Aluminum-based PCB microchannel heat dissipation control method and system

Abstract

The present application belongs to the field of heat dissipation control technology and discloses a heat dissipation control method and system for an aluminum-based PCB microchannel. The method comprises: obtaining task scheduling information and temperature information of each core area; determining the correspondence between each computing task and each core area and the heat dissipation resource allocation weight of each computing task based on the task scheduling information; solving the target coolant flow of each core area with the goal of minimizing the weighted average temperature of each core area and minimizing the total coolant flow based on the temperature information, the heat dissipation resource allocation weight, the correspondence between each computing task and each core area, and the total available coolant flow of the system; controlling each flow control unit to adjust the coolant flow of each core area based on the target coolant flow; thereby enabling forward-looking heat dissipation control to be performed in combination with the task scheduling information, giving priority to ensuring the temperature stability of high-priority task areas and optimizing resource allocation under the total resource constraint, thereby improving heat dissipation efficiency and reducing energy consumption.

Description

Technical Field

[0001] The present application relates to the field of heat dissipation control technology, and more specifically, to a method and system for controlling heat dissipation in an aluminum-based PCB microchannel. Background Art

[0002] High-performance computing devices are widely used in industrial control, data centers, artificial intelligence, and other fields. Their core configurations typically include multi-core processors. To ensure stable processor operation under high loads, an efficient heat dissipation system is crucial. Existing advanced cooling solutions utilize liquid cooling technology, where the processor is mounted on an aluminum-based printed circuit board (PCB) with excellent thermal conductivity. This aluminum-based PCB incorporates a sophisticated network of microchannels, creating pathways for the circulation of coolant. Driven by a pump, the coolant circulates within the microchannels, absorbing heat generated by the processor and dissipating it through a heat exchanger.

[0003] High-performance computing devices typically run complex multi-tasking operating systems. The operating system's task scheduler assigns different computing tasks to different core processors based on preset scheduling policies. Due to differences in the computational intensity and execution status of different tasks, as well as the dynamic migration of tasks between cores, the heat generated by each core processor and the corresponding area are subject to different heat flux densities, which can easily form local hotspots. Even more challenging, as the operating system's task scheduler continuously switches and reallocates tasks between cores, the location and intensity of these local hotspots can change rapidly and dynamically with changes in the execution status of tasks and core allocation.

[0004] To address this dynamic and uneven heat generation, aluminum-based PCB microchannel cooling systems are typically designed to be controlled in zones. By configuring independent flow control units (e.g., microvalves) in different areas of the microchannel network (usually corresponding to the core processor, hence referred to as core areas), independent regulation of the coolant flow in each area can be achieved. Existing zoned cooling control methods often employ strategies based on temperature feedback. The system collects real-time temperature values ​​from the aluminum-based PCB or the integrated temperature sensor within the processor in the area corresponding to the core processor. When the temperature in a particular area exceeds a preset threshold, the control system responsively increases the coolant flow in that area to enhance the heat dissipation effect and reduce the temperature.

[0005] However, this control method based purely on hysteresis-based temperature feedback has significant limitations. Core processors executing high-priority or high-security tasks have a low tolerance for temperature fluctuations. Even if the temperature has not yet reached the set feedback threshold, a rapid rise or sustained temperature fluctuation may adversely affect the real-time performance, execution accuracy, or stability of the task, and even increase the risk of calculation errors or system instability. Low-priority tasks, on the other hand, have a relatively high tolerance for temperature fluctuations. Traditional temperature feedback control cannot perceive and distinguish these differences in heat dissipation requirements caused by task characteristics, and cannot prioritize temperature stability for critical tasks.

[0006] At the same time, the total resources of the liquid cooling system are limited. For example, there is an upper limit to the total flow rate of the coolant circulation pump. When multiple core processor areas show a trend of rising temperatures at the same time and need to increase the coolant flow rate, if the system simply responds to the flow increase demand of all areas, it may cause the total flow demand to exceed the maximum capacity of the cooling pump, or force the cooling pump to run at high power or even full load, significantly increasing the overall power consumption of the system, and it may still not be possible to effectively control the temperature of all areas, especially the heat dissipation priority of high-priority areas. How to effectively allocate these resources (such as the total coolant flow rate) under the premise that the total resources of the cooling system are limited, give priority to ensuring the temperature stability of areas with high-priority tasks that are sensitive to temperature, while taking into account the overall heat dissipation effect and system energy consumption, is a key technical challenge that needs to be solved urgently.

[0007] Relying solely on temperature feedback, the system cannot predict which areas will soon execute high-priority tasks and, consequently, generate high heat loads. If proactive task scheduling information can be acquired and effectively utilized in real time, it would be possible to pre-adjust coolant flow to relevant areas before tasks actually start, increase computational intensity, or generate higher heat.

[0008] Implementing dynamic, proactive cooling scheduling based on task priority and scheduling information requires the system to accurately and in real time obtain the processor's task scheduling information (including task priority, assigned cores, etc.) and the task execution status of each core. A model or strategy must be established to correlate task priority with the priority or weight of cooling resource allocation. Based on this information, the cooling control system must be able to rapidly calculate and dynamically adjust the coolant flow rates across multiple microchannel zones. Furthermore, this scheduling strategy must balance the cooling requirements of different zones within the constraints of limited total cooling resources, while ensuring the temperature stability of high-priority tasks and optimizing overall system energy consumption.

[0009] In view of the above problems, the existing technology is in urgent need of improvement. Summary of the Invention

[0010] The purpose of this application is to provide an aluminum-based PCB microchannel heat dissipation control method and system, which can combine task scheduling information to perform forward-looking heat dissipation control, give priority to ensuring temperature stability in high-priority task areas, and optimize resource allocation under total resource constraints, thereby improving heat dissipation efficiency and reducing energy consumption.

[0011] In a first aspect, the present application provides an aluminum-based PCB microchannel heat dissipation control method for controlling an aluminum-based PCB microchannel heat dissipation system of a computing device having multiple core processors, wherein the aluminum-based PCB microchannel heat dissipation system has multiple core areas and multiple flow control units, each of the core processors being respectively disposed at each of the core areas, and each of the flow control units being respectively configured to adjust the coolant flow rate of each of the core areas; the method comprising the following steps:

[0012] A1 obtains task scheduling information and temperature information of each core area; the task scheduling information includes the task priority of each computing task and the corresponding task allocation core processor;

[0013] A2. Assign core processors to each computing task and determine the correspondence between each computing task and each core area;

[0014] A3. Obtain the heat dissipation resource allocation weight for each computing task based on the task priority of each computing task;

[0015] A4. Based on the temperature information of each core area, the cooling resource allocation weights of each computing task, the correspondence between each computing task and each core area, and the total available coolant flow rate of the system, a resource allocation algorithm is used to determine the target coolant flow rate for each core area, with the goal of minimizing the weighted average temperature of each core area and minimizing the total coolant flow rate.

[0016] A5. Control each flow control unit to adjust the coolant flow of each core area according to the target coolant flow of each core area.

[0017] Preferably, step A3 includes:

[0018] A301. According to the task priority of each computing task, query the task priority and heat dissipation resource allocation weight mapping table to obtain the heat dissipation resource allocation weight of each computing task.

[0019] Preferably, the task scheduling information also includes task type and task amount information of each computing task;

[0020] After step A301, the following steps are included:

[0021] A302. Determine the energy consumption level of each computing task based on the task type and task amount information of each computing task;

[0022] A303. Adjust the heat dissipation resource allocation weight of each computing task according to the energy consumption level of each computing task to obtain a final heat dissipation resource allocation weight.

[0023] Preferably, step A302 includes:

[0024] According to the task type of each computing task, query the task type and processor frequency impact factor mapping table to obtain the corresponding frequency impact factor;

[0025] Estimate the processor frequency increase value of each computing task during execution based on the task load information and frequency impact factor of each computing task;

[0026] The energy consumption level of each computing task is calculated based on the processor frequency boost value and task amount information.

[0027] Preferably, step A303 includes:

[0028] Taking each core processor as the target core in turn, obtain the set of all core processors in the neighborhood centered on the target core as the neighborhood core set;

[0029] Based on the preset heat transfer coefficient and the distance between each neighboring core and the target core in the neighborhood core set, the heat influence factor of each neighboring core on the target core is calculated; the heat influence factor of the target core on itself is 1, and the heat influence factors of other neighboring cores on the target core are proportional to the heat transfer coefficient and inversely proportional to the distance;

[0030] The sum of the products of the thermal impact factors of all neighboring cores and the energy consumption levels of the corresponding computing tasks is calculated to adjust the heat dissipation resource allocation weight of the target core and obtain the final heat dissipation resource allocation weight.

[0031] Preferably, the resource allocation algorithm is a linear programming algorithm, a quadratic programming algorithm or a model predictive control algorithm.

[0032] Preferably, step A4 includes:

[0033] A401. Construct an optimization objective function, wherein the optimization objective function includes minimizing the weighted average temperature of each core region and minimizing the total coolant flow rate; wherein the weight of each core region is determined by the heat dissipation resource allocation weight of each computing task and the corresponding relationship between each computing task and each core region, and the weighted average temperature is obtained by multiplying the temperature information of each core region by the corresponding weight, summing the result, and dividing the result by the sum of the weights;

[0034] A402. Using an optimization algorithm based on sequential quadratic programming, with the coolant flow rate of each core area as the optimization variable, solve the optimization objective function under the constraints of the constraints to obtain the target coolant flow rate of each core area; the constraints include the upper and lower limits of the coolant flow rate of each core area and the total available coolant flow rate of the system.

[0035] Preferably, step A402 includes:

[0036] B1. Initialize the parameters of the sequential quadratic programming algorithm, including the initial coolant flow rate value, Hessian matrix, Lagrange multiplier estimate, and convergence tolerance of each core area;

[0037] B2. Calculate the optimization objective function value and its gradient based on the coolant flow rate values ​​in each core area at the current iteration, as well as the constraints and their Jacobian matrix.

[0038] B3. Construct a quadratic programming subproblem, centered around the coolant flow rate values ​​in each core region of the current iteration. Based on the calculated optimization objective function value and its gradient and the Lagrange multiplier estimate, perform a quadratic approximation on the optimization objective function and a linear approximation on the constraints. The quadratic approximation uses a Hessian matrix, and the linear approximation uses a Jacobian matrix.

[0039] B4. Using the active set method to solve the quadratic programming subproblem, the search direction and search step size of the coolant flow value in each core area are obtained;

[0040] B5. Update the coolant flow value of each core area based on the search direction and search step size of the coolant flow value of each core area, and adjust the search step size according to the Armijo line search strategy to ensure that the objective function value decreases and the constraints are satisfied;

[0041] B6. Determine whether the convergence condition determined by the convergence tolerance is met. If so, use the latest coolant flow rate value of each core area as the target coolant flow rate of each core area. Otherwise, use the BFGS formula to update the Hessian matrix and Lagrange multiplier estimates, and return to step B2.

[0042] Preferably, after step A402, the method further includes:

[0043] A403 obtains the task content category of each computing task, queries the preset task content category and temperature sensitivity level mapping table, and obtains the temperature sensitivity level corresponding to each computing task;

[0044] A404. Adjust the target coolant flow rate of each core area based on the temperature sensitivity level of each computing task using preset adjustment rules.

[0045] In a second aspect, the present application provides an aluminum-based PCB microchannel heat dissipation control system for controlling an aluminum-based PCB microchannel heat dissipation system of a computing device having multiple core processors. The aluminum-based PCB microchannel heat dissipation system has multiple core areas and multiple flow control units, each of the core processors is respectively disposed at each of the core areas, and each of the flow control units is respectively used to adjust the coolant flow rate of each of the core areas; the system includes:

[0046] An information collection module is used to obtain task scheduling information and temperature information of each core area; the task scheduling information includes the task priority of each computing task and the corresponding task allocation core processor;

[0047] A mapping module is used to allocate core processors according to various computing tasks and determine the correspondence between various computing tasks and various core areas;

[0048] A weight determination module is used to obtain the heat dissipation resource allocation weight of each computing task according to the task priority of each computing task;

[0049] An optimization calculation module is used to calculate the target coolant flow rate for each core area using a resource allocation algorithm, based on the temperature information of each core area, the heat dissipation resource allocation weights of each computing task, the correspondence between each computing task and each core area, and the total available coolant flow rate of the system, with the goal of minimizing the weighted average temperature of each core area and minimizing the total coolant flow rate;

[0050] The control execution module is used to control each flow control unit to adjust the coolant flow of each core area according to the target coolant flow of each core area.

[0051] Beneficial effects: The present application provides an aluminum-based PCB microchannel heat dissipation control method and system, which combines task scheduling information and temperature information, determines the heat dissipation resource allocation weight based on task priority, and uses an optimization algorithm to solve the target coolant flow rate, thereby being able to perform forward-looking heat dissipation control in combination with task scheduling information, give priority to ensuring temperature stability in high-priority task areas, and optimize resource allocation under total resource constraints, thereby improving heat dissipation efficiency and reducing energy consumption. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 Flowchart of the aluminum-based PCB microchannel heat dissipation control method provided in an embodiment of the present application.

[0053] Figure 2 This is a schematic diagram of the structure of the aluminum-based PCB microchannel heat dissipation control system provided in an embodiment of the present application.

[0054] Explanation of reference numerals: 1. Information acquisition module; 2. Mapping module; 3. Weight determination module; 4. Optimization calculation module; 5. Control execution module. DETAILED DESCRIPTION

[0055] The technical solutions in this application will be clearly and completely described below in conjunction with the drawings in this application. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all of the embodiments. The components of the present application generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application for protection, but merely represents selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work fall within the scope of protection of this application.

[0056] It should be noted that similar reference numerals and letters represent similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and should not be understood as indicating or implying relative importance.

[0057] refer to Figure 1 This application proposes an aluminum-based PCB microchannel heat dissipation control method for controlling an aluminum-based PCB microchannel heat dissipation system of a computing device having multiple core processors. The aluminum-based PCB microchannel heat dissipation system has multiple core areas and multiple flow control units. Each core processor is respectively disposed at each core area, and each flow control unit is respectively used to adjust the coolant flow rate of each core area. The method comprises the following steps:

[0058] A1 obtains task scheduling information and temperature information of each core area; the task scheduling information includes the task priority of each computing task and the corresponding task allocation core processor;

[0059] A2. Assign core processors to each computing task and determine the correspondence between each computing task and each core area;

[0060] A3. Obtain the heat dissipation resource allocation weight for each computing task based on the task priority of each computing task;

[0061] A4. Based on the temperature information of each core area, the cooling resource allocation weights of each computing task, the correspondence between each computing task and each core area, and the total available coolant flow rate of the system, a resource allocation algorithm is used to determine the target coolant flow rate for each core area, with the goal of minimizing the weighted average temperature of each core area and minimizing the total coolant flow rate.

[0062] A5. Control each flow control unit to adjust the coolant flow of each core area according to the target coolant flow of each core area.

[0063] Among them, task scheduling information refers to the data provided by the operating system about computing tasks that are running or about to run, including the importance of each computing task and which core processor the corresponding task is assigned to execute. Its main purpose is to provide forward-looking input about task characteristics so that the system can predict the differences in heat dissipation requirements in different areas.

[0064] The cooling resource allocation weight is a numerical value determined based on factors such as the task priority of each computing task. It is used to quantify the importance or priority of the task or its region in cooling resource allocation. This can be achieved by querying a mapping table between task priorities and cooling resource allocation weights. The main purpose is to convert task characteristics into input parameters for the optimization algorithm. The mapping table between task priorities and cooling resource allocation weights can be pre-determined through statistical methods or expert experience and stored in a local database.

[0065] Among them, the resource allocation algorithm is a mathematical optimization method used to solve how to allocate limited resources to multiple demand parties under given constraints. It can be implemented using linear programming algorithms, quadratic programming algorithms, or model predictive control algorithms. Its main purpose is to find the optimal coolant flow distribution plan. Minimizing the weighted average temperature of each core area and minimizing the total coolant flow are the two optimization goals that the resource allocation algorithm needs to achieve. Minimizing the weighted average temperature aims to prioritize reducing the temperature of high-weight areas to ensure the stability of critical tasks, and minimizing the total coolant flow aims to reduce pump power consumption and improve system energy efficiency. The weighted average temperature is the temperature value obtained by multiplying the real-time temperature value of each core area by the heat dissipation resource allocation weight of the corresponding area, summing the results, and dividing by the total weight. Among them, the weight of each core area is determined by the heat dissipation resource allocation weight of each computing task and the corresponding relationship between each computing task and each core area. It is mainly to reflect the relative importance of different areas in heat dissipation assurance.

[0066] The core innovation of this application lies in combining forward-looking task scheduling information, especially task priority, with real-time core area temperature information, and utilizing a resource allocation algorithm to dynamically solve and control the target coolant flow of each core area with the goal of minimizing the weighted average temperature of each core area and minimizing the total coolant flow, under the constraint of the total available coolant flow of the system. This overcomes the lag of traditional pure temperature feedback control, achieves differentiated heat dissipation guarantees based on task importance, improves the utilization efficiency of heat dissipation resources, and reduces the overall energy consumption of the system.

[0067] Specifically, the method works as follows: First, the system obtains task scheduling information provided by the operating system, including the task priorities of each computing task and the corresponding core processors assigned to each task. It also collects real-time temperature information for each core zone. This information provides a comprehensive input foundation for subsequent intelligent decision-making. Next, the system allocates core processors based on each computing task, determines the specific physical cooling zone on which each computing task will run, and establishes a relationship between abstract task information and physical cooling zones. Then, based on the task priorities of each computing task, the system determines the importance of each computing task in cooling resource allocation and quantifies this as a cooling resource allocation weight. Furthermore, the system comprehensively utilizes the collected real-time temperature information, the determined cooling resource allocation weights, the correspondence between tasks and zones, and the current total available coolant flow rate in the system. With the optimization objectives of minimizing the weighted average temperature reflecting task importance and minimizing the total coolant flow rate to minimize system energy consumption, a resource allocation algorithm is used to calculate the optimal coolant flow rate for each core zone at the given moment, i.e., the target coolant flow rate for each core zone. Finally, based on the calculated target coolant flow rate, the system sends control instructions to the flow control unit corresponding to each core zone to precisely adjust the coolant flow through each zone. The entire process is executed periodically, enabling the cooling system to adjust the cooling intensity of each area in real time according to task scheduling and dynamic temperature changes, achieving intelligent, efficient and differentiated cooling control.

[0068] As a preferred embodiment, the solution of the present application is implemented as follows: On a computing device with multiple core processors, its aluminum-based PCB cooling system includes multiple core zones, each corresponding to a core processor, and equipped with a microvalve as a flow control unit. An interface is used to obtain a task list provided by the operating system, including each task's ID, priority, and assigned core number. Simultaneously, a temperature sensor collects the real-time temperatures of the multiple core zones. Based on the assigned core number, the core zone in which the task is running is determined. Based on the task priority, a preset priority-weight mapping table is queried to obtain the cooling resource allocation weight for each task. If a zone has multiple tasks, the weights can be aggregated (for example, taking the mean or maximum value, but not limited to these) to obtain the total zone weight. An optimization problem is constructed, with the objective function consisting of minimizing the weighted average temperature of each core zone and minimizing the total coolant flow rate. Constraints include upper and lower flow limits for each zone, as well as a total flow constraint. A sequential quadratic programming algorithm is used to solve the optimization problem, obtaining target coolant flow rates for the multiple core zones. Based on the calculated target coolant flow rates for each zone, a control execution module sends control signals to the corresponding microvalves, adjusting their openings to bring the actual flow rate close to the target flow rate.

[0069] Through the above scheme, this application realizes differentiated heat dissipation guarantees for tasks of different importance by introducing task priority information, giving priority to ensuring the temperature stability of the areas where high-priority tasks are located, and reducing the risks brought to the execution of critical tasks by temperature fluctuations. Combined with real-time temperature feedback and forward-looking task information, the system can predict and respond to changes in thermal load more promptly and accurately, improving the responsiveness and accuracy of heat dissipation control. Under the constraint of limited total coolant flow in the system, resources are intelligently allocated through optimization algorithms, avoiding resource overload or inefficient operation caused by simply responding to temperature increases in all areas, thereby improving the utilization efficiency of heat dissipation resources. By taking minimizing the total coolant flow as one of the optimization goals, the operating load of the cooling pump is reduced while meeting the heat dissipation requirements, thereby effectively reducing the overall power consumption of the system. Overall, this method can effectively cope with the dynamic and uneven heat generation characteristics of multi-core processors, and improve the heat dissipation performance, stability and energy efficiency of high-performance computing equipment.

[0070] In some embodiments, step A3 comprises:

[0071] A301. According to the task priority of each computing task, query the task priority and heat dissipation resource allocation weight mapping table to obtain the heat dissipation resource allocation weight of each computing task.

[0072] The task priority and heat dissipation resource allocation weight mapping table refers to a data structure for storing the association between task priorities and corresponding heat dissipation resource allocation weights, which can be implemented in the form of a table, array, hash table or database.

[0073] This solution provides a specific, operational, and stable implementation mechanism by defining a specific method for obtaining cooling resource allocation weights. After obtaining task scheduling information, including the task priorities of each computing task, the system no longer relies on fuzzy rules or empirical judgment to determine cooling resource allocation weights. Instead, it directly queries a pre-established mapping table of task priorities and cooling resource allocation weights based on the task priorities of each computing task. This mapping table maps different task priority levels or values ​​to preset cooling resource allocation weights. Through a simple table lookup, the system can quickly, accurately, and stably obtain the cooling resource allocation weight that matches the current task priority. The obtained cooling resource allocation weights are then used in the subsequent resource allocation algorithm as an important input to the calculation of the target coolant flow rate for each core zone. This table-based approach converts abstract task priorities into specific, quantifiable weights, providing reliable basic data for subsequent optimization calculations. In this way, the system ensures that high-priority tasks receive higher cooling resource allocation weights. This prioritizes the cooling needs of the core zones where high-priority tasks reside when the resource allocation algorithm solves the target coolant flow rate. This enables differentiated cooling control based on task priority, even when total coolant flow is limited. This clear and stable weight acquisition mechanism, combined with the subsequent resource allocation algorithm, makes the entire heat dissipation control process more accurate and reliable.

[0074] For example, task priorities can be divided into multiple levels, such as 1 to 10, where 10 represents the highest priority. The task priority and heat dissipation resource allocation weight mapping table can be preset as follows: priority 1 corresponds to a weight value of 1, priority 2 corresponds to a weight value of 2, ..., priority 10 corresponds to a weight value of 10. When the system obtains a task priority of 8 for a certain computing task, that is, based on the priority 8, the mapping table is queried to obtain a corresponding heat dissipation resource allocation weight of 8. If the task priority of another computing task is 3, the mapping table is queried to obtain a weight of 3. These obtained weight values ​​are then input into the resource allocation algorithm to calculate the target coolant flow rate for each core area.

[0075] Preferably, the task scheduling information also includes task type and task amount information of each computing task;

[0076] After step A301, the following steps may also be included:

[0077] A302. Determine the energy consumption level of each computing task based on the task type and task amount information of each computing task;

[0078] A303. Adjust the heat dissipation resource allocation weight of each computing task according to the energy consumption level of each computing task to obtain a final heat dissipation resource allocation weight.

[0079] Specifically, in addition to task priority and task allocation core processors, task scheduling information further includes task type and task quantity information for each computing task. The task type can indicate the computing characteristics of the task, such as whether it is compute-intensive, IO-intensive, or memory-intensive. The task quantity information can indicate the scale of the task, such as the amount of data that needs to be processed (corresponding to compute-intensive), the expected number of instructions to be executed (corresponding to IO-intensive), or the duration (corresponding to memory-intensive). This information provides a richer description of task attributes than a single priority, and is the basis for evaluating the impact of the task on the processor's thermal load during actual execution. Based on the task type and task quantity information of each computing task, the energy consumption level of each computing task is determined.

[0080] An energy class is a quantitative or categorized representation of the expected heat or power consumption generated by a task during execution. The energy class determination process can be based on pre-set rules, models, or table lookups, converting task type and workload information into an indicator reflecting its thermal load potential. For example, tasks can be categorized as high-energy, medium-energy, or low-energy. Based on the energy class of each computing task, the initially determined cooling resource allocation weights are adjusted to obtain the final cooling resource allocation weights. This adjustment process aims to factor in the actual thermal load potential of the task into cooling resource allocation. This adjustment can be implemented using a function, a lookup table, or a set of rules (the specific function or rules can be set based on actual needs and are not limited here). Initial priority-based weights are modified based on the task's energy class. For example, the initial weights of tasks with high energy classes may be adjusted upward, while those of tasks with low energy classes may be adjusted downward. This adjustment ensures that the final cooling resource allocation weights more comprehensively reflect the importance of the task and its actual demand on the cooling system.

[0081] This solution further refines the weight determination process by querying a mapping table based on the task priority of each computing task and obtaining preliminary cooling resource allocation weights. The task scheduling information obtained by the system not only includes task priorities and assigned core processors, but also adds information about the task type and workload for each computing task. After obtaining the preliminary weights, the system uses this newly added task type and workload information to determine the energy consumption level of each computing task through an evaluation process. The energy consumption level reflects the expected level of heat generated by the task during execution. The system then adjusts the preliminary cooling resource allocation weights previously determined based on priority based on the determined energy consumption level. For example, for tasks with the same priority but different energy consumption levels, the task with a higher energy consumption level will be assigned a higher final cooling resource allocation weight, while the task with a lower energy consumption level may be assigned a lower final weight. This adjustment ensures that the final cooling resource allocation weights more accurately reflect the importance of the task and its actual demand on the cooling system. As a result, when subsequently solving for the target coolant flow rate based on these weights, the system can more effectively prioritize the allocation of limited coolant resources to core areas where important, heat-generating tasks reside. This allows for a better balance of cooling requirements across core areas within the constraints of total coolant flow, prioritizing temperature stability for critical tasks while avoiding the waste of excessive resources allocated to low-energy tasks. This approach, which combines task priority, task type, task volume, and energy consumption level to determine cooling resource allocation weights, allows for a more refined assessment of a task's actual cooling needs than relying solely on task priority, improving the targetedness and efficiency of cooling control.

[0082] Preferably, step A302 may include:

[0083] According to the task type of each computing task, query the task type and processor frequency impact factor mapping table to obtain the corresponding frequency impact factor;

[0084] Estimate the processor frequency increase value of each computing task during execution based on the task load information and frequency impact factor of each computing task;

[0085] The energy consumption level of each computing task is calculated based on the processor frequency boost value and task amount information.

[0086] The task type and processor frequency impact factor mapping table refers to a data structure that stores the association between different computing task types and corresponding frequency impact factors, which can be implemented using a lookup table, a database, or a configuration file.

[0087] The frequency impact factor refers to a numerical indicator that quantifies the tendency of a specific task type to trigger the processor to increase the operating frequency, and can be determined by pre-calibration or empirical value.

[0088] The processor frequency boost value refers to the increase in the processor's operating frequency relative to the base frequency during task execution. It can be estimated using a calculation model or prediction algorithm based on the task load and frequency impact factor. For example, the processor frequency boost value can be estimated using the following calculation formula: △P = P0*(1+k*λ*K), where △P is the processor frequency boost value, P0 is the base frequency, k is the task load ratio (the ratio of the task load to the preset base load), λ is the frequency impact factor, and K is the scale factor (a preset adjustment parameter).

[0089] Among them, the energy consumption level refers to a graded indicator that measures the intensity of heat generation during the execution of a computing task, which can be expressed as a discrete level division or a continuous numerical value. A formula or a lookup table can be used to determine the energy consumption level of each computing task. For example, the energy consumption level can be proportional to the product of the task amount and the frequency increase value, or the task can be further divided into three energy consumption levels: low, medium, and high based on the calculation results. For example, if the calculation result falls within a certain range, it is determined to be a "high energy consumption level". In this way, an energy consumption level that reflects its expected heat generation intensity can be calculated for each computing task.

[0090] This solution further refines the specific process for determining the energy consumption level of each computing task. This process first queries a pre-defined mapping table of task types and processor frequency impact factors based on the task type to obtain the corresponding frequency impact factor. This step identifies the potential impact of different task types on dynamic processor frequency adjustments. Next, by combining the task load information for each computing task with the obtained frequency impact factor, the potential frequency increase that the processor may experience during execution of each computing task is estimated. The task load reflects the total computing demand of the task, while the frequency impact factor reflects the tendency of the task type to drive frequency increases. Combining these two factors can more accurately predict the workload the processor will likely reach to complete the task. Processor frequency increases are a key factor in significantly increasing power consumption and heat generation, so accurately estimating the frequency increase is crucial for assessing the actual heat generation potential. Finally, based on the estimated processor frequency increase and task load information, the energy consumption level of each computing task is calculated. The energy consumption level comprehensively reflects the total workload and processing intensity per unit time, providing an indicator that more closely reflects the actual heat generation during task execution.

[0091] Through the above steps, this solution can more comprehensively and accurately evaluate the energy consumption level of each computing task, because it not only considers the task type and task amount, but also further considers the potential impact of task execution on processor frequency. Applying this more accurate energy consumption level information to the subsequent cooling resource allocation weight adjustment process can enable the cooling control system to more accurately predict and respond to local hotspots, especially when the processor is running at high load or performing critical tasks, thereby optimizing the cooling effect and improving system stability and energy efficiency. This method of determining energy consumption level by considering task type, task amount, and the potential impact of the task on processor frequency can more accurately capture the heat generation characteristics of the task than methods that rely solely on task type and task amount, providing a more solid foundation for forward-looking cooling control based on task characteristics.

[0092] Preferably, step A303 may include:

[0093] Taking each core processor as the target core in turn, obtain the set of all core processors in the neighborhood centered on the target core as the neighborhood core set (including the target core itself);

[0094] Based on the preset heat transfer coefficient and the distance between each neighboring core and the target core in the neighborhood core set, the heat influence factor of each neighboring core on the target core is calculated; the heat influence factor of the target core on itself is 1, and the heat influence factors of other neighboring cores on the target core are proportional to the heat transfer coefficient and inversely proportional to the distance;

[0095] The sum of the products of the thermal impact factors of all neighboring cores and the energy consumption levels of the corresponding computing tasks is calculated to adjust the heat dissipation resource allocation weight of the target core and obtain the final heat dissipation resource allocation weight.

[0096] The neighborhood core set refers to a group of core processors physically adjacent to or close to the target core processor. The range of the neighborhood core set can be determined based on the physical layout and thermal conductivity of the processors. Specifically, the size and shape of the neighborhood core set can be predetermined based on the physical layout and thermal conductivity of the processors. For example, it can be a circular area with a preset radius.

[0097] The preset thermal conductivity coefficient refers to a parameter reflecting the thermal conductivity of a heat dissipation medium such as an aluminum-based PCB, and the parameter can be obtained through experiments or table lookup based on actual material properties.

[0098] The thermal impact factor refers to a value that quantifies the degree to which the heat generated by the neighboring core affects the temperature of the target core through heat conduction. This value reflects the efficiency of heat conduction from the neighboring core to the target core.

[0099] The technical solution of this application further adjusts the cooling resource allocation weights determined based on the energy consumption level of the task itself by taking into account the thermal coupling effects between cores in a multi-core processor system. Specifically, the solution sequentially treats each core processor as a target core. For each target core, the set of core processors within its physical neighborhood, namely the neighborhood core set, is first determined. Next, based on the thermal conductivity characteristics of the heat dissipation medium and the physical distance between each core in the neighborhood and the target core, the thermal impact factor of each neighborhood core on the target core is calculated. The target core's own thermal impact factor is set to 1, indicating that its own heat generation has the direct and greatest impact on its own temperature. The thermal impact factors of other neighborhood cores are set to be proportional to the thermal conductivity coefficient and inversely proportional to the distance, in accordance with the basic physical law of heat conduction: the better the thermal conductivity of the medium and the closer the distance, the greater the thermal conduction effect. The solution then calculates the sum of the thermal impact factors of all cores in the neighborhood core set multiplied by the energy consumption level of the task running on them. This sum of products comprehensively reflects the impact of the total heat load on the target core area, from its own heat generation and heat conduction from the neighborhood. Finally, this sum of products is used to adjust the cooling resource allocation weights initially determined for the target core, resulting in final cooling resource allocation weights that better reflect the actual thermal load distribution. For example, if the target core's own tasks are not energy-intensive, but its neighborhood is close to multiple high-energy-consuming tasks, the resulting sum of products will be larger, causing the target core's cooling resource allocation weight to be adjusted upward to account for the additional heat load from the neighborhood. This adjustment method combines task energy consumption information with the thermal coupling effect between cores, allowing the cooling resource allocation weights to take into account not only the task's own cooling needs but also the thermal environment of its physical location. This makes subsequent cooling resource allocation based on these weights more accurate and effective. Compared to adjustments based solely on task energy consumption, this thermal coupling-inclusive adjustment method more accurately reflects the actual thermal load distribution, particularly when there are local hotspots or areas where high-energy-consuming tasks are concentrated, allowing for more effective identification and response to the cooling needs of these areas.

[0100] In one specific embodiment, the neighborhood core set can be defined as all core processors whose Manhattan distance from the target core in a two-dimensional processor array is less than or equal to a preset threshold (e.g., a threshold of 2). The distance between cores can be represented by their grid distance in the processor array. The preset thermal conductivity coefficient can be set based on the thermal conductivity of the aluminum-based PCB material. The thermal impact factor can be specifically calculated as follows: for the target core itself, the thermal impact factor is 1; for other cores in the neighborhood, the thermal impact factor can be calculated as h / d, where h is a constant positively correlated with the thermal conductivity coefficient and d is the distance between the neighborhood core and the target core. After calculating the sum of the thermal impact factors of all neighboring cores and the energy consumption level of the corresponding computing tasks, this sum can be multiplied by an adjustment coefficient and then added to or multiplied by the target core's cooling resource allocation weight, which is initially determined based on the task priority and its own energy consumption, to obtain the final weight. For example, A1=A0+M*(∑(ri*Ei)), where A1 is the final heat dissipation resource allocation weight, A0 is the heat dissipation resource allocation weight before adjustment, M is the preset adjustment coefficient, ri is the thermal impact factor of the i-th neighborhood core, Ei is the energy consumption level of the computing task corresponding to the i-th neighborhood core, and ∑(ri*Ei) represents the sum of the products of the thermal impact factors of all neighborhood cores and the energy consumption levels of the corresponding computing tasks.

[0101] Through this technical solution, cooling resource allocation weights can more accurately reflect the actual thermal load distribution in a multi-core processor system, including the heat generated by tasks themselves and the thermal coupling between cores. This enables subsequent cooling resource allocation to more effectively address dynamically changing local hotspots, prioritizing temperature stability in areas with high-priority tasks and optimizing overall cooling efficiency within limited total cooling resources.

[0102] Preferably, in step A4, the resource allocation algorithm may adopt a linear programming algorithm, a quadratic programming algorithm or a model predictive control algorithm.

[0103] Linear programming is a mathematical optimization method used to solve problems where both the objective function and constraints are linear. Quadratic programming is a mathematical optimization method used to solve problems where the objective function is a quadratic function and the constraints are linear. Model predictive control is a control strategy that predicts future behavior based on a system model and solves an optimization problem in each control cycle to determine the control action.

[0104] By limiting the resource allocation algorithm to a linear programming algorithm, a quadratic programming algorithm, or a model predictive control algorithm, the present application can quickly and accurately solve the target coolant flow rate of each core area, overcoming the efficiency and accuracy issues that may be caused by unrestricted algorithm types, and enhancing the ability to handle complex constraints, thereby improving the system's real-time response capability and optimization effect.

[0105] In some possible implementations, step A4 includes:

[0106] A401. Construct an optimization objective function, wherein the optimization objective function includes minimizing the weighted average temperature of each core region and minimizing the total coolant flow rate; wherein the weight of each core region is determined by the heat dissipation resource allocation weight of each computing task and the corresponding relationship between each computing task and each core region, and the weighted average temperature is obtained by multiplying the temperature information of each core region by the corresponding weight, summing the result, and dividing the result by the sum of the weights;

[0107] A402. Using an optimization algorithm based on sequential quadratic programming, with the coolant flow rate of each core area as the optimization variable, solve the optimization objective function under the constraints of the constraints to obtain the target coolant flow rate of each core area; the constraints include the upper and lower limits of the coolant flow rate of each core area and the total available coolant flow rate of the system.

[0108] The optimization objective function is a mathematical expression that quantifies the objective to be optimized, such as minimizing temperature or flow rate. It can combine multiple objectives into a single function (such as a weighted sum) or be a multi-objective function. The weighted average temperature is a metric that measures the overall temperature state. Different weights are assigned to different regions to reflect their relative importance or sensitivity to temperature, thereby highlighting the temperature impact of key areas when calculating the average.

[0109] The sequential quadratic programming-based optimization algorithm is an iterative method for solving nonlinear constrained optimization problems. It approximates the original problem by solving a quadratic programming subproblem in each iteration. The optimization variable is the variable whose value must be determined to achieve the optimal goal during the optimization process. In this solution, it is the coolant flow rate in each core area. Constraints are restrictions that must be met for the value of the optimization variable. These restrictions can be physical (such as upper and lower flow limits) or system resource constraints (such as total available flow).

[0110] This solution builds on the basic approach by further refining the steps for determining the target coolant flow rate for each core area. First, by constructing a clear optimization objective function, the cooling resource allocation problem is transformed into a computable mathematical problem. This objective function comprehensively considers the two interrelated objectives of minimizing the weighted average temperature of each core area and minimizing the total coolant flow rate. The introduction of the weighted average temperature, combined with weights determined by factors such as task priority, allows the optimization process to prioritize the temperature stability of core areas that are more temperature-sensitive or carry more important tasks. Simultaneously, minimizing the total coolant flow rate is directly linked to reducing system energy consumption. This dual-objective design, through weight adjustment, balances ensuring the cooling requirements of critical areas with reducing overall energy consumption. The specific calculation method for the weighted average temperature is also clarified, providing a foundation for the subsequent mathematical solution. Second, the solution explicitly employs an optimization algorithm based on sequential quadratic programming to solve the constructed optimization problem. The sequential quadratic programming algorithm is an effective method for solving nonlinear constrained optimization problems. Its iterative solution process can gradually approach the optimal solution and generally exhibits good convergence. By setting the coolant flow rate for each core zone as the optimization variable, the algorithm searches for the flow rate combination that minimizes the optimization objective function, while satisfying the upper and lower limits of each zone's flow rate and the total available coolant flow rate of the system. This approach provides a structured and actionable solution process that dynamically calculates the optimal coolant flow allocation plan for each zone based on real-time temperature, task information, and system resources.

[0111] For example, in a specific implementation, the optimization objective function can be constructed in the form of a weighted summation, such as: Min(w_t*(Sum(we_i*T_i) / Sum(we_i))+w_f*Sum(F_i)), where w_t and w_f are weight coefficients used to balance the temperature target and the flow target, T_i is the temperature of the i-th core area, we_i is the weight of the i-th core area, and F_i is the coolant flow of the i-th core area. The weight of the core area we_i can be calculated based on the heat dissipation resource allocation weights of the various computing tasks assigned to the area and their corresponding relationships, such as taking the maximum value or weighted average of the task weights in the area. The optimization algorithm based on sequential quadratic programming can be implemented using an existing numerical optimization library, such as using an optimization solver that supports the sequential quadratic programming algorithm. During the solution process, the coolant flow rate F_i in each core area is used as the optimization variable, and constraints are added, such as F_i_min <= F_i <= F_i_max (where F_i_min and F_i_max are the physical limitations of the flow control unit in the i-th core area) and Sum(F_i) <= Fz (where Fz is the total available flow rate provided by the cooling pump). The optimization algorithm iterates until the flow rate value for each area that satisfies the constraints and minimizes the objective function is found, thus obtaining the target coolant flow rate for each core area.

[0112] By constructing a clear optimization objective function and solving it using an optimization algorithm based on sequential quadratic programming, this solution provides a concrete and actionable implementation of a resource allocation algorithm. This enables the system to dynamically calculate the optimal coolant flow distribution plan that meets the constraints based on real-time temperature and task information. This solution effectively balances the cooling needs of different core areas, prioritizing temperature stability in areas with high-priority tasks while minimizing the total coolant flow. This allows for more precise and efficient cooling control within limited total cooling resources, improving the overall temperature stability and energy efficiency of the system.

[0113] Preferably, step A402 may include:

[0114] B1. Initialize the parameters of the sequential quadratic programming algorithm, including the initial coolant flow rate value, Hessian matrix, Lagrange multiplier estimate, and convergence tolerance of each core area;

[0115] B2. Calculate the optimization objective function value and its gradient based on the coolant flow rate values ​​in each core area at the current iteration, as well as the constraints and their Jacobian matrix.

[0116] B3. Construct a quadratic programming subproblem, centered around the coolant flow rate values ​​in each core region of the current iteration. Based on the calculated optimization objective function value and its gradient and the Lagrange multiplier estimate, perform a quadratic approximation on the optimization objective function and a linear approximation on the constraints. The quadratic approximation uses a Hessian matrix, and the linear approximation uses a Jacobian matrix.

[0117] B4. Using the active set method to solve the quadratic programming subproblem, the search direction and search step size of the coolant flow value in each core area are obtained;

[0118] B5. Update the coolant flow value of each core area based on the search direction and search step size of the coolant flow value of each core area, and adjust the search step size according to the Armijo line search strategy to ensure that the objective function value decreases and the constraints are satisfied;

[0119] B6. Determine whether the convergence condition determined by the convergence tolerance is met. If so, use the latest coolant flow rate value of each core area as the target coolant flow rate of each core area. Otherwise, use the BFGS formula to update the Hessian matrix and Lagrange multiplier estimates, and return to step B2.

[0120] The parameters of the sequential quadratic programming algorithm are a set of values ​​required to initiate and control the sequential quadratic programming iteration process. These parameters include the initial coolant flow rate for each core region used to start the iteration, the initial Hessian matrix used to approximate the second-order information of the objective function, the initial Lagrange multiplier estimates used to handle constraints, and the convergence tolerance used to determine when the algorithm should stop. The Hessian matrix is ​​a square matrix containing the second-order partial derivatives of the objective function, which can be used to quadratically approximate the objective function at the current point. The Lagrange multiplier estimates are the current estimates of the auxiliary variables introduced in the constrained optimization problem. They are used to incorporate the constraints into the objective function, forming the Lagrangian function and reflecting the impact of the constraints on the optimal solution during the iteration process. The convergence tolerance is a small, preset positive value that defines the threshold for determining whether the algorithm has reached a sufficiently accurate solution. For example, when the iteration step size or the norm of the gradient is less than this tolerance, the algorithm is considered to have converged (i.e., the convergence criteria have been met).

[0121] The quadratic programming subproblem refers to a simplified optimization problem constructed and solved in each iteration of the sequential quadratic programming algorithm. It can have a quadratic objective function and linear constraints. By solving this subproblem, the search direction and step size of the current iteration point can be obtained.

[0122] The active set method is an iterative algorithm for solving quadratic programming problems that maintains a set of constraints that are currently considered "active," or in effect, and adjusts this set at each iteration until a solution that satisfies the KKT conditions is found. The search direction is the direction in which the variables should be moved at the current iteration point. This direction is obtained by solving the quadratic programming subproblem and points to an area where the objective function decreases and the constraints are satisfied. The search step size is the distance to move along the search direction. This step size is determined by a line search method to ensure that the objective function decreases sufficiently while satisfying the constraints.

[0123] Among them, the Armijo line search strategy refers to a non-exact line search method, which can be used to determine the appropriate search step size to strike a balance between ensuring that the objective function fully decreases and satisfying the constraints, and avoid slow convergence caused by too small a step size.

[0124] The BFGS formula refers to an update formula in a commonly used quasi-Newton method, which can be used to iteratively update the approximation of the Hessian matrix without calculating the exact second-order derivative, thereby improving the computational efficiency of the algorithm.

[0125] This technical solution provides a systematic optimization approach by detailing the iterative process for solving the target coolant flow rate using a sequential quadratic programming algorithm. First, key parameters are initialized to lay the foundation for the algorithm's startup. Next, in each iteration, first-order information about the objective function and constraints is calculated based on the current flow rate value. This information serves as the basis for constructing a local approximation model. A quadratic programming subproblem is then constructed and solved. This subproblem transforms the original nonlinear problem into a more tractable quadratic programming problem by performing a quadratic approximation on the objective function and a linear approximation on the constraints. Solving this subproblem yields a search direction and step size, indicating how to approach the optimal solution. The flow rate value is then updated based on the search direction and step size, and the step size is adjusted using a line search strategy to ensure the stability and effectiveness of the iterative process, ensuring that the objective function decreases and the constraints are satisfied. Finally, convergence conditions are checked to determine whether the optimal solution has been reached. If not, the Hessian matrix and Lagrange multiplier estimates are updated to prepare for the next iteration. This iterative process continues until convergence conditions are met, and the resulting flow rate value becomes the target coolant flow rate for each core area. Applying this detailed optimization solution process to an optimization problem constructed based on task priority, temperature, and other information can effectively determine the optimal coolant flow distribution in each core area under constraints such as the total system flow, thereby achieving precise control of the aluminum-based PCB microchannel cooling system to cope with the dynamic and uneven heat generation characteristics.

[0126] In one embodiment, the implementation of the sequential quadratic programming algorithm can be assisted by a numerical calculation library. For example, when initializing parameters, the initial coolant flow rate value can be set according to experience or preset values, the initial Hessian matrix can be set to the unit matrix, the initial Lagrange multiplier estimate can be set to the zero vector, and the convergence tolerance can be set to a small positive number. When calculating the optimization objective function value and its gradient, the constraint conditions and their Jacobian matrix, symbolic differentiation or numerical differentiation methods can be used for calculation according to the specific function and constraint expressions. When constructing a quadratic programming subproblem, a standard quadratic programming problem form can be formed based on the optimization objective function value, gradient, Hessian matrix approximation and Jacobian matrix of the current iteration point. When solving the quadratic programming subproblem, the active set method solver implemented in the existing optimization library can be called. When updating the flow value and adjusting the search step size, the search direction obtained by the solver can be combined with the Armijo line search criterion to iteratively try different step sizes to find the step size that meets the conditions. When updating the Hessian matrix and Lagrange multiplier estimates, the BFGS formula can be used to calculate the gradient, flow value, and Lagrange multiplier information of the current and previous iterations. The entire process is executed in a control loop until the convergence condition is met.

[0127] The detailed sequential quadratic programming iterative steps described above provide a specific, actionable technical solution for solving the target coolant flow rate, which helps improve algorithm performance and convergence. This solves the problem of simply stating the algorithm without specifying its specific implementation steps, which can lead to implementation difficulties and ineffective convergence or computational efficiency. This solution effectively determines the optimal coolant flow distribution for each core area while considering multiple factors such as task priority, temperature, and total flow constraints, thereby achieving precise control of the aluminum-based PCB microchannel cooling system.

[0128] Preferably, after step A402, the following steps may also be included:

[0129] A403 obtains the task content category of each computing task, queries the preset task content category and temperature sensitivity level mapping table, and obtains the temperature sensitivity level corresponding to each computing task;

[0130] A404. Adjust the target coolant flow rate of each core area based on the temperature sensitivity level of each computing task using preset adjustment rules.

[0131] Among them, the task content category refers to the classification identification of the computing task content, such as real-time control tasks, data processing tasks, background service tasks, graphics rendering tasks, etc., which can be identified by task attribute information provided by the operating system or through task name, process ID, etc.

[0132] Among them, the temperature sensitivity level refers to a quantitative or graded expression that measures the tolerance of a computing task to temperature changes or the requirements for temperature stability. For example, it can be divided into high sensitivity, medium sensitivity, low sensitivity, etc., which can be represented by a preset enumeration value or numerical range.

[0133] Among them, the task content category and temperature sensitivity level mapping table refers to a data structure that stores the association relationship between different task content categories and their corresponding temperature sensitivity levels, which can be stored and queried in the form of a lookup table, hash table or database.

[0134] Among them, the preset adjustment rule refers to the logic or algorithm for correcting the preliminary target coolant flow rate according to the task temperature sensitivity level, which can be implemented by using a lookup table, piecewise function, proportional factor or fuzzy control.

[0135] This solution, based on the preliminary target coolant flow rates for each core region obtained through an optimization algorithm, further incorporates task temperature sensitivity as an adjustment factor. First, by obtaining the task content categories of each computing task currently executing on each core region and querying a pre-set mapping between task content categories and temperature sensitivity levels, the system can identify each task's specific tolerance for temperature fluctuations or stability requirements. This process converts the abstract task content categories into temperature sensitivity information that can be used to make cooling control decisions. Subsequently, the system adjusts the preliminary target coolant flow rates for each core region, previously calculated by the optimization algorithm, based on the obtained temperature sensitivity levels of each computing task and pre-set adjustment rules. For example, for a core region carrying highly temperature-sensitive tasks, even if the preliminary optimization results already account for its priority and current temperature, the adjustment rules may further increase the coolant flow rate allocated to that region to provide greater cooling capacity and ensure that the temperature remains within a tighter range. Conversely, for core regions carrying less temperature-sensitive tasks, the adjustment rules may appropriately reduce the allocated flow rate without compromising task stability. This allows the saved coolant resources to be allocated to regions that need them more, or reduces the overall system flow requirement to optimize energy consumption.

[0136] This secondary adjustment based on task temperature sensitivity allows for a more refined coolant flow allocation plan, more accurately matching the actual cooling requirements of different tasks and, in particular, prioritizing the operational stability of temperature-sensitive critical tasks. This adjustment mechanism, combined with an optimization algorithm based on priority and temperature, creates a more comprehensive and intelligent cooling resource allocation strategy, overcoming the limitations of prioritizing and temperature alone, which can overlook the inherent temperature characteristics of tasks.

[0137] For example, in one specific embodiment, the system can obtain a list of tasks currently running on each core processor and their task content categories from the operating system. A preset mapping table of task content categories and temperature sensitivity levels can be stored in a configuration file or database table. For example, real-time control tasks are mapped to high sensitivity, data analysis tasks are mapped to medium sensitivity, and background logging tasks are mapped to low sensitivity. Based on the task allocation core processor information, the system determines the task content category currently primarily handled by each core area and queries the mapping table to obtain the corresponding temperature sensitivity level. Preset adjustment rules can be defined as follows: for core areas with high-sensitivity tasks, the initial target coolant flow rate is increased by a first preset percentage (e.g., 10%); for areas with medium-sensitivity tasks, the initial flow rate is maintained or fine-tuned; and for areas with low-sensitivity tasks, the initial flow rate is reduced by a second preset percentage (e.g., 5%), ensuring it does not fall below the minimum flow limit. The system applies these adjustment factors to the initial target coolant flow rate output by the optimization algorithm to obtain the target coolant flow rate for each core area, which is ultimately used to control the flow control unit.

[0138] refer to Figure 2 The present application provides an aluminum-based PCB microchannel heat dissipation control system for controlling an aluminum-based PCB microchannel heat dissipation system of a computing device having multiple core processors. The aluminum-based PCB microchannel heat dissipation system has multiple core areas and multiple flow control units. Each core processor is respectively disposed at each core area, and each flow control unit is respectively used to adjust the coolant flow rate of each core area. The system includes:

[0139] Information collection module 1, used to obtain task scheduling information and temperature information of each core area; the task scheduling information includes the task priority of each computing task and the corresponding task allocation core processor (the specific process can be referred to step A1 above);

[0140] Mapping module 2, used to allocate core processors according to various computing tasks and determine the correspondence between various computing tasks and various core areas (the specific process can be referred to step A2 above);

[0141] Weight determination module 3, used to obtain the heat dissipation resource allocation weight of each computing task according to the task priority of each computing task (the specific process can be referred to step A3 above);

[0142] Optimization calculation module 4 is used to calculate the target coolant flow rate for each core area using a resource allocation algorithm based on the temperature information of each core area, the heat dissipation resource allocation weights of each computing task, the correspondence between each computing task and each core area, and the total available coolant flow rate of the system, with the goal of minimizing the weighted average temperature of each core area and minimizing the total coolant flow rate (the specific process can be referred to in step A4 above);

[0143] The control execution module 5 is used to control each flow control unit to adjust the coolant flow of each core area according to the target coolant flow of each core area (the specific process can be referred to the above step A5).

[0144] The above description is merely an embodiment of the present application and is not intended to limit the scope of protection of the present application. For those skilled in the art, various modifications and variations of the present application are possible. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A method for controlling aluminum-based PCB microchannel heat dissipation, for controlling an aluminum-based PCB microchannel heat dissipation system of a computing device having multiple core processors, wherein the aluminum-based PCB microchannel heat dissipation system comprises multiple core regions and multiple flow control units, each core processor being disposed at each core region, and each flow control unit being configured to regulate the coolant flow rate at each core region; characterized in that: The steps of the method include: A1 obtains task scheduling information and temperature information of each core area; the task scheduling information includes the task priority of each computing task and the corresponding task allocation core processor; A2. Assign core processors to each computing task and determine the correspondence between each computing task and each core area; A3. Obtain the heat dissipation resource allocation weight for each computing task based on the task priority of each computing task; A4. Based on the temperature information of each core area, the cooling resource allocation weights of each computing task, the correspondence between each computing task and each core area, and the total available coolant flow rate of the system, a resource allocation algorithm is used to determine the target coolant flow rate for each core area, with the goal of minimizing the weighted average temperature of each core area and minimizing the total coolant flow rate. A5. Control each flow control unit to adjust the coolant flow rate of each core area according to the target coolant flow rate of each core area; Step A3 includes: A301. According to the task priority of each computing task, query the task priority and cooling resource allocation weight mapping table to obtain the cooling resource allocation weight of each computing task; The task scheduling information also includes task type and task amount information of each computing task; After step A301, the following steps are included: A302. Determine the energy consumption level of each computing task based on the task type and task amount information of each computing task; A303. According to the energy consumption level of each computing task, the heat dissipation resource allocation weight of each computing task is adjusted to obtain the final heat dissipation resource allocation weight; Step A303 includes: Taking each core processor as the target core in turn, obtain the set of all core processors in the neighborhood centered on the target core as the neighborhood core set; Based on the preset heat transfer coefficient and the distance between each neighboring core and the target core in the neighborhood core set, the heat influence factor of each neighboring core on the target core is calculated; the heat influence factor of the target core on itself is 1, and the heat influence factors of other neighboring cores on the target core are proportional to the heat transfer coefficient and inversely proportional to the distance; The sum of the products of the thermal impact factors of all neighboring cores and the energy consumption levels of the corresponding computing tasks is calculated to adjust the heat dissipation resource allocation weight of the target core and obtain the final heat dissipation resource allocation weight.

2. The aluminum-based PCB microchannel heat dissipation control method according to claim 1, characterized in that: Step A302 includes: According to the task type of each computing task, query the task type and processor frequency impact factor mapping table to obtain the corresponding frequency impact factor; Estimate the processor frequency increase value of each computing task during execution based on the task load information and frequency impact factor of each computing task; The energy consumption level of each computing task is calculated based on the processor frequency boost value and task amount information.

3. The aluminum-based PCB microchannel heat dissipation control method according to claim 1, characterized in that: The resource allocation algorithm is a linear programming algorithm, a quadratic programming algorithm or a model predictive control algorithm.

4. The aluminum-based PCB microchannel heat dissipation control method according to claim 1, characterized in that: Step A4 includes: A401. Construct an optimization objective function, wherein the optimization objective function includes minimizing the weighted average temperature of each core region and minimizing the total coolant flow rate; wherein the weight of each core region is determined by the heat dissipation resource allocation weight of each computing task and the corresponding relationship between each computing task and each core region, and the weighted average temperature is obtained by multiplying the temperature information of each core region by the corresponding weight, summing the result, and dividing the result by the sum of the weights; A402. Using an optimization algorithm based on sequential quadratic programming, with the coolant flow rate of each core area as the optimization variable, solve the optimization objective function under the constraints of the constraints to obtain the target coolant flow rate of each core area; the constraints include the upper and lower limits of the coolant flow rate of each core area and the total available coolant flow rate of the system.

5. The aluminum-based PCB microchannel heat dissipation control method according to claim 4, characterized in that: Step A402 includes: B1. Initialize the parameters of the sequential quadratic programming algorithm, including the initial coolant flow rate value, Hessian matrix, Lagrange multiplier estimate, and convergence tolerance of each core area; B2. Calculate the optimization objective function value and its gradient based on the coolant flow rate values ​​in each core area at the current iteration, as well as the constraints and their Jacobian matrix. B3. Construct a quadratic programming subproblem, centered around the coolant flow rate values ​​in each core region of the current iteration. Based on the calculated optimization objective function value and its gradient and the Lagrange multiplier estimate, perform a quadratic approximation on the optimization objective function and a linear approximation on the constraints. The quadratic approximation uses a Hessian matrix, and the linear approximation uses a Jacobian matrix. B4. Using the active set method to solve the quadratic programming subproblem, the search direction and search step size of the coolant flow value in each core area are obtained; B5. Update the coolant flow value of each core area based on the search direction and search step size of the coolant flow value of each core area, and adjust the search step size according to the Armijo line search strategy to ensure that the objective function value decreases and the constraints are satisfied; B6. Determine whether the convergence condition determined by the convergence tolerance is met. If so, use the latest coolant flow rate value of each core area as the target coolant flow rate of each core area. Otherwise, use the BFGS formula to update the Hessian matrix and Lagrange multiplier estimates, and return to step B2.

6. The aluminum-based PCB microchannel heat dissipation control method according to claim 4, characterized in that: After step A402, the following steps are also included: A403 obtains the task content category of each computing task, queries the preset task content category and temperature sensitivity level mapping table, and obtains the temperature sensitivity level corresponding to each computing task; A404. Adjust the target coolant flow rate of each core area based on the temperature sensitivity level of each computing task using preset adjustment rules.

7. An aluminum-based PCB microchannel heat dissipation control system for controlling an aluminum-based PCB microchannel heat dissipation system of a computing device having multiple core processors, wherein the aluminum-based PCB microchannel heat dissipation system comprises multiple core regions and multiple flow control units, wherein each core processor is disposed at each core region, and each flow control unit is configured to regulate the coolant flow rate of each core region; characterized in that: The system includes: An information collection module is used to obtain task scheduling information and temperature information of each core area; the task scheduling information includes the task priority of each computing task and the corresponding task allocation core processor; A mapping module is used to allocate core processors according to various computing tasks and determine the corresponding relationship between various computing tasks and various core areas; A weight determination module is used to obtain the heat dissipation resource allocation weight of each computing task according to the task priority of each computing task; An optimization calculation module is used to calculate the target coolant flow rate for each core area using a resource allocation algorithm, based on the temperature information of each core area, the heat dissipation resource allocation weights of each computing task, the correspondence between each computing task and each core area, and the total available coolant flow rate of the system, with the goal of minimizing the weighted average temperature of each core area and minimizing the total coolant flow rate; A control execution module for controlling each flow control unit to adjust the coolant flow of each core area according to the target coolant flow of each core area; When the weight determination module obtains the heat dissipation resource allocation weight of each computing task based on the task priority of each computing task, it executes: A301. According to the task priority of each computing task, query the task priority and cooling resource allocation weight mapping table to obtain the cooling resource allocation weight of each computing task; The task scheduling information also includes task type and task amount information of each computing task; After step A301, the following steps are included: A302. Determine the energy consumption level of each computing task based on the task type and task amount information of each computing task; A303. According to the energy consumption level of each computing task, the heat dissipation resource allocation weight of each computing task is adjusted to obtain the final heat dissipation resource allocation weight; Step A303 includes: Taking each core processor as the target core in turn, obtain the set of all core processors in the neighborhood centered on the target core as the neighborhood core set; Based on the preset heat transfer coefficient and the distance between each neighboring core and the target core in the neighborhood core set, the heat influence factor of each neighboring core on the target core is calculated; the heat influence factor of the target core on itself is 1, and the heat influence factors of other neighboring cores on the target core are proportional to the heat transfer coefficient and inversely proportional to the distance; The sum of the products of the thermal impact factors of all neighboring cores and the energy consumption levels of the corresponding computing tasks is calculated to adjust the heat dissipation resource allocation weight of the target core and obtain the final heat dissipation resource allocation weight.

Images