Fan speed regulation method, device, electronic equipment and storage medium

Abstract

The embodiment of the present application provides a fan speed regulation method, device, electronic device and storage medium, which relates to the field of computer technology. The method includes: periodically obtaining the current system status data of the electronic device; using a greedy algorithm to process the current system status data to obtain a first optimal speed regulation parameter; and adjusting the speed of the fan in the electronic device according to the first optimal speed regulation parameter. The solution can reduce the complexity of system debugging, and reduce the difficulty, time and manpower consumed in determining the optimal speed regulation parameter.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a fan speed regulation method, device, electronic device and storage medium. Background Art

[0002] In high-load data centers, exploring intelligent and refined strategies for fan speed regulation is of great practical significance. In addition, as the computing demands of data centers such as artificial intelligence (AI) servers and related computing devices, switching devices, and storage devices continue to increase, how to effectively cool electronic equipment in data centers has become a key issue.

[0003] At present, the fan speed regulation method used by electronic equipment in data centers has increased the demand for joint optimization of speed regulation parameters, increased the complexity of system debugging, and required more time for initial configuration and testing. Finding the optimal speed regulation parameters is difficult, time-consuming, and labor-intensive. Summary of the invention

[0004] The purpose of the embodiments of the present application is to provide a fan speed regulation method, device, electronic device and storage medium to reduce the complexity of system debugging and reduce the difficulty, time and manpower consumed in determining the optimal speed regulation parameters. The specific technical solution is as follows:

[0005] In a first aspect, an embodiment of the present application provides a fan speed regulation method, the method comprising:

[0006] Periodically obtain current system status data of the electronic device;

[0007] Use a greedy algorithm to process the current system state data to obtain the first optimal speed regulation parameter;

[0008] The rotation speed of the fan in the electronic device is adjusted according to the first optimal speed adjustment parameter.

[0009] In some embodiments, the method further comprises:

[0010] When the external factors of the electronic device change, the step of using the greedy algorithm to process the current system state data to obtain the first optimal speed regulation parameter is re-executed.

[0011] In some embodiments, the external factors include speed control parameters; the method further includes:

[0012] At each preset time interval, the current system state data is processed using a preset reinforcement learning model to obtain the second optimal speed regulation parameter;

[0013] The rotation speed of the fan in the electronic device is adjusted according to the second optimal speed adjustment parameter.

[0014] In some embodiments, the playback buffer area of ​​the electronic device stores historical system status data; the preset reinforcement learning model is trained using the following steps:

[0015] Acquire historical system status data stored in the replay buffer;

[0016] Inputting the historical system state data into a preset reinforcement learning model to obtain predicted system state data;

[0017] Determine a model loss value using the reward value corresponding to the historical system state data and the reward value corresponding to the predicted system state data;

[0018] With the goal of minimizing the model loss value, the parameters of the preset reinforcement learning model are optimized, and the step of inputting the historical system state data into the preset reinforcement learning model to obtain the predicted system state data is iteratively executed until the maximum cumulative reward value is obtained.

[0019] In some embodiments, the method further comprises:

[0020] When the amount of the historical system status data updated and stored in the replay buffer reaches a preset threshold, the step of obtaining the historical system status data stored in the replay buffer is performed again.

[0021] In some embodiments, the reward value is obtained by weighted average of the total power consumption of the system, the noise generated by the fan, and the outlet temperature of the electronic device.

[0022] In some embodiments, the method further comprises:

[0023] After the electronic device is initialized, a proportional-integral-differential (PID) algorithm is used to process the current system state data to obtain a predicted speed control parameter;

[0024] adjusting the speed of the fan in the electronic device according to the predicted speed regulation parameter;

[0025] When the temperature of each temperature control point in the electronic device reaches stability, the step of using a greedy algorithm to process the current system state data to obtain a first optimal speed regulation parameter is executed.

[0026] In some embodiments, the method further comprises:

[0027] In the process of using a greedy algorithm to process the current system state data to obtain the first optimal speed regulation parameter, if the temperature of the first temperature control point in the electronic device is greater than the preset temperature, the step of using the PID algorithm to process the current system state data to obtain the predicted speed regulation parameter is executed.

[0028] In a second aspect, an embodiment of the present application provides a fan speed regulating device, the device comprising:

[0029] An acquisition module, used for periodically acquiring current system status data of the electronic device;

[0030] A first prediction module is used to process the current system state data using a greedy algorithm to obtain a first optimal speed regulation parameter;

[0031] The first speed adjustment module is used to adjust the speed of the fan in the electronic device according to the first optimal speed adjustment parameter.

[0032] In some embodiments, the first prediction module is further used to re-use the greedy algorithm to process the current system state data when external factors of the electronic device change, so as to obtain the first optimal speed regulation parameter.

[0033] In some embodiments, the external factors include speed control parameters; the device further includes:

[0034] The second prediction module is used to process the current system state data using a preset reinforcement learning model at each interval of a preset duration to obtain a second optimal speed regulation parameter;

[0035] The second speed adjustment module is used to adjust the speed of the fan in the electronic device according to the second optimal speed adjustment parameter.

[0036] In some embodiments, the playback buffer area of ​​the electronic device stores historical system status data; the device also includes a training module for training a preset reinforcement learning model, specifically for:

[0037] Acquire the historical system state data stored in the replay buffer; input the historical system state data into a preset reinforcement learning model to obtain predicted system state data; determine the model loss value using the reward value corresponding to the historical system state data and the reward value corresponding to the predicted system state data; optimize the parameters of the preset reinforcement learning model with the goal of minimizing the model loss value, and iteratively execute the step of inputting the historical system state data into the preset reinforcement learning model to obtain predicted system state data until the maximum cumulative reward value is obtained.

[0038] In some embodiments, the training module is further used to reacquire the historical system status data stored in the replay buffer when the amount of historical system status data updated and stored in the replay buffer reaches a preset threshold value.

[0039] In some embodiments, the reward value is obtained by weighted average of the total power consumption of the system, the noise generated by the fan, and the outlet temperature of the electronic device.

[0040] In some embodiments, the apparatus further comprises:

[0041] A third prediction module is used to process the current system state data using a PID algorithm after the electronic device is initialized to obtain a predicted speed regulation parameter;

[0042] A third speed adjustment module, used to adjust the speed of the fan in the electronic device according to the predicted speed adjustment parameter;

[0043] The first prediction module is also used to use a greedy algorithm to process current system state data to obtain a first optimal speed regulation parameter when the temperature of each temperature control point in the electronic device reaches stability.

[0044] In some embodiments, the third prediction module is also used to use a PID algorithm to process the current system status data to obtain a predicted speed control parameter when the temperature of the first temperature control point in the electronic device is greater than a preset temperature during the process of using a greedy algorithm to process the current system status data to obtain a first optimal speed control parameter.

[0045] In a third aspect, an embodiment of the present application provides an electronic device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus;

[0046] Memory, used to store computer programs;

[0047] The processor is used to implement any method provided in the first aspect when executing a program stored in the memory.

[0048] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, any one of the methods provided in the first aspect is implemented.

[0049] In a fifth aspect, an embodiment of the present application further provides a computer program product comprising instructions, which, when executed on a computer, enables the computer to execute any of the methods provided in the first aspect.

[0050] Beneficial effects of the embodiments of the present application:

[0051] In the technical solution provided by the embodiment of the present application, a greedy algorithm is used to process the current system state data. The greedy algorithm can intelligently find the optimal speed regulation parameters with the lowest total system power consumption or the lowest total system power consumption and noise as the optimization goal, and then configure the speed of each fan in the electronic device. During the entire speed regulation process, the optimal speed regulation parameters can be obtained without manual debugging, which reduces the complexity of system debugging, and reduces the difficulty, time and manpower consumed in determining the optimal speed regulation parameters.

[0052] Of course, implementing any product or method of the present application does not necessarily require achieving all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application, and for ordinary technicians in this field, other embodiments can also be obtained based on these drawings.

[0054] Figure 1 A schematic diagram of a speed regulation curve of linear speed regulation + PID speed regulation in the prior art;

[0055] Figure 2 A first schematic diagram of a fan speed regulation method provided in an embodiment of the present application;

[0056] Figure 3 A schematic diagram of obtaining system status data provided in an embodiment of the present application;

[0057] Figure 4 A second schematic diagram of the fan speed regulation method provided in an embodiment of the present application;

[0058] Figure 5 A schematic diagram of a training method for a preset reinforcement learning model provided in an embodiment of the present application;

[0059] Figure 6 A third schematic diagram of the fan speed regulation method provided in an embodiment of the present application;

[0060] Figure 7 A schematic diagram of the overall architecture provided for an embodiment of the present application;

[0061] Figure 8 A schematic diagram of a speed regulation principle diagram provided in an embodiment of the present application;

[0062] Fig. 9 A schematic diagram of a fan speed regulating device provided in an embodiment of the present application;

[0063] Fig.10A schematic diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0064] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field based on the present application belong to the scope of protection of the present application.

[0065] In existing data centers, whether it is a platform that hosts online applications or a backend server that supports offline analysis, these electronic devices generate a lot of heat when processing Internet traffic and data analysis. Thousands of electronic devices not only need to effectively process computing tasks, but also properly manage the generated heat to ensure the efficient operation of the cooling system. Therefore, it is of great significance to study and optimize the fan speed control strategy, which is specifically reflected in the following aspects:

[0066] (1) Energy saving and efficiency improvement: Dynamically adjust the fan speed based on real-time heat distribution and workload changes to accurately match cooling requirements and significantly save energy consumption. When the data center load is low or the heat is evenly distributed, the fan can run at a lower speed to reduce unnecessary cooling power consumption.

[0067] (2) Extending the life of the equipment: Continuous high-speed operation of the fan will cause accelerated wear of the equipment. Adjusting the fan speed according to actual needs can reduce the degree of hardware wear and thus extend the service life of the equipment.

[0068] (3) Improve cooling efficiency: Dynamic adjustment of fan speed can optimize the cooling effect, especially when the temperature rises in a local area. By increasing the fan speed, the cooling of the specific area can be enhanced, effectively maintaining the temperature balance of the entire data center.

[0069] (4) Reduce noise pollution: Considering the noise generated by high-speed operation of the fan, appropriately reducing the fan speed can significantly improve the noise level of the working environment and improve the working comfort of personnel.

[0070] (5) Optimize performance: Excessive cooling not only consumes resources, but may also affect device performance. Too low a temperature may cause hardware efficiency to decline or condensation problems. Optimizing fan speed control strategies to avoid extreme temperature conditions will help maintain the device in optimal operating condition.

[0071] In general, optimizing the fan speed control strategy can not only improve the overall energy efficiency of the data center, but also extend the life of the equipment, reduce noise pollution and optimize system performance. Therefore, in high-load data centers, exploring the intelligent and refined strategies of fan speed control has extremely important practical significance. In addition, with the increasing computing demand of data centers such as artificial intelligence (AI) servers and related computing devices, switching devices, and storage devices, how to effectively cool the electronic equipment in the data center has become a key issue.

[0072] Currently, electronic equipment in data centers mainly use the following three methods to adjust fan speed.

[0073] Mode 1, linear speed regulation.

[0074] In mode 1, the electronic device adjusts the fan speed according to the temperature change, wherein the temperature and the fan speed maintain a linear relationship, that is, the fan speed increases proportionally when the temperature rises, and the fan speed decreases proportionally when the temperature drops. The linear speed regulation method is suitable for scenarios where the load changes little and the precision of the speed regulation response is not high, such as scenarios where the temperature changes relatively smoothly.

[0075] This method is easy to understand and implement, does not require complex algorithms, and temperature changes are immediately reflected in the fan speed. The control is direct, and has the advantages of simple implementation and fast response. However, this method has limited adaptability to temperature fluctuations, weak precision control, and may cause fan speed oscillation when the load changes drastically, reducing cooling efficiency. It has low accuracy and may cause oscillation problems.

[0076] Mode 2: PID speed regulation.

[0077] PID speed regulation is a feedback control technology. In method 2, the electronic device uses the PID algorithm to comprehensively consider the current temperature, past temperature trends and temperature change speed to adjust the fan speed to achieve precise temperature control. The PID speed regulation method is suitable for scenarios with rapid load fluctuations and the need for precise temperature control, such as server-intensive data centers and scenarios that require strict temperature management.

[0078] This method can effectively suppress temperature fluctuations, so that each temperature control point of the electronic device is maintained at the target temperature, and through the integration and differentiation links, it can avoid oscillation and long-term temperature deviation, and has the advantages of high adjustment accuracy and strong stability. However, this method requires adjusting the PID parameters to adapt to specific scenarios, the debugging process is more complicated, and more temperature information is recorded, which has the problem of complex implementation.

[0079] Mode 3: linear speed regulation + PID speed regulation.

[0080] In mode 3, the electronic device is pre-configured with temperature information and speed information; the temperature information includes the set point (SP) corresponding to the PID speed regulation and the minimum temperature T corresponding to the linear speed regulation. low and the maximum temperature T high , the dead zone temperature T corresponding to PID speed regulation dead ; Speed ​​information includes the fan's minimum speed N low , Intermediate speed N mid and maximum speed N high .

[0081] When the fan speed is N low ~N mid The actual temperature T of the temperature control point is less than T low When the fan speed is adjusted, the electronic device uses the PID algorithm to adjust the fan speed; in other cases, the electronic device uses linear speed regulation. For specific speed regulation curves, see Figure 1 As shown:

[0082] When the fan speed is N low ~N mid The actual temperature T of the temperature control point is less than T low When the actual temperature T of the temperature control point is less than SP-T dead , the electronic device will reduce the fan speed by 1%; if the actual temperature T of the temperature control point is within SP-T dead ~SP+T dead If the actual temperature T of the temperature control point is greater than SP+T dead , the electronic device increases the fan speed by 2%.

[0083] When the fan speed is greater than or equal to N mid , the actual temperature T of the temperature control point is greater than or equal to T low When the fan is turned on, the electronic device uses a linear speed control algorithm to adjust the speed of the fan.

[0084] In method 3, the linear speed control method and the PID speed control method are combined to fully utilize the advantages of both in different scenarios to achieve a more flexible and efficient fan speed control system. However, this method increases the demand for joint optimization of linear and PID speed control parameters, increases the complexity of system debugging, requires more time for initial configuration and testing, and is difficult, time-consuming, and labor-intensive to find the optimal speed control parameters.

[0085] To solve the above problems, the present application provides a fan speed control method, such as Figure 2 As shown, the method includes:

[0086] Step S201, periodically obtaining current system status data of the electronic device;

[0087] Step S202, using a greedy algorithm to process the current system state data to obtain a first optimal speed regulation parameter;

[0088] Step S203: adjusting the speed of the fan in the electronic device according to the first optimal speed adjustment parameter.

[0089] In the technical solution provided by the embodiment of the present application, a greedy algorithm is used to process the current system state data. The greedy algorithm can intelligently find the optimal speed regulation parameters with the lowest total system power consumption or the lowest total system power consumption and noise as the optimization goal, and then configure the speed of each fan in the electronic device. During the entire speed regulation process, the optimal speed regulation parameters can be obtained without manual debugging, which reduces the complexity of system debugging, and reduces the difficulty, time and manpower consumed in determining the optimal speed regulation parameters.

[0090] For ease of description, the following description is made with an electronic device as the execution subject. The electronic device may be an AI server or related computing device, switching device, storage device, etc. in a data center.

[0091] In the above step S201, the electronic device may pre-configure the cycle duration, and the cycle duration may be set according to actual needs. For example, the cycle duration may be 30 seconds, 1 minute, 2 minutes, etc., and this is not limited.

[0092] System status data may include, but are not limited to, the temperature of each temperature control point in the system, the speed of each fan, the noise generated by the fan, the total power consumption of the system, the system load, and the power consumption of the specified device. The temperature control points include the inlet and outlet of the electronic device, and the specified devices include the fan, the central processing unit (CPU), the memory, and the hard disk. The speed of the fan can be expressed in gears or duty cycles, where the duty cycle is the ratio of the high level time of a pulse width modulation (PWM) signal to the total cycle time.

[0093] The electronic device periodically obtains the system status data of the electronic device, that is, the current system status data, according to the period length. Figure 3 The schematic diagram of system status data acquisition shown in the figure has a cycle length of △t 1 , from t 0 Starting from time, every interval △t 1 , the electronic device obtains system status data once, such as Figure 3 t in 0 ~t 4 Data 0 to data 4 acquired at time t 0 At time t, data 0 is the current system status data. 1At time t, data 1 is the current system status data. 2 At time t, data 2 is the current system status data. 3 At time t, data 3 is the current system status data. 4 At this moment, data 4 is the current system status data, and so on.

[0094] After acquiring the system status data, the electronic device may store the acquired system status data in the replay buffer area, that is, the replay buffer area stores the historical system status data.

[0095] In the above step S202, the electronic device is pre-configured with an optimization target of a greedy algorithm (such as a Greedy algorithm), such as minimizing the total system power consumption, or obtaining a weighted average of the total system power consumption, the noise generated by the fan, and the outlet temperature of the electronic device.

[0096] The electronic device uses a greedy algorithm to process the periodically acquired system status data to find the optimal speed regulation parameter, i.e., the first optimal speed regulation parameter. The electronic device may include multiple partitions, each partition including one or more fans. Accordingly, the speed regulation mode of the electronic device includes a partition speed regulation mode and an overall speed regulation mode. Taking the optimization goal of the greedy algorithm as the lowest total system power consumption as an example, the electronic device adopts the partition speed regulation mode and the overall speed regulation mode to implement the above step S202 as follows:

[0097] (1) Partition speed regulation mode:

[0098] The electronic device randomly selects a partition that has not been speed-regulated, determines the rotation speed of each fan in the partition at the next moment, and the rotation speed at the next moment is less than the rotation speed at the current moment; adjusts the rotation speed of each fan in the partition to the rotation speed of each fan in the partition at the next moment, that is, reduces the rotation speed of each fan in the partition; after adjusting the rotation speed, obtains the total power consumption of the system, and if the total power consumption of the system is reduced, continues to reduce the rotation speed of each fan in the partition;

[0099] After adjusting the rotation speed, the total power consumption of the system is obtained. If the total power consumption of the system increases, the rotation speed of each fan in the partition is maintained unchanged. At this time, if the total power consumption of the system still increases, the rotation speed of each fan in the partition is increased until the total power consumption of the system decreases. After that, the electronic device re-executes the step of randomly selecting a partition without speed adjustment and adjusts the rotation speed of each fan in the next partition.

[0100] After the speed adjustment of the fans of all partitions is completed, the electronic device can reset the fans of all partitions to fans without speed adjustment, and re-execute the step of randomly selecting a partition without speed adjustment. After the speed adjustment of the fans of all partitions is completed for a preset number of times in this cycle, or the total power consumption of the system reaches the minimum, the electronic device can count the speeds of each fan when the total power consumption of the system is the lowest, that is, the first optimal speed adjustment parameter; send the first optimal speed adjustment parameter to the system, that is, execute step S203, and adjust the speed of the fan in the electronic device to the first optimal speed adjustment parameter.

[0101] (2) Overall speed regulation mode:

[0102] The electronic device determines the rotation speed of each fan at a next moment, and the rotation speed at the next moment is less than the rotation speed at the current moment; adjusts the rotation speed of each fan to the rotation speed of each fan at the next moment, that is, reduces the rotation speed of each fan; after adjusting the rotation speed, obtains the total power consumption of the system, and if the total power consumption of the system is reduced, continues to reduce the rotation speed of each fan;

[0103] After adjusting the rotation speed, the total power consumption of the system is obtained. If the total power consumption of the system increases, the rotation speed of each fan is maintained unchanged. At this time, if the total power consumption of the system still increases, the rotation speed of each fan is increased until the total power consumption of the system decreases. At this point, a fan speed adjustment is completed.

[0104] After that, the electronic device re-executes the step of determining the rotation speed of each fan at the next moment. After the preset number of rotation speed adjustments of all fans are completed in this cycle, or the total power consumption of the system reaches the minimum, the electronic device can count the rotation speed of each fan when the total power consumption of the system is the lowest, that is, the first optimal speed adjustment parameter; send the first optimal speed adjustment parameter to the system, that is, execute step S203, and adjust the rotation speed of the fan in the electronic device to the first optimal speed adjustment parameter.

[0105] In the embodiment of the present application, the periodic duration for the electronic device to obtain the current system status data and the interval duration for the greedy algorithm to process the system status data may be the same or different.

[0106] Still Figure 3 Take the example to illustrate. The time interval for the greedy algorithm to process system status data is △t 2 △t 2 =2×△t 1 That is, from t 1 Starting from time, every interval △t 2 ,The electronic device uses a greedy algorithm to process the system status data once to obtain the first optimal speed regulation parameter, such as Figure 3 t in 0 ~t 4 The system status data obtained at the moment, that is, data 0 to data 4, at t 0At time t, the electronic device uses the greedy algorithm to process data 0, obtains the speed control parameters, and configures the forwarding of each fan; at t 2 At time t, the electronic device uses the greedy algorithm to process data 2, obtains the speed control parameters, and configures the forwarding of each fan; at t 4 At this moment, the electronic device processes data 4 using a greedy algorithm, obtains a speed control parameter, and configures the forwarding of each fan, and so on.

[0107] In some embodiments, the electronic device can monitor external factors in real time, and the external factors may include the load, environment, power, power consumption, configuration, temperature, etc. of the electronic device. When the external factors of the electronic device are monitored to change, the electronic device re-executes step S202, that is, uses the greedy algorithm to process the current system state data to obtain the first optimal speed regulation parameter.

[0108] In the embodiment of the present application, the change of external factors can be understood as the difference of the external factors at two adjacent moments is greater than the preset difference value, or the external factors exceed the preset range. For example, if the temperature difference of the temperature control point at two adjacent moments is greater than the preset temperature difference value (i.e., the preset difference value), or the temperature of the temperature control point exceeds the preset temperature range (i.e., the preset range), it is determined that the external factors have changed.

[0109] When external factors change, it means that the load, environment, power, power consumption, configuration, or temperature of the electronic device has changed, or the fan has aged due to long-term operation, etc. The electronic device re-executes step S202 and can adjust the speed of each fan in time so that the temperature of the electronic device reaches the temperature to avoid temperature imbalance of the electronic device.

[0110] In some embodiments, the external factors may include a speed control parameter, i.e., the speed of the fan. Figure 4 As shown, the embodiment of the present application also provides a fan speed regulation method, which may include the following steps.

[0111] Step S401, periodically obtaining current system status data of the electronic device;

[0112] Step S402, processing the current system state data using a preset reinforcement learning model at each preset time interval to obtain a second optimal speed regulation parameter;

[0113] Step S403, adjusting the speed of the fan in the electronic device according to the second optimal speed adjustment parameter;

[0114] Step S404, when the external factors of the electronic device change, the current system state data is processed using a greedy algorithm to obtain a first optimal speed regulation parameter;

[0115] Step S405: adjusting the speed of the fan in the electronic device according to the first optimal speed adjustment parameter.

[0116] In the technical solution provided in the embodiment of the present application, a preset reinforcement learning model is used to perform inference once every preset time interval to adjust the rotation speed of each fan in the electronic device, so as to introduce disturbance factors into the greedy algorithm so that the greedy algorithm can run to obtain the optimal speed regulation parameters for the system as a whole, thereby preventing the greedy algorithm from stopping running after only obtaining the local optimal speed regulation parameters, thereby minimizing the total power consumption of the system and achieving energy saving and noise reduction requirements.

[0117] The above step S401 is the same as the above step S201 and will not be described again here.

[0118] In the above step S402, the preset duration can be set according to actual needs. The preset duration can be the same as the period duration of obtaining the current system status data in step S401, or it can be different.

[0119] The reinforcement learning model is pre-configured in the electronic device, that is, the preset reinforcement learning model. The preset reinforcement learning model can be a deep reinforcement learning (DRL) model or other reinforcement learning models, such as a deep deterministic policy gradient (DDPG) model, etc., and there is no limitation on this. The preset reinforcement learning model is trained using historical system state data. The training process is as follows: Figure 5 shown.

[0120] Step S501, obtaining historical system status data stored in the replay buffer;

[0121] In the embodiment of the present application, a replay buffer (Replay Buffer / ReplayMemory) is set in the electronic device.

[0122] The electronic device can store the periodically acquired system status data of the electronic device in a replay buffer (Replay Buffer / Replay Memory) to increase the amount of training data of the preset reinforcement learning model, that is, to increase the update data amount of the speed regulation model, thereby improving the accuracy of the trained speed regulation model.

[0123] The electronic device can also store periodically acquired changing system state data of the electronic device into a replay buffer. The changing system state data is system state data acquired in the process of executing the greedy algorithm to obtain the first optimal speed regulation parameter. In the embodiment of the present application, the changing system state data is an important factor in determining the speed regulation strategy. The replay buffer only stores the changing system state data, which can greatly save the space of the replay buffer, and, while ensuring the accuracy of the preset reinforcement learning model, reduces the amount of training data for training the preset reinforcement learning model, thereby improving the training efficiency of the preset reinforcement learning model.

[0124] In the embodiment of the present application, the electronic device can obtain the system status data updated and stored in the replay buffer at preset intervals, that is, the historical system status data, and can also obtain the historical system status data updated and stored in the replay buffer after the amount of system status data updated and stored in the replay buffer reaches a preset threshold value, so as to continuously optimize the preset reinforcement learning model, so that the preset reinforcement learning model meets the system requirements, and accurately assists the greedy algorithm to find the optimal speed regulation parameters. The preset time and the preset threshold value can be set according to actual needs.

[0125] Step S502, inputting historical system state data into a preset reinforcement learning model to obtain predicted system state data;

[0126] In an embodiment of the present application, the electronic device inputs multiple system state numbers (i.e., historical system state numbers) obtained from the replay buffer into a preset reinforcement learning model respectively. The preset reinforcement learning model processes the input historical system state data to obtain system state data, i.e., predicted system state data. The predicted system state data includes the fan speed, etc.

[0127] In an embodiment of the present application, a preset reinforcement learning model can output the rotational speed of each fan, and the electronic device matches the output of the preset reinforcement learning model with the historical system state data in the replay buffer. If a historical system state data matches the output of the preset reinforcement learning model, such as the historical system state data includes the output of the preset reinforcement learning model, or the fan speed included in the historical system state data has the greatest similarity to the output of the preset reinforcement learning model, then the historical system state data can be used as predicted system state data.

[0128] The preset reinforcement learning model can also directly output predicted system status data, and there is no limitation on this.

[0129] Step S503, using the reward value corresponding to the historical system state data and the reward value corresponding to the predicted system state data to determine the model loss value;

[0130] Step S504, with the goal of minimizing the model loss value, optimize the parameters of the preset reinforcement learning model, and iterate step S502 until the maximum cumulative reward value is obtained.

[0131] In the embodiment of the present application, the reward function of the preset reinforcement learning model can be determined according to performance indicators such as the total power consumption of the system, the noise generated by the fan, and the outlet temperature of the electronic device. The greater the total power consumption, noise, and outlet temperature of the system, the smaller the reward value. In one example, the reward value is the weighted average of the total power consumption of the system, the noise generated by the fan, and the outlet temperature of the electronic device. For details, please refer to the following formula.

[0132] R=w 1 ×E(P)+w 2 ×N(S)+w 3 ×T(O);

[0133] Where R represents the reward value of the reward function, E(P) represents the total power consumption of the system, N(S) represents the noise generated by the fan, T(O) represents the outlet temperature of the electronic device, and w 1 Indicates the weight value corresponding to the total power consumption of the system, w 2 Represents the weight value corresponding to the noise, w 3 Indicates the weight value corresponding to the outlet temperature. 1 、w 2 、w 3 The sum of can be -1 or other values. 1 、w 2 、w 3 The size can be set according to actual needs.

[0134] For example, if the user is concerned about the two performance indicators of system total power consumption and outlet temperature, the w 1 and w 3 , lower w 2 When users focus on the two performance indicators of system total power consumption and noise, they can increase w 1 and w 2 , lower w 3 By adjusting the weight value of the performance indicator, the purpose of emphasizing or weakening a certain performance indicator can be achieved, and a suitable reward function can be constructed to obtain a preset reinforcement learning model that meets user needs.

[0135] In the embodiment of the present application, only the three performance indicators of system total power consumption, noise and outlet temperature are used as an example to construct the reward function. In practical applications, the electronic device can use other performance indicators to construct the reward function, which is not limited.

[0136] After obtaining the predicted system state data, the electronic device obtains the performance indicators required for the reward function from the predicted system state data, and then calculates the reward value corresponding to the predicted system state data based on the obtained performance indicators using the above formula; in addition, the performance indicators required for the reward function are obtained from the historical system state data, and then calculates the reward value corresponding to the historical system state data based on the obtained performance indicators using the above formula.

[0137] The electronic device inputs the reward value corresponding to the historical system state data and the reward value corresponding to the predicted system state data into the preset loss function to obtain the model loss. With the goal of minimizing the model loss value, the electronic device can use algorithms such as gradient descent and back propagation to optimize the parameters of the preset reinforcement learning model.

[0138] Steps S502 to S504 are performed once for one iteration training. In one iteration training, the electronic device will obtain multiple historical system state data, and then obtain multiple predicted system state data. The cumulative reward value is the cumulative value of the reward values ​​corresponding to the multiple predicted system state data in one iteration training.

[0139] If the electronic device does not obtain the maximum cumulative reward value, it re-executes step S502 and continues iterative training; if it obtains the maximum cumulative reward value, it stops iterative training. The electronic device configures the parameters corresponding to the maximum cumulative reward value to the preset reinforcement learning model to obtain a trained preset reinforcement learning model.

[0140] For the trained preset reinforcement learning model, the electronic device can input the current system state data into the preset reinforcement learning model at each preset time interval. The preset reinforcement learning model infers the current system state data and outputs the speed control parameter, i.e., the second optimal speed control parameter.

[0141] In the above step S403, after the second optimal speed regulation parameter is obtained, the electronic device adjusts the rotation speed of each fan in the electronic device according to the second optimal speed regulation parameter.

[0142] For example, the electronic device includes n fans, and the second optimal speed adjustment parameter obtained by the electronic device is {F 1 , F 2 , …, F n}, F i represents the speed of fan i, i=1,2,…,n. The electronic device adjusts the speed of fan i to F i .

[0143] After the electronic device adjusts the speed of each fan in the electronic device according to the second optimal speed adjustment parameter, it can be determined that the external factors have changed, and then the greedy algorithm is started to execute steps S404 to S405. Steps S404 to S405 are similar to steps S202 to S203, and the relevant description of steps S202 to S203 can be referred to, which will not be repeated here.

[0144] In an embodiment of the present application, changes in external factors may be caused by using a preset reinforcement learning model to adjust the fan speed in steps S402 to S403, or may be caused by other factors, such as changes in the load of the electronic device, changes in the environment in which the electronic device is located, etc. Therefore, before executing steps S402 to S403, the electronic device may also execute steps S404 to S405 to eliminate the impact of changes in external factors and achieve the purpose of energy saving and noise reduction.

[0145] Corresponding to the above-mentioned fan speed regulation method, the embodiment of the present application also provides a fan speed regulation method, such as Figure 6 As shown, the method may include the following steps:

[0146] Step S601, periodically obtaining current system status data of the electronic device;

[0147] Step S602, after the electronic device is initialized, the current system state data is processed using a PID algorithm to obtain a third predicted speed regulation parameter;

[0148] In the embodiment of the present application, the electronic device is pre-configured with a PID algorithm, and the debugging parameters in the PID algorithm can be set based on experience. For example, the debugging parameters in the PID algorithm may include SP, dead zone temperature T dead 、Minimum speed N low , Intermediate speed N mid and maximum speed N high The speed reduction ratio is 1%, the speed increase ratio is 2%, etc.

[0149] After the electronic device is powered on and initialized, the electronic device uses the PID algorithm to process the current system status data, determines the speed regulation parameters of each fan, that is, predicts the speed regulation parameters, and then executes step S603.

[0150] In the embodiment of the present application, the periodic duration for the electronic device to obtain the current system status data and the interval duration for the PID algorithm to process the system status data may be the same or different.

[0151] Step S603, adjusting the speed of the fan in the electronic device according to the predicted speed adjustment parameter;

[0152] After obtaining the predicted speed control parameter, the electronic device adjusts the speed of each fan in the electronic device to the predicted speed control parameter, and then re-executes step S601 to step S603, and repeats step S601 to step S603 until the temperature of each temperature control point in the electronic device reaches stability. When the temperature of each temperature control point in the electronic device reaches stability, step S604 is executed.

[0153] Step S604, when the temperature of each temperature control point in the electronic device reaches stability, the current system state data is processed using a greedy algorithm to obtain a first optimal speed regulation parameter;

[0154] Step S605: adjusting the speed of the fan in the electronic device according to the first optimal speed adjustment parameter.

[0155] In some embodiments, during the process of adjusting the speed of the fan using the greedy algorithm, that is, during the process of looping S604, if the temperature of the first temperature control point in the electronic device is greater than the preset temperature, the process returns to step S602.

[0156] In the embodiment of the present application, the first temperature control point can be any temperature control point in the electronic device, and the preset temperature can be set according to actual needs. In the process of adjusting the speed of the fan using the greedy algorithm, the electronic device can periodically obtain the temperature of each temperature control point, and the temperature of the temperature control point can be obtained from the system status data; the electronic device detects whether the temperature of each temperature control point is greater than the preset temperature. If the temperature of each temperature control point is less than or equal to the preset temperature, the electronic device continues to use the greedy algorithm to adjust the speed of the fan, that is, executes step S604 until the first optimal speed regulation parameter is obtained; if the temperature of one of the temperature control points (that is, the first temperature control point) is greater than the preset temperature, the electronic device is forced to switch to using the PID algorithm to adjust the speed of the fan to quickly reduce the temperature of the temperature control point to avoid high temperature danger.

[0157] In the embodiment of the present application, after the electronic device is initialized, the electronic device preferentially uses the PID algorithm to adjust the fan speed. At this time, it is not necessary to adjust the fan speed to the optimum, but only to adjust the fan speed so that the temperature of each temperature control point in the electronic device is stable. Therefore, the parameter setting of the PID algorithm is relatively simple, and the debugging complexity is relatively low.

[0158] Electronic devices use PID algorithms to adjust fan speeds. After the temperature of each temperature control point reaches stability, electronic devices can use greedy algorithms to adjust the speeds of each fan in the electronic device to obtain the optimal speed control parameters. The greedy algorithm can adjust the speed control parameters by itself to obtain the speed control parameters that minimize the total power consumption of the system, that is, the optimal speed control parameters. During the entire speed control process, the optimal speed control parameters can be obtained without manual debugging, which reduces the complexity of system debugging, and reduces the difficulty, time and manpower consumed in determining the optimal speed control parameters.

[0159] In addition, in an embodiment of the present application, the electronic device preferentially uses the PID algorithm to adjust the fan speed, and then uses the greedy algorithm to adjust the fan speed, thereby avoiding the problem of excessive temperatures at each temperature control point due to the long time spent in finding the optimal speed using the greedy algorithm.

[0160] In an embodiment of the present application, the above-mentioned embodiments can be combined. For example, after the electronic device is initialized, the electronic device uses a PID algorithm to adjust the fan speed. When the temperature of each temperature control point reaches stability, a greedy algorithm is used to adjust the fan speed. When the amount of historical system status data updated and stored in the replay buffer reaches a preset threshold value, the historical system status data is used to train a preset reinforcement learning model. At every preset time interval, the preset reinforcement learning model is used to process the current system status data, introducing disturbance factors into the greedy algorithm, so that the entire system reaches the optimal state.

[0161] Combine the following Figure 7 The overall architecture diagram shown and Figure 8 The speed regulation principle diagram shown in the figure provides a detailed description of the fan speed regulation method provided in the embodiment of the present application.

[0162] like Figure 7 As shown in the figure, the speed control algorithms such as the DRL model (i.e., the preset reinforcement learning model), PID algorithm, and greedy algorithm are deployed on the agent. In this architecture, the set system status data include: fan power consumption, temperature of each temperature control point, speed of each fan in the electronic device, noise generated by the fan, total system power consumption of the electronic device, system load of the electronic device, and power consumption of specified devices in the electronic device. The set actions include the speed of each fan, which can be expressed in gear or duty cycle. The reward value can be obtained by weighted average of the total system power consumption, the noise generated by the fan, and the outlet temperature of the electronic device to achieve the purpose of energy saving and noise reduction.

[0163] The agent obtains system status data and reward values ​​from the device environment, and uses the system status data and reward values ​​to train the DRL model. The DRL model then introduces disturbance factors into the greedy algorithm, so that the greedy algorithm processes the system status data and outputs the optimal speed control parameters, such as the speed of each fan. 1 , F 2 , …, F n}, replace {F 1 , F 2 , …, F n}Configure it into the device environment (i.e. the system) to adjust the speed of each fan.

[0164] like Figure 8As shown, after the electronic device is initialized, the electronic device uses the PID algorithm to adjust the speed of each fan; after the temperature of each temperature control point in the device environment is stable, the electronic device uses the greedy algorithm to adjust the speed of each fan in a cycle, and the maximum number of cycles (such as the preset number mentioned above) is set according to actual needs. After the greedy algorithm converges, that is, after the minimum total system power consumption is obtained, the electronic device stops using the greedy algorithm to adjust the speed of each fan. At this point, the electronic device is in a stable state. In the process of using the greedy algorithm to adjust the speed of each fan, the electronic device writes the obtained system status data into the replay buffer.

[0165] When the amount of data updated and stored in the replay buffer is greater than or equal to the preset threshold value, the electronic device uses the system status data updated and stored in the replay buffer to train the DRL model. After the DRL model training is completed, the electronic device uses the DRL model to process the system status data once at a preset interval, and adjusts the speed of each fan to introduce disturbance factors to the greedy algorithm, so that the greedy algorithm continues to run to obtain the global optimal speed control parameters. In addition, in the process of using the greedy algorithm to adjust the speed of each fan, the collected new system status data can gradually cover the old system status data in the replay buffer, so as to update the stored data and facilitate the subsequent update of the DRL model.

[0166] In the embodiment of the present application, the electronic device can monitor whether external factors such as environment and power change in a stable state. When the external factors change, the electronic device switches to use the greedy algorithm to adjust the speed of each fan, that is, the greedy algorithm is used cyclically to adjust the speed of each fan to find the optimal speed adjustment parameter. Otherwise, the electronic device can keep the fan speed unchanged.

[0167] In addition, when using the greedy algorithm to adjust the speed of each fan, if the temperature of the temperature control point is detected to be higher than the preset temperature, that is, an overtemperature alarm occurs, the electronic equipment is forced to switch to using the PID algorithm to adjust the speed of each fan to avoid high temperature hazards.

[0168] The above process of adjusting the rotation speed of each fan using the PID algorithm, greedy algorithm and DRL model can be implemented by the electronic device itself, and the training and updating of the DRL model can be implemented by other electronic devices or by the electronic device itself, without limitation.

[0169] In the embodiments of the present application, by introducing intelligent algorithms (such as greedy algorithms and DRL models), more optimized speed control parameters can be found compared to traditional speed control methods; using greedy algorithms and DRL models to intelligently adjust fan speed helps to reduce power consumption and improve the competitiveness of equipment products in data centers such as servers and switching equipment. Even if the power consumption is reduced by 1%, the benefits to large data centers are also huge; for the product development stage, intelligent speed control can reduce the workload of manual speed control in the development stage; fan speed control parameters can be dynamically adjusted online to adapt to changes in the environment and configuration, and may also respond to alarms caused by aging of fans during long-term operation.

[0170] Corresponding to the above-mentioned fan speed regulation method, the embodiment of the present application also provides a fan speed regulation device, such as Fig. 9 As shown, the device comprises:

[0171] An acquisition module 901 is used to periodically acquire current system status data of an electronic device;

[0172] A first prediction module 902 is used to process the current system state data using a greedy algorithm to obtain a first optimal speed regulation parameter;

[0173] The first speed adjustment module 903 is used to adjust the speed of the fan in the electronic device according to the first optimal speed adjustment parameter.

[0174] In some embodiments, the first prediction module 902 may also be used to re-use the greedy algorithm to process the current system state data to obtain the first optimal speed regulation parameter when external factors of the electronic device change.

[0175] In some embodiments, the external factors may include speed control parameters; in this case, the fan speed control device may further include:

[0176] The second prediction module is used to process the current system state data using a preset reinforcement learning model at each interval of a preset duration to obtain a second optimal speed regulation parameter;

[0177] The second speed adjustment module is used to adjust the speed of the fan in the electronic device according to the second optimal speed adjustment parameter.

[0178] In some embodiments, the replay buffer area of ​​the electronic device stores historical system status data; the fan speed control device may further include a training module for training a preset reinforcement learning model, which may be specifically used for:

[0179] Obtain the historical system state data stored in the replay buffer; input the historical system state data into the preset reinforcement learning model to obtain the predicted system state data; determine the model loss value using the reward value corresponding to the historical system state data and the reward value corresponding to the predicted system state data; optimize the parameters of the preset reinforcement learning model with the goal of minimizing the model loss value, and iteratively execute the steps of inputting the historical system state data into the preset reinforcement learning model to obtain the predicted system state data until the maximum cumulative reward value is obtained.

[0180] In some embodiments, the training module may also be used to reacquire the historical system status data stored in the replay buffer when the amount of historical system status data updated and stored in the replay buffer reaches a preset threshold value.

[0181] In some embodiments, the reward value is obtained by weighted average of the total power consumption of the system, the noise generated by the fan, and the outlet temperature of the electronic device.

[0182] In some embodiments, the fan speed regulating device may further include:

[0183] The third prediction module is used to process the current system state data using the PID algorithm after the electronic device is initialized to obtain the predicted speed control parameters;

[0184] A third speed regulation module, used to adjust the speed of a fan in the electronic device according to the predicted speed regulation parameter;

[0185] The first prediction module 902 may also be used to process the current system state data using a greedy algorithm to obtain a first optimal speed regulation parameter when the temperature of each temperature control point in the electronic device reaches stability.

[0186] In some embodiments, the third prediction module is also used to use a PID algorithm to process the current system state data to obtain a predicted speed control parameter when the temperature of the first temperature control point in the electronic device is greater than a preset temperature during the process of using a greedy algorithm to process the current system state data to obtain a first optimal speed control parameter.

[0187] In the technical solution provided by the embodiment of the present application, a greedy algorithm is used to process the current system state data. The greedy algorithm can intelligently find the optimal speed regulation parameters with the lowest total system power consumption or the lowest total system power consumption and noise as the optimization goal, and then configure the speed of each fan in the electronic device. During the entire speed regulation process, the optimal speed regulation parameters can be obtained without manual debugging, which reduces the complexity of system debugging, and reduces the difficulty, time and manpower consumed in determining the optimal speed regulation parameters.

[0188] Corresponding to the above-mentioned fan speed adjustment method, the embodiment of the present application further provides an electronic device, such as Fig.10As shown, it includes a processor 1001 , a communication interface 1002 , a memory 1003 and a communication bus 1004 , wherein the processor 1001 , the communication interface 1002 , and the memory 1003 communicate with each other via the communication bus 1004 .

[0189] Memory 1003, used for storing computer programs;

[0190] The processor 1001 is used to implement any of the above-mentioned fan speed adjustment methods when executing the program stored in the memory 1003.

[0191] The communication bus can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The communication bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0192] The communication interface is used for communication between the above electronic device and other devices.

[0193] The memory may include a random access memory (RAM) or a non-volatile memory (NVM), such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

[0194] The processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, and discrete hardware components.

[0195] In another embodiment provided in the present application, a computer-readable storage medium is provided, in which a computer program is stored. When the computer program is executed by a processor, any of the above-mentioned fan speed regulation methods is implemented.

[0196] In another embodiment provided in the present application, a computer program product including instructions is also provided, and when the computer program product is run on a computer, the computer is enabled to execute any fan speed adjustment method in the above embodiments.

[0197] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from a website site, a computer, a server or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or a data center that includes one or more available media integrated. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive Solid State Disk (SSD)), etc.

[0198] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0199] Each embodiment in this specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device, electronic device, storage medium and program product embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.

[0200] The above description is only a preferred embodiment of the present application and is not intended to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application are included in the protection scope of the present application.

Claims

1. A fan speed regulation method, characterized in that: The method comprises: Periodically obtain current system status data of the electronic device; After the electronic device is initialized, the current system state data is processed using a PID algorithm to obtain a predicted speed regulation parameter; adjusting the speed of the fan in the electronic device according to the predicted speed regulation parameter; When the temperature of each temperature control point in the electronic device reaches stability, a greedy algorithm is used to process the current system state data to obtain a first optimal speed regulation parameter; adjusting the speed of the fan in the electronic device according to the first optimal speed regulation parameter; At each preset time interval, the current system state data is processed using a preset reinforcement learning model to obtain the second optimal speed regulation parameter; According to the second optimal speed regulation parameter, the rotation speed of the fan in the electronic device is adjusted to re-execute the step of using the greedy algorithm to process the current system state data to obtain the first optimal speed regulation parameter.

2. The method according to claim 1, characterized in that The replay buffer of the electronic device stores historical system status data; the preset reinforcement learning model is trained using the following steps: Acquire historical system status data stored in the replay buffer; Inputting the historical system state data into a preset reinforcement learning model to obtain predicted system state data; Determine a model loss value using the reward value corresponding to the historical system state data and the reward value corresponding to the predicted system state data; With the goal of minimizing the model loss value, the parameters of the preset reinforcement learning model are optimized, and the step of inputting the historical system state data into the preset reinforcement learning model to obtain the predicted system state data is iteratively executed until the maximum cumulative reward value is obtained.

3. The method according to claim 2, characterized in that The method further comprises: When the amount of the historical system status data updated and stored in the replay buffer reaches a preset threshold, the step of obtaining the historical system status data stored in the replay buffer is performed again.

4. The method according to claim 2 or 3, characterized in that: The reward value is obtained by weighted average of the total power consumption of the system, the noise generated by the fan, and the outlet temperature of the electronic equipment.

5. The method according to claim 1, characterized in that The method further comprises: In the process of using a greedy algorithm to process the current system state data to obtain the first optimal speed regulation parameter, if the temperature of the first temperature control point in the electronic device is greater than the preset temperature, the step of using the PID algorithm to process the current system state data to obtain the predicted speed regulation parameter is executed.

6. A fan speed regulating device, characterized in that: The device comprises: An acquisition module, used for periodically acquiring current system status data of the electronic device; A third prediction module is used to process the current system state data using a PID algorithm after the electronic device is initialized to obtain a predicted speed regulation parameter; A third speed adjustment module, used to adjust the speed of the fan in the electronic device according to the predicted speed adjustment parameter; A first prediction module, configured to process current system state data using a greedy algorithm to obtain a first optimal speed regulation parameter when the temperature of each temperature control point in the electronic device reaches stability; A first speed adjustment module, used for adjusting the speed of the fan in the electronic device according to the first optimal speed adjustment parameter; The second prediction module is used to process the current system state data using a preset reinforcement learning model at each interval of a preset duration to obtain a second optimal speed regulation parameter; The second speed adjustment module is used to adjust the speed of the fan in the electronic device according to the second optimal speed adjustment parameter.

7. An electronic device, characterized in that: It includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus; Memory, used to store computer programs; A processor, for implementing any of the methods described in claims 1-5 when executing a computer program stored in a memory.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

Images