling scheme is clear and explicit, including identifiers of modules to be started and shut down, and target output power, providing a reliable basis for subsequent precise control and effectively solving the problems of low module utilization, high energy consumption, and insufficient reliability in traditional power management.

[0046] In some embodiments described above in this application, a power module scheduling scheme is determined based on load forecast data, operating status data, and specification parameter data, combined with power scheduling strategies and redundancy configuration strategies. Furthermore, the total power is allocated to the selected working power modules based on the total power demand and the target number of operating modules. However, in the process of allocating total power to the working power modules, simply performing power allocation may cause the actual load rate of some power modules to deviate from their preset optimal efficiency range, thereby affecting the operating efficiency and stability of the entire power system.

[0047] To address this, this application further proposes a method for allocating total power to selected power modules based on the total power demand and the target number of modules, and determining the target output power of each power module. This method includes: acquiring the current output power data and rated power value of each power module, and calculating the current load rate of each power module; initially allocating the total power demand according to the ratio of the rated power value of each power module to obtain the initial allocated power for each power module; for power modules whose initial allocated power exceeds the upper limit of the preset optimal efficiency range, transferring the excess power to other power modules whose current load rate is lower than the lower limit of the preset optimal efficiency range, and iteratively adjusting until the actual load rate of each power module is within the preset optimal efficiency range, or stopping the adjustment when the number of iterations reaches a preset maximum iteration threshold and using the current allocated power as the target output power to obtain the target output power for each power module.

[0048] In this embodiment, when determining the target output power of each working power module, it is first necessary to obtain the current output power data and rated power value of each working power module, and calculate the current load rate of each working power module accordingly. This step aims to obtain the real-time operating information and design parameters of each power module selected for operation. The current output power data reflects the actual power provided by the power module at the current moment, which can be read in real time through sensors or management interfaces inside the power module. The rated power value is the maximum power that the power module can continuously output under normal operating conditions, which can be provided by the manufacturer and stored in the system's specification parameter data. By dividing the current output power data by the rated power value, the current load rate of each working power module can be calculated. The load rate is an important indicator for measuring the workload of the power module, and it is directly related to the operating efficiency, heat generation, and lifespan of the power module. Accurately obtaining and calculating the load rate is the foundation for subsequent refined power allocation and efficiency optimization.

[0049] Based on this, the total power demand is initially allocated according to the proportion of the rated power of each working power module, thus obtaining the initial allocated power for each working power module. After determining the number of working power modules and the total power demand, this application adopts an allocation method based on the rated power ratio to achieve preliminary power allocation. Specifically, the total power demand over a preset future period is taken as the total amount to be allocated. Then, based on the proportion of the rated power of each working power module to the total rated power of all working power modules, the initial allocated power to be borne by each power module is calculated. This allocation method ensures the fairness and rationality of power allocation, allowing power modules with larger rated power to bear more power, which is in line with their design capabilities, and provides a balanced starting point for subsequent fine-tuning.

[0050] In this embodiment, for power modules whose initial power allocation exceeds the upper limit of the preset optimal efficiency range, the excess power is transferred to other power modules whose current load rate is below the lower limit of the preset optimal efficiency range. After iterative adjustment, the actual load rate of each power module is within the preset optimal efficiency range. Alternatively, adjustment stops when the number of iterations reaches a preset maximum iteration threshold, and the current power allocation is used as the target output power, ultimately obtaining the target output power for each power module. This step is crucial for achieving efficient operation of the power modules. The preset optimal efficiency range refers to the range within which the power module can achieve the highest conversion efficiency under a specific load rate, which can be determined by the characteristic curve of the power module. After initial power allocation, the system checks whether the load rate of each power module falls within this optimal efficiency range. If the initial power allocation of a power module causes its load rate to exceed the upper limit of the optimal efficiency range, it means that the module is in an overloaded or inefficient state. In this case, the system calculates the power exceeding the upper limit of the module and transfers it to other power modules whose current load rate is below the lower limit of the preset optimal efficiency range. This power transfer process can be implemented using an iterative algorithm. For example, power can be prioritized for transfer to the module with the lowest load rate, or allocated based on the remaining capacity of the modules, until the actual load rate of all working power modules is adjusted to within a preset optimal efficiency range. When the number of iterations reaches a preset maximum iteration threshold, the adjustment stops even if not all modules are fully adjusted to the optimal efficiency range, and the currently allocated power is used as the target output power. Through this dynamic, iterative power transfer and adjustment mechanism, the target output power of each working power module is ultimately determined, ensuring that each module operates near its highest efficiency point.

[0051] In this embodiment, when determining the target output power of each working power module, a simple power allocation is no longer performed. Instead, a comprehensive consideration is introduced, taking into account the current load rate, rated power value, and preset optimal efficiency range of each power module. First, by acquiring the current output power data and rated power value of each working power module, its current load rate can be accurately calculated, providing a data foundation for subsequent refined allocation. Second, the initial allocation method based on the rated power ratio ensures the initial rationality of the power allocation. More importantly, by iteratively adjusting the initial allocated power, power exceeding the upper limit of the optimal efficiency range is transferred to modules with load rates below the lower limit of the optimal efficiency range. Simultaneously, a threshold for the number of iterations is set to avoid infinite iterations affecting system operating efficiency, thus ensuring that the actual load rate of all working power modules is adjusted to within the preset optimal efficiency range. This refined power allocation strategy effectively avoids efficiency degradation caused by excessively high or low loads on power modules, improving the overall operating efficiency of the entire server power system and reducing energy consumption. Furthermore, by balancing the load of each power module, it also helps extend the lifespan of the power modules and improves the reliability and stability of the system.

[0052] In some embodiments described above in this application, a power module scheduling scheme is proposed based on load forecast data, power module operating status data, and specification parameter data, combined with power scheduling strategies and redundancy configuration strategies, to achieve intelligent management of power modules. However, when the server load remains at a low level for an extended period, even if power modules are configured with the minimum number of modules in operation, these modules may still operate at a low load rate, leading to decreased power conversion efficiency and unnecessary energy consumption.

[0053] In response, this application further proposes that when the total power demand is lower than a preset low load threshold, the power module scheduling scheme includes marking at least one power module currently in operation as a module to be put into sleep mode, sending a sleep command to the module to be put into sleep mode to make it enter a low-power sleep mode; wherein, the power module in sleep mode maintains communication connectivity, is able to receive a wake-up command and return to the operating state; when the subsequently generated load prediction data indicates that the total power demand has risen back above the preset low load threshold, a wake-up command is sent to the power module in sleep mode to make it return to the operating state.

[0054] In this embodiment, the preset low-load threshold is a pre-set power value used to determine whether the overall load of the server is in a low-power state. When the total power demand is lower than this threshold, the system considers it appropriate to enter energy-saving mode. This threshold can be set according to the server's hardware configuration, the efficiency curve of the power module, and actual operating experience. For example, it can be set as a percentage of the total rated power (such as 10% or 15%), or a fixed power value. Its function is to serve as a condition for triggering the power module's sleep mechanism. When the total power demand is lower than the preset low-load threshold, the system needs to identify which power modules can enter sleep mode. Specifically, based on the priority ranking of power modules (such as cumulative runtime data, temperature data, etc.), modules with lower current load rates or the least impact on the overall system can be selected and marked as modules to be sleep-inducing. Marking a module as a module to be sleep-inducing means that these modules will no longer bear the load and are ready to enter a low-power state. The sleep command is a specific command issued by the control system to the marked power module, instructing it to enter energy-saving mode. Low-power sleep mode refers to a state in which the power module significantly reduces its own power consumption while maintaining basic functions (such as communication). This can be achieved by shutting down some unnecessary internal circuits, reducing the internal clock frequency, and stopping some power conversion stages. In this state, the power module no longer supplies power to the load but can still be managed by the system. Maintaining communication connectivity while in sleep mode is a key characteristic of sleep mode, ensuring that the system management module can monitor the status of the sleep module at any time and send a wake-up command when needed. Upon receiving the wake-up command, the sleep module will reactivate its internal circuits, gradually returning to normal operating status and preparing to handle the load. This recovery process requires a certain startup time, but is usually much faster than a complete cold start. When subsequent load forecast data indicates that the total power demand has risen above the preset low load threshold, the system will send a wake-up command to the power modules that previously entered sleep mode to ensure power supply capacity and redundancy, bringing them back into operation. This ensures that the system can respond flexibly to load changes, saving energy while ensuring power supply reliability.

[0055] In this embodiment, through the above technical solution, this application can intelligently identify scenarios where the server is operating under low load. When the total power demand is lower than a preset low load threshold, the system can proactively put some power modules into a low-power sleep state, thereby avoiding energy loss caused by these modules operating in low-efficiency ranges and improving the overall operating efficiency of the power system. Simultaneously, the sleep modules maintain communication connectivity and can be quickly woken up and resume operation when the load recovers, ensuring the continuity and reliability of power supply and avoiding the problem of sacrificing system stability for energy saving. Furthermore, reducing unnecessary operating time also helps extend the lifespan of the power modules.

[0056] In some embodiments described above, a method is proposed to optimize energy utilization efficiency by intelligently scheduling power modules to operate within a preset optimal efficiency range. However, during prolonged operation or under high load, the internal temperature of the power module may continuously rise. If heat dissipation management is not carried out in a timely and effective manner, it may not only affect the stability and lifespan of the power module, but may also lead to system overheating protection or failure, thereby reducing the reliability of the entire server system.

[0057] To address this, this application further proposes a method for intelligent thermal management of power modules after they are scheduled and running. This method first acquires temperature data from each power module and ambient temperature data from inside the server. Specifically, the system uses temperature sensors integrated within each power module to collect real-time temperature data from key internal points, such as power devices and transformers. Simultaneously, ambient temperature sensors are deployed inside the server to monitor the overall ambient temperature within the chassis. This sensor data can be transmitted to the main control unit for processing via an internal communication bus.

[0058] In this embodiment, the temperature data of each power module is compared with preset graded temperature thresholds to determine the current temperature level of each power module. The preset graded temperature thresholds include at least a first temperature threshold and a second temperature threshold, with the first temperature threshold being lower than the second temperature threshold. These thresholds are preset based on the power module's design specifications, reliability requirements, and actual operating experience. For example, the first temperature threshold can be set as the temperature point at which the power module begins to require enhanced heat dissipation, while the second temperature threshold is set as the temperature point at which even stronger heat dissipation is needed to avoid overheating. Through this graded comparison, continuous temperature data can be discretized into manageable temperature levels, providing a basis for refined heat dissipation control.

[0059] In this embodiment, the system determines a corresponding cooling fan speed adjustment strategy based on the current temperature level of each power module. Specifically, when the current temperature level exceeds a first temperature threshold but does not exceed a second temperature threshold, the system will increase the cooling fan speed at a first rate; while when the current temperature level exceeds the second temperature threshold, the system will increase the cooling fan speed at a second rate, wherein the second rate is greater than the first rate. This tiered adjustment strategy can be pre-stored in the control module, forming a lookup table or logical judgment rule, so that the cooling fan speed can be dynamically adjusted according to the actual heat load of the power module, achieving on-demand cooling. Finally, according to the cooling fan speed adjustment strategy, the system sends a speed adjustment command to the cooling fan control module. The main control unit generates a corresponding control signal according to the determined speed adjustment strategy and sends it to the cooling fan control module through the communication interface to drive the fan motor to run at the target speed.

[0060] In this embodiment, after the power modules adjust their output power according to the scheduling scheme and operate within their optimal efficiency range, the system can continuously monitor the temperature data of each power module and the internal ambient temperature of the server. By comparing the temperature data of the power modules with preset graded temperature thresholds, the system can accurately determine their current thermal load status and dynamically adjust the speed of the cooling fans accordingly. For example, when the temperature rises moderately, the fan increases its speed at a first rate for initial heat dissipation; when the temperature rises further, it increases its speed at a faster second rate to ensure timely and effective heat dissipation. This graded and dynamic heat dissipation strategy avoids noise and energy waste caused by excessive fan operation under low thermal loads, while responding quickly under high thermal loads, effectively preventing the power modules from overheating. This ensures the long-term stable operation and reliability of the power modules, extends their service life, and further improves the energy efficiency of the entire server system.

[0061] In some of the embodiments described above in this application, continuous monitoring of the power module's operating status is proposed, and a protection mechanism is triggered and automatic switching to a backup power module is executed upon detection of a fault. However, in actual operation, how to accurately identify the fault type, safely isolate the faulty module, and ensure that the backup module can seamlessly take over the load of the faulty module to avoid impacting the stable operation of the server remains a problem that needs to be refined and resolved. If the fault handling process is not refined enough, it may lead to switching delays or uneven load distribution, affecting system reliability.

[0062] To address this, this application further proposes a method for continuously monitoring the operational status data of each power module, comparing each parameter in the operational status data with a preset safety threshold, and determining that the power module is faulty when any parameter exceeds the corresponding preset safety threshold, triggering a protection mechanism and generating a fault alarm message. Simultaneously, the application includes the following steps for automatically switching to a backup power module according to a redundancy configuration strategy: In this embodiment, the voltage data of each power module is compared with preset overvoltage and undervoltage thresholds, the current data is compared with a preset overcurrent threshold, and the temperature data is compared with a preset overtemperature threshold to obtain detailed comparison results. This comparison process aims to accurately determine the operational health status of the power modules through multi-dimensional and refined parameter monitoring. Voltage, current, and temperature data are core indicators reflecting the operating status of the power modules, while the preset overvoltage, undervoltage, overcurrent, and overtemperature thresholds are safety boundaries set based on the power module's design specifications, safety standards, and long-term operating experience. By comparing these real-time collected data with the thresholds one by one, it is possible to quickly and accurately identify whether there are any abnormalities in the power modules, such as excessively high or low voltage, excessive current leading to overload, or excessive temperature potentially causing hardware damage. This comprehensive comparison of multiple parameters can fully cover various fault modes that may occur in the power modules, providing a reliable basis for subsequent fault determination.

[0063] In this embodiment, when the comparison result indicates that any parameter of any power module exceeds the corresponding preset safety threshold, the system will determine that the power module is a faulty power module and generate fault identification data containing the faulty power module identifier, fault parameter type, and fault parameter value. After obtaining the comparison result, once any key operating parameter of any power module is found to exceed its corresponding preset safety threshold, the power module is immediately determined to be a faulty power module. To facilitate subsequent fault handling, alarms, and maintenance, the system will immediately generate detailed fault identification data. This data not only contains the unique identification information of the faulty power module to accurately locate which module has a problem, but also clearly indicates the specific parameter type causing the fault and the actual value of that parameter. This detailed fault identification data is of great significance for quickly diagnosing the cause of the fault, guiding maintenance personnel to troubleshoot, and optimizing system operation strategies.

[0064] In this embodiment, based on the fault identification data, an output cutoff operation is performed on the faulty power module. The output power of the faulty power module before the cutoff is recorded as the power value to be switched. Simultaneously, a target module for switching is selected from the power modules in standby mode. Once the faulty power module is identified, the system immediately takes decisive measures based on the fault identification data. First, an output cutoff operation is performed on the faulty power module to isolate it from the power supply network, preventing the fault from further escalating or affecting other normally operating modules, thus ensuring the safety of the server system. Before the cutoff, the system accurately records the real-time output power of the faulty power module before the fault occurs and it is cut off; this value is defined as the power value to be switched. This is to ensure that during the subsequent switching process, the standby power module can accurately take over this part of the load, avoiding insufficient power supply to the server due to power shortages. At the same time, the system intelligently selects one or more of the most suitable power modules from the pool of power modules currently in standby mode as target modules for switching, based on preset strategies (e.g., health status, available capacity, uptime, etc.), preparing for the subsequent load transfer.

[0065] In this embodiment, a start command is finally sent to the target switching module to adjust its output power from its initial value to the power value to be switched, completing the automatic switching of the backup power module. After selecting the target switching module, the system immediately sends a start command to it, enabling it to quickly transition from standby to working state. Subsequently, the system precisely adjusts the output power of the target switching module according to the previously recorded power value to be switched. This means that the target switching module will not blindly output at maximum power, but will gradually or rapidly increase its output power from its initial value to the same power level as before the faulty module was disconnected, based on actual needs. Through this precise power adjustment, a smooth load transition can be ensured, avoiding the impact of power fluctuations on the server system, thereby achieving automatic and seamless switching of the backup power module and maximizing the continuous and stable operation of the server.

[0066] In this embodiment, by meticulously comparing multi-dimensional operating parameters such as voltage, current, and temperature with preset safety thresholds, the fault type and location of the power module can be quickly and accurately identified, avoiding fuzzy judgments and delayed processing. Secondly, after identifying the faulty module, an output cutoff operation is immediately executed, effectively isolating the fault source, preventing the fault from spreading, and ensuring the overall safety of the server system. More importantly, while cutting off the faulty module, its output power before cutoff is accurately recorded as the power value to be switched, and a backup module is intelligently selected. A start command is then sent to the backup module, and its output power is precisely adjusted to the power value to be switched. This series of meticulous operations ensures a seamless and smooth load transfer, avoiding power interruptions or power fluctuations caused by fault switching, thereby greatly improving the reliability and availability of the server power system and ensuring the continuity of server services. Compared to basic solutions that only perform fault alarms and simple switching, this solution, through meticulous management of fault identification, isolation, and load transfer, effectively solves the impact and instability problems that may occur during fault switching, enabling the power system to more intelligently and efficiently recover from module failures.

[0067] The aforementioned method effectively manages modular power supplies for servers, including continuous monitoring of the power module's operating status and triggering protection mechanisms and automatic switching to a backup power module upon detecting a fault. However, after a fault occurs and the backup power module is switched over, some residual issues may remain. For example, fault alarm information may not be cleared in a timely manner, or the newly inserted power module may not be effectively identified, verified, and managed by the system after replacing the faulty module. This could lead to unclear system status, affecting subsequent maintenance operations and the effectiveness of the overall redundancy strategy.

[0068] In response, this application further proposes that after the automatic switching of the backup power module is completed, the method further includes: after detecting that the faulty power module has been physically removed from the server, clearing the fault alarm information corresponding to the faulty power module; after detecting that a new power module has been inserted into the server, performing initialization detection on the newly inserted power module, the initialization detection including voltage calibration detection, current calibration detection and communication connectivity detection; when the initialization detection passes, configuring the newly inserted power module as a standby state and including the newly inserted power module in the scope of the redundancy configuration strategy.

[0069] In this embodiment, after detecting that a faulty power module has been physically removed from the server, the corresponding fault alarm information is cleared to ensure the accuracy of the system status and the validity of the alarm information. Once a faulty power module is physically removed, its corresponding fault alarm information loses its practical meaning; if not cleared in time, it may cause misjudgment or interference by maintenance personnel. Detecting whether a faulty power module has been physically removed can be achieved in various ways. For example, the server's power backplane integrates a hot-swap detection mechanism, which determines the module's insertion or removal status by monitoring changes in the level of specific pins on the power module slot. Alternatively, the management controller (BMC) can periodically poll the online status of each power module through communication interfaces such as I2C or PMBus. When a module is unresponsive for an extended period or its physical presence signal disappears, it can be determined that it has been removed. Once module removal is confirmed, the system management software accesses the database or log system storing fault alarm information, locates and deletes the alarm records related to the removed module, and updates the alarm display on the user interface.

[0070] In this embodiment, after a new power module is detected being inserted into the server, an initialization test is performed on the newly inserted power module. This initialization test includes voltage calibration detection, current calibration detection, and communication connectivity detection, aiming to ensure that the newly inserted power module has normal operating capability and reliability, avoiding the introduction of new potential fault points. The insertion detection mechanism for the new power module is similar to the removal detection, and can rely on the hot-swap detection signal of the power backplane. Once a new module is detected, the system immediately initiates the initialization detection process. Specifically, voltage calibration detection involves the system sending a command to the new module to output a specific voltage, and measuring the actual output voltage using internal sensors or external test points, comparing it with a preset standard value to verify its voltage output accuracy. Current calibration detection involves the system applying a simulated load or internal test circuit to make the new module output a specific current, measuring the actual output current, and comparing it with a preset standard value to verify the accuracy of its current measurement or output control. Communication connectivity detection involves the system attempting to establish a communication connection with the new module (e.g., via the PMBus protocol), sending a series of test commands (such as reading module information, setting parameters, etc.), and verifying whether the module's response is correct and timely to ensure that its communication interface function is normal.

[0071] In this embodiment, after the initialization test passes, the newly inserted power module is configured as a standby module and included in the redundancy configuration strategy. This ensures that the new module can be immediately put into use to maintain the system's redundancy capability and is correctly considered in subsequent power scheduling. When all the initialization tests pass successfully, it indicates that the newly inserted power module is functional and communicating normally. At this time, the system management program updates the module's status to "standby," meaning it is ready to take on the load or replace a faulty module when needed. Simultaneously, the system includes it in the scope of a preset redundancy configuration strategy (e.g., N+1 or N+N redundancy strategy). This means that in subsequent power scheduling decisions, the standby module will be considered an available redundant resource. When a working module fails or the total power demand increases, the system can activate and put it into operation according to the redundancy strategy.

[0072] In this embodiment, through the above technical solution, this application further improves the lifecycle management of power modules after completing the automatic switching of backup power modules. First, fault alarm information corresponding to physically removed faulty power modules is promptly cleared, avoiding interference from invalid alarms to maintenance personnel and improving the clarity and management efficiency of the system status. Second, a comprehensive initialization test is performed on the newly inserted power module, including voltage calibration test, current calibration test, and communication connectivity test, ensuring the reliability and normal function of the new module and effectively avoiding system instability caused by problems with the new module itself. Finally, the newly tested module is configured as a standby module and included in the redundancy configuration strategy, enabling the system to quickly restore and maintain its original redundancy capabilities, enhancing the overall reliability and maintainability of the server power system, and ensuring continuous stable operation of the system after module replacement.

[0073] In some embodiments described above, this application proposes an intelligent management method for modular server power supplies. This method achieves intelligent management of server power supplies through steps such as data collection, load prediction, power scheduling, and monitoring of operating status. However, during its implementation, if the power modules are not adequately initialized and configured, subsequent intelligent management operations may be based on inaccurate or incomplete information, thereby affecting the stability and reliability of the system, and may even lead to safety risks due to unknown module states or undefined protection parameters.

[0074] To address this, this application further proposes that, before the steps of collecting server load data, operating status data of each power module, and specification parameter data of each power module, and preprocessing the server load data to obtain preprocessed load data, the method further includes: performing a power-on self-test on each power module to confirm the communication connectivity and normal function of each power module; acquiring the specification parameter data of each power module, the specification parameter data including rated voltage range and rated current range; determining the initial role allocation of each power module based on the specification parameter data of each power module and the redundancy configuration strategy, the initial role allocation dividing each power module into working modules and standby modules; configuring corresponding preset safety thresholds for each power module, the preset safety thresholds including overvoltage threshold, undervoltage threshold, overcurrent threshold, and overtemperature threshold.

[0075] In this embodiment, each power module undergoes a power-on self-test to confirm its communication connectivity and normal function, aiming to ensure that each power module is in a healthy and controllable state before being put into use. Communication connectivity testing can be achieved by the main control unit sending query commands to each power module and receiving responses, verifying the integrity of the data link and the uniqueness of the module address. Normal function testing can include internal self-diagnostic programs within the modules, such as checking the internal voltage rails, clock signal, memory integrity, and whether the readings of key sensors (such as temperature sensors) are within reasonable ranges. It can even perform brief no-load output tests to verify the stability of the output voltage.

[0076] In this embodiment, the specification parameter data of each power module is acquired, including the rated voltage range and rated current range, to comprehensively understand the inherent electrical characteristics of each power module. These parameters can be stored in the non-volatile memory (such as EEPROM) inside the power module, and can be read by the main control unit through a standard communication protocol (such as PMBus). In addition to the rated power value, the rated voltage range and rated current range provide more refined operating boundary information, which is crucial for the subsequent setting of power allocation, fault detection, and protection mechanisms.

[0077] In this embodiment, the initial role allocation of each power module is determined based on the specification parameters and redundancy configuration strategy. This initial role allocation divides each power module into working modules and standby modules to establish a clear operating architecture at the beginning of system startup. The redundancy configuration strategy can be an N+1 redundancy strategy or an N+N redundancy strategy. The system calculates the required number of working modules and standby modules based on the total power demand, the rated power of each module, and the selected redundancy strategy. The initial role allocation can be based on various factors, such as prioritizing modules with shorter cumulative operating time as working modules to balance wear and tear, or allocating modules based on their health status assessment results to ensure that critical loads are powered by the most reliable modules.

[0078] In this embodiment, each power module is configured with corresponding preset safety thresholds, including overvoltage, undervoltage, overcurrent, and overtemperature thresholds, to establish the operational safety boundaries of the power modules. These thresholds can be calculated and set based on the rated voltage range, rated current range, and manufacturer-recommended safety operating specifications of the power modules. For example, the overvoltage threshold can be set to 105% of the rated voltage, the undervoltage threshold to 95% of the rated voltage, the overcurrent threshold to 120% of the rated current, and the overtemperature threshold based on the module's heat dissipation capacity and the upper limit of the ambient temperature. These thresholds serve as the benchmark for subsequent continuous monitoring and fault diagnosis, ensuring that the power modules operate within safe limits.

[0079] In this embodiment, through the above technical solution, the system can perform comprehensive initialization and configuration of all power modules before the intelligent management method of the server modular power supply is officially launched. This ensures that each power module has undergone a health check, its key operating parameters and safety boundaries are clearly defined, and a clear initial role assignment is established at the system level. This provides a solid and reliable foundation for subsequent intelligent management steps such as load data collection, load prediction, power scheduling, and fault monitoring. Specifically, through power-on self-test, faulty modules can be avoided from being included in the management scope, ensuring the accuracy of data collection; by obtaining more comprehensive specification parameters, subsequent power allocation and scheduling decisions are more accurate; through initial role assignment, the foundation for the effective implementation of redundancy strategies is laid, improving system availability; and by configuring preset safety thresholds, subsequent operational status monitoring and fault judgment can trigger protection mechanisms in a timely and accurate manner, thereby improving the overall stability, security, and operating efficiency of the server modular power supply system.

[0080] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the intelligent management method of the modular power supply of the server in this application. Any simple modifications based on this technical concept are within the protection scope of this application.

[0081] This application also provides an intelligent management system for a server modular power supply, see reference. Figure 2 The intelligent management system for the server modular power supply includes: a memory 10, a processor 20, and an intelligent management program for the server modular power supply stored on the memory 10 and executable on the processor 20. The intelligent management program for the server modular power supply is configured to implement the steps of the intelligent management method for the server modular power supply.

[0082] The intelligent management system for modular server power supplies provided in this application, employing the intelligent management method for modular server power supplies in the above embodiments, can improve the fault tolerance and energy efficiency of modular server power supplies. Compared with the prior art, the beneficial effects of the intelligent management system for modular server power supplies provided in this application are the same as those of the intelligent management method for modular server power supplies provided in the above embodiments, and other technical features in the intelligent management system for modular server power supplies are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0083] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0084] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. All equivalent structural transformations made under the technical concept of this application using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included within the scope of patent protection of this application.

Claims

1. A method for intelligent management of modular power supplies for servers, characterized in that, The method includes: The system collects server load data, operating status data of each power module, and specification parameter data of each power module. It then preprocesses the server load data to obtain preprocessed load data. The server load data includes processor utilization data, memory utilization data, and network traffic data. The operating status data includes voltage data, current data, temperature data, current output power data, and cumulative runtime data for each power module. The specification parameter data includes the rated power value of each power module. The preprocessed load data is matched with the historical load data for the corresponding time period in the historical load data. Based on the matching result, the direction and magnitude of load change in the future preset time period are determined, and load prediction data is generated. Based on the load forecast data, the operating status data of each power module, and the specification parameter data, combined with the preset power scheduling strategy and redundancy configuration strategy, a power module scheduling scheme is determined; the power module scheduling scheme includes the identification information of the power module to be adjusted, the adjustment type, and the target output power of each power module after adjustment. According to the power module scheduling scheme, control commands are sent to the corresponding power modules to adjust the output power and operating status of each power module, so that each power module operates within the preset optimal efficiency range. The operating status data of each power module is continuously monitored, and each parameter in the operating status data is compared with a preset safety threshold. When any parameter exceeds the corresponding preset safety threshold, the power module is determined to be a faulty power module, triggering the protection mechanism and generating a fault alarm message. At the same time, the automatic switching of the backup power module is performed according to the redundancy configuration strategy.

2. The intelligent management method for modular power supplies of servers as described in claim 1, characterized in that, The steps of performing trend matching between the preprocessed load data and historical load data for the corresponding time period in the historical load data, and determining the direction and magnitude of load change within a future preset time period based on the matching results to generate load prediction data include: Load peak features, load average features, and load fluctuation features are extracted from the preprocessed load data, and the load peak features, load average features, and load fluctuation features are combined into load feature data. Historical load peak features, historical load average features, and historical load fluctuation features for the corresponding time period are extracted from the historical load data and combined into historical load features. The load feature data is then trend-matched with the historical load features, and the similarity between the load feature data and the historical load features is calculated. The subsequent change trend associated with the historical load feature corresponding to the highest similarity is selected as a reference trend. Based on the degree of deviation between the load feature data and the reference trend, the direction and magnitude of load change within a preset time period are determined, and the load prediction data is generated.

3. The intelligent management method for modular power supplies of servers as described in claim 1, characterized in that, Based on the load prediction data, the operating status data of each power module, and the specification parameter data, combined with the preset power scheduling strategy and redundancy configuration strategy, a power module scheduling scheme is determined. The power module scheduling scheme includes the identification information of the power modules to be adjusted, the adjustment type, and the target output power of each power module after adjustment. The steps include: The total power demand value for a future preset period is calculated based on the load forecast data, and the total power demand value is compared with the rated power value of each power module to determine the minimum number of working power modules required to meet the total power demand value. The number of redundant power modules is determined based on the redundancy configuration strategy, and the minimum number is added to the number of redundant power modules to obtain the target number of working modules; wherein, the redundancy configuration strategy is an N+1 redundancy strategy or an N+N redundancy strategy. The current output power data, temperature data, and cumulative running time data of each power module in the operation status data are comprehensively analyzed to determine the priority ranking of each power module, and the power modules are selected as working power modules in descending order of priority. Based on the total power demand and the target number of working modules, the total power is allocated to the selected working power modules, and the target output power of each working power module is determined. The power module scheduling scheme is generated by summarizing the information. The power module scheduling scheme includes: the identifier of the power module to be started, the identifier of the power module to be shut down, and the target output power value of each working power module.

4. The intelligent management method for modular power supplies of servers as described in claim 3, characterized in that, Based on the total power demand and the target number of operating modules, the steps of allocating the total power to the selected operating power modules and determining the target output power of each operating power module include: Obtain the current output power data and rated power value of each working power module, and calculate the current load rate of each working power module; The total power demand is initially allocated according to the ratio of the rated power value of each working power module to obtain the initial allocated power of each working power module. For a power supply module whose initial power allocation exceeds the upper limit of the preset optimal efficiency range, the excess power is transferred to other power supply modules whose current load rate is lower than the lower limit of the preset optimal efficiency range. After iterative adjustment, the actual load rate of each power supply module is within the preset optimal efficiency range. Alternatively, when the number of iterative adjustments reaches the preset maximum iteration threshold, the adjustment is stopped and the current power allocation is used as the target output power to obtain the target output power of each power supply module.

5. The intelligent management method for modular power supplies of servers as described in claim 3, characterized in that, When the total power demand is lower than a preset low load threshold, the power module scheduling scheme further includes: At least one power module currently in operation is marked as a module to be hibernated, and a hibernation command is sent to the module to be hibernated to make it enter a low-power hibernation state; wherein, the power module in hibernation state maintains communication connectivity and is able to receive a wake-up command and return to the operating state; When the subsequently generated load prediction data indicates that the total power demand value rises above the preset low load threshold, a wake-up command is sent to the power module in sleep mode to restore it to working state.

6. The intelligent management method for modular power supplies of servers as described in claim 1, characterized in that, After the step of sending control commands to the corresponding power modules according to the power module scheduling scheme to adjust the output power and operating status of each power module so that each power module operates within a preset optimal efficiency range, the method further includes: Obtain the temperature data of each power module and the ambient temperature data inside the server; The temperature data of each power module is compared with a preset graded temperature threshold to determine the current temperature level of each power module; the preset graded temperature threshold includes at least a first temperature threshold and a second temperature threshold, and the first temperature threshold is lower than the second temperature threshold. Based on the current temperature level of each power module, a corresponding cooling fan speed adjustment strategy is determined; wherein, when the current temperature level exceeds the first temperature threshold but does not exceed the second temperature threshold, the cooling fan speed is increased at a first rate; when the current temperature level exceeds the second temperature threshold, the cooling fan speed is increased at a second rate, wherein the second rate is greater than the first rate. According to the cooling fan speed adjustment strategy, a speed adjustment command is sent to the cooling fan control module.

7. The intelligent management method for modular power supplies of servers as described in claim 1, characterized in that, The operation status data of each power module is continuously monitored, and each parameter in the operation status data is compared with a preset safety threshold. When any parameter exceeds the corresponding preset safety threshold, the power module is determined to be a faulty power module, triggering the protection mechanism and generating a fault alarm message. At the same time, the automatic switching of the backup power module is performed according to the redundancy configuration strategy, including the following steps: The voltage data of each power module is compared with the preset overvoltage threshold and the preset undervoltage threshold, the current data is compared with the preset overcurrent threshold, and the temperature data is compared with the preset overtemperature threshold to obtain the comparison results. When the comparison result indicates that any parameter of any power module exceeds the corresponding preset safety threshold, the power module is determined to be a faulty power module, and fault identification data containing the faulty power module identifier, fault parameter type and fault parameter value is generated. Based on the fault identification data, an output cut-off operation is performed on the faulty power module, and the output power of the faulty power module before the cut-off is recorded as the power value to be switched. At the same time, a target module for switching is selected from the power modules in standby state. A start command is sent to the target switching module to adjust the output power of the target switching module from the initial value to the power value to be switched, thereby completing the automatic switching of the backup power module.

8. The intelligent management method for modular power supplies of servers as described in claim 7, characterized in that, After the automatic switching of the backup power module is completed, the method further includes: After detecting that the faulty power module has been physically removed from the server, the fault alarm information corresponding to the faulty power module is cleared. After a new power module is detected being inserted into the server, an initialization test is performed on the newly inserted power module. The initialization test includes voltage calibration test, current calibration test, and communication connectivity test. Once the initialization test passes, the newly inserted power module is configured as a standby module and brought under the jurisdiction of the redundancy configuration strategy.

9. The intelligent management method for modular power supplies of servers as described in claim 1, characterized in that, Before the steps of collecting server load data, operating status data of each power module, and specification parameter data of each power module, and preprocessing the server load data to obtain preprocessed load data, the method further includes: Perform a power-on self-test on each power module to confirm that the communication connectivity and functions of each power module are normal. Obtain the specification parameter data of each power module, which also includes the rated voltage range and rated current range; Based on the specification parameters and redundancy configuration strategy of each power module, the initial role assignment of each power module is determined, and the initial role assignment divides each power module into working modules and standby modules. Each power module is configured with a corresponding preset safety threshold, which includes an overvoltage threshold, an undervoltage threshold, an overcurrent threshold, and an overtemperature threshold.

10. An intelligent management system for a server modular power supply, characterized in that, The intelligent management system for the server modular power supply includes: a memory, a processor, and an intelligent management program for the server modular power supply stored in the memory and executable on the processor. The intelligent management program for the server modular power supply is configured to implement the steps of the intelligent management method for the server modular power supply as described in any one of claims 1 to 9.

Images