A heat dissipation control method and device, a storage medium and an electronic device

Abstract

The embodiment of the present disclosure provides a heat dissipation control method and device, a storage medium and an electronic device, and relates to the technical field of computers. The method is applied to an electronic device, the electronic device comprises a heat dissipation device, and the method comprises the following steps: acquiring a set of physical parameters of the electronic device and noise generated by the heat dissipation device; determining a control parameter for heat dissipation according to the set of physical parameters and the noise; and controlling the heat dissipation device of the electronic device according to the control parameter. In this way, the control parameter of the heat dissipation device is dynamically adjusted according to the temperature of each key component of the electronic device and the noise generated by the heat dissipation device, so that the overall heat dissipation demand of the electronic device and the power consumption of the electronic device are considered, and the resource utilization efficiency is improved.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of computer technology, and particularly relates to a heat dissipation control method and device, a storage medium and an electronic device. BACKGROUND

[0002] The baseboard management controller (BMC) is a controller used for monitoring and managing servers, and its main functions include: 1) device information management: records detailed information of the server, such as model, manufacturer, production date, and production and technical information of each component, chassis and motherboard information, etc. At the same time, it also records the information of the BMC itself, such as the server host name, IP address and BMC firmware version, etc. 2) Server state monitoring management: real-time detection of the health status of each key component of the server (such as CPU, memory, hard disk, fan and frame, etc.) such as temperature and voltage. According to the data of the temperature collection point, the BMC adjusts the fan speed to ensure that the server does not overheat, and controls the overall power consumption within a reasonable range. If any component of the server appears abnormal, the BMC will report the relevant information to the upper network administrator in time through various industry general specifications (such as SNMP protocol, SMTP protocol and Redfish protocol). 3) Remote control management of the server: allows administrators to remotely perform server power on, reboot, maintenance, firmware update and system installation, etc. 4) Maintenance management: including log management, user management, BIOS management and alarm management, etc., to ensure the stable operation of the server. In summary, the BMC plays a key role in the operation and management of the server, and the BMC is mainly responsible for the monitoring, management and remote control of the server to ensure the stable operation and efficient management of the server.

[0003] Server heat dissipation is crucial to maintaining stable hardware operation, especially in high-load situations where overheating problems may occur. Heat dissipation includes passive heat dissipation and active heat dissipation, passive heat dissipation may include heat sinks, heat pipe technology, while active heat dissipation mainly relies on fans. The fan is controlled by the BMC, the key is how to optimize the fan control strategy to balance the heat dissipation effect and noise, energy consumption. Common techniques may include pulse width modulation (PWM) control fan speed, or use temperature sensor feedback to achieve dynamic adjustment.

[0004] In the related art, server active cooling mainly relies on the BMC to read the temperature data of the temperature sensor of the CPU, memory, network card and other key components, and calculates the maximum PWM according to the preset parameters (such as fixed threshold) through the PID (Proportional Integral Derivative) algorithm, and controls the fan rotation speed by using the PWM signal. This scheme relies on single component temperature reading and control source, and there is a risk of long-term full-speed operation of the fan. Specifically, 1) the temperature sampling dimension is single, and a multi-sensor data fusion mechanism is not established, which is easy to cause false judgment of local hot spots and lead to frequent full-speed operation of the fan; 2) the static configuration of the PID parameter lacks dynamic adaptability, and when facing sudden cooling pressure, it can only passively trigger the limit speed, which aggravates mechanical wear and tear; 3) the control strategy is completely coupled with the hardware temperature, and lacks environmental perception ability, and cannot realize acoustic optimization while ensuring cooling efficiency.

[0005] This scheme will directly cause the fan to be in high-load working condition for a long time, the bearing life is shortened to 60%-70% of the nominal value, and the noise when running at full speed generally exceeds the 55dB limit value specified in the TIA-942 standard. It is impossible to meet the differentiated noise reduction needs of customers: balancing cooling and silence during the day, and reducing speed to below 48dB at night.

[0006] In addition, the original scheme lacks redundant control channel design, and single-point sensor failure will cause system misjudgment, further increasing the risk of abnormal wear of the fan. This extensive temperature control mechanism cannot meet the complex needs of modern IDC machine rooms for acoustic comfort, device reliability and energy efficiency ratio. SUMMARY

[0007] The present disclosure provides a heat dissipation control method, device, storage medium and electronic equipment to at least solve the above technical problems in the prior art.

[0008] The technical scheme of the embodiment of the present disclosure is as follows:

[0009] In a first aspect, the embodiment of the present disclosure provides a heat dissipation control method, which is applied to an electronic equipment including a heat dissipation device, and the method comprises:

[0010] Obtaining a set of physical parameters of the electronic equipment and noise generated by the heat dissipation device;

[0011] Determining a control parameter for heat dissipation according to the set of physical parameters and the noise;

[0012] Controlling the heat dissipation device of the electronic equipment according to the control parameter.

[0013] In the above scheme, the set of physical parameters of the electronic equipment is obtained, comprising:

[0014] acquire at least one physical parameter of the electronic device, determine the set of physical parameters according to the acquired at least one physical parameter;

[0015] The at least one physical parameter comprises at least one of a component temperature and a fan rotating speed, and the component comprises a CPU, a memory and a PCIE device.

[0016] In the above solution, the control parameter for heat dissipation is determined according to the set of physical parameters and the noise, comprising:

[0017] According to a first rule, an evaluation coefficient of at least one component is determined according to a temperature parameter of the at least one component and the noise.

[0018] According to a second rule, the control parameter is determined according to the evaluation coefficient of the at least one component.

[0019] In the above solution, according to a first rule, at least one control index is determined according to a temperature parameter of at least one component and the noise, comprising:

[0020] A temperature weight and a temperature threshold corresponding to each component are determined according to each component.

[0021] An evaluation coefficient of each component is determined according to a temperature parameter of each component, the temperature threshold, the temperature weight, the noise and a noise weight.

[0022] In the above solution, an evaluation coefficient of each component is determined according to a temperature parameter of each component, the noise, the temperature weight and a noise weight, comprising:

[0023] A temperature control index is determined according to a temperature parameter of each component and the temperature threshold.

[0024] A noise control index is determined according to the noise and a reference noise.

[0025] An evaluation coefficient of each component is determined according to the temperature control index, the noise control index, the temperature weight and the noise weight.

[0026] In the above solution, the control parameter is determined according to the evaluation coefficient of the at least one component based on a second rule, comprising:

[0027] An evaluation coefficient vector is determined according to the evaluation coefficient of the at least one component.

[0028] The evaluation coefficient vector is input into a first model to obtain an output result of the first model, and the output result is used to indicate an adjustment range of a PID parameter.

[0029] In the above scheme, before determining the control parameters for heat dissipation based on the set of physical parameters and the noise, the method further includes:

[0030] Based on the set of physical parameters and the noise, determine whether a heat dissipation failure exists;

[0031] If a heat dissipation failure is determined, a fault alarm is generated, which is used to indicate the type of failure.

[0032] If it is determined that there is no heat dissipation fault, control parameters for heat dissipation are determined based on the set of physical parameters and the noise.

[0033] In the above scheme, determining whether a heat dissipation fault exists based on the set of physical parameters and the noise includes:

[0034] Time-series data of rotational speed and / or temperature are generated based on the set of physical parameters;

[0035] A first feature set is extracted based on the noise, and the first feature set includes: BPF and HAR;

[0036] The first feature set, the time-series data of the rotational speed and / or temperature are input into the second model to obtain the output of the second model. The output is used to indicate the fault type and fault anomaly level.

[0037] In the above scheme, generating fault alarms includes:

[0038] If the fault type belongs to the target fault type and the fault anomaly level exceeds the level threshold, a fault alarm is generated based on the fault type and the fault anomaly level.

[0039] In the above scheme, the method further includes: generating the second model; the generation of the second model includes:

[0040] Obtain a training dataset; the training dataset includes: at least one training data and a label for each training data, each training data including: time-series data of sample noise, sample rotation speed and / or temperature; the label is the fault type and fault anomaly level;

[0041] The hybrid neural network is trained using the training dataset to obtain the trained hybrid neural network, which serves as the second model.

[0042] The hybrid neural network includes: a front-end network, a feature fusion layer, and an output layer;

[0043] The front-end network employs a first branch network and a second branch network. The first branch network is used to extract the features of the sample noise, and the second branch network is used to extract the features of the time-series data of the sample rotation speed and / or temperature.

[0044] The feature fusion layer is used to process data using an attention mechanism based on the features extracted by the front-end network;

[0045] The output layer is used to obtain the output result based on the data processing result of the feature fusion layer.

[0046] The method in the above scheme further includes:

[0047] Obtain a fault database of at least one fan, and fine-tune the second model based on the fault database to obtain a second model corresponding to the electronic device using the fan.

[0048] In the above scheme, obtaining the noise generated by the heat dissipation device includes:

[0049] The first noise generated by the fan and / or the second noise from the air duct are respectively obtained;

[0050] The first noise and the second noise are combined to obtain the noise generated by the heat dissipation device.

[0051] Secondly, embodiments of this disclosure provide a heat dissipation control device, which is applied to an electronic device, the electronic device including a heat dissipation component, and the device comprising:

[0052] An acquisition module is used to acquire the set of physical parameters of the electronic device and the noise generated by the heat dissipation device;

[0053] The first processing module is used to determine the control parameters for heat dissipation based on the set of physical parameters and the noise.

[0054] The second processing module is used to control the heat dissipation device of the electronic device according to the control parameters.

[0055] Thirdly, embodiments of this disclosure provide an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform any of the heat dissipation control methods described above.

[0056] Fourthly, embodiments of this disclosure provide a non-transitory computer-readable storage medium storing computer instructions for causing a computer to perform the heat dissipation control method according to any one of the claims.

[0057] The embodiments disclosed herein have the following beneficial effects:

[0058] The heat dissipation control method, apparatus, storage medium, and electronic device provided in this disclosure are applied to an electronic device, which includes a heat dissipation device. The method includes: acquiring a set of physical parameters of the electronic device and noise generated by the heat dissipation device; determining control parameters for heat dissipation based on the set of physical parameters and the noise; and controlling the heat dissipation device of the electronic device according to the control parameters. Thus, by dynamically adjusting the control parameters of the heat dissipation device based on the temperature of each key component of the electronic device and the noise generated by the heat dissipation device, both the overall heat dissipation requirements of the electronic device and the power consumption of the electronic device are taken into account, thereby improving resource utilization efficiency.

[0059] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0060] Figure 1 A schematic flowchart of a heat dissipation control method provided in an embodiment of this disclosure;

[0061] Figure 2 A schematic diagram of a hybrid neural network architecture provided in an embodiment of this disclosure;

[0062] Figure 3 A flowchart illustrating a training method for a second model provided in an embodiment of this disclosure;

[0063] Figure 4 This is a schematic diagram of the structure of a heat dissipation control device provided in an embodiment of the present disclosure;

[0064] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this disclosure. Detailed Implementation

[0065] To make the objectives, features, and advantages of this disclosure more apparent and understandable, the technical solutions in the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this disclosure, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this disclosure without creative effort are within the scope of protection of this disclosure.

[0066] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0067] If the application documents contain similar descriptions such as "first / second", the following explanation shall be added: In the following description, the terms "first / second / third" are used only to distinguish similar objects and do not represent a specific order of objects. It is understood that "first / second / third" may be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0068] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0069] Figure 1 This is a flowchart illustrating a heat dissipation control method provided in an embodiment of the present disclosure, as shown below. Figure 1 As shown, the method is applied to an electronic device, the electronic device including a heat dissipation device, and the method includes:

[0070] Step 101: Obtain the physical parameter set of the electronic device and the noise generated by the heat dissipation device;

[0071] Step 102: Determine the control parameters for heat dissipation based on the set of physical parameters and the noise;

[0072] Step 103: Control the heat dissipation device of the electronic device according to the control parameters.

[0073] The electronic device can be a server, computer, projector, or other device with heat dissipation components.

[0074] In some embodiments, obtaining the physical parameter set of the electronic device includes:

[0075] Collect at least one physical parameter of the electronic device, and determine the set of physical parameters based on the collected at least one physical parameter;

[0076] The at least one physical parameter includes at least one of the following: component temperature, fan speed (such as the fan's PWM value).

[0077] The component can be any part of an electronic device that is affected by or influenced by temperature. For example, the component may include at least one of the following: CPU (Central Processing Unit), RAM (Random Access Memory), PCIe (Peripheral Component Interconnect Express) device, etc. Accordingly, component temperatures include: CPU temperature, memory temperature, and PCIe device temperature.

[0078] Here, PCIe is a high-speed data transmission interface used to connect a computer motherboard to various external hardware devices. PCIe devices can include graphics cards, network cards, storage devices, etc.

[0079] Based on the design of different electronic devices, the processing device may also include: GPU (Graphics Processing Unit), TPU (Tensor Processing Unit), FPGA (Field-Programmable Gate Array), ASIC (Application-Specific Integrated Circuit), DSP (Digital Signal Processor), NPU (Neural Processing Unit), VPU (Vision Processing Unit), etc. Correspondingly, the component temperature may also include the temperature of the above-mentioned devices.

[0080] The physical parameters may also include: ambient temperature, ambient humidity, and internal air pressure of the chassis.

[0081] Here, the electronic device has a Base Control Module (BMC), which is designed with data acquisition capabilities to collect and record the corresponding times of the above physical parameters. For example, the BMC monitors the temperature sensors, humidity sensors, etc., corresponding to the above components to determine the various physical parameters.

[0082] In some embodiments, acquiring the noise generated by the heat dissipation device includes:

[0083] The first noise generated by the fan and / or the second noise from the air duct are respectively obtained;

[0084] The first noise and the second noise are combined to obtain the noise generated by the heat dissipation device.

[0085] Here, the heat dissipation device may include: a fan and an air duct; the first noise generated by the fan is the noise produced when the fan is running. The second noise generated by the air duct is the noise produced when airflow flows in the air duct.

[0086] The noise generated by the heat dissipation device is obtained by combining the first and second noise sources using mathematical models (such as logarithmic weighting, spectrum analysis, etc.) to integrate the contributions of different noise sources and obtain a result reflecting the overall noise level. The specific merging method should be selected based on the actual situation (such as the frequency characteristics and intensity of the noise sources).

[0087] In one example, a weighted merging method is used. This takes into account that the noise generated by the fan and the air duct may have different effects on the total noise, and can be merged using weighting coefficients. For example, if it is verified that the fan noise is dominant, a higher weight can be given to the fan noise and a lower weight to the air duct noise.

[0088] In another example, spectral analysis is used. Considering that noise sources may have different frequency characteristics, spectral analysis can be used to overlay the spectra of each noise source to more accurately reflect the contribution of each noise source to different frequency bands. For the noise intensity of each frequency band, noise levels can be graded and assigned different weights, and a weighted merging method can be used to obtain the spectrum of the entire noise signal.

[0089] Of course, other methods can be used for merging, which are not limited here.

[0090] Here, the electronic device has a BMC (Browser Control Center), which is designed with data acquisition capabilities to collect the above-mentioned noises and record the corresponding acquisition time. For example, the first noise generated by the fan and / or the second noise of the air duct are collected in real time using a high-precision microphone array (e.g., signal-to-noise ratio ≥ 60dB) (e.g., designed sampling rate ≥ 48kHz, resolution 24bit).

[0091] After obtaining the above physical parameters and noise, the timestamps of each data can be aligned with the BMC clock to obtain a data set in the following form: time-noise-physical parameter set, so as to clearly and accurately record the dataset.

[0092] In some embodiments, determining the control parameters for heat dissipation based on the set of physical parameters and the noise includes:

[0093] Based on the first rule, the evaluation coefficient of at least one component is determined according to the temperature parameter of at least one component and the noise.

[0094] Based on the second rule, the control parameters are determined according to the evaluation coefficients of the at least one component.

[0095] Here, the first rule, used to analyze the temperature parameters and noise of at least one component to determine or reflect the heat dissipation situation, specifically indicates an analysis method that obtains an evaluation coefficient for each component by processing or calculating the temperature parameters and noise of at least one component accordingly. This evaluation coefficient is used to measure the working state of the heat dissipation device (e.g., a fan), or in other words, to reflect the control indicators that currently require strengthening or weakening heat dissipation.

[0096] The second rule is used to indicate how to determine control parameters based on the evaluation coefficients after obtaining them.

[0097] For example, in a cooling system, a PID controller is used to maintain electronic equipment within an ideal temperature range by adjusting control variables such as fan speed in real time. The control parameters can be the PID parameter adjustment range, namely proportional (P), integral (I), and derivative (D). The proportional (P) parameter affects the system's response speed; a suitable P parameter can bring the temperature close to the target value in a shorter time. The integral (I) parameter can eliminate the system's steady-state error; an appropriate I parameter helps eliminate temperature drift, but an excessively high integral parameter may lead to over-adjustment and oscillation. The derivative (D) parameter reduces system oscillation and over-adjustment; the D parameter helps the system react quickly and reduces excessive temperature fluctuations. These three parameters determine how to adjust the fan output to achieve more precise cooling control.

[0098] For PID parameters, their adjustment range can be discretized into multiple actions. For example, the adjustment range of the P parameter can be discretized into 5 actions: -10%, -5%, 0%, +5%, and +10%, representing the adjustment magnitude of the P parameter relative to the current value. Similarly, the I and D parameters are also discretized. In this way, the control parameters are designed as a discrete action space, with each action corresponding to a set of PID parameter adjustment ranges.

[0099] In some embodiments, based on a first rule, at least one control index is determined according to the temperature parameter of at least one component and the noise, including:

[0100] Determine the temperature weight and temperature threshold corresponding to each component;

[0101] The evaluation coefficient for each component is determined based on its temperature parameters, temperature threshold, temperature weight, noise, and noise weight.

[0102] Here, the first rule can include the temperature weight and temperature threshold corresponding to each component.

[0103] Temperature weights represent the degree to which the temperature of a component affects the overall system control. Different components can have different effects on the system, therefore, different weights can be assigned to different components.

[0104] A temperature threshold is a boundary value for the temperature of a component. When the temperature reaches or exceeds this value, certain control measures or alarms can be triggered. For example, the temperature threshold for a CPU can be designed to be 85°C.

[0105] In some embodiments, determining an evaluation coefficient for each component based on its temperature parameter, the noise, the temperature weight, and the noise weight includes:

[0106] The temperature control parameters are determined based on the temperature parameters of each component and the temperature threshold.

[0107] Based on the noise level and the reference noise level, noise control parameters are determined;

[0108] The evaluation coefficient for each component is determined based on the temperature control index, the noise control index, the temperature weight, and the noise weight.

[0109] Here, an evaluation function is provided to calculate the evaluation coefficient, as shown in the following formula:

[0110]

[0111] Where Tcr is the temperature threshold (as in the example above, assuming the CPU evaluation coefficient is calculated, it is 85℃), Tcu is the current temperature; Nba is the reference noise (e.g., 45dB), Ncu is the current noise, a is the temperature weight, b is the noise weight, and a+b=1.

[0112] The calculated evaluation coefficients can be used to measure the fan's effectiveness in controlling the temperature of this component. Among them, This indicates the temperature control parameters. This indicates noise control indicators.

[0113] The larger the value of the temperature control index, the closer the current temperature Tcu is to the temperature threshold Tcr (i.e., the higher the temperature). In the evaluation function, the larger this value is, the worse the temperature control effect is.

[0114] The larger the value of the noise control index, the closer the current noise Ncu is to the reference noise Nba (i.e., the louder the noise). In the evaluation function, the larger this part is, the worse the noise control effect is.

[0115] The weighting coefficients 'a' and 'b' are used to balance the importance of temperature control and noise control. By adjusting the values ​​of 'a' and 'b', you can prioritize either temperature control or noise control based on actual needs. For example, if more emphasis is placed on temperature control, 'a' can be set to be greater than 'b', such as 'a=0.7' and 'b=0.3'; if more emphasis is placed on noise control, 'a' can be set to be less than 'b', such as 'a=0.3' and 'b=0.7'.

[0116] Considering the different importance of different components in the system and their varying sensitivities to temperature and noise, the application of a and b for each component can differ.

[0117] For example, taking CPU, memory, and PCIe devices as examples, since the CPU is the core component of a computer, high temperatures can lead to performance degradation or even damage. Therefore, CPU temperature control is a primary concern, and in the CPU's evaluation function, 'a' can be set relatively high, for example, a=0.8 and b=0.2. Memory, on the other hand, may have relatively higher requirements for noise levels, so in the memory's evaluation function, 'b' can be set relatively high, for example, a=0.4 and b=0.6. For PCIe devices, temperature and noise levels may be equally important; therefore, in the PCIe device's evaluation function, 'a' and 'b' are set the same, i.e., a=0.5 and b=0.5.

[0118] In some embodiments, determining the control parameters based on the evaluation coefficients of the at least one component according to the second rule includes:

[0119] Based on the evaluation coefficients of the at least one component, determine the evaluation coefficient vector;

[0120] The evaluation coefficient vector is input into the first model to obtain the output result of the first model, which is used to indicate the control parameters.

[0121] Here, the second rule can include the first model, which is used to obtain control parameters based on the evaluation coefficient vector.

[0122] In one example, the first model uses the DDQN model. DDQN (Double Deep Q-Network) is an algorithm in deep reinforcement learning used to solve decision problems in a discrete action space. It approximates the Q function through a neural network, thereby guiding the agent (here referring to the fan speed control system) to select the optimal action (here referring to the control parameters, i.e., the adjustment range of the PID parameters).

[0123] In practical applications, the evaluation coefficient of each component is used as the input to the DDQN model. The DDQN model processes the input and outputs the control parameters, i.e., the adjustment range of the PID parameters. Assuming the components include: CPU, memory, and PCIe device, the evaluation coefficients corresponding to the CPU, memory, and PCIe device are obtained respectively. These three evaluation coefficients are combined into an evaluation coefficient vector, which is then used as the input to the DDQN model.

[0124] Accordingly, the method further includes: training a DDQN model as the first model. A training method is provided, specifically including the following steps:

[0125] 1. Initialize the network: Initialize a target network and an evaluation network. These two networks have the same structure but different parameters. The evaluation network is used to calculate the Q-value in real time (which determines the agent's behavior), and the target network is used to calculate the target Q-value to help stabilize training.

[0126] 2. Environmental Interaction: At each time step, the agent selects an action (the adjustment range of the PID parameters) based on the current state (here, the evaluation coefficient of each component). This action is applied to the environment (e.g., adjusting the PID parameters of the fan), and the environment then provides feedback on the new state and corresponding reward.

[0127] 3. Calculate the reward: The reward function can be designed according to the system requirements. If adjusting the fan speed results in a decrease in temperature and no significant increase in noise, a positive reward can be given; conversely, if the temperature increases or the noise becomes excessive, a negative reward can be given.

[0128] 4. Store experience: Store the current state, action, reward, and next state in the experience replay buffer.

[0129] 5. Sampling Training: Randomly sample a batch of samples from the experience replay buffer, use the evaluation network to calculate the current Q-value, and use the target network to calculate the target Q-value. Update the parameters of the evaluation network by minimizing the error between the current Q-value and the target Q-value.

[0130] 6. Regularly update the target network: At regular intervals, copy the parameters of the evaluation network into the target network to stabilize the training process.

[0131] Through the above steps, the DDQN model can learn the optimal PID parameters based on environmental feedback, thereby achieving optimized control.

[0132] Thus, the trained DDQN model has an action space that is the PID parameter adjustment range (e.g., P ± 20%, I ± 15%, D ± 10%); its state space includes noise feature vectors, temperature gradient, and rotational speed deviation; and its reward function focuses on temperature rise suppression (a reward of +5 for every 1°C decrease) and noise penalty (a penalty of -3 for every 1dB increase). The optimal PID parameters are then output using this DDQN model.

[0133] The method provided in this disclosure takes into account that the operating conditions of a fan are constantly changing during actual operation. For example, factors such as temperature and load size will affect the optimal fan speed. In order to achieve dynamic adjustment of the fan speed, the method provides a method that uses a DDQN model to calculate the optimal PID parameters in real time based on real-time collected physical data (such as vibration (manifested through noise), temperature, speed, etc.), and dynamically adjusts the PID parameters to precisely control the fan speed and improve the heat dissipation effect.

[0134] In some embodiments, before determining the control parameters for heat dissipation based on the set of physical parameters and the noise, the method further includes:

[0135] Based on the set of physical parameters and the noise, determine whether a heat dissipation failure exists;

[0136] If a heat dissipation failure is determined, a fault alarm is generated, which is used to indicate the type of failure.

[0137] If it is determined that there is no heat dissipation fault, control parameters for heat dissipation are determined based on the set of physical parameters and the noise.

[0138] This paper provides a method for determining whether a heat dissipation fault has occurred based on physical parameters and noise levels. If a heat dissipation fault is detected, a fault alarm signal is issued to inform the user of the fault and indicate the type of fault, thus preventing damage to the device due to overheating. If no fault is detected, the heat dissipation control parameters are determined by adjusting them (e.g., adjusting fan speed) based on physical parameters and noise data to ensure good heat dissipation.

[0139] In some embodiments, generating a fault alarm includes:

[0140] If the fault type belongs to the target fault type and the fault anomaly level exceeds the level threshold, a fault alarm is generated based on the fault type and the fault anomaly level.

[0141] Here, fault types and fault severity levels can be predefined so that fault alarms can be generated by judging the fault type and its severity.

[0142] In one example, the fault type may include at least one of the following:

[0143] Impeller imbalance: The impeller vibrates or is damaged due to imbalance during operation;

[0144] Bearing wear: Wear on bearings can cause machines to malfunction or be damaged.

[0145] Foreign object interference: Foreign objects inside the machine affect its normal operation.

[0146] Each type of fault is specifically classified into fault severity levels to reflect its severity. These levels are usually represented by 0 to 5, where level 0 indicates no fault and level 5 indicates the most severe fault.

[0147] For example, if the detected fault type matches the system's predetermined target type (such as impeller imbalance, bearing wear, etc.), the system will then determine whether the fault severity level exceeds a threshold. If the fault severity level exceeds a set threshold (such as a threshold of 3 or 4), a fault alarm will be generated. The generated fault alarm will include the fault type and fault severity level, serving as alert information to help maintenance personnel or the system take appropriate action.

[0148] It should be noted that the above is only one example of fault type and fault severity classification. In actual applications, different designs may be used for different electronic devices and their designs. The above example is not intended to limit the classification.

[0149] In some embodiments, determining whether a heat dissipation fault exists based on the set of physical parameters and the noise includes:

[0150] Time-series data of rotational speed and / or temperature are generated based on the set of physical parameters;

[0151] A first feature set is extracted based on the noise, and the first feature set includes at least one of the following: Blade Passing Frequency (BPF) and Harmonic Amplitude Ratio (HAR).

[0152] The first feature set, the time-series data of the rotational speed and / or temperature are input into the second model to obtain the output of the second model. The output is used to indicate the fault type and fault anomaly level.

[0153] Here, the time-series data of rotation speed and / or temperature is time-series data including rotation speed and various temperatures established based on the acquisition time; each temperature includes: the temperature of each component, ambient temperature, etc.; if parameters such as humidity and air pressure are acquired, it may also include: ambient humidity and air pressure inside the chassis.

[0154] Specifically, BPF (Blog Frequency Filter) refers to the frequency at which each blade passes a fixed point during rotation. It is generally used in rotating equipment, such as fan motors. Because the blades interact with the surrounding air or other objects during rotation, they generate periodic noise or vibration. This frequency is usually proportional to the blade's rotational speed. BPF can be used to analyze the operating status of a fan. If abnormal BPF noise is detected, it may indicate blade imbalance, wear, or damage. This analysis considers the noise frequencies that blades may generate during rotation and uses BPF extraction to analyze the blade's condition (such as abnormal rotational speed, missing blades, or damage). For example, a sudden change in BPF may indicate abnormal blade rotational speed or a reduction in the number of blades.

[0155] Specifically, HAR (Hyperradius Amplitude) is an indicator describing the amplitude ratio of each frequency component in a noise signal. Noise signals typically consist of multiple frequency components, including the fundamental frequency (original frequency) and its harmonic components. HAR assesses the characteristics of noise by calculating the amplitude ratio of each harmonic (such as the second harmonic, third harmonic, etc.) relative to the fundamental frequency. Changes in HAR can also reflect the performance status of equipment. For example, as equipment is used over time, HAR may change due to wear, aging, and other factors. By monitoring the trend of HAR changes, the performance degradation of the equipment can be assessed, allowing for early scheduling of maintenance and repairs. For instance, if the amplitude of a harmonic at a certain frequency increases abnormally, it may indicate a fault in a component. Therefore, HAR analysis can be used to identify abnormal vibrations or other signs of malfunction.

[0156] In practical applications, the extracted noise feature set (BPF and / or HAR) is input into the second model along with time-series data of other physical parameters (such as rotational speed, temperature, etc.). The time-series data represents the physical parameters as they change over time, i.e., they can reflect the changes during operation. The second model generates output results based on the input data.

[0157] The following is an application example: BMC has a data preprocessing function. After acquiring the noise generated by the heat dissipation device, it uses an adaptive filter (such as the LMS algorithm) to eliminate environmental background noise (such as the sound of the computer room air conditioner) to obtain the noise signal to be analyzed. Then, it performs a short-time Fourier transform (STFT) on the noise signal to generate a time-spectrum graph (e.g., a time window length of 20ms and an overlap rate of 50%). Finally, it uses a specific physical model to extract features BPF and / or HAR. BPF and / or HAR can help BMC detect the changing trend of server fans and issue alarms.

[0158] In some embodiments, the method further includes: generating the second model; the generation of the second model includes:

[0159] Obtain a training dataset; the training dataset includes: at least one training data and a label for each training data, each training data including: time-series data of sample noise, sample rotation speed and / or temperature; the label is the fault type and fault anomaly level;

[0160] The hybrid neural network is trained using the training dataset to obtain the trained hybrid neural network, which serves as the second model.

[0161] The hybrid neural network includes: a front-end network, a feature fusion layer, and an output layer;

[0162] The front-end network employs a first branch network and a second branch network. The first branch network is used to extract noise features, and the second branch network is used to extract features of time-series data of rotation speed and / or temperature.

[0163] The feature fusion layer is used to process data using an attention mechanism based on the features extracted by the front-end network;

[0164] The output layer is used to obtain the output result based on the data processing result of the feature fusion layer.

[0165] Specifically, the second model can employ a hybrid neural network architecture, such as... Figure 2 The diagram illustrates a hybrid neural network architecture, which includes:

[0166] 1) The front-end network can use a multi-branch CNN (such as a parallel first-branch network (1D-CNN) and a second-branch network (2D-CNN)).

[0167] 2) Feature fusion layer: This layer dynamically weights multimodal features (time-series data of noise, rotational speed, and / or temperature) using an attention mechanism. Here, the feature fusion layer calculates the similarity between each modal feature and the query vector using the attention mechanism, obtaining attention weights. These weights are then used to weight and sum the multimodal features, fusion to highlight important information and aid in the identification of abnormal vibrations.

[0168] 3) Output layer: outputs the fault type and fault level.

[0169] Specifically, when acquiring the training dataset, anomalies can be simulated by applying random frequency domain perturbations (±10% harmonic shift) to normal noise. Acquiring the training dataset can specifically include:

[0170] 1) Data preprocessing includes: collecting normal noise data from actually operating and normal-functioning equipment using a BMC (Browser Control Center). This data represents a series of sampling points in the time domain, and its spectral distribution can be obtained through Fourier transform in the frequency domain. The spectrum of the normal noise data is analyzed to identify the main harmonic frequency components. For example, the time-domain signal is converted into a frequency-domain signal using a Fast Fourier Transform (FFT), and then the peak frequencies in the spectrum are identified; these peak frequencies are the main harmonic frequencies.

[0171] 2) Apply random frequency domain perturbation, including: generating a random offset within ±10% for each major harmonic frequency. For example, if a harmonic frequency is f0, the random offset Δf would range from -0.1f0 ≤ Δf ≤ 0.1f0. This offset can be generated using a random number generator. Add the generated random offset to the original harmonic frequency to obtain the adjusted harmonic frequency f′ = f0 + Δf. Reconstruct the spectrum based on the adjusted harmonic frequency. The perturbation can be simulated by increasing or decreasing the amplitude at the adjusted harmonic frequency. For example, a Gaussian function or other appropriate function can be used to generate the amplitude at the new harmonic frequency while keeping the amplitudes of other frequency components unchanged or making appropriate adjustments.

[0172] 3) Frequency domain to time domain conversion, including: converting the reconstructed spectrum back to the time domain signal through inverse Fourier transform (IFFT) to obtain simulated abnormal vibration signal data.

[0173] Specifically, during the training of the second model, when the model confidence level falls below a threshold, a robotic arm is triggered to simulate a real anomaly in a laboratory environment, generating labeled data. For example... Figure 3 As shown, the training steps include:

[0174] Step 301: Model confidence assessment, including: setting a confidence threshold (denoted as T) in advance based on task requirements and model performance; in the fault identification task, if the maximum classification probability of a certain sample by the model is lower than the threshold T, it is considered that the model's prediction result for that sample is not reliable enough, triggering the robotic arm to simulate a real anomaly in the laboratory environment, and proceeding to step 302.

[0175] The collected physical parameters are input into the trained model, which then outputs a probability distribution for each possible fault type. For example, for three anomaly types—impeller imbalance, bearing wear, and foreign object interference—the model will output three probability values, P1, P2, and P3, corresponding to the probabilities of these three anomaly types, respectively. The confidence level is typically calculated by taking the maximum value from the probability distribution as the model's confidence level C, i.e., C = max(P1, P2, P3).

[0176] Step 302: Trigger the robot arm to simulate an anomaly and generate a dataset to be labeled, including:

[0177] Identify low-confidence samples: When C < T, determine that the sample is a low-confidence sample.

[0178] Determine the type of anomaly: Although the prediction result of the model for low-confidence samples is unreliable, the possible direction of the anomaly type can still be roughly judged based on the probability distribution. For example, if P1 is relatively large, it can be initially considered that there may be an anomaly related to impeller imbalance.

[0179] Robot arm simulates an anomaly: Control the robot arm to operate on the device in a laboratory environment to simulate the corresponding abnormal situation. For example, if it is initially judged as impeller imbalance, the robot arm can simulate the working condition of impeller imbalance by adjusting the position of the counterweight of the impeller or changing the installation angle of the impeller. During the process of the robot arm simulating an anomaly, use BMC and related sensors to collect multi-modal data such as the vibration, speed, and temperature of the device as the generated labeled data.

[0180] Step 303: Perform data labeling on the dataset to be labeled to obtain a labeled dataset, including: Label the data generated by the robot arm simulating an anomaly. For example, for the data simulating impeller imbalance, the labeling personnel can determine that it is an impeller imbalance anomaly and label the severity level (level 3) according to the degree of imbalance. Integrate the labeling results with the collected multi-modal data to form a complete labeled dataset.

[0181] Step 304: Use the labeled dataset for model update and optimization, including: Add the newly generated labeled data to the original training dataset to expand the scale and diversity of the training data. Retrain the neural network model using the expanded training dataset. An incremental learning method can be adopted, that is, on the basis of the original model parameters, continue to use the new data for training to retain the model's learning ability for the original data and at the same time improve the model's recognition ability for new abnormal situations.

[0182] Evaluate the updated model on an independent test set to check whether the performance of the model has been improved. If indicators such as the accuracy rate and recall rate of the model have increased, it means that the active learning strategy is effective; otherwise, the reason needs to be further analyzed, and the strategy or model structure needs to be adjusted.

[0183] In some embodiments, the method further includes:

[0184] Obtain a failure database of at least one fan, and fine-tune the second model according to the failure database to obtain the second model corresponding to the electronic device using the fan.

[0185] Here, after obtaining the second model mentioned above, it can be fine-tuned based on the fault databases of various fan manufacturers (such as the EBMPapstNBR dataset) to adapt to various fan models, thus obtaining a second model for each fan model. In this way, the second model used by BMC can be obtained by matching the fan model of the electronic device, and using the matched second model for identification can improve the accuracy of identification.

[0186] The method provided in this disclosure dynamically adjusts the control parameters of the cooling fan based on the temperature and real-time noise of each key component of the electronic device. This approach balances the overall heat dissipation needs of the electronic device with effective management and control of its power consumption, thereby improving resource utilization efficiency and reducing operating costs. It represents a leap from "passive response" to "intelligent prediction" and from "single temperature control" to "multi-objective optimization," making it particularly suitable for demanding noise-sensitive environments such as high-density data centers and medical equipment rooms. It combines economic efficiency with technological foresight.

[0187] During the control process, the second model can be used to accurately identify the fan's operating status (including normal and various abnormal states), and the first model can be used to dynamically control the fan speed to ensure that the fan can operate efficiently and stably under different operating conditions.

[0188] Among them, by identifying abnormal states, potential faults (such as fan imbalance, hard drive mechanical failure, or loose heat sinks) can be detected, providing maintenance personnel with accurate diagnostic basis, predictive maintenance reduces the need for emergency repairs, optimizes spare parts inventory and manpower scheduling, and extends equipment life cycle (such as avoiding premature replacement of expensive components).

[0189] By calculating the optimal control parameters, the BMC can dynamically control the fan speed, while optimizing multiple dimensions such as heat dissipation efficiency, noise control, energy consumption, and hardware lifespan. This breaks through the limitations of traditional single-objective optimization (such as only reducing temperature). Through reinforcement learning, it continuously explores the optimal control strategy, adapts to long-term environmental changes (such as seasonal temperature fluctuations) and hardware aging (such as fan efficiency decline), and achieves precise heat dissipation control to meet customer noise requirements.

[0190] Figure 4 This is a schematic diagram of the structure of a heat dissipation control device provided in an embodiment of this disclosure; as shown below. Figure 4 The device shown is used in an electronic device, the electronic device including a heat dissipation device, and the device includes:

[0191] An acquisition module is used to acquire the set of physical parameters of the electronic device and the noise generated by the heat dissipation device;

[0192] The first processing module is used to determine the control parameters for heat dissipation based on the set of physical parameters and the noise.

[0193] The second processing module is used to control the heat dissipation device of the electronic device according to the control parameters.

[0194] In some embodiments, the acquisition module is used to acquire at least one physical parameter of the electronic device and determine the physical parameter set based on the acquired at least one physical parameter;

[0195] The at least one physical parameter includes at least one of the following: component temperature, fan speed; the component includes: CPU, memory, PCIe device.

[0196] In some embodiments, the first processing module is configured to determine an evaluation coefficient for at least one component based on a first rule, according to the temperature parameters of at least one component and the noise.

[0197] Based on the second rule, the control parameters are determined according to the evaluation coefficients of the at least one component.

[0198] In some embodiments, the first processing module is configured to determine a temperature weight and a temperature threshold corresponding to each component;

[0199] The evaluation coefficient for each component is determined based on its temperature parameters, temperature threshold, temperature weight, noise, and noise weight.

[0200] In some embodiments, the first processing module is configured to determine a temperature control index based on the temperature parameters of each component and the temperature threshold.

[0201] Based on the noise level and the reference noise level, noise control parameters are determined;

[0202] The evaluation coefficient for each component is determined based on the temperature control index, the noise control index, the temperature weight, and the noise weight.

[0203] In some embodiments, the first processing module is configured to determine an evaluation coefficient vector based on the evaluation coefficients of the at least one component;

[0204] The evaluation coefficient vector is input into the first model to obtain the output result of the first model, which is used to indicate the adjustment range of the PID parameters.

[0205] In some embodiments, the first processing module is configured to determine whether a heat dissipation fault exists based on the physical parameter set and the noise before determining the control parameters for heat dissipation based on the physical parameter set and the noise.

[0206] If a heat dissipation failure is determined, a fault alarm is generated, which is used to indicate the type of failure.

[0207] If it is determined that there is no heat dissipation fault, control parameters for heat dissipation are determined based on the set of physical parameters and the noise.

[0208] In some embodiments, the first processing module is configured to generate time-series data of rotational speed and / or temperature based on the set of physical parameters;

[0209] A first feature set is extracted based on the noise, and the first feature set includes: BPF and HAR;

[0210] The first feature set, the time-series data of the rotational speed and / or temperature are input into the second model to obtain the output of the second model. The output is used to indicate the fault type and fault anomaly level.

[0211] In some embodiments, the first processing module is configured to generate a fault alarm based on the fault type and the fault anomaly level if the fault type belongs to a target fault type and the fault anomaly level exceeds a level threshold.

[0212] In some embodiments, the first processing module is further configured to acquire a training dataset; the training dataset includes: at least one training data and a label for each training data, each training data including: time-series data of sample noise, sample rotation speed and / or temperature; the label is a fault type and a fault anomaly level;

[0213] The hybrid neural network is trained using the training dataset to obtain the trained hybrid neural network, which serves as the second model.

[0214] The hybrid neural network includes: a front-end network, a feature fusion layer, and an output layer;

[0215] The front-end network employs a first branch network and a second branch network. The first branch network is used to extract the features of the sample noise, and the second branch network is used to extract the features of the time-series data of the sample rotation speed and / or temperature.

[0216] The feature fusion layer is used to process data using an attention mechanism based on the features extracted by the front-end network;

[0217] The output layer is used to obtain the output result based on the data processing result of the feature fusion layer.

[0218] In some embodiments, the first processing module is further configured to acquire a fault database of at least one fan, and fine-tune the second model according to the fault database to obtain a second model corresponding to an electronic device using the fan.

[0219] In some embodiments, the acquisition module is configured to acquire a first noise generated by the fan and / or a second noise generated by the air duct, respectively;

[0220] The first noise and the second noise are combined to obtain the noise generated by the heat dissipation device.

[0221] It is understood that, when implementing the corresponding heat dissipation control method, the heat dissipation control device provided in the above embodiments can allocate the above processing to different program modules as needed to complete all or part of the processing described above. Furthermore, the device and the corresponding method embodiments provided in the above embodiments belong to the same concept, and their specific implementation process is detailed in the method embodiments, which will not be repeated here.

[0222] This application provides a computer program product or computer program that includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform a heat dissipation control method.

[0223] This application provides a computer-readable storage medium storing executable instructions, wherein the executable instructions are stored and when executed by a processor, they will cause the processor to execute the heat dissipation control method provided in this application.

[0224] In some embodiments, the computer-readable storage medium may be a memory such as FRAM, ROM, PROM, EPROM, EEPROM, flash memory, magnetic surface memory, optical disk, or CD-ROM; or it may be a variety of devices including one or any combination of the above-mentioned memories.

[0225] In some embodiments, executable instructions may take the form of a program, software, software module, script, or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as a standalone program or as a module, component, subroutine, or other unit suitable for use in a computing environment.

[0226] As an example, executable instructions may, but do not necessarily, correspond to files in a file system. They may be stored as part of a file that holds other programs or data, for example, in one or more scripts in a Hyper Text Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple collaborating files (e.g., a file that stores one or more modules, subroutines, or code sections).

[0227] As an example, executable instructions can be deployed to execute on a single computing device, or on multiple computing devices located in one location, or on multiple computing devices distributed across multiple locations and interconnected via a communication network.

[0228] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this disclosure; as shown below. Figure 5 As shown, the electronic device 50 includes: a processor 501, and a memory 502 communicatively connected to the processor 501; the memory 502 stores instructions executable by the processor 501. The instructions are executed by the processor 501 to enable the processor 501 to perform:

[0229] Obtain the set of physical parameters of the electronic device and the noise generated by the heat dissipation device;

[0230] Based on the set of physical parameters and the noise, determine the control parameters for heat dissipation;

[0231] The heat dissipation device of the electronic device is controlled according to the control parameters.

[0232] The electronic devices and corresponding control methods provided in the above embodiments belong to the same concept, and their specific implementation process can be found in the method embodiments, which will not be repeated here.

[0233] In practical applications, the electronic device 50 may further include at least one network interface 503. The various components of the electronic device 50 are coupled together via a bus system 504. It is understood that the bus system 504 is used to implement communication between these components. In addition to a data bus, the bus system 504 also includes a power bus, a control bus, and a status signal bus. However, for clarity, in... Figure 5 All buses are labeled as bus system 504. The number of processors 501 and the number of memories 502 can be at least one. The network interface 503 is used for wired or wireless communication between the electronic device 50 and other devices.

[0234] The memory 502 in this embodiment is used to store various types of data to support the operation of the electronic device 50.

[0235] The methods disclosed in the above embodiments of this disclosure can be applied to processor 501, or implemented by processor 501. Processor 501 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the integrated logic circuit of the hardware in processor 501 or by instructions in the form of software. The processor 501 may be a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Processor 501 can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this disclosure. A general-purpose processor may be a microprocessor or any conventional processor, etc. The steps of the methods disclosed in the embodiments of this disclosure can be directly manifested as being executed by a hardware decoding processor, or being executed by a combination of hardware and software modules in the decoding processor. The software modules may be located in a storage medium, which is located in memory 502. Processor 501 reads the information in memory 502 and combines it with its hardware to complete the steps of the aforementioned control method.

[0236] In some embodiments, the electronic device 50 may be implemented by one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), general-purpose processors, controllers, microcontrollers (MCUs), microprocessors, or other electronic components to perform the aforementioned methods.

[0237] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0238] In the above description, the term "some embodiments" refers to a subset of all possible embodiments. However, it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0239] Unless otherwise defined, all technical and scientific terms used in this disclosure have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs. The terminology used in this disclosure is for the purpose of describing embodiments of this disclosure only and is not intended to be limiting of this disclosure.

[0240] It should be understood that in the various embodiments of this disclosure, the sequence number of each implementation process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this disclosure.

[0241] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this disclosure, "a plurality of" means two or more, unless otherwise explicitly specified.

[0242] The above description is merely a specific embodiment of this disclosure, but the scope of protection of this disclosure is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this disclosure should be included within the scope of protection of this disclosure. Therefore, the scope of protection of this disclosure should be determined by the scope of the claims.

Claims

1. A heat dissipation control method, characterized in that, The method is applied to an electronic device, the electronic device including a heat dissipation device, and the method includes: The electronic device acquires a set of physical parameters and noise generated by its heat dissipation devices; the set of physical parameters includes at least one physical parameter, which includes at least one of the following: component temperature, fan speed; the components include: a central processing unit (CPU), memory, and peripheral components interconnected by a standard PCIe device. Based on the set of physical parameters and the noise, determine the control parameters for heat dissipation; The heat dissipation device of the electronic device is controlled according to the control parameters; The step of determining the control parameters for heat dissipation based on the set of physical parameters and the noise includes: Based on the first rule, the evaluation coefficient of at least one component is determined according to the temperature parameter of at least one component and the noise. Based on the second rule, the control parameters are determined according to the evaluation coefficients of the at least one component; the second rule is used to indicate that after obtaining the evaluation coefficients, the control parameters are determined based on the evaluation coefficient analysis. Based on the first rule, at least one control index is determined according to the temperature parameter of at least one component and the noise, including: Determine the temperature weight and temperature threshold corresponding to each component; An evaluation coefficient for each component is determined based on its temperature parameters, temperature threshold, temperature weight, noise, and noise weight. The evaluation coefficient is used to measure the operating status of the heat dissipation device, or to reflect the control indicators that currently require strengthening or weakening heat dissipation. The determination of the control parameters based on the evaluation coefficients of the at least one component, according to the second rule, includes: Based on the evaluation coefficients of the at least one component, determine the evaluation coefficient vector; The evaluation coefficient vector is input into the first model to obtain the output result of the first model, and the output result is used to indicate the adjustment range of the PID parameters; Before determining the control parameters for heat dissipation based on the set of physical parameters and the noise, the method further includes: Based on the set of physical parameters and the noise, determine whether a heat dissipation failure exists; If a heat dissipation failure is determined, a fault alarm is generated, which is used to indicate the type of failure. If it is determined that there is no heat dissipation fault, the control parameters for heat dissipation are determined based on the set of physical parameters and the noise. Determining whether a heat dissipation fault exists based on the set of physical parameters and the noise includes: Time-series data of rotational speed and / or temperature are generated based on the set of physical parameters; A first feature set is extracted based on the noise, and the first feature set includes: blade passing frequency (BPF) and harmonic component amplitude ratio (HAR). The first feature set, the time-series data of the rotation speed and / or temperature are input into the second model to obtain the output of the second model. The output is used to indicate the fault type and fault anomaly level. The generation of fault alarms includes: if the fault type belongs to the target fault type and the fault anomaly level exceeds the level threshold, generating a fault alarm based on the fault type and the fault anomaly level; Acquiring the noise generated by the heat dissipation device includes: The first noise generated by the fan and the second noise from the air duct were obtained respectively; The first noise and the second noise are combined using weighted merging or spectrum analysis to obtain the noise generated by the heat dissipation device.

2. The method according to claim 1, characterized in that, The acquisition of the physical parameter set of the electronic device includes: Collect at least one physical parameter of the electronic device, and determine the set of physical parameters based on the collected at least one physical parameter.

3. The method according to claim 1, characterized in that, Based on the temperature parameters of each component, the noise, the temperature weight, and the noise weight, an evaluation coefficient for each component is determined, including: The temperature control parameters are determined based on the temperature parameters of each component and the temperature threshold. Based on the noise level and the reference noise level, noise control parameters are determined; The evaluation coefficient for each component is determined based on the temperature control index, the noise control index, the temperature weight, and the noise weight.

4. The method according to claim 1, characterized in that, The method further includes: generating the second model; the generation of the second model includes: Obtain a training dataset; the training dataset includes: at least one training data and a label for each training data, each training data including: time-series data of sample noise, sample rotation speed and / or temperature; the label is the fault type and fault anomaly level; The hybrid neural network is trained using the training dataset to obtain the trained hybrid neural network, which serves as the second model. The hybrid neural network includes: a front-end network, a feature fusion layer, and an output layer; The front-end network employs a first branch network and a second branch network. The first branch network is used to extract the features of the sample noise, and the second branch network is used to extract the features of the time-series data of the sample rotation speed and / or temperature. The feature fusion layer is used to process data using an attention mechanism based on the features extracted by the front-end network; The output layer is used to obtain the output result based on the data processing result of the feature fusion layer.

5. The method according to claim 4, characterized in that, The method further includes: Obtain a fault database of at least one fan, and fine-tune the second model based on the fault database to obtain a second model corresponding to the electronic device using the fan.

6. A heat dissipation control device, characterized in that, The device is used in an electronic device, the electronic device including a heat dissipation device, and the device includes: An acquisition module is used to acquire the physical parameter set of the electronic device and the noise generated by the heat dissipation device; the physical parameter set includes at least one physical parameter, and the at least one physical parameter includes at least one of the following: component temperature, fan speed; the components include: central processing unit (CPU), memory, peripheral components, and a PCIe standard interconnect device; The first processing module is used to determine the control parameters for heat dissipation based on the set of physical parameters and the noise. The second processing module is used to control the heat dissipation device of the electronic device according to the control parameters; The first processing module is configured to determine an evaluation coefficient for at least one component based on a first rule, according to the temperature parameters of at least one component and the noise; and to determine the control parameters based on the evaluation coefficients of the at least one component based on a second rule; the second rule is configured to instruct that the control parameters be determined based on the evaluation coefficients after obtaining them. The first processing module is configured to determine the temperature weight and temperature threshold corresponding to each component; and to determine the evaluation coefficient of each component based on the temperature parameters, the temperature threshold, the temperature weight, the noise, and the noise weight. The evaluation coefficient is used to measure the working state of the heat dissipation device, or to reflect the control indicators that currently require strengthening or weakening heat dissipation. The first processing module is configured to determine an evaluation coefficient vector based on the evaluation coefficients of the at least one component; input the evaluation coefficient vector into a first model to obtain the output result of the first model, wherein the output result is used to indicate the adjustment range of the PID parameters; The first processing module is used to determine whether a heat dissipation fault exists based on the physical parameter set and the noise before determining the control parameters for heat dissipation based on the physical parameter set and the noise. If a heat dissipation failure is determined, a fault alarm is generated, which is used to indicate the type of failure. If it is determined that there is no heat dissipation fault, the control parameters for heat dissipation are determined based on the set of physical parameters and the noise. The first processing module is used to generate time-series data of rotational speed and / or temperature based on the set of physical parameters; A first feature set is extracted based on the noise, and the first feature set includes: BPF and HAR; The first feature set, the time-series data of the rotation speed and / or temperature are input into the second model to obtain the output of the second model. The output is used to indicate the fault type and fault anomaly level. The first processing module is used to generate a fault alarm based on the fault type and the fault anomaly level if the fault type belongs to the target fault type and the fault anomaly level exceeds the level threshold. The acquisition module is used to acquire the first noise generated by the fan and the second noise of the air duct respectively; and to combine the first noise and the second noise by weighted merging or spectrum analysis to obtain the noise generated by the heat dissipation device.

7. An electronic device, characterized in that, include: At least one processor; And a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of any one of claims 1 to 5.

8. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, The computer instructions are used to cause the computer to perform the method according to any one of claims 1 to 5.

Images