WO2023040165A1 - 一种服务器风扇转速控制方法、装置、设备及介质 - Google Patents

一种服务器风扇转速控制方法、装置、设备及介质 Download PDF

Info

Publication number
WO2023040165A1
WO2023040165A1 PCT/CN2022/074611 CN2022074611W WO2023040165A1 WO 2023040165 A1 WO2023040165 A1 WO 2023040165A1 CN 2022074611 W CN2022074611 W CN 2022074611W WO 2023040165 A1 WO2023040165 A1 WO 2023040165A1
Authority
WO
WIPO (PCT)
Prior art keywords
rotational speed
historical
server
fan
value
Prior art date
Application number
PCT/CN2022/074611
Other languages
English (en)
French (fr)
Inventor
闫玉婕
于泉泉
王焕超
刘闻禹
韩煦
Original Assignee
苏州浪潮智能科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 苏州浪潮智能科技有限公司 filed Critical 苏州浪潮智能科技有限公司
Publication of WO2023040165A1 publication Critical patent/WO2023040165A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F1/00—Details not covered by groups G06F3/00 - G06F13/00 and G06F21/00
    • G06F1/16—Constructional details or arrangements
    • G06F1/20—Cooling means
    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F04—POSITIVE - DISPLACEMENT MACHINES FOR LIQUIDS; PUMPS FOR LIQUIDS OR ELASTIC FLUIDS
    • F04D—NON-POSITIVE-DISPLACEMENT PUMPS
    • F04D27/00—Control, e.g. regulation, of pumps, pumping installations or pumping systems specially adapted for elastic fluids
    • F04D27/004—Control, e.g. regulation, of pumps, pumping installations or pumping systems specially adapted for elastic fluids by varying driving speed
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F1/00—Details not covered by groups G06F3/00 - G06F13/00 and G06F21/00
    • G06F1/26—Power supply means, e.g. regulation thereof
    • G06F1/32—Means for saving power
    • G06F1/3203—Power management, i.e. event-based initiation of a power-saving mode
    • G06F1/3206—Monitoring of events, devices or parameters that trigger a change in power modality
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F1/00—Details not covered by groups G06F3/00 - G06F13/00 and G06F21/00
    • G06F1/26—Power supply means, e.g. regulation thereof
    • G06F1/32—Means for saving power
    • G06F1/3203—Power management, i.e. event-based initiation of a power-saving mode
    • G06F1/3234—Power saving characterised by the action undertaken
    • G06F1/325—Power saving in peripheral device
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F11/00—Error detection; Error correction; Monitoring
    • G06F11/30—Monitoring
    • G06F11/3055—Monitoring arrangements for monitoring the status of the computing system or of the computing system component, e.g. monitoring if the computing system is on, off, available, not available
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F11/00—Error detection; Error correction; Monitoring
    • G06F11/30—Monitoring
    • G06F11/3058—Monitoring arrangements for monitoring environmental properties or parameters of the computing system or of the computing system component, e.g. monitoring of power, currents, temperature, humidity, position, vibrations
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/243—Classification techniques relating to the number of classes
    • G06F18/2433—Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G06N20/20—Ensemble learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management

Definitions

  • the present application relates to the server field, and in particular to a server fan speed control method, device, equipment and medium.
  • the heat dissipation of the server mainly depends on the fan.
  • the fan is mainly directly controlled by the BMC (Baseboard Manager Controller) or CPLD (Complex Programmable Logic Device) on the motherboard.
  • BMC Baseboard Manager Controller
  • CPLD Complex Programmable Logic Device
  • the above-mentioned server fan cooling methods have relatively large disadvantages.
  • the temperature of the motherboard jumps rapidly, and the fan speed will be directly adjusted to a large extent.
  • the fan will run rapidly to reduce the temperature.
  • large power consumption and noise will also be generated, which will have a great impact on the power supply.
  • the current heat dissipation method of the server fan has the problems of high noise, high power consumption, and low performance.
  • the purpose of this application is to provide a server fan speed control method, device, equipment and media, which can effectively reduce the rapid adjustment of the server fan speed, thereby reducing fan noise and fan power consumption, and improving fan performance.
  • the efficiency makes the fan run more stable.
  • the present application discloses a server fan speed control method, including:
  • the inputting the current operating state data into the pre-created rotational speed prediction model includes:
  • the training sample data includes the historical operating state data of the server and the fan speed value corresponding to the historical operating state data.
  • the server fan speed control method further includes:
  • At a plurality of first historical moments respectively collect the historical operating state data of the server and the corresponding fan speed value to obtain a sample set including a plurality of initial sample data; wherein the historical operating state data includes the server The first historical chassis temperature at the first historical moment;
  • the collecting historical running status data of the server includes:
  • the inputting the current running state data into multiple pre-created rotation speed prediction models respectively includes:
  • the current running state data is respectively input into the first rotational speed prediction model constructed in advance based on the XGBoost algorithm, the second rotational speed prediction model constructed based on the support vector machine algorithm, and the third rotational speed prediction model constructed based on the artificial neural network algorithm.
  • the predicting the fan speed value corresponding to the current operating state data through the speed prediction model to obtain an initial speed prediction value includes:
  • using the target historical rotational speed value to optimize and correct the initial rotational speed prediction value includes:
  • a server fan speed control device including:
  • the data acquisition module is used to collect the running status data of the server at the current moment, and obtain the current running status data
  • a prediction module configured to input the current operating state data into a pre-created rotational speed prediction model, and predict the fan rotational speed value corresponding to the current operating state data through the rotational speed prediction model, to obtain an initial rotational speed prediction value;
  • a data acquisition module configured to acquire the fan speed value of the server within a historical time period separated by a preset time from the current time, and obtain a corresponding target historical speed value
  • the rotation speed control module is configured to use the target historical rotation speed value to optimize and correct the initial predicted rotation speed value, and control the rotation speed of the fan of the server according to the optimized predicted rotation speed value.
  • the present application discloses an electronic device, including a processor and a memory; wherein, when the processor executes a computer program stored in the memory, the aforementioned server fan speed control method is realized.
  • the present application discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, the foregoing server fan speed control method is implemented.
  • this application first collects the running status data of the server at the current moment to obtain the current running status data, and then inputs the current running status data into the pre-created rotation speed prediction model, and uses the rotation speed prediction model to predict and match the current running status data.
  • the fan speed value corresponding to the state data is obtained to obtain the initial speed prediction value, and then obtain the fan speed value of the server in the historical time period separated from the current time by a preset time period to obtain the corresponding target historical speed value, and then use the
  • the target historical rotational speed value is optimized and corrected to the initial predicted rotational speed value, and the rotational speed of the fan of the server is controlled according to the optimized rotational speed predicted value.
  • this application predicts the fan speed of the server based on the running status data of the server at the current moment and the pre-created speed prediction model to obtain the initial speed prediction value, and further uses the historical time period of the server fan at a preset time interval from the current time.
  • the historical speed value of the above-mentioned initial speed prediction value is optimized and corrected, instead of directly using the above-mentioned initial speed prediction value to control the speed of the server fan, which can effectively reduce the rapid adjustment of the subsequent fan speed adjustment to a certain extent. Thereby, the noise of the fan and the power consumption of the fan are reduced, the performance of the fan is improved, and the operation of the fan is more stable.
  • FIG. 1 is a flow chart of a server fan speed control method disclosed in the present application
  • FIG. 2 is a flow chart of a specific server fan speed control method disclosed in the present application.
  • FIG. 3 is a flowchart of a method for obtaining training sample data disclosed in the present application.
  • FIG. 4 is a flow chart of a specific server fan speed control method disclosed in the present application.
  • FIG. 5 is a schematic structural diagram of a server fan speed control device disclosed in the present application.
  • FIG. 6 is a structural diagram of an electronic device disclosed in the present application.
  • FIG. 7 is a schematic structural diagram of a computer-readable storage medium disclosed in the present application.
  • the present application provides a server fan speed control scheme, which can effectively reduce the rapid adjustment of the server fan speed, thereby reducing fan noise and fan power consumption, improving fan performance, and making fan operation more stable.
  • the embodiment of the present application discloses a server fan speed control method, as shown in FIG. 1 , the method includes:
  • Step S11 Collect the running status data of the server at the current moment to obtain the current running status data.
  • the BMC on the mainboard of the server can collect the running state data of the server at the current moment to obtain the current running state data; wherein, the current running state data includes but not limited to CPU usage, memory Utilization, hard disk read and write speed, power supply and chassis temperature, etc.
  • Step S12 Input the current operating state data into a pre-created rotational speed prediction model, and predict the fan rotational speed value corresponding to the current operating state data through the rotational speed prediction model, to obtain an initial rotational speed prediction value.
  • the collected above-mentioned current running state data can be input into the pre-created rotation speed prediction model.
  • the input data will be processed accordingly, and then the fan speed prediction value corresponding to the above current running state data, that is, the initial speed prediction value will be output.
  • there will be a plurality of output fan speed values and the multiple fan speed values will be processed accordingly, such as screened and/or calculated, to obtain the predicted initial speed value.
  • Step S13 Acquiring the rotation speed value of the fan of the server within a historical time period separated by a preset time from the current moment, and obtaining a corresponding target historical rotation speed value.
  • the current running state data is input into the pre-created speed prediction model, and the fan speed value corresponding to the current running state data is predicted by the speed prediction model, and after the initial speed prediction value is obtained, the The BMC on the main board of the server collects the fan speed value of the server in the historical period of the preset time before the above current moment, and obtains the historical fan speed value corresponding to the historical time period, that is, the target historical rotational speed value.
  • the aforementioned preset duration can be determined by manually setting parameters.
  • Step S14 Optimizing and correcting the initial predicted rotational speed value by using the target historical rotational speed value, and controlling the rotational speed of the fan of the server according to the optimized predicted rotational speed value.
  • the above-mentioned initial rotation speed can be predicted by using the above-mentioned target historical rotation speed value. It can be understood that the above-mentioned initial speed prediction value is predicted based on the current operating state data. In specific applications, in order to prevent the fan speed from rapidly increasing due to a relatively large deviation between the predicted value and the current fan speed value, the above-mentioned initial rotation speed prediction value can be further optimized and corrected by using the above-mentioned target historical rotation speed value. And it can generate the corresponding speed control command according to the optimized speed prediction value, and send the above speed control command to the BMC. After the BMC obtains the above speed control command, it can control the server fan in PWM mode to adjust the current server fan speed. To the above target historical speed value.
  • the running status data of the server at the current moment is collected to obtain the current running status data, and then the current running status data is input into the pre-created rotation speed prediction model, and the rotation speed prediction model is used to predict and match the
  • the fan speed value corresponding to the current operating state data is obtained to obtain the initial speed prediction value, and then the fan speed value of the server in the historical time period separated from the current time by a preset time is obtained to obtain the corresponding target historical speed value, and then use
  • the target historical rotational speed value is optimized and corrected to the initial predicted rotational speed value, and the rotational speed of the fan of the server is controlled according to the optimized rotational speed predicted value.
  • the embodiment of the present application is based on the running status data of the server at the current moment and the pre-created rotation speed prediction model to predict the fan rotation speed of the server and obtain the initial rotation speed prediction value, and further use the historical time with a preset time interval from the current time
  • the historical speed value of the server fan in the segment optimizes and corrects the above-mentioned initial speed prediction value, instead of directly using the above-mentioned initial speed prediction value to control the speed of the server fan. Adjustment, thereby reducing fan noise and fan power consumption, improving fan performance, and making fan operation more stable.
  • the embodiment of the present application discloses a specific server fan speed control method, as shown in Figure 2, the method includes:
  • Step S21 Collect the running status data of the server at the current moment to obtain the current running status data.
  • Step S22 Input the current running state data into multiple pre-created rotation speed prediction models; wherein, the multiple rotation speed prediction models are to be trained based on different model construction algorithms using training sample data respectively.
  • a model is obtained after the model is trained, and the training sample data includes the historical operating state data of the server and the fan speed value corresponding to the historical operating state data.
  • the collected above-mentioned current running state data can be respectively input into multiple rotational speed prediction models created in advance based on different model construction algorithms .
  • the above-mentioned multiple predictive models use the training sample data including the historical operating state data of the server and the fan speed value corresponding to the historical operating state data, respectively to a plurality of pre-built based on different model construction algorithms to be trained
  • the model obtained after the model is trained; the model construction algorithm includes but is not limited to XGBoost (Extreme Gradient Boosting, extreme gradient boosting) algorithm, support vector machine (SVM, Support Vector Machine) algorithm, artificial neural network (ANN, Artificial Neural Network ) algorithm, etc.
  • this embodiment further discloses a process of obtaining the above training sample data, which may specifically include:
  • Step S31 At a plurality of first historical moments, respectively collect historical operating state data of the server and corresponding fan speed values to obtain a sample set including a plurality of initial sample data; wherein the historical operating state data includes the The first historical chassis temperature of the server at the first historical moment;
  • Step S32 Add the preset time interval to the multiple first historical moments respectively to obtain multiple second historical moments corresponding to the multiple first historical moments, and respectively collect the Chassis temperatures at multiple second historical moments to obtain multiple second historical chassis temperatures;
  • Step S33 judging whether the difference between the second historical chassis temperature and the corresponding first historical chassis temperature is smaller than a preset threshold, and if so, the sample set corresponding to the first historical chassis temperature
  • the initial sample data is normalized to obtain the training sample data, otherwise, the initial sample data corresponding to the first historical chassis temperature is eliminated from the sample set.
  • the historical operation status data of the server at the above historical time and the fan speed value at the corresponding historical time can be collected on a large scale based on any number of historical time points to obtain the historical
  • the sampling data that is, the initial sample data
  • the historical operating status data includes the chassis temperature at the first historical moment, that is, the first historical chassis temperature; it should be pointed out that, in addition to the first historical chassis temperature, the historical operating status data Including but not limited to CPU occupancy rate, memory utilization rate, hard disk read/write rate, power supply, etc. of the server at the historical moment.
  • the sampling number of the initial sample data in this embodiment may be determined based on actual application requirements, for example, it may be 200,000.
  • first historical moments collect the historical operating status data of the server and the corresponding fan speed values respectively to obtain a sample set including multiple initial sample data, and obtain the above-mentioned data at preset time intervals.
  • the first historical moment corresponds to a plurality of second historical moments after the delay of the above-mentioned time interval, and respectively collects the chassis temperature of the server at the second historical moment to obtain a plurality of corresponding second historical chassis temperatures.
  • the preset time interval is a
  • the temperature of the server case at time t+a is collected and marked as P t+a .
  • the parameter a is usually set to 10, and the unit is second (s).
  • the initial sample data corresponding to the above-mentioned first historical chassis temperature is normalized, so as to obtain training sample data for model training.
  • the above normalization methods include but not limited to linear function conversion method, logarithmic function conversion method, arctangent function conversion method and the like.
  • the value of the preset threshold above is not fixed, but is a custom value based on the chassis temperature of the server chassis in normal operation state and the temperature of the chassis that is too high and reaches a temperature that affects the normal operation of the server. For example, if the temperature of the server chassis under normal operation is taken as 25 degrees, and according to the actual server application conditions, the temperature affecting the normal operation of the server is taken as 55 degrees, then the above-mentioned temperature affecting the normal operation of the server can be compared with the above-mentioned normal operation state The temperature difference is used as the preset threshold, that is, the preset threshold is 30.
  • the difference between the second historical chassis temperature and the first historical chassis temperature is not less than a preset threshold, it is determined that the The initial sample data is invalid, and the initial sample data corresponding to the first historical chassis temperature is removed from the sample set.
  • the order of the above-mentioned legal sample data can be disturbed to obtain the reordered training samples data.
  • Step S23 Predict the fan speed values corresponding to the current operating state data through the multiple speed prediction models, and obtain corresponding multiple predicted speed values.
  • the above-mentioned multiple rotation speed prediction models will perform corresponding processing on the above-mentioned current running state data, and then output multiple The fan speed prediction value corresponding to the current running state data, that is, the multiple speed prediction values.
  • the plurality of rotational speed prediction models include but not limited to a first rotational speed prediction model constructed in advance based on the XGBoost algorithm, a second rotational speed prediction model constructed based on a support vector machine algorithm, and a third rotational speed prediction model constructed based on an artificial neural network algorithm wait.
  • Step S24 Using an outlier detection algorithm to determine the predicted rotational speed value to be eliminated from the multiple predicted rotational speed values, and then remove the predicted rotational speed value to be eliminated from the multiple predicted rotational speed values to obtain the remaining rotational speed prediction value.
  • the outlier detection algorithm can be used to obtain the fan speed values from the above multiple speeds. Determine the value with a large deviation from other predicted values in the predicted value, that is, the predicted value of the rotational speed to be deleted, and then remove the numerical value with a large deviation from other predicted values from the above-mentioned multiple predicted values of the rotational speed, and then obtain the remaining speed prediction.
  • Step S25 performing an average calculation on the predicted remaining rotational speed to obtain the predicted initial rotational speed.
  • the predicted value of the rotational speed after using the outlier detection algorithm to determine the predicted rotational speed value to be eliminated from the multiple predicted rotational speed values, and then to remove the predicted rotational speed value to be eliminated from the multiple predicted rotational speed values, to obtain After the predicted value of the remaining rotational speed, the predicted value of the rotational speed, that is, the predicted value of the initial rotational speed, can be determined by performing an average calculation on the predicted value of the remaining rotational speed.
  • Step S26 Acquiring the rotation speed value of the fan of the server within a historical time period separated by a preset duration from the current moment, to obtain a corresponding target historical rotation speed value.
  • Step S27 Perform weighted average processing on the target historical rotational speed value and the initial rotational speed prediction value, so as to complete the optimization correction of the initial rotational speed prediction value, and adjust the fan of the server according to the optimized rotational speed prediction value Rotational speed control; wherein, the value after the weighted average processing is the optimized predicted value of the rotational speed.
  • the weight of the initial rotational speed prediction value can be set to 3/5
  • the weight of the target historical rotational speed value can be set to 2/5
  • the target historical rotational speed value and the initial rotational speed prediction value can be calculated using the weights set above.
  • Weighted average processing and use the value after weighted average processing to control the fan speed of the server.
  • the same weight may be configured for the above-mentioned target historical rotational speed value and the above-mentioned initial rotational speed predicted value.
  • the collected current operating state data are respectively input into multiple pre-created rotation speed prediction models, and then the outlier detection algorithm is used to eliminate the predicted values with large deviations from the outputted multiple predicted rotation speed values.
  • the initial rotation speed prediction value is obtained, and the above initial rotation speed prediction value is optimized and corrected by obtaining the fan rotation speed value of the server in the historical time period separated from the current time by a preset time period, which can be based on the current operating status data in real time
  • Obtaining the predicted value of the fan speed can reduce the noise caused by the sudden increase of the fan speed during operation, reduce the power consumption of the fan, increase the performance of the fan, make the change of the fan speed more smooth, and prolong the life of the fan.
  • the embodiment of the present application discloses a specific server fan speed control method, as shown in Figure 4, the method includes:
  • Step S41 Collect the running status data of the server at the current moment to obtain the current running status data.
  • Step S42 Input the current operating state data into the first rotational speed prediction model constructed in advance based on the XGBoost algorithm, the second rotational speed prediction model constructed based on the support vector machine algorithm, and the third rotational speed prediction model constructed based on the artificial neural network algorithm .
  • the collected above-mentioned current running state data can be respectively input into the first rotational speed prediction model constructed in advance based on the XGBoost algorithm, based on the support
  • the second rotational speed prediction model constructed by the vector machine algorithm and the third rotational speed prediction model constructed based on the artificial neural network algorithm can be respectively input into the first rotational speed prediction model constructed in advance based on the XGBoost algorithm, based on the support
  • the second rotational speed prediction model constructed by the vector machine algorithm and the third rotational speed prediction model constructed based on the artificial neural network algorithm.
  • the artificial neural network may use a three-layer perceptron, and the modeling method may adopt a backpropagation algorithm.
  • Step S43 Predict the fan speed value corresponding to the current operating state data through the first speed prediction model, the second speed prediction model and the third speed prediction model, and obtain the corresponding first speed prediction value, The second predicted rotational speed value and the third predicted rotational speed value.
  • the second rotational speed prediction model constructed based on the support vector machine algorithm
  • the third rotational speed prediction model constructed based on the artificial neural network algorithm
  • the rotational speed prediction model predict the fan rotational speed value corresponding to the current operating state data through the first rotational speed prediction model, the second rotational speed prediction model and the third rotational speed prediction model, and output three corresponding rotational speed prediction values, That is, the first predicted rotational speed value, the second predicted rotational speed value and the third predicted rotational speed value are marked as S m1 , S m2 , and S m3 respectively.
  • Step S44 Using an outlier detection algorithm to determine the predicted rotational speed value to be eliminated from the first predicted rotational speed value, the second predicted rotational speed value, and the third predicted rotational speed value, and then predict the rotational speed to be eliminated The value is eliminated from the plurality of predicted rotation speed values to obtain the remaining predicted rotation speed values.
  • the fan speed value corresponding to the current operating state data is predicted by the first speed prediction model, the second speed prediction model, and the third speed prediction model to obtain the corresponding first speed
  • the outlier detection algorithm can be used to determine the predicted rotational speed to be eliminated from the above S m1 , S m2 , and S m3
  • S m3 is determined to be the predicted value of the rotational speed to be eliminated
  • S m3 is removed from the above S m1 , S m2 , and S m3 to obtain the remaining predicted value of the rotational speed, that is, S m1 , S m2 .
  • the average of the three predicted rotational speed values may be directly calculated.
  • Step S45 performing an average calculation on the predicted remaining rotational speed to obtain the predicted initial rotational speed.
  • an outlier detection algorithm is used to determine the predicted rotational speed value to be eliminated from the first predicted rotational speed value, the second predicted rotational speed value, and the third predicted rotational speed value, and then the After the predicted value of the eliminated rotational speed is removed from the plurality of predicted rotational speeds to obtain the predicted residual rotational speed, the average calculation may be performed on the predicted remaining rotational speed, and the calculation result may be used as the predicted initial rotational speed. Specifically, for example, by performing average calculation on the remaining above-mentioned S m1 and S m2 , the average calculation result can be obtained and marked as S m .
  • Step S46 Acquiring the rotation speed value of the fan of the server within a historical time period separated by a preset duration from the current moment, to obtain a corresponding target historical rotation speed value.
  • the fans of the server in the historical time period that is separated from the current moment by a preset duration may be collected.
  • Speed value get the corresponding target historical speed value. For example, taking the current collection time as a reference, collecting the historical fan speed values within the period b before the current time, and collecting the historical fan speed values within the time period b according to a preset sampling period, to obtain multiple
  • Step S47 Optimizing and correcting the initial predicted rotational speed value by using the target historical rotational speed value, and controlling the rotational speed of the fan of the server according to the optimized predicted rotational speed value.
  • the target historical rotation speed value can be used to calculate the above initial rotation speed value.
  • the predicted value is optimized and corrected. For example, by calculating the average value of V and S m above, the above-mentioned initial speed prediction value is optimized and corrected, and the speed of the fan of the server is controlled based on the PWM square wave based on the speed prediction value calculated by the average value.
  • the collected current operating state data is respectively input into the first rotational speed prediction model constructed based on the XGBoost algorithm, the second rotational speed prediction model constructed based on the support vector machine algorithm, and the first rotational speed prediction model constructed based on the artificial neural network algorithm.
  • the three-speed prediction model and then use the outlier detection algorithm to remove the prediction value with a large deviation from the multiple output prediction values of the rotation speed, and obtain the initial prediction value of the rotation speed through calculation, and obtain the preset time interval from the current time.
  • the fan speed value of the server in the historical time period is optimized and corrected for the above-mentioned initial speed prediction value.
  • the fan speed can be predicted in real time based on the server's current operating status data, which can make the fan speed change more stable and reduce the fan speed.
  • the noise generated by the sudden increase of the speed during operation reduces the power consumption of the fan, increases the efficiency of the fan, and prolongs the life of the fan.
  • the embodiment of the present application also discloses a server fan speed control device, as shown in Fig. 5, the device includes:
  • the data collection module 11 is used to collect the running state data of the server at the current moment, and obtains the current running state data
  • a prediction module 12 configured to input the current operating state data into a pre-created rotational speed prediction model, and predict the fan rotational speed value corresponding to the current operating state data through the rotational speed prediction model, to obtain an initial rotational speed prediction value;
  • a data acquisition module 13 configured to acquire the fan rotation speed value of the server within a historical time period separated by a preset time from the current time, and obtain a corresponding target historical rotation speed value;
  • the rotation speed control module 14 is configured to use the target historical rotation speed value to optimize and correct the initial predicted rotation speed value, and control the rotation speed of the fan of the server according to the optimized predicted rotation speed value.
  • the embodiment of the present application first collect the running status data of the server at the current moment to obtain the current running status data, and then input the current running status data into the pre-created rotation speed prediction model, and use the rotation speed prediction model to predict and
  • the fan speed value corresponding to the current operating state data is obtained by obtaining an initial speed prediction value, and then obtaining the fan speed value of the server within a historical time period separated from the current time by a preset time length to obtain a corresponding target historical speed value, Then, the target historical rotational speed value is used to optimize and correct the initial predicted rotational speed value, and control the rotational speed of the fan of the server according to the optimized rotational speed predicted value.
  • the fan rotation speed of the server is predicted to obtain the initial rotation speed prediction value, and further use the historical time period separated by a preset time from the current time.
  • the historical speed value of the server fan optimizes and corrects the above-mentioned initial speed prediction value, instead of directly using the above-mentioned initial speed prediction value to control the server fan speed, which can effectively reduce the rapid adjustment of subsequent fan speed adjustments to a certain extent. In this way, the fan noise and fan power consumption are reduced, the performance of the fan is improved, and the fan operation is more stable.
  • the prediction module 12 may specifically include:
  • the first input unit is configured to respectively input the current operating state data into a plurality of pre-created rotation speed prediction models; wherein, the plurality of rotation speed prediction models are pre-constructed based on different model construction algorithms using training sample data respectively.
  • a model obtained after a plurality of models to be trained are trained, and the training sample data includes historical operating state data of the server and fan speed values corresponding to the historical operating state data.
  • the server fan speed control device may specifically include:
  • the first data collection unit is configured to separately collect the historical operating status data of the server and the corresponding fan speed values at multiple first historical moments, so as to obtain a sample set including multiple initial sample data; wherein, the The historical running state data includes a first historical chassis temperature of the server at the first historical moment;
  • a temperature acquisition unit configured to add a preset time interval to a plurality of first historical moments to obtain a plurality of second historical moments respectively corresponding to a plurality of first historical moments, and collect the respective The chassis temperature of the server at multiple second historical moments to obtain multiple second historical chassis temperatures;
  • a judging unit configured to judge whether the difference between the second historical chassis temperature and the corresponding first historical chassis temperature is smaller than a preset threshold
  • a normalization unit configured to collect the samples corresponding to the first historical chassis temperature if the difference between the second historical chassis temperature and the corresponding first historical chassis temperature is less than a preset threshold Normalize the initial sample data to obtain the training sample data;
  • a first eliminating unit configured to remove the first historical chassis temperature from the sample set if the difference between the second historical chassis temperature and the corresponding first historical chassis temperature is not less than a preset threshold corresponding to the initial sample data.
  • the first data collection unit may specifically include:
  • the second data collection unit is configured to collect CPU occupancy rate, memory utilization rate, hard disk read/write rate, power supply and temperature of the first historical chassis of the server.
  • the first input unit may specifically include:
  • the second input unit is used to respectively input the current operating state data into the first rotational speed prediction model constructed in advance based on the XGBoost algorithm, the second rotational speed prediction model constructed based on the support vector machine algorithm, and the first rotational speed prediction model constructed based on the artificial neural network algorithm.
  • the prediction module 12 may specifically include:
  • a predicting unit configured to predict fan speed values corresponding to the current operating state data through the multiple speed prediction models, and obtain corresponding multiple predicted speed values
  • the second eliminating unit is configured to use an outlier detection algorithm to determine a predicted rotational speed value to be eliminated from the plurality of predicted rotational speed values, and then delete the predicted rotational speed value to be eliminated from the plurality of predicted rotational speed values, Obtain the predicted value of the residual speed;
  • a calculation unit configured to perform an average calculation on the predicted residual rotational speed to obtain the predicted initial rotational speed.
  • the rotational speed control module 14 may specifically include:
  • a correction unit configured to perform weighted average processing on the target historical rotational speed value and the initial rotational speed prediction value, so as to complete the optimization correction of the initial rotational speed prediction value; wherein, the value after the weighted average processing is optimized predicted speed of .
  • FIG. 6 is a structural diagram of an electronic device 20 according to an exemplary embodiment.
  • the content in the figure should not be regarded as any limitation on the application scope of the present application.
  • FIG. 6 is a schematic structural diagram of an electronic device 20 provided by an embodiment of the present application.
  • the electronic device 20 may specifically include: at least one processor 21 , at least one memory 22 , a power supply 23 , a communication interface 24 , an input/output interface 25 and a communication bus 26 .
  • the memory 22 is used to store a computer program, and the computer program is loaded and executed by the processor 21 to implement relevant steps in the server fan speed control method disclosed in any of the above-mentioned embodiments.
  • the electronic device 20 in this embodiment may specifically be an electronic computer.
  • the power supply 23 is used to provide working voltage for each hardware device on the electronic device 20;
  • the communication interface 24 can create a data transmission channel between the electronic device 20 and external devices, and the communication protocol it follows is applicable Any communication protocol in the technical solution of the present application is not specifically limited here;
  • the input and output interface 25 is used to obtain external input data or output data to the external, and its specific interface type can be selected according to specific application needs, here Not specifically limited.
  • the memory 22, as a resource storage carrier can be a read-only memory, random access memory, magnetic disk or optical disk, etc., and the resources stored thereon can include operating system 221, computer program 222, etc., and the storage method can be temporary storage or permanent storage. .
  • the operating system 221 is used to manage and control each hardware device on the electronic device 20 and a computer program 222, which may be Windows Server, Netware, Unix, Linux, etc.
  • the computer program 222 may further include a computer program capable of completing other specific tasks in addition to the computer program capable of implementing the method for controlling the fan speed of the server performed by the electronic device 20 disclosed in any of the foregoing embodiments.
  • FIG. 7 is a schematic structural diagram of a computer-readable storage medium disclosed in the present application. As shown in FIG. 7 , further, the present application also discloses a computer-readable storage medium 601 for storing a computer program 610; wherein, when the computer program 610 is executed by a processor, the server fan speed control disclosed above is realized. method. Regarding the specific steps of the method, reference may be made to the corresponding content disclosed in the foregoing embodiments, and details are not repeated here.
  • each embodiment in this specification is described in a progressive manner, each embodiment focuses on the difference from other embodiments, and the same or similar parts of each embodiment can be referred to each other.
  • the description is relatively simple, and for the related information, please refer to the description of the method part.
  • RAM random access memory
  • ROM read-only memory
  • EEPROM electrically programmable ROM
  • EEPROM electrically erasable programmable ROM
  • registers hard disk, removable disk, CD-ROM, or any other Any other known storage medium.

Abstract

本申请公开了一种服务器风扇转速控制方法、装置、设备及介质，包括：采集当前时刻服务器的运行状态数据，得到当前运行状态数据；将当前运行状态数据输入至预先创建的转速预测模型，并通过转速预测模型预测与当前运行状态数据对应的风扇转速值得到初始转速预测值；获取与当前时刻相隔预设时长的历史时间段内服务器的风扇转速值得到目标历史转速值；利用目标历史转速值对初始转速预测值进行优化修正，并根据优化后的转速预测值对风扇进行转速控制。本申请基于预先创建的转速预测模型得到初始转速预测值后，利用服务器风扇的历史转速值对初始转速预测值进行优化修正，有效减少了风扇急速调整转速的情况，降低了风扇噪声、功耗并提升了风扇稳定性。

Description

一种服务器风扇转速控制方法、装置、设备及介质
本申请要求在2021年09月14日提交中国专利局、申请号为202111071724.1、发明名称为“一种服务器风扇转速控制方法、装置、设备及介质”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及服务器领域，特别涉及一种服务器风扇转速控制方法、装置、设备及介质。
背景技术
随着信息技术的快速发展，对于服务器的运算能力和数据处理能力提出了更高的要求，服务器在对数据进行处理的过程中会导致服务器内机箱温度升高，当温度过高时可能会影响服务器的正常运转，进而影响用户的体验。
目前服务器散热主要依靠风扇，风扇主要由主板上的BMC(Baseboard Manager Controller，基板管理控制器)或者CPLD(Complex Programmable Logic Device，复杂可编程逻辑器件)直接控制，通过采集主板上温度传感器的温度值，并通过与预设阈值进行比较来判断机箱内温度是否过热，若达到预设阈值，则通过PWM(Pulse Width Modulation，脉宽调制)控制方式来调整风扇的转速。
然而，上述服务器风扇散热方法存在较大的弊端，例如，当服务器开机启动时，主板温度快速跳跃，风扇转速将直接进行大幅度调整，此时风扇噪音较大，功耗较高，以及当机箱内温度达到设定阈值时，风扇会进行急速运转以降低温度，此时同样会产生较大功耗及噪音，从而对电源产生较大影响。综上所述，可以看出目前服务器风扇散热的方法存在噪声大、功耗高、效能低的问题。
发明内容
有鉴于此，本申请的目的在于提供一种服务器风扇转速控制方法、装置、设备及介质，能够有效减少服务器风扇的转速出现急速调整的情况，从而降低了风扇噪声和风扇功耗，提升了风扇的效能，使得风扇运行更加稳定。其具体方案如下：
第一方面，本申请公开了一种服务器风扇转速控制方法，包括：
采集当前时刻服务器的运行状态数据，得到当前运行状态数据；
将所述当前运行状态数据输入至预先创建的转速预测模型，并通过所述转速预测模型预测与所述当前运行状态数据对应的风扇转速值，得到初始转速预测值；
获取与所述当前时刻相隔预设时长的历史时间段内所述服务器的风扇转速值，得到相应的目标历史转速值；
利用所述目标历史转速值对所述初始转速预测值进行优化修正，并根据优化后的转速预测值对所述服务器的风扇进行转速控制。
可选的，所述将所述当前运行状态数据输入至预先创建的转速预测模型，包括：
将所述当前运行状态数据分别输入至预先创建的多个转速预测模型；其中，所述多个转速预测模型为利用训练样本数据分别对预先基于不同模型构建算法构建的多个待训练模型进行训练后得到的模型，所述训练样本数据包括所述服务器的历史运行状态数据以及与所述历史运行状态数据对应的风扇转速值。
可选的，所述服务器风扇转速控制方法，还包括：
在多个第一历史时刻下，分别采集所述服务器的历史运行状态数据以及相应的风扇转速值，以得到包括多个初始样本数据的样本集；其中，所述历史运行状态数据包括所述服务器在所述第一历史时刻下的第一历史机箱温度；
将预设时间间隔分别与多个所述第一历史时刻进行相加，以得到与多个所述第一历史时刻分别对应的多个第二历史时刻，并分别采集所述服务器在多个所述第二历史时刻下的机箱温度，以得到多个第二历史机箱温度；
判断所述第二历史机箱温度与相应的所述第一历史机箱温度之间的差值是否小于预设阈值，如果是则对所述样本集中与所述第一历史机箱温度对应的所述初始样本数据进行归一化，以得到所述训练样本数据，如果否则从所述样本集中剔除所述第一历史机箱温度对应的所述初始样本数据。
可选的，所述采集所述服务器的历史运行状态数据，包括：
采集所述服务器的CPU占用率、内存利用率、硬盘读写速率、电源功率和所述第一历史机箱温度。
可选的，所述将所述当前运行状态数据分别输入至预先创建的多个转速预测模型，包括：
将所述当前运行状态数据分别输入至预先基于XGBoost算法构建的第一转速预测模型、基于支持向量机算法构建的第二转速预测模型和基于人工神经网络算法构建的第三转速预测模型。
可选的，所述通过所述转速预测模型预测与所述当前运行状态数据对应的风扇转速值，得到初始转速预测值，包括：
通过所述多个转速预测模型预测与所述当前运行状态数据对应的风扇转速值，得到相应的多个转速预测值；
利用离群点检测算法从所述多个转速预测值中确定出待剔除转速预测值，然后将所述待剔除转速预测值从所述多个转速预测值中剔除，得到剩余转速预测值；
对所述剩余转速预测值进行平均计算，以得到所述初始转速预测值。
可选的，所述利用所述目标历史转速值对所述初始转速预测值进行优化修正，包括：
对所述目标历史转速值和所述初始转速预测值进行加权平均处理，以完成对所述初始转速预测值的优化修正；其中，所述加权平均处理后的数值为优化后的转速预测值。
第二方面，本申请公开了一种服务器风扇转速控制装置，包括：
数据采集模块，用于采集当前时刻服务器的运行状态数据，得到当前运行状态数据；
预测模块，用于将所述当前运行状态数据输入至预先创建的转速预测模型，并通过所述转速预测模型预测与所述当前运行状态数据对应的风扇转速值，得到初始转速预测值；
数据获取模块，用于获取与所述当前时刻相隔预设时长的历史时间段内所述服务器的风扇转速值，得到相应的目标历史转速值；
转速控制模块，用于利用所述目标历史转速值对所述初始转速预测值进行优化修正，并根据优化后的转速预测值对所述服务器的风扇进行转速控制。
第三方面，本申请公开了一种电子设备，包括处理器和存储器；其中，所述处理器执行所述存储器中保存的计算机程序时实现前述的服务器风扇转速控制方法。
第四方面，本申请公开了一种计算机可读存储介质，用于存储计算机程序；其中，所述计算机程序被处理器执行时实现前述的服务器风扇转速控制方法。
可见，本申请先采集当前时刻服务器的运行状态数据，得到当前运行状态数据，然后将所述当前运行状态数据输入至预先创建的转速预测模型，并通过所述转速预测模型预测与所述当前运行状态数据对应的风扇转速值，得到初始转速预测值，再获取与所述当前时刻相隔预设时长的历史时间段内所述服务器的风扇转速值，得到相应的目标历史转速值，然后利用所述目标历史转速值对所述初始转速预测值进行优化修正，并根据优化后的转速预测值对所述服务器的风扇进行转速控制。可见，本申请基于当前时刻服务器的运行状态数据和预先创建的转速预测模型，对服务器的风扇转速进行预测得到初始转速预测值后，进一步利用与当前时刻相隔预设时长的历史时间段内服务器风扇的历史转速值对上述初始转速预测值进行优化修正，而并非是直接利用上述初始转速预测值对服务器风扇进行转速控制，这样一定程度上能够有效减少后续进行风扇转速调整时出现急速调整的情况，从而降低了风扇噪声和风扇功耗，提升了风扇的效能，使得风扇运行更加稳定。
附图说明
为了更清楚地说明本申请实施例或现有技术中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本申请的实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据提供的附图获得其他的附图。
图1为本申请公开的一种服务器风扇转速控制方法流程图；
图2为本申请公开的一种具体的服务器风扇转速控制方法流程图；
图3为本申请公开的一种训练样本数据的获取方法流程图；
图4为本申请公开的一种具体的服务器风扇转速控制方法流程图；
图5为本申请公开的一种服务器风扇转速控制装置结构示意图；
图6为本申请公开的一种电子设备结构图；
图7为本申请公开的一种计算机可读存储介质结构示意图。
具体实施方式
下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本申请一部分实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本申请保护的范围。
目前，当服务器开机启动时，主板温度快速跳跃，风扇转速将直接进行大幅度调整，此时风扇噪音较大，功耗较高，以及当机箱内温度达到设定阈值时，风扇会进行急速运转以降低温度，此时同样会产生较大功耗及噪音，从而对电源产生较大影响。为此，本申请提供了一种服务器风扇转速控制方案，能够有效减少服务器风扇的转速出现急速调整的情况，从而降低了风扇噪声和风扇功耗，提升了风扇的效能，使得风扇运行更加稳定。
本申请实施例公开了一种服务器风扇转速控制方法，参见图1所示，该方法包括：
步骤S11：采集当前时刻服务器的运行状态数据，得到当前运行状态数据。
本实施例中，通过服务器主板上的BMC可以对当前时刻下的服务器的运行状态数据进行采集，得到所述当前运行状态数据；其中，所述当前运行状态数据包括但不限于CPU占用率、内存利用率、硬盘读写速率、电源功率和机箱温度等。
步骤S12：将所述当前运行状态数据输入至预先创建的转速预测模型，并通过所述转速预测模型预测与所述当前运行状态数据对应的风扇转速值，得到初始转速预测值。
本实施例中，在采集到当前时刻服务器的运行状态数据，得到当前运行状态数据之后，可以将采集到的上述当前运行状态数据输入到预先创建的转速预测模型，上述转速预测模型在获取到上述当前运行状态数据之后，会将输入的数据进行相应的处理，进而输出与上述当前运行状态数据对应的风扇转速预测值，即所述初始转速预测值。需要指出的是，在将所述当前运行状态数据输入到预先创建的转速预测模型的过程中，所述转速预测模型既可以有一个也可以有多个，相应的，当有多个转速预测模型时，输出的风扇转速值也会对应有多个，并将上述多个风扇转速值进行相应的处理，如进行筛选和/或计算，进而得到所述初始转速预测值。
步骤S13：获取与所述当前时刻相隔预设时长的历史时间段内所述服务器的风扇转速值，得到相应的目标历史转速值。
本实施例中，将所述当前运行状态数据输入至预先创建的转速预测模型，并通过所述转速预测模型预测与所述当前运行状态数据对应的风扇转速值，得到初始转速预测值之后，可以通过服务器主板上的BMC对上述当前时刻之前的经过预设时长的历史时间段内的服务器的风扇转速值进行采集，得到与上述历史时间段相对应的风扇历史转速值，即所述目标历史转速值。可以理解的是，本实施例可以通过人工设置参数的方式来确定上述预设时长。
步骤S14：利用所述目标历史转速值对所述初始转速预测值进行优化修正，并根据优化后的转速预测值对所述服务器的风扇进行转速控制。
本实施例中，在获取与所述当前时刻相隔预设时长的历史时间段内所述服务器的风扇转速值，得到相应的目标历史转速值之后，可以利用上述目标历史转速值对上述初始转速预测值进行优化修正，可以理解的是，上述初始转速预测值是根据当前运行状态数据预测出来的，在具体的应用中为防止预测值与风扇的当前转速值偏差相对较大导致风扇转速的激增，可以利用上述目标历史转速值对上述初始转速预测值进行进一步的优化修正。并且可以根据优化后的转速预测值生成相应的转速控制指令，并将上述转速控制指令发送至BMC，BMC在获取到上述转速控制指令之后，可以以PWM的方式控制服务器风扇将当前服务器风扇转速调整至上述目标历史转速值。
可见，本申请实施例先采集当前时刻服务器的运行状态数据，得到当前运行状态数据，然后将所述当前运行状态数据输入至预先创建的转速预测模型，并通过所述转速预测模型预测与所述当前运行状态数据对应的风扇转速值，得到初始转速预测值，再获取与所述当前时刻相隔预设时长的历史时间段内所述服务器的风扇转速值，得到相应的目标历史转速值，然后利用所述目标历史转速值对所述初始转速预测值进行优化修正，并根据优化后的转速预测值对所述服务器的风扇进行转速控制。由上可知，本申请实施例基于当前时刻服务器的运行状态数据和预先创建的转速预测模型，对服务器的风扇转速进行预测得到初始转速预测值后，进一步利用与当前时刻相隔预设时长的历史时间段内服务器风扇的历史转速值对上述初始转速预测值进行优化修正，而并非是直接利用上述初始转速预测值对服务器风扇进行转速控 制，这样一定程度上能够有效减少后续进行风扇转速调整时出现急速调整的情况，从而降低了风扇噪声和风扇功耗，提升了风扇的效能，使得风扇运行更加稳定。
本申请实施例公开了一种具体的服务器风扇转速控制方法，参见图2所示，该方法包括：
步骤S21：采集当前时刻服务器的运行状态数据，得到当前运行状态数据。
步骤S22：将所述当前运行状态数据分别输入至预先创建的多个转速预测模型；其中，所述多个转速预测模型为利用训练样本数据分别对预先基于不同模型构建算法构建的多个待训练模型进行训练后得到的模型，所述训练样本数据包括所述服务器的历史运行状态数据以及与所述历史运行状态数据对应的风扇转速值。
本实施例中，在采集当前时刻服务器的运行状态数据，得到当前运行状态数据之后，具体可以将采集到的上述当前运行状态数据分别输入到预先基于不同模型构建算法创建的多个转速预测模型中。其中，上述多个预测模型为利用包括所述服务器的历史运行状态数据以及与所述历史运行状态数据对应的风扇转速值的训练样本数据，分别对预先基于不同模型构建算法构建的多个待训练模型进行训练后得到的模型；所述模型构建算法包括但不限于XGBoost(Extreme Gradient Boosting，极端梯度提升)算法、支持向量机(SVM，Support Vector Machine)算法、人工神经网络(ANN，Artificial Neural Network)算法等。
需要指出的是，参见图3所示，本实施例中进一步公开了一种上述训练样本数据的获取过程，具体可以包括：
步骤S31：在多个第一历史时刻下，分别采集服务器的历史运行状态数据以及相应的风扇转速值，以得到包括多个初始样本数据的样本集；其中，所述历史运行状态数据包括所述服务器在所述第一历史时刻下的第一历史机箱温度；
步骤S32：将预设时间间隔分别与多个所述第一历史时刻进行相加，以得到与多个所述第一历史时刻分别对应的多个第二历史时刻，并分别采集所述服务器在多个所述第二历史时刻下的机箱温度，以得到多个第二历史机箱温度；
步骤S33：判断所述第二历史机箱温度与相应的所述第一历史机箱温度之间的差值是否小于预设阈值，如果是则对所述样本集中与所述第一历史机箱温度对应的所述初始样本数据进行归一化，以得到所述训练样本数据，如果否则从所述样本集中剔除所述第一历史机箱温度对应的所述初始样本数据。
本实施例中，在获取上述训练样本数据之前，可以以任意多个历史时刻为基准，大规模地采集上述历史时刻下的服务器的历史运行状态数据以及对应历史时刻下的风扇转速值，得到历史采样数据，即所述初始样本数据，进而可以得到包括多个所述初始样本数据的样本集。其中，所述历史运行状态数据种包括所述第一历史时刻下的机箱温度，即所述第一历史机箱温度；需要指出的是，除了上述第一历史机箱温度以外，所述历史运行状态数据包括但不限于所述服务器在所述历史时刻下的CPU占用率、内存利用率、硬盘读写速率、电源功率等。例如，采集t时刻下的历史运行状态数据以及t时刻下的风扇转速值，得到单一样本d t＝[d t1,d t2,d t3,d t4,d t5]和S t，其中d t1、d t2、d t3、d t4、d t5分别对应t时刻下的服务器的CPU占用率、内存利用率、硬盘读写速率、电源功率和机箱温度。另外，本实施例中所述初始样本数据的采样数量可以基于实际应用需求来进行确定，例如可以取为200000。
在多个第一历史时刻下，分别采集所述服务器的历史运行状态数据以及相应的风扇转速值，以得到 包括多个初始样本数据的样本集之后，将按照预设的时间间隔，获取与上述第一历史时刻对应的并经过上述时间间隔延迟后的多个第二历史时刻，并分别对上述第二历史时刻下的服务器的机箱温度进行采集，得到对应的多个第二历史机箱温度。例如，预设时间间隔为a，采集t+a时刻下的服务器的机箱温度，并标记为P t+a，可选的，参数a通常取为10，单位为秒(s)。
进一步的，分别判断上述第二历史机箱温度与上述第一历史机箱温度之间的差值是否小于预设阈值，如果上述第二历史机箱温度与上述第一历史机箱温度之间的差值小于预设阈值，则将上述第一历史机箱温度对应的初始样本数据进行归一化处理，从而得到用于模型训练的训练样本数据。其中，上述归一化的方法包括但不限于线性函数转换法、对数函数转换法、反正切函数转换法等。
需要指出的是，上述预设阈值的值不是固定，而是根据服务器机箱在正常运转状态下的机箱温度及机箱温度过高并达到影响服务器正常运转的温度而自定义的值。例如，将服务器机箱温度正常运转状态的温度取为25度，并且根据实际服务器应用情况，将影响服务器正常运转的温度取为55度，则可以将上述影响服务器正常运转的温度与上述正常运转状态的温度之差作为预设阈值，即预设阈值为30。
在另一种具体的实施例中，如果上述第二历史机箱温度与上述第一历史机箱温度之间的差值不小于预设阈值，则判定为与所述第一历史机箱温度对应的所述初始样本数据不合法，并从所述样本集中将与所述第一历史机箱温度对应的所述初始样本数据剔除。本实施例中，在进行模型训练之前，为了提升模型训练效果，还可以在所有合法的初始样本数据被挑选出来之后，对上述合法样本数据的顺序进行打乱，以得到重新排序后的训练样本数据。
步骤S23：通过所述多个转速预测模型预测与所述当前运行状态数据对应的风扇转速值，得到相应的多个转速预测值。
本实施例中，在将所述当前运行状态数据分别输入至预先创建的多个转速预测模型之后，上述多个转速预测模型会对上述当前运行状态数据进行相应的处理，进而输出多个与上述当前运行状态数据对应的风扇转速预测值，即所述多个转速预测值。其中，所述多个转速预测模型包括但不限于预先基于XGBoost算法构建的第一转速预测模型、基于支持向量机算法构建的第二转速预测模型和基于人工神经网络算法构建的第三转速预测模型等。
步骤S24：利用离群点检测算法从所述多个转速预测值中确定出待剔除转速预测值，然后将所述待剔除转速预测值从所述多个转速预测值中剔除，得到剩余转速预测值。
本实施例中，在通过所述多个转速预测模型预测与所述当前运行状态数据对应的风扇转速值，得到相应的多个转速预测值之后，可以利用离群点检测算法从上述多个转速预测值中确定出与其他预测值偏差较大的数值，即所述待剔除转速预测值，然后将上述与其他预测值偏差较大的数值从上述多个转速预测值中剔除，进而得到剩余的转速预测值。另外，需要指出的是，除了上述离群点检测算法以外，还可以采用其他剔除算法对上述多个转速预测值进行剔除，所述其他剔除算法包括但不限于K-means(k-means clustering algorithm，k均值聚类算法)、KNN(k-Nearest Neighbor，k最近邻)算法及SVM算法等。
步骤S25：对所述剩余转速预测值进行平均计算，以得到所述初始转速预测值。
本实施例中，在利用离群点检测算法从所述多个转速预测值中确定出待剔除转速预测值，然后将所述待剔除转速预测值从所述多个转速预测值中剔除，得到剩余转速预测值之后，可以通过对上述剩余转速预测值进行平均计算的方式确定出转速预测值，即所述初始转速预测值。
步骤S26：获取与所述当前时刻相隔预设时长的历史时间段内所述服务器的风扇转速值，得到相应的目标历史转速值。
步骤S27：对所述目标历史转速值和所述初始转速预测值进行加权平均处理，以完成对所述初始转速预测值的优化修正，并根据优化后的转速预测值对所述服务器的风扇进行转速控制；其中，所述加权平均处理后的数值为优化后的转速预测值。
本实施例中，在获取到与所述当前时刻相隔预设时长的历史时间段内所述服务器的风扇转速值，得到相应的目标历史转速值之后，为了进一步地优化上述初始转速预测值，可以对上述目标历史转速值和上述初始转速预测值进行加权平均处理。具体的，本实施例可以基于实际应用需要为上述目标历史转速值和上述初始转速预测值配置不同的权重。例如，可以将初始转速预测值的权重设置为3/5，将目标历史转速值的权重设置为2/5，并利用上述设置的权重对所述目标历史转速值和所述初始转速预测值进行加权平均处理，并利用加权平均处理后的值对服务器的风扇进行转速控制。当然，本实施例也可以为上述目标历史转速值和上述初始转速预测值配置相同的权重。
其中，关于上述步骤S21、S26更加具体的处理过程可以参考前述实施例中公开的相应内容，在此不再进行赘述。
可见，本申请实施例将采集到的当前运行状态数据分别输入至预先创建的多个转速预测模型，然后利用离群点检测算法从输出的多个转速预测值中剔除偏差较大的预测值，经过计算得到初始转速预测值，并通过获取与所述当前时刻相隔预设时长的历史时间段内所述服务器的风扇转速值对上述初始转速预测值进行优化修正，能够实时的根据当前运行状态数据获取到风扇转速预测值，能够减小风扇运转过程中转速突然变大而产生的噪声，降低风扇功耗，增加风扇效能，风扇转速变化更平滑，延长风扇寿命。
本申请实施例公开了一种具体的服务器风扇转速控制方法，参见图4所示，该方法包括：
步骤S41：采集当前时刻服务器的运行状态数据，得到当前运行状态数据。
步骤S42：将所述当前运行状态数据分别输入至预先基于XGBoost算法构建的第一转速预测模型、基于支持向量机算法构建的第二转速预测模型和基于人工神经网络算法构建的第三转速预测模型。
本实施例中，在采集到当前时刻服务器的运行状态数据，得到当前运行状态数据之后，可以将采集到的上述当前运行状态数据分别输入到预先基于XGBoost算法构建的第一转速预测模型、基于支持向量机算法构建的第二转速预测模型和基于人工神经网络算法构建的第三转速预测模型中。需要指出的是，在预先基于人工神经网络算法构建第三转速预测模型时，可选的，人工神经网络可以使用3层感知机，并且建模方法可以采用反向传播算法。
步骤S43：通过所述第一转速预测模型、所述第二转速预测模型及所述第三转速预测模型预测与所述当前运行状态数据对应的风扇转速值，得到相应的第一转速预测值、第二转速预测值及第三转速预测值。
本实施例中，在将所述当前运行状态数据分别输入至预先基于XGBoost算法构建的第一转速预测模型、基于支持向量机算法构建的第二转速预测模型和基于人工神经网络算法构建的第三转速预测模型之后，通过上述第一转速预测模型、上述第二转速预测模型及上述第三转速预测模型预测与所述当前运行状态数据对应的风扇转速值，可以输出相应的三个转速预测值，即所述第一转速预测值、所述第二转速预测值及所述第三转速预测值，并分别标记为S m1、S m2、S m3。
步骤S44：利用离群点检测算法从所述第一转速预测值、所述第二转速预测值及所述第三转速预测 值中确定出待剔除转速预测值，然后将所述待剔除转速预测值从所述多个转速预测值中剔除，得到剩余转速预测值。
本实施例中，在通过所述第一转速预测模型、所述第二转速预测模型及所述第三转速预测模型预测与所述当前运行状态数据对应的风扇转速值，得到相应的第一转速预测值、第二转速预测值及第三转速预测值，即S m1、S m2、S m3之后，可以利用离群点检测算法从上述S m1、S m2、S m3中确定出待剔除转速预测值，在一种具体实施例中，当S m3被确定为待剔除转速预测值时，则将S m3从上述S m1、S m2、S m3中剔除，得到剩余的转速预测值，即S m1、S m2。
在另一种具体实施例中，若上述S m1、S m2、S m3具有相同的转速预测值，则可以直接对上述三个转速预测值计算平均。
步骤S45：对所述剩余转速预测值进行平均计算，以得到所述初始转速预测值。
本实施例中，在利用离群点检测算法从所述第一转速预测值、所述第二转速预测值及所述第三转速预测值中确定出待剔除转速预测值，然后将所述待剔除转速预测值从所述多个转速预测值中剔除，得到剩余转速预测值之后，可以对上述剩余转速预测值进行平均计算，并将计算结果作为所述初始转速预测值。具体的，例如，通过对剩余的上述S m1、S m2进行平均计算，可以得到平均计算结果，并将其标记为S m。
步骤S46：获取与所述当前时刻相隔预设时长的历史时间段内所述服务器的风扇转速值，得到相应的目标历史转速值。
本实施例中，在对所述剩余转速预测值进行平均计算，得到所述初始转速预测值，即S m之后，可以采集与上述当前时刻相隔预设时长的历史时间段内所述服务器的风扇转速值，得到相应的目标历史转速值。例如，以所述当前采集时刻为基准，采集上述当前时刻之前的经过b时长内的历史风扇转速值，并按照预设的采样周期对上述b时长内的历史风扇转速值进行采集，得到多个历史风扇转速值，即所述目标历史转速值，并标记为V＝{v 1,v 2,......v n}。
步骤S47：利用所述目标历史转速值对所述初始转速预测值进行优化修正，并根据优化后的转速预测值对所述服务器的风扇进行转速控制。
本实施例中，在获取与所述当前时刻相隔预设时长的历史时间段内所述服务器的风扇转速值，得到相应的目标历史转速值之后，可以利用所述目标历史转速值对上述初始转速预测值进行优化修正。例如，通过计算上述V与S m的均值实现对上述初始转速预测值进行优化修正，并利用均值计算后的转速预测值并基于PWM方波对服务器的风扇进行转速控制。
其中，关于上述步骤S41、S47更加具体的处理过程可以参考前述实施例中公开的相应内容，在此不再进行赘述。
可见，本申请实施例将采集到的当前运行状态数据分别输入至预先基于XGBoost算法构建的第一转速预测模型、基于支持向量机算法构建的第二转速预测模型和基于人工神经网络算法构建的第三转速预测模型，然后利用离群点检测算法从输出的多个转速预测值中剔除偏差较大的预测值，经过计算得到初 始转速预测值，并通过获取与所述当前时刻相隔预设时长的历史时间段内所述服务器的风扇转速值对上述初始转速预测值进行优化修正，通过上述方法可以实时的根据服务器当前运行状态数据预测出风扇的转速，能够使风扇转速变化更平稳，减小风扇运转过程中转速突然变大产生的噪声，降低了风扇的功耗，增加了风扇的效能，延长风扇寿命。
相应的，本申请实施例还公开了一种服务器风扇转速控制装置，参见图5所示，该装置包括：
数据采集模块11，用于采集当前时刻服务器的运行状态数据，得到当前运行状态数据；
预测模块12，用于将所述当前运行状态数据输入至预先创建的转速预测模型，并通过所述转速预测模型预测与所述当前运行状态数据对应的风扇转速值，得到初始转速预测值；
数据获取模块13，用于获取与所述当前时刻相隔预设时长的历史时间段内所述服务器的风扇转速值，得到相应的目标历史转速值；
转速控制模块14，用于利用所述目标历史转速值对所述初始转速预测值进行优化修正，并根据优化后的转速预测值对所述服务器的风扇进行转速控制。
其中，关于上述各个模块的具体工作流程可以参考前述实施例中公开的相应内容，在此不再进行赘述。
可见，本申请实施例中，先采集当前时刻服务器的运行状态数据，得到当前运行状态数据，然后将所述当前运行状态数据输入至预先创建的转速预测模型，并通过所述转速预测模型预测与所述当前运行状态数据对应的风扇转速值，得到初始转速预测值，再获取与所述当前时刻相隔预设时长的历史时间段内所述服务器的风扇转速值，得到相应的目标历史转速值，然后利用所述目标历史转速值对所述初始转速预测值进行优化修正，并根据优化后的转速预测值对所述服务器的风扇进行转速控制。可见，本申请实施例基于当前时刻服务器的运行状态数据和预先创建的转速预测模型，对服务器的风扇转速进行预测得到初始转速预测值后，进一步利用与当前时刻相隔预设时长的历史时间段内服务器风扇的历史转速值对上述初始转速预测值进行优化修正，而并非是直接利用上述初始转速预测值对服务器风扇进行转速控制，这样一定程度上能够有效减少后续进行风扇转速调整时出现急速调整的情况，从而降低了风扇噪声和风扇功耗，提升了风扇的效能，使得风扇运行更加稳定。
在一些具体实施例中，所述预测模块12，具体可以包括：
第一输入单元，用于将所述当前运行状态数据分别输入至预先创建的多个转速预测模型；其中，所述多个转速预测模型为利用训练样本数据分别对预先基于不同模型构建算法构建的多个待训练模型进行训练后得到的模型，所述训练样本数据包括所述服务器的历史运行状态数据以及与所述历史运行状态数据对应的风扇转速值。
在一些具体实施例中，所述服务器风扇转速控制装置，具体还可以包括：
第一数据采集单元，用于在多个第一历史时刻下，分别采集所述服务器的历史运行状态数据以及相应的风扇转速值，以得到包括多个初始样本数据的样本集；其中，所述历史运行状态数据包括所述服务器在所述第一历史时刻下的第一历史机箱温度；
温度采集单元，用于将预设时间间隔分别与多个所述第一历史时刻进行相加，以得到与多个所述第一历史时刻分别对应的多个第二历史时刻，并分别采集所述服务器在多个所述第二历史时刻下的机箱温度，以得到多个第二历史机箱温度；
判断单元，用于判断所述第二历史机箱温度与相应的所述第一历史机箱温度之间的差值是否小于预设阈值；
归一化单元，用于如果所述第二历史机箱温度与相应的所述第一历史机箱温度之间的差值小于预设阈值，则对所述样本集中与所述第一历史机箱温度对应的所述初始样本数据进行归一化，以得到所述训练样本数据；
第一剔除单元，用于如果所述第二历史机箱温度与相应的所述第一历史机箱温度之间的差值不小于预设阈值，则从所述样本集中剔除所述第一历史机箱温度对应的所述初始样本数据。
在一些具体实施例中，所述第一数据采集单元，具体可以包括：
第二数据采集单元，用于采集所述服务器的CPU占用率、内存利用率、硬盘读写速率、电源功率和所述第一历史机箱温度。
在一些具体实施例中，所述第一输入单元，具体可以包括：
第二输入单元，用于将所述当前运行状态数据分别输入至预先基于XGBoost算法构建的第一转速预测模型、基于支持向量机算法构建的第二转速预测模型和基于人工神经网络算法构建的第三转速预测模型。
在一些具体实施例中，所述预测模块12，具体可以包括：
预测单元，用于通过所述多个转速预测模型预测与所述当前运行状态数据对应的风扇转速值，得到相应的多个转速预测值；
第二剔除单元，用于利用离群点检测算法从所述多个转速预测值中确定出待剔除转速预测值，然后将所述待剔除转速预测值从所述多个转速预测值中剔除，得到剩余转速预测值；
计算单元，用于对所述剩余转速预测值进行平均计算，以得到所述初始转速预测值。
在一些具体实施例中，所述转速控制模块14，具体可以包括：
修正单元，用于对所述目标历史转速值和所述初始转速预测值进行加权平均处理，以完成对所述初始转速预测值的优化修正；其中，所述加权平均处理后的数值为优化后的转速预测值。
进一步的，本申请实施例还公开了一种电子设备，图6是根据一示例性实施例示出的电子设备20结构图，图中的内容不能认为是对本申请的使用范围的任何限制。
图6为本申请实施例提供的一种电子设备20的结构示意图。该电子设备20，具体可以包括：至少一个处理器21、至少一个存储器22、电源23、通信接口24、输入输出接口25和通信总线26。其中，所述存储器22用于存储计算机程序，所述计算机程序由所述处理器21加载并执行，以实现前述任一实施例公开的服务器风扇转速控制方法中的相关步骤。另外，本实施例中的电子设备20具体可以为电子计算机。
本实施例中，电源23用于为电子设备20上的各硬件设备提供工作电压；通信接口24能够为电子设备20创建与外界设备之间的数据传输通道，其所遵循的通信协议是能够适用于本申请技术方案的任意通信协议，在此不对其进行具体限定；输入输出接口25，用于获取外界输入数据或向外界输出数据，其具体的接口类型可以根据具体应用需要进行选取，在此不进行具体限定。
另外，存储器22作为资源存储的载体，可以是只读存储器、随机存储器、磁盘或者光盘等，其上所存储的资源可以包括操作系统221、计算机程序222等，存储方式可以是短暂存储或者永久存储。
其中，操作系统221用于管理与控制电子设备20上的各硬件设备以及计算机程序222，其可以是 Windows Server、Netware、Unix、Linux等。计算机程序222除了包括能够用于完成前述任一实施例公开的由电子设备20执行的服务器风扇转速控制方法的计算机程序之外，还可以进一步包括能够用于完成其他特定工作的计算机程序。
图7为本申请公开的一种计算机可读存储介质结构示意图。如图7所示，进一步的，本申请还公开了一种计算机可读存储介质601，用于存储计算机程序610；其中，所述计算机程序610被处理器执行时实现前述公开的服务器风扇转速控制方法。关于该方法的具体步骤可以参考前述实施例中公开的相应内容，在此不再进行赘述。
本说明书中各个实施例采用递进的方式描述，每个实施例重点说明的都是与其它实施例的不同之处，各个实施例之间相同或相似部分互相参见即可。对于实施例公开的装置而言，由于其与实施例公开的方法相对应，所以描述的比较简单，相关之处参见方法部分说明即可。
专业人员还可以进一步意识到，结合本文中所公开的实施例描述的各示例的单元及算法步骤，能够以电子硬件、计算机软件或者二者的结合来实现，为了清楚地说明硬件和软件的可互换性，在上述说明中已经按照功能一般性地描述了各示例的组成及步骤。这些功能究竟以硬件还是软件方式来执行，取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能，但是这种实现不应认为超出本申请的范围。
结合本文中所公开的实施例描述的方法或算法的步骤可以直接用硬件、处理器执行的软件模块，或者二者的结合来实施。软件模块可以置于随机存储器(RAM)、内存、只读存储器(ROM)、电可编程ROM、电可擦除可编程ROM、寄存器、硬盘、可移动磁盘、CD-ROM、或技术领域内所公知的任意其它形式的存储介质中。
最后，还需要说明的是，在本文中，诸如第一和第二等之类的关系术语仅仅用来将一个实体或者操作与另一个实体或操作区分开来，而不一定要求或者暗示这些实体或操作之间存在任何这种实际的关系或者顺序。而且，术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、方法、物品或者设备不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、方法、物品或者设备所固有的要素。在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括所述要素的过程、方法、物品或者设备中还存在另外的相同要素。
以上对本申请所提供的一种服务器风扇转速控制方法、装置、设备及介质进行了详细介绍，本文中应用了具体个例对本申请的原理及实施方式进行了阐述，以上实施例的说明只是用于帮助理解本申请的方法及其核心思想；同时，对于本领域的一般技术人员，依据本申请的思想，在具体实施方式及应用范围上均会有改变之处，综上所述，本说明书内容不应理解为对本申请的限制。

Claims (10)

  1. 一种服务器风扇转速控制方法，其特征在于，包括：
    采集当前时刻服务器的运行状态数据，得到当前运行状态数据；
    将所述当前运行状态数据输入至预先创建的转速预测模型，并通过所述转速预测模型预测与所述当前运行状态数据对应的风扇转速值，得到初始转速预测值；
    获取与所述当前时刻相隔预设时长的历史时间段内所述服务器的风扇转速值，得到相应的目标历史转速值；
    利用所述目标历史转速值对所述初始转速预测值进行优化修正，并根据优化后的转速预测值对所述服务器的风扇进行转速控制。
  2. 根据权利要求1所述的服务器风扇转速控制方法，其特征在于，所述将所述当前运行状态数据输入至预先创建的转速预测模型，包括：
    将所述当前运行状态数据分别输入至预先创建的多个转速预测模型；其中，所述多个转速预测模型为利用训练样本数据分别对预先基于不同模型构建算法构建的多个待训练模型进行训练后得到的模型，所述训练样本数据包括所述服务器的历史运行状态数据以及与所述历史运行状态数据对应的风扇转速值。
  3. 根据权利要求2所述的服务器风扇转速控制方法，其特征在于，还包括：
    在多个第一历史时刻下，分别采集所述服务器的历史运行状态数据以及相应的风扇转速值，以得到包括多个初始样本数据的样本集；其中，所述历史运行状态数据包括所述服务器在所述第一历史时刻下的第一历史机箱温度；
    将预设时间间隔分别与多个所述第一历史时刻进行相加，以得到与多个所述第一历史时刻分别对应的多个第二历史时刻，并分别采集所述服务器在多个所述第二历史时刻下的机箱温度，以得到多个第二历史机箱温度；
    判断所述第二历史机箱温度与相应的所述第一历史机箱温度之间的差值是否小于预设阈值，如果是则对所述样本集中与所述第一历史机箱温度对应的所述初始样本数据进行归一化，以得到所述训练样本数据，如果否则从所述样本集中剔除所述第一历史机箱温度对应的所述初始样本数据。
  4. 根据权利要求3所述的服务器风扇转速控制方法，其特征在于，所述采集所述服务器的历史运行状态数据，包括：
    采集所述服务器的CPU占用率、内存利用率、硬盘读写速率、电源功率和所述第一历史机箱温度。
  5. 根据权利要求2所述的服务器风扇转速控制方法，其特征在于，所述将所述当前运行状态数据分别输入至预先创建的多个转速预测模型，包括：
    将所述当前运行状态数据分别输入至预先基于XGBoost算法构建的第一转速预测模型、基于支持向量机算法构建的第二转速预测模型和基于人工神经网络算法构建的第三转速预测模型。
  6. 根据权利要求2所述的服务器风扇转速控制方法，其特征在于，所述通过所述转速预测模型预测与所述当前运行状态数据对应的风扇转速值，得到初始转速预测值，包括：
    通过所述多个转速预测模型预测与所述当前运行状态数据对应的风扇转速值，得到相应的多个转速预测值；
    利用离群点检测算法从所述多个转速预测值中确定出待剔除转速预测值，然后将所述待剔除转速预测值从所述多个转速预测值中剔除，得到剩余转速预测值；
    对所述剩余转速预测值进行平均计算，以得到所述初始转速预测值。
  7. 根据权利要求1至6任一项所述的服务器风扇转速控制方法，其特征在于，所述利用所述目标历史转速值对所述初始转速预测值进行优化修正，包括：
    对所述目标历史转速值和所述初始转速预测值进行加权平均处理，以完成对所述初始转速预测值的优化修正；其中，所述加权平均处理后的数值为优化后的转速预测值。
  8. 一种服务器风扇转速控制装置，其特征在于，包括：
    数据采集模块，用于采集当前时刻服务器的运行状态数据，得到当前运行状态数据；
    预测模块，用于将所述当前运行状态数据输入至预先创建的转速预测模型，并通过所述转速预测模型预测与所述当前运行状态数据对应的风扇转速值，得到初始转速预测值；
    数据获取模块，用于获取与所述当前时刻相隔预设时长的历史时间段内所述服务器的风扇转速值，得到相应的目标历史转速值；
    转速控制模块，用于利用所述目标历史转速值对所述初始转速预测值进行优化修正，并根据优化后的转速预测值对所述服务器的风扇进行转速控制。
  9. 一种电子设备，其特征在于，包括处理器和存储器；其中，所述处理器执行所述存储器中保存的计算机程序时实现如权利要求1至7任一项所述的服务器风扇转速控制方法。
  10. 一种计算机可读存储介质，其特征在于，用于存储计算机程序；其中，所述计算机程序被处理器执行时实现如权利要求1至7任一项所述的服务器风扇转速控制方法。
PCT/CN2022/074611 2021-09-14 2022-01-28 一种服务器风扇转速控制方法、装置、设备及介质 WO2023040165A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202111071724.1 2021-09-14
CN202111071724.1A CN113534936B (zh) 2021-09-14 2021-09-14 一种服务器风扇转速控制方法、装置、设备及介质

Publications (1)

Publication Number Publication Date
WO2023040165A1 true WO2023040165A1 (zh) 2023-03-23

Family

ID=78092479

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/074611 WO2023040165A1 (zh) 2021-09-14 2022-01-28 一种服务器风扇转速控制方法、装置、设备及介质

Country Status (2)

Country Link
CN (1) CN113534936B (zh)
WO (1) WO2023040165A1 (zh)

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116050030A (zh) * 2023-04-03 2023-05-02 亿昇(天津)科技有限公司 一种鼓风机转子轴心位置确定方法、装置及设备
CN116432542A (zh) * 2023-06-12 2023-07-14 国网江西省电力有限公司电力科学研究院 一种基于误差序列修正的开关柜母排温升预警方法及系统
CN116499529A (zh) * 2023-06-25 2023-07-28 北京电科智芯科技有限公司 设备运行状态监测方法、装置、管理终端和存储介质
CN116956471A (zh) * 2023-09-19 2023-10-27 中国空气动力研究与发展中心计算空气动力研究所 大型运输机的气动力预测方法、装置、设备及介质
CN117311159A (zh) * 2023-10-30 2023-12-29 上海新华控制技术集团科技有限公司 控制系统的自适应调节方法、装置、存储介质及电子设备
CN117521423A (zh) * 2024-01-05 2024-02-06 山东鑫林纸制品有限公司 瓦楞纸加工的机器速度智能预测方法、装置及存储介质
CN117529069A (zh) * 2024-01-08 2024-02-06 浙江德塔森特数据技术有限公司 一种数据中心峰谷电应用控制方法以及管控服务器
CN117782647A (zh) * 2024-02-23 2024-03-29 杭州中港地铁装备维护有限公司 一种地铁行走部轴承健康状态监测系统

Families Citing this family (17)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113534936B (zh) * 2021-09-14 2021-12-17 苏州浪潮智能科技有限公司 一种服务器风扇转速控制方法、装置、设备及介质
CN114020121B (zh) * 2021-11-02 2023-11-07 北京百度网讯科技有限公司 控制云手机服务器温度的方法、装置以及设备
CN114546073A (zh) * 2021-12-31 2022-05-27 广东省工业边缘智能创新中心有限公司 服务器的风扇控制方法及其系统、温控系统
CN114320970B (zh) * 2021-12-31 2023-09-22 浪潮(北京)电子信息产业有限公司 一种风扇的异常预警方法、装置、设备及介质
CN114442794B (zh) * 2022-01-20 2023-07-18 苏州浪潮智能科技有限公司 服务器功耗控制方法、系统、终端及存储介质
CN114490275B (zh) * 2022-04-11 2022-07-19 江苏嘉擎信息技术有限公司 一种基于mcu的主机监测控制方法、系统、电子设备及介质
CN114922842B (zh) * 2022-06-28 2023-07-21 武汉烽火技术服务有限公司 风扇调速控制方法、装置、设备及可读存储介质
CN115079738B (zh) * 2022-07-28 2024-01-19 苏州浪潮智能科技有限公司 一种服务器风扇转速调控方法、装置、设备及存储介质
CN115097920B (zh) * 2022-07-29 2023-08-08 苏州浪潮智能科技有限公司 一种调节服务器机箱的温度的方法、装置、服务器及介质
CN115169251B (zh) * 2022-09-06 2022-11-25 新华三人工智能科技有限公司 一种风扇转速控制方法、装置及设备
CN116221160A (zh) * 2023-01-06 2023-06-06 歌尔股份有限公司 风扇噪声调整方法、装置、头戴显示设备及存储介质
CN115915734B (zh) * 2023-03-02 2023-05-23 苏州浪潮智能科技有限公司 风扇运行控制方法、装置、电子设备及存储介质
CN116483180A (zh) * 2023-03-10 2023-07-25 苏州振畅智能科技有限公司 一种主动散热装置的智能控制方法、系统、设备及介质
CN116661574B (zh) * 2023-07-13 2023-12-12 之江实验室 一种计算设备散热的方法、装置、存储介质及电子设备
CN116906362A (zh) * 2023-08-31 2023-10-20 肇庆晟辉电子科技有限公司 一种散热风扇转速控制系统及方法
CN117406844B (zh) * 2023-11-14 2024-04-19 珠海世宁达科技有限公司 一种基于神经网络的显卡风扇控制方法及相关装置
CN117258138B (zh) * 2023-11-21 2024-03-12 安徽通灵仿生科技有限公司 一种心室辅助系统的冲洗设备的控制方法及装置

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9846444B1 (en) * 2017-06-07 2017-12-19 Aic Inc. Method for controlling and adjusting fans of electronic apparatus
CN108194398A (zh) * 2017-12-29 2018-06-22 青岛海信医疗设备股份有限公司 风扇控制方法及装置
CN108691795A (zh) * 2017-04-05 2018-10-23 营邦企业股份有限公司 电子设备的风扇控制与修正方法
CN110134208A (zh) * 2019-04-08 2019-08-16 上海闻泰信息技术有限公司 风扇转速自动调节方法、系统及服务器
CN110725806A (zh) * 2018-07-16 2020-01-24 研祥智能科技股份有限公司 风扇转速控制方法、装置及服务器
CN113534936A (zh) * 2021-09-14 2021-10-22 苏州浪潮智能科技有限公司 一种服务器风扇转速控制方法、装置、设备及介质

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104863881B (zh) * 2014-02-21 2018-10-23 南京中兴新软件有限责任公司 电子设备中风扇转速的控制方法及装置
WO2018047238A1 (ja) * 2016-09-06 2018-03-15 三菱電機株式会社 冷凍サイクル装置
CN106837838A (zh) * 2016-12-29 2017-06-13 郑州云海信息技术有限公司 服务器风扇转速的控制系统及方法
CN112393390B (zh) * 2020-07-15 2021-07-06 上海有孚智数云创数字科技有限公司 云计算数据中心精密空调节能控制方法
CN112395046B (zh) * 2020-07-30 2021-06-04 上海有孚智数云创数字科技有限公司 虚拟机迁移规划调度方法及其系统与介质

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108691795A (zh) * 2017-04-05 2018-10-23 营邦企业股份有限公司 电子设备的风扇控制与修正方法
US9846444B1 (en) * 2017-06-07 2017-12-19 Aic Inc. Method for controlling and adjusting fans of electronic apparatus
CN108194398A (zh) * 2017-12-29 2018-06-22 青岛海信医疗设备股份有限公司 风扇控制方法及装置
CN110725806A (zh) * 2018-07-16 2020-01-24 研祥智能科技股份有限公司 风扇转速控制方法、装置及服务器
CN110134208A (zh) * 2019-04-08 2019-08-16 上海闻泰信息技术有限公司 风扇转速自动调节方法、系统及服务器
CN113534936A (zh) * 2021-09-14 2021-10-22 苏州浪潮智能科技有限公司 一种服务器风扇转速控制方法、装置、设备及介质

Cited By (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116050030A (zh) * 2023-04-03 2023-05-02 亿昇(天津)科技有限公司 一种鼓风机转子轴心位置确定方法、装置及设备
CN116050030B (zh) * 2023-04-03 2023-07-28 亿昇(天津)科技有限公司 一种鼓风机转子轴心位置确定方法、装置及设备
CN116432542A (zh) * 2023-06-12 2023-07-14 国网江西省电力有限公司电力科学研究院 一种基于误差序列修正的开关柜母排温升预警方法及系统
CN116432542B (zh) * 2023-06-12 2023-10-20 国网江西省电力有限公司电力科学研究院 一种基于误差序列修正的开关柜母排温升预警方法及系统
CN116499529A (zh) * 2023-06-25 2023-07-28 北京电科智芯科技有限公司 设备运行状态监测方法、装置、管理终端和存储介质
CN116499529B (zh) * 2023-06-25 2023-09-22 北京电科智芯科技有限公司 设备运行状态监测方法、装置、管理终端和存储介质
CN116956471A (zh) * 2023-09-19 2023-10-27 中国空气动力研究与发展中心计算空气动力研究所 大型运输机的气动力预测方法、装置、设备及介质
CN116956471B (zh) * 2023-09-19 2024-01-12 中国空气动力研究与发展中心计算空气动力研究所 大型运输机的气动力预测方法、装置、设备及介质
CN117311159A (zh) * 2023-10-30 2023-12-29 上海新华控制技术集团科技有限公司 控制系统的自适应调节方法、装置、存储介质及电子设备
CN117521423A (zh) * 2024-01-05 2024-02-06 山东鑫林纸制品有限公司 瓦楞纸加工的机器速度智能预测方法、装置及存储介质
CN117521423B (zh) * 2024-01-05 2024-04-05 山东鑫林纸制品有限公司 瓦楞纸加工的机器速度智能预测方法、装置及存储介质
CN117529069A (zh) * 2024-01-08 2024-02-06 浙江德塔森特数据技术有限公司 一种数据中心峰谷电应用控制方法以及管控服务器
CN117529069B (zh) * 2024-01-08 2024-04-05 浙江德塔森特数据技术有限公司 一种数据中心峰谷电应用控制方法以及管控服务器
CN117782647A (zh) * 2024-02-23 2024-03-29 杭州中港地铁装备维护有限公司 一种地铁行走部轴承健康状态监测系统

Also Published As

Publication number Publication date
CN113534936B (zh) 2021-12-17
CN113534936A (zh) 2021-10-22

Similar Documents

Publication Publication Date Title
WO2023040165A1 (zh) 一种服务器风扇转速控制方法、装置、设备及介质
EP3274827B1 (en) Technologies for offloading and on-loading data for processor/coprocessor arrangements
US9557792B1 (en) Datacenter power management optimizations
US7792597B2 (en) Control systems and method using a shared component actuator
US20030023885A1 (en) Automated power management system for a network of computers
KR101990411B1 (ko) 클라우드 시스템에서의 우선 순위 기반 자원 스케일링 시스템, 장치 및 방법
TW202001101A (zh) 動態溫度控制設備與動態溫度控制方法
CN115915734B (zh) 风扇运行控制方法、装置、电子设备及存储介质
CN109324679A (zh) 一种服务器能耗控制方法及装置
CN111290831A (zh) 一种云计算基于强化学习的虚拟机迁移方法
CN116560475A (zh) 一种服务器风扇控制方法及计算机设备
CN114330099A (zh) 一种网卡功耗调整方法、装置、设备及可读存储介质
JP4782082B2 (ja) パケット処理装置、方法、およびプログラム
Hammami et al. On-policy vs. off-policy deep reinforcement learning for resource allocation in open radio access network
US20230112534A1 (en) Artificial intelligence planning method and real-time radio access network intelligence controller
CN108241526B (zh) 基于模式预测的自适应控制策略
Moghaddam et al. Dynamic energy optimization in chip multiprocessors using deep neural networks
US8521855B2 (en) Centralized server-directed power management in a distributed computing system
Aghasi et al. A decentralized adaptation of model-free Q-learning for thermal-aware energy-efficient virtual machine placement in cloud data centers
US7925873B2 (en) Method and apparatus for controlling operating parameters in a computer system
US20170208121A1 (en) Server pool management
KR102273908B1 (ko) 산업용 빅데이터 서버를 위한 냉각 시스템
CN117193500A (zh) 基于lstm的服务器动态管理方法、系统、终端及存储介质
Zhao et al. Gradient-based adaptive particle swarm optimizer with improved extremal optimization
CN115263793B (zh) 一种服务器风扇的调速方法、调速装置、设备及介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22868545

Country of ref document: EP

Kind code of ref document: A1