Server fan card heat dissipation power regulation and control system in load environment

Abstract

The invention discloses a server fan card heat dissipation power regulation and control system in a load environment, and relates to the technical field of automatic control. The system comprises an information acquisition module used for acquiring historical temperature data, historical load data and a heat dissipation power curve of a target server; the demand degree prediction module is used for predicting a first heat dissipation demand degree of the target server at each sampling moment in the target future time period based on the historical temperature data and the historical load data; the interval division module is used for dividing the target future time period into a plurality of target time intervals based on the first heat dissipation demand degree of each sampling moment in the target future time period; the strategy determination module is used for determining a heat dissipation strategy of the target time interval based on each first heat dissipation demand degree in the target time interval, the fluctuation characteristic of the first heat dissipation demand degree and the heat dissipation power curve; and the instruction generation module is used for generating a fan card regulation and control instruction for the target server based on the heat dissipation strategy of each target time interval.

Description

Technical Field

[0001] The present invention relates to the field of automatic control technology, and in particular to a heat dissipation power control system for a server fan card under a load environment. Background Art

[0002] As server hardware performance continues to improve, computing power is significantly enhanced, but power consumption is also rising significantly. Given the high-density architecture of servers, heat dissipation becomes increasingly critical. If heat dissipation measures are inadequate, hardware may be damaged by overheating, and in severe cases, even fail. Given that server loads vary, intelligent control of the server's internal cooling system is essential to precisely match the cooling requirements under varying load conditions.

[0003] At present, the conventional practice for server cooling regulation is to dynamically adjust the power of the fan card in the cooling system based on the real-time load changes and temperature fluctuations of the server.

[0004] However, during actual server operation, if the load fluctuates frequently, this regulation method will cause the fan card power to be constantly adjusted, thereby increasing the overall energy consumption of the cooling system. Summary of the Invention

[0005] The embodiment of the present invention provides a server fan card heat dissipation power control system under a load environment, which can reduce unnecessary power adjustments of the heat dissipation system and effectively reduce the overall energy consumption of the heat dissipation system.

[0006] The present invention provides a server fan card heat dissipation power control system under a load environment, the system comprising: An information acquisition module is used to obtain historical temperature data, historical load data, and a heat dissipation power curve of the target server. The heat dissipation power curve is used to represent the heat dissipation power of the fan card corresponding to the target server being able to recover to the standard temperature under different temperature conditions. A demand prediction module is used to predict the first heat dissipation demand of the target server at each sampling time in a target future time period based on historical temperature data and historical load data; An interval division module is configured to divide the target future time period into a plurality of target time intervals based on the first heat dissipation demand degrees at each sampling moment in the target future time period, wherein the first heat dissipation demand degrees within the target time intervals are within the same demand degree range; a strategy determination module, configured to determine, for each target time interval, a heat dissipation strategy for the target time interval based on each first heat dissipation requirement within the target time interval, a fluctuation characteristic of the first heat dissipation requirement, and a heat dissipation power curve; The instruction generation module is used to generate a fan card control instruction for the target server based on the heat dissipation strategy of each target time interval, so as to control the heat dissipation power of the fan card of the target server.

[0007] Furthermore, the present invention also proposes that the demand prediction module includes the following units: a demand calculation unit, configured to determine a second heat dissipation demand at each historical sampling moment based on each historical temperature data and each historical load data; a heat dissipation demand ordering unit, configured to order the second heat dissipation demand orders in chronological order to obtain a heat dissipation demand order sequence; The demand prediction unit is used to predict the first heat dissipation demand of the target server at each sampling moment in the target future time period based on the heat dissipation demand sequence through an autoregressive integral moving average model.

[0008] Furthermore, the present invention also proposes a demand calculation unit, which is used to: Obtain target historical temperature data and first historical load data at a target historical sampling moment, as well as second historical load data at a sampling moment before the target historical sampling moment, where the target historical sampling moment is any historical sampling moment; A second heat dissipation requirement at the target historical sampling moment is determined by using the ratio of the first historical load data to the second historical load data and the target historical temperature data.

[0009] Furthermore, the present invention also proposes that the interval division module includes the following units: a time period determination unit, configured to determine a target future time period as a time period to be divided; A result determination unit is used to determine multiple candidate division results by taking any sampling moment within the time period to be divided as a division boundary; a distinguishability determining unit, configured to determine, for each candidate partitioning result, the distinguishability of the heat dissipation requirements of the two candidate time intervals based on the first heat dissipation requirements of the two candidate time intervals in the candidate partitioning result; The result determination unit is further configured to determine the candidate partition result with the smallest cumulative value of the heat dissipation demand discrimination between the two candidate time intervals as the target partition result; The loop execution unit is used to update the time period to be divided into each target time interval in the target division result, and return to the loop execution to use any sampling moment in the time period to be divided as the division boundary, determine multiple candidate division results, until the recursive stop condition is met, and obtain multiple target time intervals.

[0010] Furthermore, the present invention also proposes a distinguishability determination unit, configured to: Obtaining each first heat dissipation demand degree in the candidate time interval; Collecting statistics on each first heat dissipation demand in the candidate time interval to obtain a heat dissipation demand statistical result for the candidate time interval, where the heat dissipation demand statistical result includes at least one of a peak heat dissipation demand, a valley heat dissipation demand, an average heat dissipation demand, a maximum heat dissipation demand, and a minimum heat dissipation demand; The heat dissipation demand statistical results are used to determine the heat dissipation demand differentiation of the candidate time intervals.

[0011] Furthermore, the present invention also proposes that the strategy determination module includes the following units: a demand analysis unit, configured to determine an average heat dissipation demand based on each first heat dissipation demand within a target time interval, and determine a fluctuation influence coefficient based on a fluctuation characteristic of the first heat dissipation demand; a heat dissipation demand correction unit, configured to correct the average heat dissipation demand by using the fluctuation influence coefficient to obtain a first interval heat dissipation demand in a target time interval; The curve matching unit is used to match the heat dissipation demand in the first interval with the heat dissipation power curve to determine the heat dissipation strategy in the target time interval.

[0012] Furthermore, the present invention also proposes a demand analysis unit for: Obtaining a peak heat dissipation demand and a valley heat dissipation demand in each first heat dissipation demand; Based on each peak heat dissipation demand and each valley heat dissipation demand, determine the time interval between adjacent peak heat dissipation demand, the average peak heat dissipation demand, and the average valley heat dissipation demand; The average peak heat dissipation demand is subtracted from the average valley heat dissipation demand to determine the average peak-valley difference; The fluctuation impact coefficient of the target time interval is determined by using the differences between each time interval and the average peak-to-valley value.

[0013] Furthermore, the present invention also proposes that after the average heat dissipation requirement is corrected using the fluctuation influence coefficient to obtain the heat dissipation requirement for the first interval of the target time interval, the strategy determination module further includes: A difference acquisition unit, configured to acquire the difference in heat dissipation requirements between adjacent target time intervals; An interval merging unit, configured to merge adjacent target time intervals to obtain a merged time interval when the difference in heat dissipation requirements is less than or equal to a preset difference threshold; The demand analysis unit is further configured to determine a second interval heat dissipation demand of the combined time interval based on the two first interval heat dissipation demand corresponding to the combined time interval; The curve matching unit is used to match the heat dissipation demand in the second interval with the heat dissipation power curve to determine the heat dissipation strategy for the combined time interval.

[0014] Furthermore, the present invention also proposes that, for each target time interval, after determining the heat dissipation strategy for the target time interval based on each first heat dissipation requirement within the target time interval, the fluctuation characteristics of the first heat dissipation requirement, and the heat dissipation power curve, the system further includes: A power value interpolation module is used to interpolate the transition heat dissipation power value between two adjacent target time intervals by linear interpolation; The instruction generation module is used to generate a fan card control instruction for the target server based on the heat dissipation strategy of each target time interval and each transition heat dissipation power value.

[0015] Furthermore, the present invention also proposes that the heat dissipation strategy includes the heat dissipation power of the fan card in the target time interval; Instruction generation module, used to: Match the fan card heat dissipation power of the target time interval with the fan speed curve to obtain the fan speed of the target time interval. The fan speed curve is used to represent the fan speed corresponding to different fan card heat dissipation powers of the target server; The fan speeds in each target time interval are converted into fan card control instructions for the target server.

[0016] The present invention has the following beneficial effects: In a server fan card heat dissipation power control system under a load environment, an embodiment of the present invention provides a method for first obtaining historical temperature data, historical load data, and a heat dissipation power curve for a target server. Based on the historical temperature and load data, a first heat dissipation demand at each sampling time within a target future time period is predicted. The future time period is then divided into multiple target time intervals based on the first heat dissipation demand, ensuring that the first heat dissipation demand within each interval is within the same range, thereby avoiding frequent power adjustments due to frequent load fluctuations. A heat dissipation strategy is then determined for each target time interval by comprehensively considering the first heat dissipation demand, fluctuation characteristics, and heat dissipation power curve within the interval, rather than simply making immediate adjustments based on real-time load and temperature. Finally, control instructions for the fan card are generated based on the heat dissipation strategy for each target time interval, thereby controlling the fan card's heat dissipation power. This method of determining a heat dissipation strategy based on historical data prediction, time interval division, and comprehensive consideration of multiple factors avoids constant adjustment of the fan card power due to frequent load fluctuations, thereby reducing unnecessary power adjustments in the heat dissipation system and effectively lowering the overall energy consumption of the heat dissipation system. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0018] Figure 1 A schematic diagram of the structure of a server fan card heat dissipation power control system under a load environment provided by one embodiment of the present invention; Figure 2 A schematic diagram of the structure of a demand prediction module provided by one embodiment of the present invention; Figure 3 A schematic diagram of the structure of an interval division module provided by an embodiment of the present invention; Figure 4 A schematic diagram of the structure of a policy determination module provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0019] To further illustrate the technical means and effectiveness of the present invention in achieving its intended objectives, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effectiveness of a server fan card heat dissipation power control system under load conditions proposed by the present invention. In the following description, references to "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0020] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0021] It should be noted that the acquisition, storage, use, and processing of data in the technical solution of the present invention comply with the relevant provisions of laws and regulations.

[0022] It should be noted that in the embodiments of the present invention, certain software, components, models and other existing solutions in the industry may be mentioned. They should be regarded as exemplary and their purpose is only to illustrate the feasibility of implementing the technical solution of the present invention, but it does not mean that the applicant has or will necessarily use the solution.

[0023] In traditional high-density server cooling systems, dynamic fan card power adjustment relies on a feedback mechanism based on real-time load and temperature data. When server loads fluctuate dramatically over short periods, this feedback mechanism generates high-frequency power adjustment commands, causing the fan cards to operate in an unstable mode for extended periods. This adjustment mode, when faced with sudden load changes, not only generates a large amount of ineffective power switching but also causes a misalignment between cooling power and load demand, causing power adjustments to lag behind actual cooling demand changes.

[0024] When faced with the above problems, the present application first recognized the essential contradiction of the traditional method that relies on real-time feedback to cause frequent power adjustments. The root cause is that it only focuses on the immediate status and lacks forward-looking judgment on the load change trend. In this regard, the present application attempts to approach from the perspective of time series prediction, and by analyzing the correlation characteristics between historical temperature and load, establishes a prediction model for heat dissipation demand, so as to predict the changes in heat dissipation demand in future time periods. It was further discovered that simply predicting the heat dissipation demand at a single point in time will still lead to strategy fragmentation. For this reason, a time interval division mechanism is introduced to merge time periods with similar demand characteristics, and combine the static characteristics of the heat dissipation power curve with the dynamic influence of the fluctuation characteristics to finally form an interval-level control strategy that takes into account both stability and adaptability.

[0025] In this regard, Figure 1 As shown, the present invention provides a schematic diagram of a system for regulating the heat dissipation power of a server fan card under a load environment. The system 100 comprises an information acquisition module 110, a demand prediction module 120, an interval division module 130, a strategy determination module 140, and an instruction generation module 150.

[0026] An information acquisition module 110 is configured to acquire historical temperature data, historical load data, and a heat dissipation power curve of a target server, wherein the heat dissipation power curve is used to represent the heat dissipation power of a fan card corresponding to the target server being able to recover to a standard temperature under different temperature conditions; The demand prediction module 120 is configured to predict the first heat dissipation demand of the target server at each sampling time within a target future time period based on the historical temperature data and the historical load data; An interval division module 130 is configured to divide the target future time period into a plurality of target time intervals based on the first heat dissipation demand degrees at each sampling moment in the target future time period, wherein each first heat dissipation demand degree in the target time interval is within the same demand degree range; A strategy determination module 140 is configured to determine, for each target time interval, a heat dissipation strategy for the target time interval based on each first heat dissipation requirement within the target time interval, a fluctuation characteristic of the first heat dissipation requirement, and a heat dissipation power curve; The instruction generation module 150 is configured to generate a fan card control instruction for a target server based on the heat dissipation strategy of each target time interval, so as to control the heat dissipation power of the fan card of the target server.

[0027] In this embodiment, the server fan card is a hardware component used to improve server cooling. It typically integrates multiple sensors and control functions, intelligently adjusting fan speed, temperature monitoring, and other cooling measures to ensure stable server operation and prevent overheating under high loads. Multiple temperature sensors are integrated within the card to monitor the temperatures of key server components in real time. This temperature data can be used to determine fan speed, activate cooling systems, and more.

[0028] Historical temperature data refers to the temperature information recorded during server operation. This data can be collected periodically through temperature sensors and used to reflect the temperature trends of the server at different points in time. Historical load data refers to the resource usage of the server when processing tasks. This data can be collected through a monitoring system to collect the utilization of the central processing unit, memory, and hard disk, and used to analyze the correlation between load and temperature. The heat dissipation power curve is a mapping of the fan card power required to restore the server to the standard temperature under different temperature conditions. This curve can be generated through experimental testing or thermodynamic model calculations and is used to establish a quantitative correspondence between temperature and heat dissipation power.

[0029] The first heat dissipation requirement refers to a quantitative indicator for predicting the heat dissipation capacity required by the server in the future time period. It can be achieved by performing time series analysis on historical data through an autoregressive model or a machine learning algorithm to predict the changing trend of heat dissipation demand in advance.

[0030] The demand range refers to a standard for classifying heat dissipation demand into numerical intervals. Specifically, this can be achieved by setting a threshold or dynamically dividing by a clustering algorithm, and is used to group time periods with similar heat dissipation demand into the same interval.

[0031] The fluctuation characteristics refer to the amplitude and frequency of changes in the first heat dissipation demand in the time series, which can be achieved by calculating the variance, range or Fourier transform analysis to evaluate the stability of the heat dissipation demand.

[0032] A cooling strategy is a fan card power control scheme tailored to a specific time interval. This strategy maps cooling requirements to a cooling power curve using a lookup table or interpolation method, allowing for dynamic adjustment of cooling system operating parameters.

[0033] The core innovation of this application is to generate a phased cooling strategy by predicting future cooling needs and dividing the dynamic time intervals, combining the cooling power curve with fluctuation characteristic analysis, thereby reducing the repeated adjustment of cooling power due to frequent load fluctuations and achieving optimal control of cooling energy consumption.

[0034] As an example, the information acquisition module 110 first acquires temperature data and load rate data of the target server sampled every 5 minutes over the past 30 days as historical temperature data and historical load data. It also acquires a heat dissipation power curve corresponding to the target server, which reflects the fan card heat dissipation power required to cool the server to a standard temperature of 38°C at different temperatures.

[0035] Next, the demand prediction module 120 uses an autoregressive integrated moving average model to predict the first heat dissipation demand every 5 minutes within the next 24 hours based on historical data.

[0036] The interval division module 130 further divides the next 24 hours into a plurality of target time intervals based on the first heat dissipation demand every 5 minutes in the next 24 hours, so that the fluctuation of the first heat dissipation demand in each target time interval does not exceed 10%.

[0037] Furthermore, the strategy determination module 140 calculates the average heat dissipation demand for each target time interval and determines a fluctuation impact coefficient based on the amplitude and frequency of demand fluctuations within the interval. The average heat dissipation demand is multiplied by the fluctuation impact coefficient to obtain a corrected interval heat dissipation demand. The interval heat dissipation demand is then matched with the heat dissipation power curve to obtain the target heat dissipation power for each target time interval.

[0038] Finally, the instruction generation module 150 generates a corresponding fan card control instruction sequence based on the target heat dissipation power of each target time interval, and executes the corresponding instruction at the beginning of each time interval to achieve segmented control of the fan card heat dissipation power.

[0039] Through this embodiment, the target server's historical temperature data, historical load data, and heat dissipation power curve are first obtained. Based on the historical temperature data and historical load data, the first heat dissipation demand at each sampling moment in the target future time period is predicted. The future time period is then divided into multiple target time intervals based on the first heat dissipation demand, ensuring that the first heat dissipation demand within the same interval is within the same range, thereby avoiding frequent power adjustments due to frequent load fluctuations. Next, for each target time interval, a heat dissipation strategy is determined by comprehensively considering the first heat dissipation demand, fluctuation characteristics, and heat dissipation power curve within the interval, rather than simply making immediate adjustments based on real-time load and temperature. Finally, based on the heat dissipation strategy for each target time interval, control instructions are generated for the fan card to regulate the fan card's heat dissipation power. This method of determining a heat dissipation strategy based on historical data prediction, dividing time intervals, and comprehensively considering multiple factors avoids the need for constant adjustment of the fan card's power due to frequent load fluctuations, thereby reducing unnecessary power adjustments in the heat dissipation system and effectively reducing the overall energy consumption of the heat dissipation system.

[0040] In some of the above-mentioned schemes of the present application, it is proposed to predict the first heat dissipation requirement of the target server at each sampling moment in the future time period based on historical temperature data and historical load data. However, when predicting the future heat dissipation requirement directly based on historical data, the time series characteristics of the historical data may not be fully exploited, resulting in insufficient prediction accuracy, thereby affecting the accuracy of subsequent time interval division and heat dissipation strategy.

[0041] In this regard, Figure 2 As shown, the present application further proposes a demand prediction module 120, which includes the following units: The demand calculation unit 121 is used to determine the second heat dissipation demand at each historical sampling moment based on each historical temperature data and each historical load data; The heat dissipation demand ordering unit 122 is configured to order the second heat dissipation demand orders in chronological order to obtain a heat dissipation demand order sequence; The demand prediction unit 123 is configured to predict the first heat dissipation demand of the target server at each sampling moment in a target future time period based on the heat dissipation demand sequence by using an autoregressive integral moving average model.

[0042] In this embodiment, the second heat dissipation requirement can be calculated by multiplying the historical temperature data by the historical load data to reflect the actual heat dissipation demand at the historical moment; the heat dissipation requirement sequence is formed into time series data through time sorting to provide input for subsequent models; the autoregressive integral moving average model improves the prediction accuracy by capturing the trend and periodic characteristics in the time series.

[0043] Specifically, when calculating the second cooling requirement, the product of historical temperature data and historical load data at each historical sampling moment forms an indicator reflecting the instantaneous cooling requirement. After arranging multiple second cooling requirements in chronological order, the resulting time series data can be analyzed using an autoregressive integrated moving average model. The autoregressive term captures internal correlations in the series, differencing eliminates non-stationarity, and the moving average term fits random noise, thereby accurately predicting the first cooling requirement at each future sampling moment.

[0044] As an example, the demand calculation unit 121 determines the second heat dissipation demand at each historical sampling moment based on the historical temperature data and the historical load data. For example, a temperature sensor may collect historical temperature data from the server every 10 minutes over the past week, while also recording the historical load data at the corresponding moment. For each historical sampling moment, the corresponding second heat dissipation demand is determined based on the product of the historical temperature data and the historical load data.

[0045] Then, the demand ordering unit 122 orders the second heat dissipation demand degrees in chronological order to obtain a heat dissipation demand degree sequence. Specifically, the second heat dissipation demand degrees of 1008 sampling points in a week can be arranged in chronological order to form a time sequence of length 1008.

[0046] Finally, based on the cooling requirement sequence, the cooling requirement prediction unit 123 uses an autoregressive integrated moving average model to predict the first cooling requirement of the target server at each sampling time within the target future time period. Specifically, the model parameters can be fitted using the past 1008 data points. The cooling requirement is then predicted every 10 minutes for the next 24 hours, resulting in 144 predicted values ​​as the first cooling requirement.

[0047] This embodiment enables accurate prediction of future server cooling requirements based on historical data, providing a reliable basis for subsequent cooling strategy development. This allows for precise regulation of server cooling power, avoiding energy waste caused by frequent adjustments. Furthermore, this method fully leverages the time series characteristics inherent in historical data, capturing the cyclical and trending nature of cooling requirements through an autoregressive integrated moving average model, thereby improving the accuracy and reliability of the predictions.

[0048] In some of the above-mentioned schemes of the present application, in the process of predicting future heat dissipation requirements through historical temperature data and historical load data, if the heat dissipation requirements are calculated only based on the load data at a single time point, it is easy to ignore the impact of the load change rate at adjacent moments on the heat dissipation requirements, resulting in insufficient accuracy in the calculation of the heat dissipation requirements.

[0049] In this regard, the present application further proposes a demand calculation unit 121, which is specifically used to: Obtain target historical temperature data and first historical load data at a target historical sampling moment, as well as second historical load data at a sampling moment before the target historical sampling moment, where the target historical sampling moment is any historical sampling moment; A second heat dissipation requirement at the target historical sampling moment is determined by using the ratio of the first historical load data to the second historical load data and the target historical temperature data.

[0050] In this embodiment, the first and second historical load data correspond to the server load status at two adjacent sampling moments, respectively. A ratio calculation quantifies the dynamic trend of load changes. The target historical temperature data reflects the cooling demand base at the target historical sampling moment. Combined with the load change rate, this data can form a multi-dimensional cooling demand assessment.

[0051] Specifically, at the target historical sampling time, temperature data T and load data L1 are collected, while load data L2 from the previous sampling time is extracted. The ratio of L1 to L2 is used as the load dynamic factor, which represents the rate of load change. The temperature data T is multiplied by the load dynamic factor to obtain the second heat dissipation demand at the target historical sampling time.

[0052] As an example, the second heat dissipation requirement at the target historical sampling moment can be determined by the following formula 1: Formula 1 In formula 1, Used to characterize the second heat dissipation requirement at sampling time i, Used to represent the temperature data at sampling time i, Used to characterize the load data at sampling time i, Used to represent the load data at sampling time i-1.

[0053] in, Represents the change of load at sampling time i. The larger the value, the greater the relative load of server i at sampling time i, and the greater the heat generated. Combined with the temperature at sampling time i , that is, the greater the heat generation at sampling time i and the higher the temperature accumulation, the greater the corresponding heat dissipation demand.

[0054] It should be noted that in order to ensure that the calculation results are meaningful, when performing fractional operations in the embodiments of the present invention, when encountering a situation where the denominator is 0, it is necessary to add a parameter adjustment factor greater than 0 to the denominator to prevent the denominator from being 0. The value of the parameter adjustment factor is set by the implementer according to actual conditions, and this application does not impose any special restrictions.

[0055] This embodiment accurately calculates the second cooling requirement at the historical sampling moment. By taking into account load variations between adjacent moments, this method can more accurately reflect the actual cooling requirements of the server under varying load and temperature conditions. This calculation method facilitates more accurate prediction of cooling requirements for future time periods, providing a reliable basis for developing appropriate cooling strategies. Furthermore, by accurately understanding cooling requirements, over- or under-cooling can be avoided, effectively improving server cooling efficiency, extending hardware life, and reducing energy consumption.

[0056] In some of the above-mentioned schemes of the present application, when the target future time period is divided into multiple target time intervals, if the division method fails to fully consider the inherent consistency of the heat dissipation demand within the interval, it may lead to large differences in heat dissipation demand between intervals, thereby causing frequent heat dissipation power adjustments and increasing system energy consumption.

[0057] In this regard, Figure 3 As shown, the present application further proposes an interval division module 130, which includes the following units: The time period determination unit 131 is configured to determine the target future time period as the time period to be divided; A result determination unit 132 is configured to determine multiple candidate division results using any sampling time within the time period to be divided as a division boundary; a distinguishability determining unit 133 for determining, for each candidate partitioning result, the distinguishability of the heat dissipation requirements of the two candidate time intervals based on the first heat dissipation requirements of the two candidate time intervals in the candidate partitioning result; The result determination unit 134 is further configured to determine the candidate partitioning result with the smallest cumulative value of the heat dissipation demand discrimination between the two candidate time intervals as the target partitioning result; The loop execution unit 135 is used to update the time period to be divided into each target time interval in the target division result, and return to the loop execution to use any sampling time in the time period to be divided as the division boundary to determine multiple candidate division results until the recursive stop condition is met to obtain multiple target time intervals.

[0058] In this embodiment, candidate partition results are generated by traversing all possible sampling moments within the time period to be partitioned as partition boundaries. Each candidate partition result divides the time period into two consecutive subintervals. For example, if the sampling moments are 1 second, 3 seconds, 5 seconds, 7 seconds, 9 seconds, 11 seconds, and 13 seconds, and the 7th second is used as the partition boundary, then the 1st to 5th seconds can be divided into one subinterval, and the 9th to 13th seconds can be divided into another subinterval; alternatively, the 1st to 7th seconds can be divided into one subinterval, and the 7th to 13th seconds can be divided into another subinterval.

[0059] Heat dissipation discrimination is calculated by counting the heat dissipation statistics for candidate time intervals, such as peak heat dissipation, valley heat dissipation, or average heat dissipation. The lower the heat dissipation discrimination, the greater the correlation between the heat dissipation demands within the candidate time intervals. The candidate partition with the smallest cumulative heat dissipation discrimination value is selected as the target partition, ensuring minimal differences in the heat dissipation distribution between the two sub-intervals. The recursive stopping condition is achieved by setting a minimum time interval length, a threshold for the number of partitions, or a heat dissipation discrimination threshold to avoid over-partitioning.

[0060] Specifically, the time period to be divided is recursively split into smaller subintervals. Each time, the impact of all possible division methods on the consistency of cooling requirements within the interval is calculated. By summing the heat requirement differentiation of the two subintervals, the division method with the highest correlation is selected, minimizing the fluctuation range of the cooling requirement within each subinterval. When the recursive stopping condition is triggered, the resulting target time interval has a high degree of cohesion in terms of cooling requirement, thereby reducing the frequency of switching cooling strategies between different intervals.

[0061] For example, if a candidate partitioning result divides a time period into the first half and the second half, and the peak-to-valley difference in the heat dissipation demand in the first half is 15 and the peak-to-valley difference in the second half is 10, the cumulative value is 25. Another candidate partitioning result has two sub-intervals with peak-to-valley differences of 12 and 8, respectively, with a cumulative value of 20. This latter sub-interval is selected as the target partitioning result. By recursively executing this process, the granularity of the time interval divisions is gradually refined until the preset stopping condition is met.

[0062] As an example, time period determination unit 131 determines the target future time period as the time period to be divided. Result determination unit 132 uses any sampling time within the time period to be divided as a division boundary and determines multiple candidate division results. For example, each sampling time within the time period to be divided can be selected as a division boundary to obtain multiple candidate division results.

[0063] Then, for each candidate partition result, the distinguishability determining unit 133 determines the distinguishability of the heat dissipation requirements of the two candidate time intervals based on the first heat dissipation requirements of the two candidate time intervals in the candidate partition result. Specifically, the non-correlation coefficient of the first heat dissipation requirements in the two candidate time intervals can be calculated as a measure of the heat dissipation requirement distinguishability.

[0064] Then, the result determination unit 134 determines the candidate partitioning result with the smallest cumulative value of the heat dissipation requirement differentiation between the two candidate time intervals as the target partitioning result, so that the heat dissipation requirement in the divided time intervals has a high similarity.

[0065] Finally, the loop execution unit 135 updates the time period to be divided into target time intervals in the target division result and returns to the loop execution, using any sampling time within the time period to be divided as the division boundary, to determine multiple candidate division results until the recursive stopping condition is met, thereby obtaining multiple target time intervals. The recursive stopping condition can be set to the maximum cumulative value of the heat dissipation requirement discrimination being less than or equal to the heat dissipation requirement discrimination threshold.

[0066] This embodiment allows the target future time period to be divided into multiple target time intervals with similar cooling requirements. This allows for the development of appropriate cooling strategies for each target time interval, avoiding frequent adjustments to fan card power and reducing cooling system energy consumption. Furthermore, this method, through recursive partitioning, can adaptively generate reasonable time interval divisions, improving the flexibility and accuracy of cooling control.

[0067] In some of the above-mentioned schemes of the present application, when dividing the candidate time intervals and calculating their distinctiveness to determine the optimal division result, if only relying on a single statistical indicator to evaluate the heat dissipation demand distribution characteristics, it may lead to deviations in the correlation calculation results, thereby affecting the rationality of the time interval division.

[0068] In this regard, the present application further proposes a distinguishability determination unit 133, which is specifically configured to: Obtaining each first heat dissipation demand degree in the candidate time interval; Collecting statistics on each first heat dissipation demand in the candidate time interval to obtain a heat dissipation demand statistical result for the candidate time interval, where the heat dissipation demand statistical result includes at least one of a peak heat dissipation demand, a valley heat dissipation demand, an average heat dissipation demand, a maximum heat dissipation demand, and a minimum heat dissipation demand; The heat dissipation demand statistical results are used to determine the heat dissipation demand differentiation of the candidate time intervals.

[0069] In this embodiment, the peak heat dissipation demand is used to reflect the instantaneous maximum intensity of the heat dissipation demand within the interval, the valley heat dissipation demand represents the instantaneous minimum intensity of the heat dissipation demand, the average heat dissipation demand is used to quantify the overall demand level of the interval, and the maximum heat dissipation demand and the minimum heat dissipation demand further expand the distribution range of the demand amplitude within the interval.

[0070] By combining multiple statistical indicators, a multi-dimensional heat dissipation demand distribution feature vector can be constructed. For example, when calculating the distinctiveness of a candidate time interval, the peak heat dissipation demand, valley heat dissipation demand, average heat dissipation demand, maximum heat dissipation demand, and minimum heat dissipation demand can be extracted and mapped into a five-dimensional vector for quantitative evaluation using Euclidean distance or cosine similarity algorithms.

[0071] Specifically, when dividing the time intervals, the first cooling demand data for all sampling moments within the candidate time interval is first extracted. Extreme values ​​are identified within this data set to determine the peak and valley cooling demand values, and the arithmetic mean is calculated as the average cooling demand value. Furthermore, the maximum and minimum cooling demand values ​​are determined by traversing the data set, generating a statistical result encompassing five dimensions. This statistical result allows for a comprehensive assessment of the cooling demand fluctuation amplitude, central trend, and dispersion within the candidate time interval.

[0072] As an example, the heat dissipation requirement differentiation of the candidate time interval can be determined by the following formula 2: Formula 2 In formula 2, P is used to characterize the differentiation of heat dissipation requirements. Used to characterize the peak heat dissipation demand corresponding to the k-th fluctuation peak, It is used to represent the valley heat dissipation demand corresponding to the k-th fluctuation peak, and n is used to represent the number of fluctuation peaks in the candidate time interval. Used to characterize the maximum heat dissipation requirement, Used to characterize the minimum heat dissipation requirement, Used to represent the average cooling requirement.

[0073] in, It represents the degree of fluctuation of the heat dissipation demand. The smaller the value, the more similar the peak-to-valley differences of all fluctuation peaks are. This means that the fluctuation degrees of the heat dissipation demand at different times are more similar and the correlation is greater. It represents the difference in the level of heat dissipation demand. The smaller the value, the smaller the difference in the level of heat dissipation demand at different times. The flatter the overall value, the greater the correlation.

[0074] This embodiment conducts a comprehensive statistical analysis of the heat dissipation demand data within the candidate time intervals, generating multi-dimensional statistical indicators. These indicators comprehensively reflect the overall distribution characteristics and changing patterns of heat dissipation demand, helping to more accurately assess the rationality of candidate partitioning schemes. Quantitative analysis of the discriminability of heat dissipation demand provides an objective basis for selecting the optimal partitioning scheme, thereby achieving more refined and efficient time interval partitioning.

[0075] In some of the above-mentioned solutions of this application, if the heat dissipation strategy is determined only based on the average heat dissipation demand, the fluctuation characteristics of the heat dissipation demand are not fully considered, resulting in insufficient matching between the heat dissipation strategy and actual demand, affecting the heat dissipation efficiency or increasing energy consumption.

[0076] In this regard, Figure 4 As shown, the present application further proposes a strategy determination module 140, comprising the following units: a demand analysis unit 141 configured to determine an average heat dissipation demand based on the first heat dissipation demand values ​​within a target time interval, and to determine a fluctuation influence coefficient based on a fluctuation characteristic of the first heat dissipation demand values; The heat dissipation demand correction unit 142 is configured to correct the average heat dissipation demand by using the fluctuation influence coefficient to obtain a first interval heat dissipation demand in a target time interval; The curve matching unit 143 is configured to match the heat dissipation demand in the first interval with the heat dissipation power curve to determine a heat dissipation strategy for the target time interval.

[0077] In this embodiment, the average heat dissipation requirement is obtained by calculating the arithmetic mean of the first heat dissipation requirement at all sampling moments within the target time interval; the fluctuation influence coefficient is generated by extracting the peak and valley characteristics of the first heat dissipation requirement and performing statistical analysis, which may specifically include calculating the peak and valley value differences and the fluctuation frequency; the correction process uses a weighted superposition method to linearly or nonlinearly combine the fluctuation influence coefficient and the average heat dissipation requirement; the heat dissipation power curve is stored in the form of discrete data points or continuous functions, and the heat dissipation power value corresponding to the heat dissipation requirement in the first interval is matched by an interpolation or fitting algorithm.

[0078] Specifically, the average of each first heat dissipation demand degree within the target time interval is calculated to generate an average heat dissipation demand degree as the basic heat dissipation demand indicator; the extreme value detection is simultaneously performed on the heat dissipation demand degree sequence corresponding to the target time interval, the peak and valley value data are extracted, and the peak and valley value difference and the fluctuation frequency are calculated to generate a fluctuation influence coefficient reflecting the dynamic change of heat dissipation demand; the fluctuation influence coefficient is then superimposed with the average heat dissipation demand degree according to a preset weight ratio to generate the first interval heat dissipation demand degree of the target time interval that comprehensively considers the steady-state demand and dynamic fluctuations; finally, the first interval heat dissipation demand degree of the target time interval is compared with the heat dissipation power curve, and the heat dissipation power value corresponding to the first interval heat dissipation demand degree is selected as the heat dissipation strategy for the target time interval.

[0079] As an example, the demand analysis unit 141 determines a heat dissipation strategy for a target time interval based on each first heat dissipation demand degree within the target time interval, the fluctuation characteristics of the first heat dissipation demand degrees, and the heat dissipation power curve. Specifically, the average heat dissipation demand degree is first determined based on each first heat dissipation demand degree within the target time interval, and the fluctuation impact coefficient is determined based on the fluctuation characteristics of the first heat dissipation demand degrees. The average heat dissipation demand degree can be obtained by calculating the arithmetic mean of all first heat dissipation demand degrees within the target time interval. The fluctuation impact coefficient can be determined by analyzing the amplitude and frequency of changes in the first heat dissipation demand degrees within the target time interval. For example, a statistical indicator such as standard deviation or coefficient of variation can be used to quantify the fluctuation impact coefficient.

[0080] Furthermore, the demand correction unit 142 corrects the average heat dissipation demand using the fluctuation influence coefficient to obtain a first interval heat dissipation demand of the target time interval. Specifically, the first interval heat dissipation demand can be determined by the following formula 3: Formula 3 In formula 3, Used to characterize the heat dissipation demand of the first interval of the zth target time interval, Used to represent the average heat dissipation demand in the zth target time interval, Used to characterize the fluctuation impact coefficient of the z-th target time interval.

[0081] Among them, the average heat dissipation demand is weightedly adjusted with the fluctuation influence coefficient as the weight, and the heat dissipation demand of the first interval of the target time interval can be obtained. Then, its size not only represents the average level of the entire target time interval, but also combines the influence of the fluctuation peak, thereby improving its extensiveness.

[0082] Finally, the curve matching unit 143 matches the heat dissipation requirement in the first interval with the heat dissipation power curve to determine the heat dissipation strategy for the target time interval. The matching process can be achieved by finding the point on the heat dissipation power curve that is closest to the heat dissipation requirement in the first interval.

[0083] This embodiment enables the development of targeted cooling strategies based on the server's cooling requirements during different time periods. This approach not only considers the average cooling requirement but also the fluctuating characteristics of cooling requirements, making the cooling strategy more accurate and efficient. By carefully analyzing and rationally modifying cooling requirements, it is possible to avoid issues such as insufficient or excessive cooling, effectively balancing the server's cooling performance and energy consumption. Furthermore, because cooling strategies are developed for specific time periods, they can reduce energy waste caused by frequent fan adjustments, improving the stability and reliability of the overall cooling system.

[0084] In some of the above-mentioned schemes of the present application, when the heat dissipation strategy is determined by the average heat dissipation demand and the fluctuation influence coefficient, the calculation of the fluctuation influence coefficient does not take into account the time distribution characteristics of the peak heat dissipation demand and the valley heat dissipation demand, resulting in insufficient matching accuracy between the corrected first-interval heat dissipation demand and the heat dissipation power curve, affecting the control effect of the heat dissipation strategy.

[0085] In this regard, the present application further proposes a demand analysis unit for: Obtaining a peak heat dissipation demand and a valley heat dissipation demand in each first heat dissipation demand; Based on each peak heat dissipation demand and each valley heat dissipation demand, determine the time interval between adjacent peak heat dissipation demand, the average peak heat dissipation demand, and the average valley heat dissipation demand; The average peak heat dissipation demand is subtracted from the average valley heat dissipation demand to determine the average peak-valley difference; The fluctuation impact coefficient of the target time interval is determined by using the differences between each time interval and the average peak-to-valley value.

[0086] In this embodiment, the peak heat dissipation demand can be extracted using a sliding window method to find local maxima, and the valley heat dissipation demand can be extracted using a sliding window method to find local minima. The time interval between adjacent peaks can be calculated by timestamp difference. The average peak heat dissipation demand can be calculated using the arithmetic mean, and the average valley heat dissipation demand can be calculated using the arithmetic mean. The average peak-valley difference is calculated using the absolute value of the difference between the average peak heat dissipation demand and the average valley heat dissipation demand.

[0087] As an example, the fluctuation impact coefficient can be determined by the following formula 4: Formula 4 In formula 4, Used to characterize the fluctuation impact coefficient of the z-th target time interval, Used to represent the time corresponding to the k-th peak heat dissipation demand, Used to represent the time corresponding to the k+1th peak heat dissipation demand, A quantity used to represent the peak cooling demand. Used to characterize the average peak-to-valley value difference of the zth target time interval, Used to characterize the average peak-to-valley value difference of other target time intervals except the z-th target time interval.

[0088] in, Represents the frequency of fluctuation peaks in the zth target time interval. If the peak moments of any adjacent fluctuation peaks are closer, The smaller it is, the closer the fluctuation peaks are and the more frequent the fluctuations are. It represents the degree of fluctuation of the fluctuation peak in the z-th target time interval (i.e., the peak-to-valley difference). The smaller the value, the greater the degree of fluctuation of the fluctuation peak in the z-th target time interval.

[0089] This embodiment accurately calculates the fluctuating characteristics of cooling demand within the target time interval, providing an important basis for determining the subsequent cooling strategy. This allows for more precise control of fan card cooling power, avoiding increased cooling system energy consumption due to frequent load fluctuations. Furthermore, by considering the differences between peak and valley values ​​and the fluctuation period, this solution comprehensively reflects the dynamic changes in cooling demand, facilitating the development of more appropriate cooling strategies.

[0090] In some of the above-mentioned solutions of the present application, in the process of determining the cooling strategy based on the cooling demand in the first interval, if the difference in cooling demand between adjacent target time intervals is small, the cooling strategy may be frequently adjusted, thereby increasing the overall energy consumption of the cooling system.

[0091] In this regard, the present application further proposes to use the fluctuation influence coefficient to correct the average heat dissipation requirement. After obtaining the first interval heat dissipation requirement of the target time interval, the strategy determination module 140 further includes: A difference acquisition unit, configured to acquire the difference in heat dissipation requirements between adjacent target time intervals; An interval merging unit, configured to merge adjacent target time intervals to obtain a merged time interval when the difference in heat dissipation requirements is less than or equal to a preset difference threshold; The demand analysis unit 141 is further configured to determine a second interval heat dissipation demand of the combined time interval based on the two first interval heat dissipation demand corresponding to the combined time interval; The curve matching unit 143 is specifically configured to: The heat dissipation demand in the second interval is matched with the heat dissipation power curve to determine the heat dissipation strategy for the combined time interval.

[0092] In this embodiment, the difference in heat dissipation requirements can be obtained by calculating the difference in the first-interval heat dissipation requirements of two adjacent target time intervals. The difference can be calculated based on an absolute value or a relative ratio. The preset difference threshold is derived from historical data statistics, and the specific numerical range can be set to a relative difference ratio of 5% to 15%. The second-interval heat dissipation requirements of the combined time interval can be generated using a weighted average method, with weights allocated based on the duration ratio of the adjacent target time intervals. The heat dissipation power curve matching process compares the heat dissipation requirements of the second interval with a pre-established heat dissipation power curve and selects the corresponding heat dissipation power value as the heat dissipation strategy output.

[0093] Specifically, when the difference in the heat dissipation requirements of two adjacent target time intervals does not exceed the preset difference threshold, the merging operation integrates the originally separate control periods into a continuous period. The heat dissipation requirement of the second interval generated by the weighted average method takes into account the heat dissipation demand levels of adjacent time periods, and generates a continuous and stable heat dissipation strategy through the matching process of the heat dissipation power curve. This process effectively reduces the number of heat dissipation power adjustments and reduces the energy consumption of the cooling system while ensuring the heat dissipation effect. For example, when the heat dissipation requirements of two adjacent 10-minute periods are 85 and 89 respectively, if the preset difference threshold is 10%, the system will be merged into a 20-minute control period and 87 will be used as the unified heat dissipation requirement. The fan card power only needs to be adjusted once instead of twice.

[0094] As an example, after the target server is divided into multiple target time intervals in the target future time period, the first interval heat dissipation requirements corresponding to two adjacent target time intervals are 75 and 78, respectively. The difference acquisition unit calculates the difference in heat dissipation requirements between the two, and using the cosine similarity algorithm, the difference in heat dissipation requirements is 0.98, with a preset difference threshold of 0.99. Because the difference in heat dissipation requirements does not exceed the preset difference threshold, the time interval merging mechanism is triggered, and the interval merging unit merges the two adjacent target time intervals to form a merged time interval.

[0095] The demand analysis unit 141 further calculates the second-interval heat demand for the combined time interval using a weighted average algorithm. The first time interval includes data from four sampling moments, with a weight of 0.6; the second time interval includes data from two sampling moments, with a weight of 0.4. The resulting second-interval heat demand is 76.2. Finally, the curve matching unit matches the second-interval heat demand with the heat dissipation power curve to determine the heat dissipation strategy for the combined time interval.

[0096] As another example, three consecutive target time intervals may be directly merged. Specifically, the difference in heat dissipation requirements between the three consecutive target time intervals is first determined using the following formula 5: Formula 5 In formula 5, It is used to characterize the difference in heat dissipation demand between the zth target time interval and the two target time intervals before and after it. Used to characterize the heat dissipation demand of the first interval of the zth target time interval, It is used to characterize the heat dissipation demand of the first interval of the z-1th target time interval. Used to represent the first interval heat dissipation demand of the z+1th target time interval.

[0097] When the difference in heat dissipation demand between the zth target time interval and the two previous and next target time intervals is less than or equal to the preset difference threshold, the z-1th target time interval, the zth target time interval and the z+1th target time interval can be directly merged into a merged time interval.

[0098] Through this embodiment, the frequent switching of strategies caused by slight differences in heat dissipation requirements between adjacent time intervals is effectively reduced. By merging time intervals with similar heat dissipation requirements, the frequency of fan card power adjustment is significantly reduced. While maintaining the heat dissipation efficiency of the server, the energy consumption of the heat dissipation system is optimized and controlled, solving the energy waste problem caused by repeated adjustment of fan power due to load fluctuations in the prior art.

[0099] In this embodiment, the heat dissipation strategies between adjacent target time intervals may vary greatly. If the heat dissipation power suddenly changes significantly during actual operation, the stability of the system may be reduced, and the normal operation of the internal components of the server may also be affected.

[0100] In this regard, the present application further proposes that, for each target time interval, after determining the heat dissipation strategy for the target time interval based on each first heat dissipation requirement within the target time interval, the fluctuation characteristics of the first heat dissipation requirement, and the heat dissipation power curve, the system further includes: A power value interpolation module is used to interpolate the transition heat dissipation power value between two adjacent target time intervals by linear interpolation; The instruction generation module 150 is specifically configured to: Based on the heat dissipation strategy of each target time interval and each transition heat dissipation power value, a fan card control instruction for the target server is generated.

[0101] In this embodiment, the transition heat dissipation power value can be specifically determined by the following formula 6: Formula 6 In formula 6, It is used to characterize the transition heat dissipation power value at time t between two adjacent target time intervals. Used to characterize the heat dissipation power of the zth target time interval, Used to represent the heat dissipation power in the z+1th target time interval. Used to represent the end time of the zth target time interval, Used to represent the starting time of the z+1th target time interval.

[0102] Among them, with The closer the time At this moment, the corresponding heat dissipation power is Gradually Gradual change ensures smoother connection of heat dissipation power between different target time intervals, avoids temperature fluctuations caused by sudden changes, and improves system stability and efficiency.

[0103] As an example, the server's heat dissipation power in the previous time interval is 1200W, and in the next time interval it is 800W. The power value interpolation module divides the time window between the two intervals into three consecutive time points using linear interpolation, and calculates the three transitional heat dissipation power values ​​of 1100W, 1000W, and 900W using the above formula 5. When generating fan card control instructions, the execution sequence of the heat dissipation strategy is constructed as a stepped power curve of 1200W, 1100W, 1000W, 900W, and 800W, where the transitional heat dissipation power values ​​are linearly interpolated between adjacent intervals.

[0104] This embodiment effectively avoids sudden changes in cooling power during target time intervals, ensuring a smooth transition in fan speed adjustment. This reduces the instantaneous current surge in the cooling system caused by sudden power changes, reduces stress on the fan's mechanical structure, and avoids temperature rebound caused by sudden changes in cooling power, thereby improving the stability and energy efficiency of the cooling control process.

[0105] In some of the above-mentioned solutions of this application, when the heat dissipation strategy directly generates fan card control instructions, the correspondence between heat dissipation power and fan speed is not considered, resulting in the adjustment of heat dissipation power being unable to accurately correspond to the actual changes in fan speed, thereby affecting the heat dissipation efficiency and increasing system energy consumption.

[0106] In this regard, the present application further proposes a heat dissipation strategy including fan card heat dissipation power in a target time interval; The instruction generation module 150 is specifically configured to: Match the fan card heat dissipation power of the target time interval with the fan speed curve to obtain the fan speed of the target time interval. The fan speed curve is used to represent the fan speed corresponding to different fan card heat dissipation powers of the target server; The fan speeds in each target time interval are converted into fan card control instructions for the target server.

[0107] In this embodiment, a fan speed curve is generated through experimental calibration or fitting of historical operating data, and includes a one-to-one correspondence between heat dissipation power and fan speed. The matching process is implemented using a lookup table or interpolation algorithm. For example, the closest heat dissipation power point on the curve is matched based on the heat dissipation power value, and the corresponding fan speed is extracted. If the heat dissipation power does not directly match, the target speed is calculated through linear interpolation of adjacent heat dissipation power values. When converting the fan speed to a control command, a pulse width modulation signal or digital control command is used, and the command includes a target fan speed value or a speed percentage parameter.

[0108] Specifically, when generating fan card control instructions, a mapping relationship between heat dissipation power and fan speed is established through a predefined fan speed curve, so that the heat dissipation power determined in each target time interval can be directly mapped to the corresponding fan speed parameter. For example, when the heat dissipation strategy determines that a heat dissipation power of 500 watts needs to be maintained in a certain time interval, by querying the speed curve, it is found that the fan speed corresponding to the heat dissipation power is 2500 rpm, and then a control instruction containing the speed value is generated. This process avoids the frequent fluctuations in speed caused by relying solely on real-time load adjustment in traditional methods, reduces the number of adjustments through the deterministic association between heat dissipation power and fan speed, and reduces the dynamic energy consumption of the cooling system. Furthermore, the fan speed curve can be dynamically updated according to the server model or environmental parameters, such as increasing the speed value corresponding to the same power in a high temperature environment, so as to adapt to the heat dissipation requirements under different working conditions.

[0109] As an example, when generating a fan card control instruction, the instruction generation module 150 first obtains a pre-established fan speed curve, which is obtained through experimental testing and includes a mapping relationship between sixteen sets of heat dissipation power values ​​and corresponding fan speeds. When the target time interval is assigned a heat dissipation power strategy value of 380W, 380W is input into the fan speed curve for linear interpolation calculation to obtain the corresponding standard speed value of 1850 rpm. The standard speed value is converted into a pulse signal with a duty cycle of 62%, and finally a fan control instruction set including a start timestamp, duration and duty cycle parameters is generated. The control instruction set is sent to the baseboard management controller through the protocol interface to drive the fan card to run continuously at a constant speed.

[0110] Through this embodiment, the frequent step-by-step adjustment of fan speed caused by sudden load changes in the prior art is effectively avoided. By establishing a static mapping relationship between heat dissipation power and fan speed, dynamic power regulation is converted into steady-state speed control, thereby reducing the frequency of command switching during control signal transmission, reducing the additional power loss of the heat dissipation system caused by command oscillation, and maintaining a dynamic balance between heat dissipation power and heat dissipation demand.

[0111] It should be understood that the present invention is not limited to the specific configurations and processes described above and illustrated in the figures. For the sake of brevity, a detailed description of known methods is omitted. In the above embodiments, several specific steps are described and illustrated as examples. However, the method of the present invention is not limited to the specific steps described and illustrated. Those skilled in the art may make various changes, modifications, and additions, or change the order of the steps after understanding the spirit of the present invention.

[0112] It should also be noted that the exemplary embodiments described herein describe methods or systems based on a series of steps or devices. However, the present invention is not limited to the order of the steps described above. In other words, the steps may be performed in the order described in the embodiments, or in a different order, or several steps may be performed simultaneously.

[0113] The above description is only a specific embodiment of the present invention. Those skilled in the art will clearly understand that for the convenience and brevity of description, the specific working processes of the systems, modules and units described above can refer to the corresponding processes in the aforementioned method embodiments, and will not be repeated here. It should be understood that the protection scope of the present invention is not limited to this. Any person skilled in the art can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention.

Claims

1. A server fan card heat dissipation power control system under load environment, characterized in that: The system comprises: An information acquisition module is used to obtain historical temperature data, historical load data, and a heat dissipation power curve of a target server, wherein the heat dissipation power curve is used to represent the heat dissipation power of a fan card corresponding to the target server being able to recover to a standard temperature under different temperature conditions; a demand prediction module, configured to predict a first heat dissipation demand of the target server at each sampling moment in a target future time period based on the historical temperature data and the historical load data; an interval division module, configured to divide the target future time period into a plurality of target time intervals based on the first heat dissipation demand degrees at each sampling moment in the target future time period, wherein the first heat dissipation demand degrees within the target time intervals are within the same demand degree range; a strategy determination module, configured to determine, for each target time interval, a heat dissipation strategy for the target time interval based on each first heat dissipation requirement within the target time interval, a fluctuation characteristic of the first heat dissipation requirement, and the heat dissipation power curve; The instruction generation module is used to generate a fan card control instruction for the target server based on the heat dissipation strategy of each target time interval, so as to control the heat dissipation power of the fan card of the target server.

2. The server fan card heat dissipation power control system under load environment according to claim 1, characterized in that: The demand prediction module includes the following units: a demand calculation unit, configured to determine a second heat dissipation demand at each historical sampling moment based on each of the historical temperature data and each of the historical load data; a heat dissipation demand ordering unit, configured to order the second heat dissipation demand orders in chronological order to obtain a heat dissipation demand order sequence; The demand prediction unit is used to predict the first heat dissipation demand of the target server at each sampling moment in a target future time period based on the heat dissipation demand sequence through an autoregressive integral moving average model.

3. The server fan card heat dissipation power control system under load environment according to claim 2, characterized in that: The demand calculation unit is configured to: Obtain target historical temperature data and first historical load data at a target historical sampling moment, and second historical load data at a sampling moment before the target historical sampling moment, where the target historical sampling moment is any one of the historical sampling moments; A second heat dissipation requirement at the target historical sampling moment is determined by using a ratio of the first historical load data to the second historical load data and the target historical temperature data.

4. The server fan card heat dissipation power control system under load environment according to claim 1, characterized in that: The interval division module includes the following units: a time period determination unit, configured to determine the target future time period as the time period to be divided; A result determination unit, configured to determine a plurality of candidate division results by taking any sampling moment within the time period to be divided as a division boundary; a distinguishability determining unit, configured to determine, for each candidate partitioning result, a distinguishability of the heat dissipation requirements of the two candidate time intervals based on each of the first heat dissipation requirements of the two candidate time intervals in the candidate partitioning result; The result determination unit is further configured to determine the candidate partitioning result with the smallest cumulative value of the heat dissipation demand differentiation between the two candidate time intervals as the target partitioning result; A loop execution unit is used to update the time period to be divided into each target time interval in the target division result, and return to the loop execution to use any sampling moment in the time period to be divided as the division boundary, determine multiple candidate division results, until the recursive stop condition is met, and obtain multiple target time intervals.

5. The server fan card heat dissipation power control system under load environment according to claim 4, characterized in that: The distinguishability determination unit is configured to: Obtaining each first heat dissipation demand in the candidate time interval; performing statistics on each of the first heat dissipation requirements in the candidate time interval to obtain a heat dissipation requirement statistical result for the candidate time interval, the heat dissipation requirement statistical result including at least one of a peak heat dissipation requirement, a valley heat dissipation requirement, an average heat dissipation requirement, a maximum heat dissipation requirement, and a minimum heat dissipation requirement; The heat dissipation requirement statistical result is used to determine the heat dissipation requirement differentiation of the candidate time interval.

6. The server fan card heat dissipation power control system under load environment according to claim 1, characterized in that: The strategy determination module includes the following units: a demand analysis unit, configured to determine an average heat dissipation demand based on each of the first heat dissipation demand levels within the target time interval, and determine a fluctuation influence coefficient based on a fluctuation characteristic of the first heat dissipation demand levels; a heat dissipation demand correction unit, configured to correct the average heat dissipation demand by using the fluctuation influence coefficient to obtain a first interval heat dissipation demand in the target time interval; A curve matching unit is used to match the heat dissipation demand in the first interval with the heat dissipation power curve to determine the heat dissipation strategy in the target time interval.

7. The server fan card heat dissipation power control system under load environment according to claim 6, characterized in that: The demand analysis unit is used to: Obtaining a peak heat dissipation demand and a valley heat dissipation demand in each of the first heat dissipation demand; Based on each of the peak heat dissipation requirements and each of the valley heat dissipation requirements, determine the time interval between adjacent peak heat dissipation requirements, the average peak heat dissipation requirement, and the average valley heat dissipation requirement; Subtracting the average peak heat dissipation demand from the average valley heat dissipation demand to determine an average peak-to-valley difference; The fluctuation influence coefficient of the target time interval is determined by using each of the time intervals and the average peak-to-valley value difference.

8. The server fan card heat dissipation power control system under load environment according to claim 6, characterized in that: After the average heat dissipation requirement is corrected by using the fluctuation influence coefficient to obtain the first interval heat dissipation requirement of the target time interval, the strategy determination module further includes: a difference obtaining unit, configured to obtain the difference in heat dissipation requirements between adjacent target time intervals; an interval merging unit, configured to merge adjacent target time intervals to obtain a merged time interval when the difference in the heat dissipation requirements is less than or equal to a preset difference threshold; The demand analysis unit is further configured to determine a second interval heat dissipation demand of the combined time interval based on the two first interval heat dissipation demand levels corresponding to the combined time interval; The curve matching unit is configured to match the heat dissipation demand in the second interval with the heat dissipation power curve to determine a heat dissipation strategy for the combined time interval.

9. The server fan card heat dissipation power control system under load environment according to any one of claims 1 to 8, characterized in that: After determining, for each target time interval, the heat dissipation strategy for the target time interval based on each first heat dissipation requirement within the target time interval, the fluctuation characteristics of the first heat dissipation requirement, and the heat dissipation power curve, the system further includes: A power value interpolation module, configured to interpolate the transition heat dissipation power value between two adjacent target time intervals by linear interpolation; The instruction generation module is configured to generate a fan card control instruction for the target server based on the heat dissipation strategy of each target time interval and each transition heat dissipation power value.

10. The server fan card heat dissipation power control system under a load environment according to any one of claims 1 to 8, wherein the heat dissipation strategy includes the fan card heat dissipation power in the target time interval; It is characterized by: The instruction generation module is used to: Matching the fan card heat dissipation power of the target time interval with a fan speed curve to obtain the fan speed of the target time interval, wherein the fan speed curve is used to represent the fan speed corresponding to the target server under different fan card heat dissipation powers; The fan speeds in each target time interval are converted into fan card control instructions for the target server.

Images