A control method and device of a temperature adjustment module and a computing device

Abstract

The embodiment of the application discloses a temperature adjusting module control method, device and computing equipment, the method comprises the following steps: determining a temperature control error and a temperature index parameter according to the temperature of a plurality of temperature control points in the computing equipment; if the temperature index parameter satisfies an adjustment condition, determining a target control parameter according to the temperature control error and a control model; and controlling the operation of the temperature adjusting module based on the target control parameter. In this way, flexible temperature control can be realized.

Description

Technical Field

[0001] This application relates to the field of computing devices, and more particularly to a control method, apparatus, and computing device for a temperature regulation module. Background Technology

[0002] As a high-performance computer, a server provides users with a variety of functions through networks, such as data analysis, multimedia data publishing, information interaction, and storage. With the increasing functionality of servers, the performance requirements also become higher. The integration of diverse processing chips such as CPUs, GPUs, and IPUs into a single server transforms it into a multi-heat-generating, highly dynamic, and high-concurrency heat-generating system. Furthermore, with modular design, servers are compatible with various peripherals and functional modules, leading to even more complex and diverse heat generation issues.

[0003] Current heat dissipation solutions include air cooling, water cooling, etc. These solutions simply adjust the heat dissipation based on the internal temperature of the server. The temperature control method is singular, and the heat dissipation effect is poor for servers with many heat sources. Summary of the Invention

[0004] This application provides a control method, device, and computing equipment for a temperature regulation module, which can flexibly regulate and control the temperature of the device.

[0005] In a first aspect, embodiments of this application provide a control method for a temperature regulation module, including:

[0006] Based on the temperatures of multiple temperature control points within the computing device, the temperature control error and temperature index parameters are determined; if the temperature index parameters meet the adjustment conditions, the target control parameters are determined based on the temperature control error and the control model; and the operation of the temperature regulation module is controlled based on the target control parameters.

[0007] In this technical solution, temperature index parameters related to the collected temperature are used to determine whether to perform adaptive adjustments to parameters such as PID. When the temperature index parameters meet the adjustment conditions for adaptive adjustment, control parameters can be generated by combining temperature control error and control model. Then, the operation of temperature regulation modules such as fans can be controlled through these control parameters. This allows for flexible and more accurate temperature control of computing devices based on the measured temperature, enhancing the dynamic response of temperature control and enabling temperature control to adapt to faster temperature changes.

[0008] In one implementation of this embodiment, determining the target control parameter based on the temperature control error and control model if the temperature index parameter meets the adjustment conditions includes: if the temperature index parameter meets the adjustment conditions, iterating the process control parameter at least once based on the temperature control error and control model; wherein the process control parameter is configured to control the temperature adjustment module to work for a preset duration; if the temperature adjustment module operates based on the latest process control parameter, causing the temperature index parameter to not meet the adjustment conditions, then the latest process control parameter is determined as the target control parameter.

[0009] In this implementation, an appropriate control parameter is determined through iterative processing and continuous trial and error to control the operating temperature of computing devices such as servers. This approach can effectively adapt to temperature control in various scenarios, improve the automation and intelligence of temperature control, and achieve more precise heat dissipation control under changes in the working environment, components, and applications of computing devices.

[0010] In one implementation of this embodiment, after controlling the operation of the temperature regulation module based on the target control parameters, the method further includes: S1, acquiring the temperatures of multiple temperature control points within the computing device again, and determining the temperature control update error and temperature index update parameters based on the acquired temperatures of the multiple temperature control points within the computing device; S2, if the temperature index update parameters meet the adjustment conditions, acquiring updated control parameters based on the temperature control update error and the control model; S3, controlling the operation of the temperature regulation module based on the updated control parameters; and after controlling the operation of the temperature regulation module based on the updated control parameters, iterative processing is performed according to S1, S2, and S3 until the temperature control update error no longer meets the adjustment conditions.

[0011] In this implementation, after the temperature regulation module has been controlled, the system can again use the temperature index parameters related to the newly acquired temperature (i.e., the temperature index update parameters mentioned above) to determine whether to continue adaptive parameter adjustment. This process continues until the temperature index parameters determine that adaptive parameter adjustment is not needed (the adjustment conditions are not met). After one or more cycles, suitable PID parameters are obtained, which in turn provide appropriate control parameters to control the temperature regulation module, adjusting the temperature of the computing device to a suitable range or below a certain temperature threshold. This provides a more flexible PID parameter adjustment method and is applicable to more heat dissipation and cooling scenarios.

[0012] In one implementation of this embodiment, the control model is obtained based on a PID control strategy and a neural network learning algorithm. Determining the target control parameters based on the temperature control error and the control model includes: determining the PID parameter adjustment based on recorded historical control parameters, recorded historical PID parameter adjustments used to determine the historical control parameters, and historical temperature control errors; and obtaining the control parameters using a neural network learning algorithm based on the configured PID parameters and PID parameter adjustments. The aforementioned target control parameters or process control parameters can all be obtained in this manner.

[0013] In this implementation, the PID control strategy and the neural network learning algorithm are combined. The control parameters at the current moment are obtained from the PID parameter adjustment, control parameters, temperature control error at the historical moment and the temperature control error at the current moment. The actual test results show that the PID parameters can be adapted more quickly, and the temperature regulation module such as the fan can be controlled more quickly, realizing temperature regulation in a variety of scenarios.

[0014] In one implementation of this embodiment, determining the temperature control error based on the temperatures of multiple temperature control points within the computing device includes: acquiring the temperature of each temperature control point within the computing device, and obtaining the temperature deviation of each temperature control point based on the temperature of each temperature control point and the desired temperature; and performing a weighted summation based on the weight values ​​corresponding to each temperature control point and the temperature deviations to obtain the temperature control error. The desired temperature is a preset value; the server can maintain good performance when operating at the desired temperature.

[0015] In this implementation, the temperature control error calculated using the above method can better reflect the temperature status of the computing device. The obtained temperature control error can be better used to calculate subsequent control parameters, so as to better control the temperature and achieve cooling control within the computing device.

[0016] In one implementation of this embodiment, before obtaining the temperature control error by weighted summation based on the weight values ​​corresponding to each temperature control point and each temperature deviation, the method further includes: setting the weight values ​​corresponding to each temperature control point based on the attribute information of the heat source module corresponding to each temperature control point; wherein, the attribute information of the heat source module includes any one or two of the following: the type of the heat source module, and the distance between the heat source module and the temperature regulation module in the calculation device.

[0017] In this implementation, different calculation weights are set for different heat sources, which allows for more targeted temperature control error calculation of the heat sources. This leads to better control parameters for cooling the computing device and better cooling of the computing device.

[0018] In one implementation of this embodiment, the temperature index parameter is used to represent the degree of deviation between the temperature corresponding to each temperature control point and the desired temperature.

[0019] In this implementation, the deviation between the temperature at the temperature control point and the desired temperature is used as a criterion to determine whether to perform adaptive adjustment of the PID parameters. This can promptly trigger the acquisition of the required control parameters through the control model, which is beneficial for better cooling within the computing device.

[0020] In one implementation of this embodiment, the temperature index parameter includes: the variance calculated based on the temperature corresponding to each temperature control point and the desired temperature.

[0021] In this implementation, the variance between the temperature at the control point and the desired temperature is used as a criterion to determine whether to perform adaptive adjustment of the PID parameters. This can promptly trigger the acquisition of the required control parameters through the control model, which is beneficial for better cooling within the computing device.

[0022] In one implementation of this embodiment, the neuron learning algorithm is a single-neuron learning algorithm. The input of the single-neuron learning algorithm is the configured PID parameters, the adjustment amount of the PID parameters is the weight value, and the output of the single-neuron learning algorithm is the control parameter. The single-neuron learning algorithm sums the proportional P parameter included in the configured PID parameters with the proportional P parameter adjustment amount in the PID parameter adjustment amount, sums the integral I parameter included in the configured PID parameters with the integral I parameter adjustment amount in the PID parameter adjustment amount, and sums the derivative D parameter included in the configured PID parameters with the derivative D parameter adjustment amount in the PID parameter adjustment amount. Then, it predicts the target control parameter through a prediction function. The control parameter output by the single-neuron learning algorithm can be the target control parameter or a process control parameter, from which the target control parameter is further obtained.

[0023] In this implementation, the PID control strategy and the single-neuron learning algorithm are further combined to deploy a computational method to obtain control parameters, making the calculation simpler and more efficient.

[0024] In one implementation of this embodiment, the method further includes: if the temperature index parameter does not meet the adjustment conditions, generating control parameters according to the configured PID parameters to control the operation of the temperature regulation module of the computing device.

[0025] This implementation method can perform adaptive temperature control based on PID parameters, and can also avoid adaptive adjustment of PID parameters when the adjustment conditions are not met, thereby improving the efficiency of temperature adjustment and meeting the requirements for automation and intelligence of adaptive adjustment.

[0026] In one implementation of this embodiment, determining the target control parameters based on the temperature control error and the control model includes: obtaining the target control parameters based on the temperature control error and the neural network model; wherein the neural network model is obtained after supervised training based on temperature training parameters under multiple scenarios and control parameters determined experimentally in each scenario. Different scenarios correspond to different equipment temperature control requirements, such as scenarios where servers operate under different workloads (full load or half load, etc.), scenarios with different aging levels, etc.

[0027] In this implementation, the neural network model trained using supervised data obtained from experiments can obtain control parameters more quickly, efficiently, and accurately, thereby achieving temperature control of the computing device.

[0028] In one implementation of this embodiment, the method further includes: acquiring temperature change information, the temperature change information including the change value before and after controlling the operation of the temperature regulation module; if the temperature change information meets the temperature rise condition of the computing device, then shortening the preset duration; or, if the temperature change information meets the temperature decrease condition of the computing device, then increasing the preset duration.

[0029] In this implementation, the preset duration for controlling the operation of the temperature regulation module can be dynamically adjusted as needed. This can, to a certain extent, meet the adaptive temperature control requirements of the computing device, while also reducing the computational load and saving energy by increasing the time interval.

[0030] In one implementation of this embodiment, each temperature control point within the computing device corresponds to at least one heat source module, and the heat source module includes any one or more of the central processing unit, graphics processing unit, and image processing unit disposed within the computing device.

[0031] In this implementation, more comprehensive temperature monitoring of computing devices can be achieved. Starting from more comprehensive temperature monitoring, better adaptive adjustment of control parameters can be realized, and better temperature control of computing devices can be achieved.

[0032] Secondly, embodiments of this application provide a control device for a temperature regulation module, comprising:

[0033] The acquisition unit is used to acquire the temperature of the temperature control point in the computing device;

[0034] The processing unit is used to determine the temperature control error and temperature index parameters based on the temperatures of multiple temperature control points within the computing device; if the temperature index parameters meet the adjustment conditions, it determines the target control parameters based on the temperature control error and the control model; and controls the operation of the temperature regulation module based on the target control parameters.

[0035] Thirdly, embodiments of this application provide a computing device, which includes a processor for implementing the control method of the temperature regulation module as described in the first aspect above.

[0036] Fourthly, this application also provides a computing device, including: a state detection module, an adaptive learning module, a PID control module, and a temperature regulation module, wherein: the state detection module is used to determine a temperature control error and a temperature index parameter based on the temperatures of multiple temperature control points within the computing device, and to notify the adaptive learning module via a trigger signal when the temperature index parameter meets the adjustment conditions; the adaptive learning module is used to obtain a PID parameter adjustment amount based on the temperature control error when receiving the trigger signal; the PID control module is used to determine a target control parameter based on the PID parameter adjustment amount and a control model, and to control the temperature regulation module based on the target control parameter; the temperature regulation module is used to perform cooling processing under the control of the target control parameter.

[0037] Fifthly, embodiments of this application provide a computer-readable storage medium storing a computer program, the computer program including program instructions, which, when executed, cause the control method of the temperature regulation module described in the first aspect to be implemented.

[0038] In a sixth aspect, embodiments of this application provide a computer program product including a computer program or instructions, which, when executed on a computer, causes the computer to perform the control method of the temperature regulation module as described in the first aspect. Attached Figure Description

[0039] Figure 1a This is a schematic diagram of the system architecture of a computing device provided in an embodiment of this application;

[0040] Figure 1b This is a schematic diagram of the architecture of a temperature control system provided in an embodiment of this application;

[0041] Figure 2 This is a flowchart illustrating a control method for a temperature regulation module provided in an embodiment of this application;

[0042] Figure 3This is a schematic diagram illustrating the acquisition of control parameters through a single-neuron learning algorithm;

[0043] Figure 4 This is a flowchart illustrating another control method for a temperature regulation module according to an embodiment of this application;

[0044] Figure 5 This is a flowchart illustrating another control method for a temperature regulation module according to an embodiment of this application;

[0045] Figure 6 This is a schematic diagram of the structure of a control device for a temperature regulation module according to an embodiment of this application;

[0046] Figure 7 This is a schematic diagram of the structure of a computing device provided in this application;

[0047] Figure 8 This is a schematic diagram of the structure of a chip provided in this application. Detailed Implementation

[0048] To better understand the technical solutions provided in the embodiments of this application, some technical terms involved in the embodiments of this application will be introduced first.

[0049] (1) PID (Proportional, Integral, Differential)

[0050] PID control algorithm: A control algorithm that combines proportional, integral, and differential functions. Based on the input deviation, it performs calculations according to the proportional, integral, and differential functional relationships, and the result is used to control the output. A PID control module based on the PID algorithm consists of a proportional unit, an integral unit, and a differential unit, and uses the settings of PID parameters (Xp, Xi, and Xd) to control target objects such as fans.

[0051] (2) Equipment Management Unit

[0052] The device management unit (DMU) performs functions such as component management and asset management within the computer system, including power-on / off control, fan speed regulation, and electronic tag management. The DMU can integrate numerous platform management functions. For example, platform management functions may include a series of monitoring and control functions that operate on system hardware. This includes monitoring system temperature, voltage, fans, power supply, etc., and making corresponding adjustments to ensure the system is in a healthy state. Platform management functions may also include recording various hardware information and logging data to alert users and help locate subsequent problems.

[0053] It should be noted that different equipment manufacturers may use different names for the device management unit. In this embodiment of the application, the device management unit may be referred to as the Baseboard Management Controller (BMC). It is understood that the device management unit may also have other names.

[0054] The temperature regulation module control method of this application embodiment can be applied to computing devices with heat sources, such as servers including multiple heat sources. For example, a system architecture diagram of the computing device can be shown as follows: Figure 1a As shown, a computing device is an electronic device with data processing, data transmission and reception, and data storage capabilities. For example, a computing device can be a rack-mount server, a full-rack server, or other similar server. The computing device 10 includes a rack 101, which contains, but is not limited to, a device management unit 102 and multiple power supplies 103. Figure 1a Taking three power supplies as an example, the device management unit 102 is electrically connected to multiple power supplies 103 via a bus 104. The device management unit 102 is primarily used for monitoring and managing the computing device 10. For example, the device management unit 102 can monitor the status of various hardware devices within the computing device (such as temperature and voltage). Furthermore, the device management unit 102 can perform system configuration, firmware upgrades, and fault diagnosis.

[0055] Power supply 103 can be a power supply unit (PSU).

[0056] The power supply 103 may include a processor, which may include various controllers with processing functions such as a microcontroller unit (MCU), a field-programmable gate array (FPGA), and a complex programmable logic device (CPLD). An MCU is a microcontroller unit that integrates a processing unit and peripheral modules. The processing unit is a functional component that interprets and executes instructions; in this embodiment, the processing unit may refer to a central processing unit (CPU). Peripheral modules may include, but are not limited to, random access memory (RAM), read-only memory (ROM), timers / counters, and input / output (I / O) ports. The processing unit and peripheral modules can be connected via a bus. An MCU may also be called a single-chip microcomputer or a microcontroller.

[0057] Bus 104 may include, but is not limited to: Controller Area Network (CAN) bus, InterIntegrated Circuit (I2C) bus, System Management Bus (SMBus), etc.

[0058] The system architecture of the computing device may also include: a motherboard 105; various chips may be configured on the motherboard 105, such as one or more CPUs (central processing unit), one or more GPUs (graphics processing unit), one or more IPUs (image processing unit), etc.

[0059] The system architecture of a computing device may also include, but is not limited to: basic hardware such as temperature control modules and fans; configurable components such as hard drives, network cards, and redundant arrays of independent disks (RAID); and various controllers with processing functions such as FPGAs and CPLDs.

[0060] In this embodiment of the application, the computing device is equipped with temperature sensors for heat sources, such as the power supply 103 and various chips on the motherboard 105. These temperature sensors are connected to the device management unit 102, collect the temperature of each heat source and send it to the device management unit 102, so that the device management unit 102 can perform cooling control on the temperature inside the computing device based on the received temperature of each heat source.

[0061] The control method for the temperature regulation module within a computing device provided in this application is described below. The control method for the temperature regulation module can be executed by a device management unit such as a BMC or a controller.

[0062] When controlling the heat dissipation of computing devices with multiple heat sources, PID control algorithms can be used to control the heat dissipation of temperature regulation modules such as air cooling and water cooling, keeping the computing device's temperature within a relatively suitable range. The process of heat dissipation control based on PID control algorithms can include: first, temperature information is collected by a temperature sensor and sent to the BMC (Body Control Module); then, after receiving the temperature information, if temperature control is required, the BMC outputs a PWM (Pulse Width Modulation) signal to the fan through the PID control algorithm, thereby adjusting the fan speed, accelerating heat dissipation, and lowering the temperature. The PID parameters used in this PID control algorithm can be parameters obtained after PID heat dissipation control and debugging for specific computing devices in a laboratory environment. In other words, the heat dissipation of computing devices can be controlled by a static PID control algorithm based on fixed parameters. Based on the heat dissipation control scheme based on PID control algorithms, the PID parameters can be further dynamically adjusted, allowing the PID heat dissipation control scheme to adapt to more scenarios. For example, when the internal components of the computing device experience increased temperature rise due to relative aging, the PID parameters can be adaptively adjusted instead of using fixed PID parameters, enabling the PID control algorithm to better achieve the cooling control of the computing device.

[0063] Please see Figure 1b This is a schematic diagram of the architecture of a temperature control system provided in an embodiment of this application. This embodiment of the application can achieve adaptive control of temperature regulation modules such as fans in computing devices, such as… Figure 1b As shown, the system includes a state detection module, an adaptive learning module, and a PID control module. In one possible implementation, the PID control module includes two parts: a unit that executes the PID control algorithm based on PID parameters and an execution unit. The execution unit can directly control the temperature regulation module, or it can be omitted. The PID control module outputs control parameters calculated based on the PID parameters, which are then used to indirectly control the temperature regulation module through an external unit.

[0064] The state detection module can determine the temperature control error and temperature index parameters based on the temperatures of multiple temperature control points within the computing device, and notify the adaptive learning module via a trigger signal when the temperature index parameters meet the adjustment conditions. The adaptive learning module, upon receiving the trigger signal, can obtain the PID parameter adjustment amount based on the temperature control error. The PID control module can determine the target control parameter based on the PID parameter adjustment amount and the control model, and control the temperature regulation module based on the target control parameter. The temperature regulation module within the computing device is used to perform cooling operations under the control of the target control parameter. Specifically, the temperature regulation module can refer to air-cooled, water-cooled, or other cooling modules used to lower the temperature within the computing device.

[0065] In one implementation of this embodiment, during the system operation of the computing device, the temperature control error e(t) can be obtained by detecting the measured temperature and the desired temperature per unit time through the state detection module, and temperature index parameters will also be obtained. These temperature index parameters represent the degree of deviation between the temperature corresponding to the temperature control point and the desired temperature. Specifically, the temperature index parameters can refer to the temperature variance obtained after calculating the variance of the temperatures and desired temperatures of each currently collected temperature control point. This temperature variance is used to determine the adaptability of the current PID parameters to temperature regulation modules such as fans. For example, by using configured PID parameters (which can be the current PID parameters or...),... If the temperature variance obtained after the control parameters generated by the user-configured PID parameters control the temperature regulation module are large and exceed the set range, it indicates that the configured PID parameters are not suitable for the server's temperature control in the current environment and have not generated suitable control parameters. In this case, it can be determined that the temperature index parameters meet the adjustment conditions. Therefore, the adaptive learning module can be activated to calculate the PID parameter adjustment amount (i.e., the proportional P parameter adjustment amount ωp(t), integral I parameter adjustment amount ωi(t), and D parameter adjustment amount ωd(t)) based on the temperature control error e(t), and adjust the configured PID parameters to obtain new control parameters u(t) for controlling the fan and other temperature regulation modules. The execution unit then obtains the control signal T(t) that can ultimately control the fan speed based on u(t). If the temperature variance obtained after controlling the temperature regulation module with the control parameters generated by the configured PID parameters is within the set range, it indicates that the temperature index parameters do not meet the adjustment conditions. This means that the configured PID parameters are suitable for server operation and can effectively control the temperature of the computing device. The configured PID parameters can be maintained to continue generating control parameters to control the temperature regulation module.

[0066] In other words, if the configured PID parameters can effectively control the temperature of the computing device and adapt to its operation, then the PID parameters will remain in operation. If the configured PID parameters cannot effectively control the temperature of the computing device and are not suitable for its operation, the adaptive learning module can be activated by a trigger signal to dynamically adjust the PID parameters and finally output appropriate control parameters to control the temperature regulation module, so that the corresponding temperature variance decreases to the set range, thereby ultimately achieving the goal of controlling the temperature in the computing device within a suitable temperature range.

[0067] It should be noted that, Figure 1b For illustrative purposes only, in one implementation of this embodiment, the functions of the state detection module, adaptive learning module, PID control module, and execution unit can each be implemented in a controller or BMC, or in other words, the state detection module, adaptive learning module, PID control module, and execution unit can be located in the controller or BMC.

[0068] Please see Figure 2 , Figure 2 This is a flowchart illustrating a control method for a temperature regulation module according to an embodiment of this application. The method can be executed by a controller or a BMC. The method includes, but is not limited to, the following steps:

[0069] S201: Determine the temperature control error and temperature index parameters based on the temperatures of multiple temperature control points within the computing device. In this embodiment, the temperature control error can be calculated based on the temperatures collected at each temperature control point at the current moment. For a computing device, there may be multiple heat source modules. Each temperature control point within the computing device corresponds to at least one heat source module. The heat source module includes any one or more of the following: central processing unit (CPU), graphics processing unit (GPU), and image processing unit (IPU). For example, some servers contain one or more CPUs (central processing unit), one or more GPUs (graphics processing unit), one or more IPUs (image processing unit), etc., which are heat source modules. Each heat source module can be determined as a temperature control point and its temperature can be collected as needed. For ease of description, the temperature control error and temperature index parameters obtained in S201 are defined as the temperature control error at the first moment and the temperature index parameters at the first moment. The calculation process for temperature control error can be referred to the description below. The temperature index parameter is used to represent the degree of deviation between the temperature corresponding to each temperature control point and the desired temperature. In one implementation, it can specifically be the variance calculated based on the temperature corresponding to each temperature control point and the desired temperature. The desired temperature is a preset temperature, which can be considered as the suitable temperature for the server to operate at.

[0070] In one implementation of this embodiment, determining the temperature control error based on the temperatures of multiple temperature control points within the computing device may specifically include: acquiring the temperatures of each temperature control point within the computing device, and obtaining the temperature deviation of each temperature control point based on the temperature of each temperature control point and the desired temperature; and performing a weighted summation based on the weight values ​​corresponding to each temperature control point and the temperature deviations to obtain the temperature control error. In other words, based on the temperature collected from each temperature control point, the temperature deviation between each collected temperature and the desired temperature can first be calculated, where the desired temperature can be a uniform temperature. Then, the temperature control error of each temperature control point can be statistically calculated using a temperature weighting method.

[0071] Before obtaining the temperature control error through weighted summation, it is necessary to determine the weight values ​​for the weighted calculation. The weight values ​​for each temperature control point can be set according to the attribute information of the heat source module corresponding to each temperature control point. The attribute information of the heat source module includes any one or two of the following: the type of the heat source module, and the distance between the heat source module and the temperature regulation module in the calculation device.

[0072] Table 1 below illustrates the temperature deviations of multiple heat source modules collected at time t relative to the desired temperature, such as e11(t), e22(t), and emn(t), as well as the weighting coefficients configured for each heat source module, such as p11, p22, and pmn. These weighting coefficients are empirical values. Users can pre-set the weight levels for different types of heat source modules according to their actual needs. Different weight levels will automatically configure corresponding weighting coefficients. For example, users can set the CPU to the first weight level, the GPU to the second weight level, and the IPU to the third weight level. This automatically assigns a first weighting coefficient to the CPU based on the first weight level, a second weighting coefficient to the GPU based on the second weight level, and a third weighting coefficient to the CPU based on the third weight level. The first, second, and third weighting coefficients are not the same. For instance, if the user's focus is on the CPU temperature, the CPU can be set to the first weight level, increasing the CPU's temperature weight, resulting in a larger first weighting coefficient for the CPU. In one implementation of this embodiment, weighting coefficients can be automatically assigned to temperature regulation modules at various levels within the computing device based on user-defined weighting levels. This allows for the weighted summation of temperature deviations at temperature control points corresponding to each heat source module, resulting in the temperature control error of the computing device.

[0073] For computing devices, weighting coefficients can be further determined based on the distance between the heat source modules and the temperature control modules within the device. Heat sources closer to the temperature control module are assigned higher weighting coefficients, while those farther away are assigned lower weighting coefficients. For each heat source module within the computing device, the user can record its identifier in a distance table or identifier sequence based on its distance from the temperature control module (either from farthest to nearest or from nearest to farthest). This allows for the automatic assignment of weighting coefficients to each temperature control module based on the distance table or identifier sequence, facilitating the weighted summation of temperature deviations at the corresponding temperature control points for each heat source module to obtain the temperature control error of the computing device.

[0074] In one implementation of this embodiment, the weighting levels set for each type of heat source module and the aforementioned distance table or identifier sequence can be comprehensively considered to automatically assign weighting coefficients to each temperature regulation module, so as to perform weighted summation calculation on the temperature deviation of the temperature control point corresponding to each heat source module and obtain the temperature control error of the calculation device.

[0075] The weighting coefficient can be a proportional relationship, and the final weight value can be obtained through normalization.

[0076] Table 1

[0077]

[0078] Based on the table above and the following formula, the temperature control error, namely the temperature control error e(t), can be obtained.

[0079]

[0080] S202: If the temperature index parameter meets the adjustment conditions, the target control parameter is determined based on the temperature control error and the control model. The temperature index parameter includes the aforementioned temperature variance of the calculation device, which can be calculated based on the difference between the temperature at each temperature control point and the desired temperature. Whether the temperature index parameter meets the adjustment conditions is mainly determined by comparing the temperature variance with a variance threshold, i.e., comparing whether the temperature variance is greater than or equal to the set variance threshold. The comparison result determines whether the temperature index parameter meets the adjustment conditions. If the comparison result is that the temperature variance at the first moment is greater than or equal to the set variance threshold, then the temperature index parameter is considered to meet the adjustment conditions. A new PID parameter adjustment amount needs to be adaptively generated to adjust the PID parameter. Based on the adjusted PID parameter, control parameters that can control temperature regulation modules such as fans are obtained. Conversely, if the temperature variance is less than the set variance threshold, then the temperature index parameter is considered not to meet the adjustment conditions. No adjustment to the PID parameter is required, and no new PID parameter adjustment amount needs to be calculated. A control parameter can be directly generated using the PID parameter corresponding to the current temperature index parameter not meeting the adjustment conditions, thereby controlling temperature regulation modules such as fans.

[0081] In one implementation of this embodiment, the control model can predict the current control parameters for controlling the fan based on relevant parameters from previous times (which can be considered as some stored historical parameters) and the current temperature control error. The control model can be obtained based on a PID control strategy and a neural network learning algorithm. The step of determining the target control parameters based on the temperature control error and the control model can include: determining the PID parameter adjustment based on recorded historical control parameters, recorded historical PID parameter adjustments used to determine the historical control parameters, and historical temperature control errors; and obtaining the control parameters through a neural network learning algorithm based on the configured PID parameters and PID parameter adjustments. The final control parameters obtained through the neural network learning algorithm can be the target control parameters or a process control parameter; the target control parameters need to be derived from the process control parameters.

[0082] Simultaneously, after each generation of control parameters, the generated control parameters, their corresponding PID parameter adjustments, and temperature control errors are stored as historical records in memory. This allows for the direct retrieval of the required historical control parameters, corresponding historical PID parameter adjustments, and historical temperature control errors from this historical record when new control parameters are needed. The most recently acquired control parameter is stored as the historical control parameter, its corresponding PID parameter adjustment is stored as the historical PID parameter adjustment, and its corresponding temperature control error is stored as the historical temperature control error.

[0083] The temperature control error and temperature index parameters obtained in S201 are defined as the temperature control error and temperature index parameters at the first moment. Therefore, S202 specifically includes: obtaining the PID parameter adjustment amount at the first moment based on the PID parameter adjustment amount at the third moment (historical PID parameter adjustment amount), the control parameters predicted at the third moment (historical control parameters), and the temperature control error at the first moment and the temperature control error at the third moment (historical temperature control error); and obtaining the control parameters at the first moment through a neural network learning algorithm based on the configured PID parameters and the PID parameter adjustment amount at the first moment. The third moment is before the first moment. The PID parameter adjustment amount at the first moment is calculated as follows.

[0084] Taking time t as the first time point, the corresponding PID parameter adjustment and control parameters will also be calculated at the time point before time t, i.e., the third time point (t-1 time point). Based on these data at time t-1 time point, the PID parameter adjustment at time t time point is calculated. Then, based on the PID parameter adjustment at time t time point and the existing PID parameters (i.e. the configured PID parameters), the control parameters at time t time point are obtained.

[0085] In one implementation of this embodiment, the neuron learning algorithm can be a single-neuron learning algorithm. The input of the single-neuron learning algorithm is the configured PID parameters, the PID parameter adjustment at the first time step is the weight value, and the output of the single-neuron learning algorithm is the control parameters at the first time step. The configured PID parameters are either default PID parameters or PID parameters calculated at the third time step. The PID parameters at the third time step can be obtained by summing the PID parameters calculated at the previous time step and the PID parameter adjustment calculated at the third time step, and so on. After summing the proportional P parameter included in the configured PID parameters with the proportional P parameter adjustment at the first time step, the integral I parameter included in the configured PID parameters with the integral I parameter adjustment at the first time step, and the derivative D parameter included in the configured PID parameters with the derivative D parameter adjustment at the first time step, the single-neuron learning algorithm then predicts the control parameters at the first time step using a prediction function.

[0086] The PID parameter adjustment mentioned above can be calculated using the following formula:

[0087] ωp(t)=ωp(t-1)+ηp*e(t)*u(t-1)*x1(t) Formula 2;

[0088] ωi(t)=ωi(t-1)+ηi*e(t)*u(t-1)*x2(t) Formula 3;

[0089] ωd(t)=ωd(t-1)+ηd*e(t)*u(t-1)*x3(t) Formula 4;

[0090] Among them, x1(t)=e(t)-e(t-1), x2(t)=e(t), x3(t)=e(t)-2e(t-1)+e(t-2).

[0091] ηp, ηi, and ηd represent the learning rates of the proportional, integral, and differential terms, respectively, and are considered as learning factors that can be customized based on empirical values. In one implementation of this embodiment, a very small value can be set, such as 0.1 or a smaller value. The smaller the value, the slower the learning rate, but the less prone it is to divergence. If a fast learning rate is not required, a small value can be set, but not too small.

[0092] e(t) is the temperature control error mentioned above, which represents the difference between the measured temperature and the desired temperature at time t. It can be calculated using Formula 1. e(t-1) represents the difference between the measured temperature and the desired temperature at time t-1, and e(t-2) represents the difference between the measured temperature and the desired temperature at time t-2.

[0093] u(t-1) represents the control parameter at time t-1. This control parameter is an adjustment used to control the fan speed. Fan systems are generally controlled via PWM, with the fan speed determined by the duty cycle. For example, a 40% duty cycle corresponds to 2000 RPM, and a 60% duty cycle corresponds to 3000 RPM. u(t) is the parameter related to the duty cycle. It's important to understand that this doesn't mean u(t) = 60%; u(t) is just a numerical value, such as 126, which is subsequently converted to the corresponding duty cycle using a function.

[0094] ωp(t) is the proportional P parameter adjustment at time t, while ωp(t-1) is the proportional P parameter adjustment at time t-1; ωi(t) is the integral I parameter adjustment at time t, while ωi(y-1) is the integral I parameter adjustment at time t-1; ωd(t) is the D parameter adjustment at time t, while ωd(t-1) is the D parameter adjustment at time t-1.

[0095] In one implementation of this embodiment, the PID parameter adjustment is first calculated based on the temperature control error e(t) at the first moment. Then, based on the existing PID parameters, the control parameters are obtained through a single-neuron learning algorithm using the PID parameter adjustment as weight values. For details, please refer to [link to relevant documentation]. Figure 3 , Figure 3 This diagram illustrates the acquisition of control parameters using a single-neuron learning algorithm. The algorithm calculates the control parameter u(t) by applying the configured PID parameters and using the PID parameter adjustments as weights. From... Figure 3 It can be seen that the input to the single-neuron learning algorithm is the configured PID parameters ( Figure 3 The PID parameters (Xp, Xi, and Xd) are calculated using the PID parameter adjustments obtained through formulas 2, 3, and 4 above, and these adjustments are used as weight values ​​for calculation. The configured PID parameters can be default PID parameters (default proportional P parameter, integral I parameter, and D parameter). These default PID parameters can be the initially set PID parameters. Each time the PID parameter adjustment is obtained, the default PID parameter is added to the PID parameter adjustment to obtain the new default PID parameter, overwriting the previous default PID parameter.

[0096] In the embodiments of this application, if the temperature index parameter meets the adjustment conditions, the PID parameter adjustment amount needs to be obtained through Formula 1, and then through Formula 2, Formula 3, and Formula 4. Based on the configured PID parameters, the control parameters are obtained through a single neuron learning algorithm with the PID parameter adjustment amount as the weight value. However, if the temperature index parameter does not meet the adjustment conditions, the PID parameter adjustment amount calculated by the above formulas is not required. Instead, the control parameters are directly generated based on the PID parameters of the previous time step and the PID parameters obtained by adjusting the PID parameters of the previous time step.

[0097] Alternatively, if the temperature parameter meets the adjustment conditions, the PID parameter adjustment amount needs to be obtained through Formula 1, followed by Formulas 2, 3, and 4. Based on the configured PID parameters, the PID parameter adjustment amount is used as the weight value to obtain the control parameters through a single-neuron learning algorithm. However, if the temperature parameter does not meet the adjustment conditions, the PID parameter adjustment amount calculated using the above formulas is not needed. Instead, the PID parameters from the previous time step are used as input, and the PID parameter adjustment amount obtained from the previous time step is used as the weight value to obtain the control parameters through a single-neuron learning algorithm. This is equivalent to directly generating the control parameters using the original PID parameters.

[0098] In one implementation of this embodiment, determining the target control parameter based on the temperature control error and the control model may also include: obtaining the target control parameter based on the temperature control error at the first moment and the neural network model; wherein, the neural network model is obtained after supervised training based on temperature training parameters under multiple scenarios and control parameters measured experimentally in each scenario. That is, the control model can directly obtain a control parameter based on the temperature control error at the first moment. For example, this control model corresponds to a neural network model trained using temperature control errors under numerous scenarios and supervised labels measured experimentally in laboratories and other locations. After training, this neural network model can take temperature control error, temperature, and other data as input and output the corresponding control parameter at that moment.

[0099] S203: Control the temperature regulation module to operate based on the target control parameters. After obtaining the control parameter u(t), use the control parameter u(t) as the adjustment amount for temperature control to control the temperature regulation module to reduce the temperature of the computing device as desired. For example, adjust the fan speed based on the temperature control adjustment amount to increase the fan speed in order to reduce the temperature of the computing device.

[0100] In one implementation of this embodiment, after executing S201 to S203, the control method of the temperature regulation module of this application can be realized. This method dynamically generates relatively suitable control parameters based on the control model, starting from the temperatures of various temperature control points in the computing device, to control the operation of the temperature regulation module. In other embodiments, step one and subsequent steps described below can be further executed to further improve the adaptability of the PID parameter adjustment.

[0101] In one implementation of this embodiment, the process of determining the target control parameter in S202 can be a continuous optimization and adjustment process. S202 may include: if the temperature index parameter meets the adjustment conditions, then at least once, based on the temperature control error and the control model, iterating the process control parameter; wherein the process control parameter is configured to control the temperature regulating module to work for a preset duration; if the temperature regulating module operates based on the latest process control parameter, causing the temperature index parameter to no longer meet the adjustment conditions, then the latest process control parameter is determined as the target control parameter. Before determining the final target control parameter, the process control parameter is iteratively obtained according to steps S201-S203. When the temperature index parameter obtained after controlling the temperature regulating module to work with a certain process control parameter no longer meets the adjustment conditions, then the process control parameter at this time is the target control parameter.

[0102] Taking the first time point (t) and the second time point (t+1) as examples, after executing S201, S202 and S203 at time t, if the temperature index parameter obtained again still meets the adjustment conditions, then the control parameter obtained at time t is a process control parameter. At time t+1, the above S201, S202 and S203 are executed again until the temperature index parameter no longer meets the adjustment conditions after executing S201, S202 and S203 again. At this time, the control parameter obtained at time t+1 can be determined as the target control parameter, and the temperature regulation module is used for cooling control.

[0103] In another implementation of this embodiment, after step one completes S203, the method may further include: Step one, acquiring the temperatures of multiple temperature control points within the computing device again, and determining the temperature control update error and temperature index update parameters based on the acquired temperatures of the multiple temperature control points within the computing device again. For ease of understanding, the time corresponding to acquiring the temperatures of the multiple temperature control points within the computing device again can be understood as the second time, which is after the first time. That is, after executing S203, for example, after increasing the fan speed to cool down, the temperatures collected by each temperature control point are acquired again, and the temperature control error at the second time (i.e., the temperature control update error in step one) and the temperature index parameters at the second time (temperature index update parameters) are acquired in the same manner as described above, so as to determine whether to continue generating new control parameters to adjust the temperature regulation module.

[0104] Step two: If the temperature index update parameters meet the adjustment conditions, then update control parameters are obtained based on the temperature control update error and the control model. In other words, if the temperature index parameters at the second time moment meet the adjustment conditions, new control parameters are obtained again based on the temperature control error and the control model at the second time moment, resulting in the control parameters for the second time moment (updated control parameters). These updated control parameters are then used to control the temperature regulation module of the equipment.

[0105] Step three involves controlling the operation of the temperature regulation module based on the updated control parameters. After controlling the operation of the temperature regulation module based on the updated control parameters, iterative processing is performed according to steps one, two, and three until the temperature control update error no longer meets the adjustment conditions.

[0106] In one implementation of this embodiment, the process of obtaining updated control parameters based on the temperature control update error and control model can be referred to the foregoing description. Specifically, it can be understood as taking the second moment as the new first moment and repeatedly executing the steps of determining the PID parameter adjustment amount and finally obtaining the control parameters. Simultaneously, after executing step three, step one is executed again at a moment after the second moment to determine whether to continue executing subsequent steps such as step two. This process is repeated until the temperature index parameter obtained at a certain moment does not meet the adjustment conditions, at which point it is no longer necessary to execute the aforementioned steps such as calculating and processing related parameters using formulas 2, 3, and 4. In one implementation of this embodiment, when the temperature index parameter or the updated temperature index parameter does not meet the adjustment conditions, a fourth control parameter can be directly generated according to the configured PID parameters to control the operation of the temperature regulation module of the computing device.

[0107] In one implementation of this embodiment, after controlling the operation of the temperature regulation module based on the target control parameters, the method triggers the execution of determining the temperature control update error and temperature index update parameters based on the temperatures of multiple temperature control points within the computing device, obtained again, according to a set time interval parameter. The method may further include: acquiring temperature change information, which includes changes before and after controlling the operation of the temperature regulation module; if the temperature change information meets the temperature rise condition of the computing device, then shortening the time interval indicated by the time interval parameter; or, if the temperature change information meets the temperature decrease condition of the computing device, then increasing the time interval indicated by the time interval parameter. In other words, by simply calculating the temperature change, it is quickly determined whether a new round of PID parameter adjustment is needed as soon as possible, i.e., whether to execute steps one and two above again as soon as possible. The temperature of the computing device can be collected and the temperature change information can be determined by a separate temperature acquisition module, or the temperature change information can be determined by the temperature collected in real time or periodically by a module in the computing device used to collect temperature data from multiple temperature control points. Once it is determined that the temperature of the computing device has increased or decreased based on the temperature change information, the time interval parameter can be modified, and the time point for starting step one can be determined according to the modified time interval parameter.

[0108] Taking the first and second moments mentioned above as examples, the second moment is determined based on the first moment and the time interval parameter. The parameter value of the time interval parameter between different moments is allowed to be dynamically set. When the temperature change information meets the temperature increase condition, the time interval parameter is reduced. For example, the temperature change information may include temperature difference, temperature variance, temperature control error, etc. If the temperature difference obtained by subtracting the temperature before control from the temperature after controlling the operation of the temperature adjustment module is greater than the preset temperature threshold, or the difference between the measured temperature and the expected temperature is greater than the preset temperature threshold, or the temperature variance is greater than the preset variance threshold, or the temperature control error is greater than the preset error threshold, then the set time interval is reduced; otherwise, the set time interval is increased.

[0109] Understandably, the main solution of this application is as follows: It determines whether adaptive adjustment of controller parameters such as PID controllers is needed by collecting temperature-related temperature index parameters. Specifically, it determines whether adaptive adjustment is required, i.e., whether the PID parameter adjustment amount needs to be updated to obtain new control parameters for temperature regulation. If adaptive adjustment is required, the PID parameter adjustment amount is determined based on the currently collected temperature-related temperature control error to obtain new control parameters. These new control parameters are then used to control the operation of temperature regulation modules such as fans. This application can use PID control strategies and neural network learning algorithms to obtain a control model for predicting control parameters. Alternatively, it can directly use some trained neural network models as control models to predict a control parameter. For example, the input of the neural network model includes the temperature control error (and may also include the aforementioned configured PID parameters, directly collected temperature, etc., the specific input being related to the training of the neural network model), and the output is the control parameters used to control the temperature regulation module. During training, the neural network model can acquire temperature control errors and / or other data from multiple scenarios (such as multiple scenarios with different numbers of temperature control points, multiple scenarios with different degrees of fan aging, and multiple scenarios with different types of temperature regulation modules) as training data. The model can also pre-set the corresponding supervision label (a manually adjusted control supervision parameter suitable for the corresponding scenario) for each scenario through manual methods such as laboratory manual debugging. This will train the initial neural network model and optimize it to obtain a neural network model that can be used to predict the control parameters at the first moment based on the temperature control error at the first moment.

[0110] In this embodiment, on the one hand, temperature control of computing devices with multiple heat sources can be achieved through PID control strategy. On the other hand, since computing devices may experience aging of temperature regulation modules such as fans, changes in ambient temperature, changes in device configuration, and changes in operating services during use, these changes may exceed the effective temperature control capabilities of the preset temperature control scheme in the computing device. Therefore, on the basis of static PID control strategy, adaptive adjustment of PID parameters and generation of control parameters can also be achieved.

[0111] Furthermore, scenarios beyond laboratory simulations during computing equipment use may lead to ineffective temperature control, triggering temperature alarms, and causing performance degradation or even component damage. Manual parameter adjustments by service personnel to adapt to different environments are inefficient and, in some complex cases, impossible, failing to achieve effective temperature control. This application dynamically adjusts PID parameters based on the actual ambient temperature, enabling adaptive controller parameters to effectively respond to changing scenarios. In the current scenario, through repeated trial and prediction, appropriate PID parameter adjustments are obtained to calculate control parameters, thereby controlling the operation of temperature regulation modules such as fans. This improves the adaptability of heat dissipation control to component, environmental, and heat source changes, providing a stable and reliable temperature environment for computing equipment and increasing the efficiency of temperature control in different scenarios.

[0112] Next, let's combine... Figure 1b and Figure 4 The control method described in this application is explained. Figure 4 This is a flowchart illustrating another control method for a temperature regulation module according to an embodiment of this application. The method described in this embodiment can be implemented using a BMC or a controller. In one implementation of this embodiment, the BMC or controller can be configured with features such as... Figure 1b The modules shown implement the control method for the temperature regulation module. This application primarily addresses situations where the originally fixed-parameter temperature control scheme cannot adequately adapt to scenarios such as aging of temperature regulation modules (e.g., fans), component changes due to damage, or environmental changes. The method described in this application includes the following steps.

[0113] S401: Obtain the temperature control error and temperature index parameters at the first moment. In this embodiment, the temperature control error is calculated based on the temperatures collected from multiple temperature control points. Figure 1bThis is achieved through a state detection module. Simultaneously, the calculated temperature index parameters can be used to determine whether subsequent parameter adaptive processing is required. In one implementation of this embodiment, a variance range can be set. The temperature index parameter is the temperature variance calculated based on the temperatures of each temperature control point. If the temperature variance exceeds the variance threshold (the maximum value of the variance range), the temperature index parameter at the first moment is considered to meet the adjustment conditions, and S402 is executed. If the temperature variance is within the variance range, S404 is executed directly. If the temperature variance is less than the minimum value of the variance range, no processing is required, and S401 is executed again after a certain time interval (e.g., after the time interval indicated by the time interval parameter mentioned in the above embodiment). S401 corresponds to S201 in the aforementioned embodiment.

[0114] S402: Start the adaptive learning module; the adaptive learning module can ultimately calculate the PID parameter adjustment amount through the above formulas 1, 2, 3 and 4.

[0115] S403: Obtain the PID parameter adjustment amount calculated by the adaptive learning module. S403 to S405 can be in the following order: Figure 1b The PID control module shown is performing the operation.

[0116] S404: The configured PID parameters are combined with the PID parameter adjustment to obtain new PID parameters. It can be understood that for S404, if the temperature parameters obtained from S401 determine that S404 can be executed directly, then in this case, the configured PID parameters can continue to be used; however, if according to S401, S402 and S403 need to be executed first, then in this case, the PID parameter adjustment is the PID parameter adjustment obtained in S403.

[0117] S405: For the obtained new PID parameters, calculate the target control parameters, which are the control parameters u(t) at the first moment, and output u(t).

[0118] It is understood that S402 to S405 above are related to S202 in the aforementioned embodiments. Specifically, the specific implementation of S402 and S403 can refer to the acquisition process of ωp(t), ωi(t), and ωd(t) in the aforementioned embodiments, while the implementation of S404 and S405 refers to the process of obtaining control parameters through a single neuron learning algorithm.

[0119] S406: Controls the fan speed based on target control parameters via the execution unit. S405 and S406 can be implemented in the PID control module.

[0120] At the next moment after S406 is executed, i.e., when the second moment arrives, the second moment is taken as the new first moment, and the relevant steps of S401-S406 above are repeated. That is, temperature monitoring and sampling are performed at the second moment to obtain the new temperature of each temperature control point. Then, new processing is performed through the state detection module, adaptive learning module, and PID control module. Iterative learning continues until the final collected temperature index parameters no longer meet the adjustment conditions and the new PID parameters can meet the new scenario. Under the new scenario, the temperature of the computing device can be effectively controlled so that the temperature in the computing device is within a reasonable temperature range.

[0121] The proposed solution dynamically adjusts PID parameters to generate control parameters that adapt to changing scenarios. Under the current scenario, through repeated trial and prediction, it finds the most suitable control parameters to control temperature regulation modules such as fans. This enhances the adaptability of the server's cooling system to changes in components, environment, and heat sources, providing the server with a stable and reliable temperature environment. It improves the efficiency of temperature control in different scenarios, reduces human resource costs, and increases user satisfaction when using the computing equipment.

[0122] Next, let's combine... Figure 1b and Figure 5 The control method described in this application is explained. Figure 5 This is a flowchart illustrating another control method for a temperature regulation module according to an embodiment of this application. The method described in this embodiment can be implemented using a BMC or a controller. In one implementation of this embodiment, the BMC or controller can be configured with features such as... Figure 1b The modules shown implement the control method for the temperature regulation module. This application embodiment primarily addresses the scenario where new computing devices such as servers are configured and running, and adaptively configures PID parameters starting from an initially set PID parameter to cope with future temperature control issues such as fan aging mentioned in the above embodiment. The method of this application embodiment includes the following steps.

[0123] S501: Obtain the temperature control error and temperature index parameters at the first moment.

[0124] S502: Start the adaptive learning module; the adaptive learning module can calculate the PID parameter adjustment amount using the above formulas 1, 2, 3 and 4.

[0125] S503: Obtain the PID parameter adjustment amount calculated by the adaptive learning module. S503 to S505 can be in the following order: Figure 1b The PID control module shown is performing the operation.

[0126] S504: New PID parameters are obtained by combining the configured PID parameters with the PID parameter adjustment amount.

[0127] S505: Calculate the target control parameters, i.e., the output u(t), from the obtained new PID parameters.

[0128] S506: The fan speed is controlled by the execution unit based on the target control parameters.

[0129] When the second moment arrives after S506 has been executed, the second moment is taken as the new first moment, and the relevant steps of S501-S506 are repeated.

[0130] It is understood that S501 to S506 are the same as S401 to S406 in the previous embodiment. The difference is that the following steps are also included in the embodiments of this application.

[0131] S507: Does the temperature index parameter at the new moment meet the adjustment conditions? The temperature index parameter at the new moment can refer to the temperature index parameter obtained at the next moment after executing S501 to S506. If the judgment result of S507 is negative, then execute S508 below; if the judgment result is positive, then continue to execute S502 and its subsequent related steps.

[0132] S508: Get new PID parameters.

[0133] S509: The newly obtained PID parameters are set as the configured PID parameters so that when the temperature index parameters are determined to meet the adjustment conditions in the next time, the configured PID parameters can be directly obtained.

[0134] S510: Configure the PID parameters in the PID control module, and then perform the steps of combining the configured PID parameters with the PID parameter adjustment according to the configured PID parameters.

[0135] Existing static parameter adjustment methods require thermal testing of each configuration in a laboratory, followed by manual diagnosis based on the test results, adjustment of PID parameters, and then testing again. This process is repeated multiple times to adjust the PID parameters until suitable parameters are obtained. This process consumes significant manpower and resources, and the quality of parameters can vary due to differences in personnel skills. The embodiment of this application, however, only requires configuring an initial value. Through automated testing, the PID parameters are adaptively adjusted, reducing manpower input and improving configuration efficiency.

[0136] Corresponding to the methods described in the above embodiments, this application also provides corresponding apparatus, including modules or units for executing the corresponding methods described in the above embodiments. The modules or units may be software, hardware, or a combination of software and hardware.

[0137] Please see Figure 6 This is a schematic diagram of the structure of a control device for a temperature regulation module according to an embodiment of this application. This device can be applied in a controller or in EMC (Electronic Control System). The device includes:

[0138] The acquisition unit 601 is used to acquire the temperature of the temperature control point in the computing device;

[0139] The processing unit 602 is used to determine the temperature control error and temperature index parameters based on the temperatures of multiple temperature control points within the computing device; if the temperature index parameters meet the adjustment conditions, then the target control parameters are determined based on the temperature control error and the control model; and the operation of the temperature regulation module is controlled based on the target control parameters.

[0140] In one implementation of this embodiment, after controlling the operation of the temperature regulation module based on the target control parameters, the processing unit 602 is further configured to acquire the temperatures of multiple temperature control points within the computing device, and determine the temperature control update error and temperature index update parameters based on the re-acquired temperatures of the multiple temperature control points within the computing device; if the temperature index update parameters meet the adjustment conditions, then update control parameters are acquired based on the temperature control update error and the control model; and the operation of the temperature regulation module is controlled based on the update control parameters.

[0141] In one implementation of this embodiment, the processing unit 602, while determining the target control parameter based on the temperature control error and the control model if the temperature index parameter meets the adjustment conditions, is further configured to iterate the process control parameter at least once based on the temperature control error and the control model if the temperature index parameter meets the adjustment conditions; wherein the process control parameter is configured to control the temperature adjustment module to work for a preset duration; if the temperature adjustment module operates based on the latest process control parameter, causing the temperature index parameter to not meet the adjustment conditions, then the latest process control parameter is determined as the target control parameter.

[0142] In one implementation of this embodiment, the control model is obtained based on a PID control strategy and a neural network learning algorithm. When the processing unit 602 determines the target control parameters based on the temperature control error and the control model, it determines the PID parameter adjustment amount based on the recorded historical control parameters, the recorded historical adjustment amount of the PID parameters used to determine the historical control parameters, and the historical temperature control error, and based on the temperature control error; and obtains the control parameters through the neural network learning algorithm based on the configured PID parameters and the PID parameter adjustment amount.

[0143] In one implementation of this embodiment, when the processing unit 602 is used to determine the temperature control error based on the temperatures of multiple temperature control points within the computing device, it is used to obtain the temperature of each temperature control point within the computing device, and obtain the temperature deviation of each temperature control point based on the temperature of each temperature control point and the desired temperature; and perform a weighted summation based on the weight value corresponding to each temperature control point and each temperature deviation to obtain the temperature control error.

[0144] In one implementation of this embodiment, before the processing unit 602 performs a weighted summation based on the weight values ​​corresponding to each temperature control point and each temperature deviation to obtain the temperature control error, it is further configured to set the weight values ​​corresponding to each temperature control point based on the attribute information of the heat source module corresponding to each temperature control point; wherein, the attribute information of the heat source module includes any one or two of the following: the type of the heat source module, and the distance between the heat source module and the temperature regulation module in the computing device.

[0145] In one implementation of this embodiment, the temperature index parameter is used to represent the degree of deviation between the temperature corresponding to each temperature control point and the desired temperature.

[0146] In one implementation of this embodiment, the temperature index parameter includes: the variance calculated based on the temperature corresponding to each temperature control point and the desired temperature.

[0147] In one implementation of this embodiment, the neuron learning algorithm is a single-neuron learning algorithm. The input of the single-neuron learning algorithm is the configured PID parameters, the adjustment amount of the PID parameters is the weight value, and the output of the single-neuron learning algorithm is the control parameters. After the single-neuron learning algorithm sums the proportional P parameter included in the configured PID parameters with the proportional P parameter adjustment amount in the PID parameter adjustment amount, sums the integral I parameter included in the configured PID parameters with the integral I parameter adjustment amount in the PID parameter adjustment amount, and sums the derivative D parameter included in the configured PID parameters with the derivative D parameter adjustment amount in the PID parameter adjustment amount, it then predicts the target control parameters through a prediction function.

[0148] In one implementation of this embodiment, the processing unit 602 is further configured to generate a fourth control parameter according to the configured PID parameter to control the operation of the temperature regulation module of the computing device if the temperature index parameter does not meet the adjustment conditions.

[0149] In one implementation of this embodiment, when the processing unit 602 determines the target control parameters based on the temperature control error and the control model, it is used to obtain the target control parameters based on the temperature control error at the first moment and the neural network model; wherein, the neural network model is obtained after supervised training based on temperature training parameters in multiple scenarios and control parameters determined experimentally in each scenario.

[0150] In one implementation of this embodiment, after controlling the operation of the temperature regulation module based on the target control parameters, the processing unit 602 triggers the execution of determining the temperature control update error and temperature index update parameters based on the temperature of multiple temperature control points within the computing device obtained again, according to a set time interval parameter. The processing unit 602 is also used to acquire temperature change information, which includes the change value before and after controlling the operation of the temperature regulation module. If the temperature change information meets the temperature rise condition of the computing device, the time interval indicated by the time interval parameter is shortened; or, if the temperature change information meets the temperature decrease condition of the computing device, the time interval indicated by the time interval parameter is increased.

[0151] In one implementation of this embodiment, each temperature control point within the computing device corresponds to at least one heat source module, and the heat source module includes any one or more of the central processing unit, graphics processing unit, and image processing unit disposed within the computing device.

[0152] Based on the same inventive concept, the principle and beneficial effects of the temperature regulation module control device provided in the embodiments of this application are similar to the principle and beneficial effects of the solution in the various method embodiments of this application. Please refer to the principle and beneficial effects of the various method embodiments. For the sake of brevity, they will not be repeated here.

[0153] Please see Figure 7 This is a schematic diagram of the structure of a computing device provided in this application. The computing device includes at least a processor, which is used to implement the control method for the temperature regulation module in the above embodiments. In one implementation of this embodiment, Figure 7 The computing device 70 shown includes at least one processor 701 and a transceiver 702, and optionally, a memory 703.

[0154] Memory 703 may be volatile memory, such as random access memory; memory may also be non-volatile memory, such as read-only memory, flash memory, hard disk drive (HDD) or solid-state drive (SSD); or memory 703 may be any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto. Memory 703 may be a combination of the above-described memories.

[0155] This application embodiment does not limit the specific connection medium between the processor 701, transceiver 702, and memory 703 described above. This application embodiment... Figure 7 The processor 701, transceiver 702, and memory 703 are connected via bus 704, and bus 704 is in... Figure 7 The connections between other components are shown in bold lines only and are not intended to be limiting. This 704 bus can be divided into address bus, data bus, control bus, etc. For ease of illustration, Figure 7 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0156] Processor 701 may have data transmission and reception capabilities, enabling it to communicate with other devices or equipment (such as communication between CPLD and BMC via local bus), in situations such as... Figure 7 In addition, an independent data communication unit, such as transceiver 702, can be set up to send and receive data; when the processor 701 communicates with other devices or equipment, it can transmit data through transceiver 702.

[0157] In one example, when the computing device is Figure 7 When in the form shown, Figure 7 The processor in the system can execute any of the above method embodiments.

[0158] Specifically, Figure 6 The functions / implementation processes of the acquisition unit and the processing unit can all be accessed through... Figure 7 The processor 701 in the memory calls computer execution instructions stored in memory 703 to implement the function. Alternatively, Figure 6 The function / implementation process of the acquisition unit is achieved through Figure 7 The transceiver 702 in the middle is used to implement the function / implementation process of the processing unit. Figure 7 The processor 701 in the memory calls computer execution instructions stored in the memory 703 to implement the function.

[0159] In one implementation, the computing device 70 may further include circuits that can implement the relevant functions described in the foregoing method embodiments. The processor described in this application can be implemented as an integrated circuit (IC), analog IC, radio frequency integrated circuit (RFIC), mixed-signal IC, application-specific integrated circuit (ASIC), printed circuit board (PCB), computing device, etc. The processor can also be manufactured using IC process technologies such as complementary metal oxide semiconductor (CMOS), n-metal-oxide-semiconductor (NMOS), positive-channel metal oxide semiconductor (PMOS), bipolar junction transistor (BJT), bipolar CMOS (BiCMOS), silicon germanium (SiGe), gallium arsenide (GaAs), etc.

[0160] The computing device described in the above embodiments may include a controller (CPLD) or a device management unit (BMC), and the structure of the computing device may be independent of... Figure 7 The limitations apply. The computing device can be a standalone device or part of a larger device. For example, the computing device could be:

[0161] (1) Independent integrated circuit IC, or chip, or chip system or subsystem;

[0162] (2) A collection of one or more ICs, optionally including storage components for storing data and computer programs;

[0163] (3) ASIC, such as modem;

[0164] (4) Modules that can be embedded in other devices;

[0165] (5) Receivers, terminals, smart terminals, cellular phones, wireless devices, handheld devices, mobile units, vehicle-mounted devices, network devices, cloud devices, artificial intelligence devices, etc.

[0166] (6) Others, etc.

[0167] For cases where the computing device can be a chip or a chip system, see [link to relevant documentation]. Figure 8The diagram shows the structure of the chip. Figure 8 The chip shown includes a processor 801 and an interface 802, and there may be one or more processors 801. The processor 801 is used to perform the relevant steps of the foregoing method embodiments.

[0168] Optionally, the chip may also include a memory 803 for storing necessary computer programs and data. The memory 803 may be provided separately or integrated with the processor 801.

[0169] It is understood that some optional features in the embodiments of this application can be implemented independently in certain scenarios without relying on other features, such as the current solution on which they are based, to solve the corresponding technical problems and achieve the corresponding effects. Alternatively, they can be combined with other features as needed in certain scenarios. Correspondingly, the apparatus given in the embodiments of this application can also implement these features or functions, which will not be elaborated here.

[0170] Those skilled in the art will also understand that the various illustrative logical blocks and steps listed in the embodiments of this application can be implemented by electronic hardware, computer software, or a combination of both. Whether such functionality is implemented through hardware or software depends on the specific application and the overall system design requirements. Those skilled in the art can use various methods to implement the described functionality for corresponding applications, but such implementation should not be construed as exceeding the scope of protection of the embodiments of this application.

[0171] The solutions described in this application can be implemented in various ways. For example, these technologies can be implemented in hardware, software, or a combination of hardware. For hardware implementation, the processing unit for executing these technologies at the control device of the temperature regulation module can be implemented in one or more general-purpose processors, digital signal processors (DSPs), digital signal processing devices, application-specific integrated circuits (ASICs), programmable logic devices, field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic, discrete hardware components, or any combination thereof. The general-purpose processor can be a microprocessor; alternatively, it can also be any conventional processor, controller, microcontroller, or state machine. The processor can also be implemented through a combination of computing devices, such as a digital signal processor and a microprocessor, multiple microprocessors, one or more microprocessors combined with a digital signal processor core, or any other similar configuration.

[0172] This application also provides a computer-readable medium having a computer program stored thereon, which, when executed by a computer, implements the functions of any of the above method embodiments.

[0173] This application also provides a computer program product that, when executed by a computer, implements the functions of any of the above method embodiments.

[0174] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., high-density digital video discs (DVDs)), or semiconductor media (e.g., solid-state drives (SSDs)).

[0175] It is understood that the term "embodiment" used throughout the specification means that a specific feature, structure, or characteristic related to an embodiment is included in at least one embodiment of this application. Therefore, various embodiments throughout the specification do not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. It is understood that in the various embodiments of this application, the sequence number of the above-described processes 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 application.

[0176] It is understood that in this application, "when," "if," and "if" all refer to the device making a corresponding action under certain objective circumstances, and are not time-limited, nor do they require the device to make a judgment when it is implemented, nor do they imply any other limitations.

[0177] In this application, the use of singular pronouns to denote "one or more" rather than "one and only one," unless otherwise specified. In this application, unless otherwise specified, "at least one" is intended to mean "one or more," and "more than" is intended to mean "two or more."

[0178] In addition, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that there can be three kinds of relationships. For example, A and / or B can represent three cases: A exists alone, A and B exist simultaneously, and B exists alone. A can be singular or plural, and B can be singular or plural.

[0179] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0180] The same or similar parts between the various embodiments in this application can be referred to mutually. In the various embodiments of this application, and in the various implementation methods / methods / implementations within each embodiment, unless otherwise specified or logically conflicting, the terminology and / or descriptions between different embodiments and between the various implementation methods / methods / implementations within each embodiment are consistent and can be mutually referenced. The technical features in different embodiments and the various implementation methods / methods / implementations within each embodiment can be combined according to their inherent logical relationships to form new embodiments, implementation methods, methods, or implementation approaches. The above-described embodiments of this application do not constitute a limitation on the scope of protection of this application.

[0181] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A control method of a temperature adjustment module, characterized by, The method includes: Based on the temperatures of multiple temperature control points within the computing device, determine the temperature control error and temperature index parameters; If the temperature index parameter meets the adjustment conditions, the process control parameter is iterated at least once based on the temperature control error and the control model; wherein, the process control parameter is configured to control the temperature regulation module to work for a preset duration; If the temperature regulation module operates based on the latest process control parameters, causing the temperature index parameters to fail to meet the adjustment conditions, then the latest process control parameters will be determined as the target control parameters. The operation of the temperature regulation module is controlled based on the target control parameters; Acquire temperature change information, which includes temperature change values ​​before and after controlling the operation of the temperature regulation module; If the temperature change information meets the temperature rise condition of the computing device, then the preset duration is shortened; If the temperature change information meets the temperature reduction condition of the computing device, then the preset duration is increased.

2. The method of claim 1, wherein, The control model is obtained based on a proportional-integral-derivative (PID) control strategy and a neural network learning algorithm. The determination of the target control parameters based on the temperature control error and the control model includes: Based on the recorded historical control parameters, the recorded historical adjustment of PID parameters used to determine the historical control parameters, and the historical temperature control error, the PID parameter adjustment is determined based on the temperature control error. Based on the configured PID parameters and the PID parameter adjustment amount, control parameters are obtained through a neural network learning algorithm.

3. The method as described in claim 1, characterized in that, The step of determining the temperature control error based on the temperatures of multiple temperature control points within the computing device includes: The temperature of each temperature control point within the computing device is obtained, and the temperature deviation of each temperature control point is obtained based on the temperature of each temperature control point and the desired temperature. The temperature control error is obtained by weighting and summing the weight values ​​corresponding to each temperature control point and each temperature deviation.

4. The method as described in claim 3, characterized in that, Before obtaining the temperature control error by weighted summation based on the weight values ​​corresponding to each temperature control point and each temperature deviation, the following steps are also included: Set the weight value corresponding to each temperature control point according to the attribute information of the heat source module corresponding to each temperature control point; The attribute information of the heat source module includes: the type of the heat source module, and any one or two of the distances between the heat source module and the temperature regulation module within the computing device.

5. The method as described in claim 1, characterized in that, The temperature index parameters include: the variance calculated based on the temperature corresponding to each temperature control point and the desired temperature.

6. The method as described in claim 2, characterized in that, The neuron learning algorithm is a single neuron learning algorithm. The input of the single neuron learning algorithm is the configured PID parameter, the adjustment amount of the PID parameter is the weight value, and the output of the single neuron learning algorithm is the control parameter. The single-neuron learning algorithm sums the proportional P parameter included in the configured PID parameters with the proportional P parameter adjustment amount in the PID parameter adjustment amount, sums the integral I parameter included in the configured PID parameters with the integral I parameter adjustment amount in the PID parameter adjustment amount, and sums the derivative D parameter included in the configured PID parameters with the derivative D parameter adjustment amount in the PID parameter adjustment amount, and then predicts the control parameters through a prediction function.

7. The method according to any one of claims 1-6, characterized in that, Each temperature control point within the computing device corresponds to at least one heat source module, and the heat source module includes any one or more of the central processing unit, graphics processing unit, and image processing unit located within the computing device.

8. A computing device, characterized in that, The computing device includes a processor for implementing the control method of the temperature regulation module as described in any one of claims 1-7.

Images