Cold plate liquid cooling system control method, device, equipment, medium and product
By optimizing the control of the cold plate liquid cooling system using the Informer time series prediction model and the dung beetle algorithm, the shortcomings of real-time control of the cold plate liquid cooling system are solved, precise heat dissipation and energy consumption optimization are achieved, and the system stability and energy efficiency are improved.
Patent Information
- Application Number
- CN202411085696.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-08
- Publication Date
- 2024-11-29
AI Technical Summary
Existing cold plate liquid cooling systems lack real-time control solutions, resulting in unmet heat dissipation requirements and deficiencies in energy consumption and fault diagnosis.
By employing the Informer time series prediction model combined with the dung beetle algorithm, the system acquires historical operating status data of the cold plate liquid cooling system and the cooling server to predict future operating status. Based on a multi-objective optimization control model, the system seeks the optimal control scheme to optimize temperature control, fault diagnosis, and energy consumption.
It achieves real-time and precise heat dissipation of the cold plate liquid cooling system, improves energy efficiency, reduces heat dissipation capacity loss, enhances system stability and reliability, and meets the requirements of green and low-carbon development.
Smart Images

Figure CN119045322A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of liquid cooling system control, and in particular to a cold plate type liquid cooling system control method, device, equipment, medium and product. Background Art
[0002] With the development of the digital economy, the scale of data centers, as a new type of infrastructure, has continued to expand, resulting in an overall upward trend in the average single-rack power of data centers. Air cooling can no longer meet the heat dissipation needs, and cold plate liquid cooling systems have been widely promoted.
[0003] Existing methods for improving the operating state of cold plate refrigeration systems mostly involve upgrading and optimizing cold plate liquid cooling hardware and coolant, but there is no technical solution for optimizing the real-time control scheme during the operation of the cold plate liquid cooling system. Summary of the invention
[0004] The present invention provides a cold plate liquid cooling system control method, device, equipment, storage medium and program product to solve the defect that the prior art does not optimize the real-time control mode during the operation of the cold plate liquid cooling system, and optimize the real-time control mode during the operation of the cold plate liquid cooling system.
[0005] The present invention provides a cold plate type liquid cooling system control method, comprising: Acquire multiple control schemes for the cold plate type liquid cooling system to be controlled, each of the control schemes including values of multiple controllable parameters of the cold plate type liquid cooling system to be controlled; Inputting input data into a trained operation status prediction model to obtain operation status prediction data output by the operation status prediction model, wherein the input data includes the control scheme, the cold plate type liquid cooling system to be controlled, and historical operation status data of a cooling server corresponding to the cold plate type liquid cooling system to be controlled, and the operation status prediction data reflects the operation status of the cold plate type liquid cooling system to be controlled and the cooling server at a future moment; Based on the operation state prediction data respectively corresponding to the control schemes, multiple control schemes are optimized to obtain a target control scheme for the cold plate liquid cooling system to be controlled; The running status prediction model is obtained by training based on multiple sets of training data, and each set of training data includes sample input data and running status prediction data labels corresponding to the sample input data.
[0006] According to the cold plate liquid cooling system control method provided by the present invention, the optimization of multiple control schemes is performed based on the operation state prediction data respectively corresponding to the control schemes to obtain the target control scheme of the cold plate liquid cooling system to be controlled, including: Based on the running state prediction data corresponding to the control scheme, an objective function value corresponding to the control scheme is obtained, wherein the objective function value reflects the temperature control effect, fault information and energy consumption state of the cold plate type liquid cooling system to be controlled corresponding to the running state prediction data; Based on the objective function values corresponding to the control schemes, multiple control schemes are optimized to obtain the target control scheme.
[0007] According to the cold plate liquid cooling system control method provided by the present invention, the objective function value corresponding to the control scheme is obtained based on the operating state prediction data corresponding to the control scheme, including: Determine a first function value based on the server shell temperature and the cold plate inlet temperature in the operation state prediction data, wherein the first function value reflects the temperature control effect of the cold plate type liquid cooling system to be controlled corresponding to the operation state prediction data; Determine a second function value based on each operating parameter in the operating state prediction data and actual statistical values of each operating parameter, wherein the second function value reflects fault information of the cold plate type liquid cooling system to be controlled corresponding to the operating state prediction data; A third function value is determined based on the difference between the server shell temperature and the ambient temperature in the operating status prediction data and the server power consumption in the operating status prediction data, and the third function value reflects the energy consumption status of the cold plate liquid cooling system to be controlled and the cooling server corresponding to the operating status prediction data.
[0008] According to the cold plate liquid cooling system control method provided by the present invention, the optimization of multiple control schemes based on the objective function value corresponding to the control scheme to obtain the target control scheme of the cold plate liquid cooling system to be controlled includes: updating each individual in the first population based on the dung beetle algorithm to obtain a second population, wherein the individuals in the first population correspond to the control scheme; performing non-dominated sorting on the individuals in the second population based on the objective function value corresponding to each individual in the second population, and selecting the individuals in the second population based on the sorting result to obtain a third population; Selecting individuals from the third population based on the reference point to obtain a fourth population; Clearing the first population, and adding individuals in the fourth population to the first population; Repeat the step of updating each individual in the first population based on the dung beetle algorithm to obtain the second population until a preset end condition is met, wherein the total number of individuals in the fourth population reaches a preset number or the number of times the fourth population is generated reaches a preset number; The target control scheme is determined based on each individual in the fourth population.
[0009] According to the cold plate liquid cooling system control method provided by the present invention, the updating of each individual in the first population based on the dung beetle algorithm to obtain the second population includes: Based on the dung beetle algorithm, each individual in the first population is updated to obtain the intermediate population; Based on the operating status prediction data corresponding to the individuals in the intermediate population, the penalty function value corresponding to each individual in the intermediate population is determined, and the individuals in the intermediate population are updated based on the penalty function value to obtain the second population, wherein the penalty function value reflects the thermal performance and flow resistance performance of the cold plate in the cold plate liquid cooling system to be controlled corresponding to the state prediction data.
[0010] According to the cold plate liquid cooling system control method provided by the present invention, the penalty function value corresponding to each individual in the intermediate population is determined based on the operating state prediction data corresponding to the individuals in the intermediate population, including: When the cold plate inlet flow rate in the operating state prediction data corresponding to the individual in the intermediate population is not within a first preset range, determining the penalty function value based on the actual value of the cold plate outlet flow rate and the cold plate inlet flow rate in the operating state prediction data; When the server power consumption in the operating status prediction data corresponding to the individual in the intermediate population is not within the second preset range, the penalty function value is determined based on the cold plate thermal resistance of the cold plate liquid cooling system to be controlled, the service shell temperature in the operating status prediction data, the cold plate inlet temperature, and the server power consumption.
[0011] The present invention also provides a cold plate type liquid cooling system control device, comprising: A scheme generating module, used for acquiring a plurality of control schemes for the cold plate type liquid cooling system to be controlled, each of the control schemes comprising values of a plurality of controllable parameters of the cold plate type liquid cooling system to be controlled; A state prediction module, used for inputting input data into a trained operation state prediction model, and obtaining operation state prediction data output by the operation state prediction model, wherein the input data includes the control scheme, the cold plate type liquid cooling system to be controlled, and historical operation state data of a cooling server corresponding to the cold plate type liquid cooling system to be controlled, and the operation state prediction data reflects the operation state of the cold plate type liquid cooling system to be controlled and the cooling server at a future moment; An optimization module, used for optimizing a plurality of control schemes based on the operation state prediction data respectively corresponding to the control schemes, to obtain a target control scheme for the cold plate type liquid cooling system to be controlled; The running status prediction model is obtained by training based on multiple sets of training data, and each set of training data includes sample input data and running status prediction data labels corresponding to the sample input data.
[0012] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, any of the above-mentioned cold plate liquid cooling system control methods is implemented.
[0013] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the cold plate liquid cooling system control method as described in any one of the above is implemented.
[0014] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the cold plate liquid cooling system control method described above is implemented.
[0015] The cold plate liquid cooling system control method, device, equipment, storage medium and program product provided by the present invention obtain multiple control schemes for the cold plate liquid cooling system to be controlled, input the historical operating status data of the control scheme, the cold plate liquid cooling system to be controlled and the corresponding cooling server into a trained operating status prediction model, obtain operating status prediction data reflecting the operating status of the cold plate liquid cooling system to be controlled and the cooling server at a future moment, optimize multiple control schemes based on the operating status prediction data corresponding to the control scheme, and obtain a target control scheme. In this process, the future operating status of the cold plate liquid cooling system to be controlled and the cooling server to be controlled is first predicted based on the historical operating status data of the cold plate liquid cooling system to be controlled and the cooling server, and optimize multiple control schemes based on the prediction results to determine the target control scheme for the cold plate liquid cooling system to be controlled, thereby optimizing the real-time control method during the operation of the cold plate liquid cooling system. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0017] Figure 1 It is a flow chart of a cold plate type liquid cooling system control method provided by the present invention.
[0018] Figure 2It is a structural schematic diagram of a cold plate type liquid cooling system to be controlled in a cold plate type liquid cooling system control method provided by the present invention.
[0019] Figure 3 It is a schematic diagram of an optimization process in a cold plate type liquid cooling system control method provided by the present invention.
[0020] Figure 4 It is a schematic diagram of the fourth group in a cold plate liquid cooling system control method provided by the present invention.
[0021] Figure 5 It is a structural schematic diagram of the cold plate type liquid cooling system control device provided by the present invention.
[0022] Figure 6 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION
[0023] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0024] Combine the following Figure 1-4 The cold plate liquid cooling system control method provided by the present invention is described as follows: Figure 1 As shown, the present invention provides a cold plate liquid cooling system control method, comprising the steps of: S110, obtaining multiple control schemes for the cold plate type liquid cooling system to be controlled, each control scheme including multiple control parameters of the cold plate type liquid cooling system to be controlled; S120, inputting input data into a trained operation status prediction model, obtaining operation status prediction data output by the operation status prediction model, wherein the input data includes a control scheme, a cold plate liquid cooling system to be controlled, and historical operation status data of a cooling server corresponding to the cold plate liquid cooling system to be controlled, and the operation status prediction data reflects the operation status of the cold plate liquid cooling system to be controlled and the cooling server at a future moment; S130, optimizing multiple control schemes based on the operation status prediction data corresponding to the control schemes, and obtaining a target control scheme for the cold plate liquid cooling system to be controlled.
[0025] The cooling server is the cooling target of the cold plate liquid cooling system to be controlled, that is, the cold plate liquid cooling system to be controlled is used to dissipate heat and cool the cooling server. Figure 2As shown, the cold plate liquid cooling system is divided into primary circulation equipment and secondary circulation equipment. The primary circulation equipment mainly includes cooling towers, plate heat exchangers, and chillers. The secondary circulation equipment mainly includes cooling capacity distribution units (CDU) and liquid cooling cabinets (cold plate type). In order to better utilize the resources of the liquid cooling system, a closed cooling tower is used to assist mechanical refrigeration devices. The cold source adopts a combined cooling method of chillers and cooling towers, and can be flexibly adjusted according to outdoor temperature conditions. It can be seen that there are many controllable parameters of the cold plate refrigeration system. In the prior art, there is no solution for real-time control of the controllable parameters of the cold plate refrigeration system based on the actual conditions of the cold plate refrigeration system and the cooling server.
[0026] The method provided by the present invention predicts the effect of the control scheme of the cold plate refrigeration system to be controlled in the future, optimizes based on the effect, and obtains a better target control scheme for the controllable parameters of the cold plate refrigeration system to be controlled, that is, the control scheme includes multiple values of the controllable parameters of the cold plate liquid cooling system to be controlled. There are many controllable parameters and state parameters of the cold plate refrigeration system, which are coupled with each other and coupled with the parameters of the cooling server, and there is a complex relationship. The method provided by the present invention adopts a neural network model to predict the operating status of the cold plate liquid cooling system to be controlled and the cooling server. In other words, the operating status prediction model is obtained based on training of multiple groups of training data, and each group of training data includes sample input data and operating status prediction data labels corresponding to the sample input data.
[0027] Specifically, the historical operating status data of the cold plate liquid cooling system to be controlled and the cooling server in the input data can be collected by sensors. The sensor is used to monitor the data center ambient temperature, coolant temperature, coolant flow rate, cooling water temperature, cooling water flow rate, fan speed, spray pump speed, circulating water pump flow rate, plate heat exchanger flow rate, cold plate inlet temperature, cold plate inlet flow, server shell temperature and server power consumption in real time, and transmit the monitored data to the device that executes the method provided by the present invention for analysis and decision-making. The device can be the controller of the cold plate liquid cooling system to be controlled. The controller is the core of the intelligent group control of the cold plate liquid cooling system, responsible for receiving data from the sensor, analyzing the data and making decisions, controlling the operating status of the liquid cooling equipment to realize real-time monitoring and control of the data center server, and making full use of the liquid cooling system to achieve energy-saving refrigeration, green and low-carbon. The controller can use an embedded system industrial computer.
[0028] The operating status prediction model adopted in the present invention is the Informer time series prediction model. The Informer time series prediction model is improved on the basis of the Transformer. By modifying the structure of the transformer, the speed of the transformer is improved. The Informer time series prediction model proposes a sparse self-attention mechanism. The traditional self-attention mechanism contains three important elements: Q, K, and V. The Informer constructs a sparse self-attention model by utilizing the relationship between input samples to screen out the most important Q and reduce the computational complexity. The Informer time series prediction model proposes a self-attention distillation mechanism to perform downsampling operations to reduce the number of computational dimensions and network parameters. The Informer time series prediction model proposes a generative decoder. The decoder part consists of two identical multi-head attention layers, which can avoid the rapid decrease in speed during long-term reasoning and prediction.
[0029] Collect the time series data of parameters such as data center ambient temperature, coolant temperature, coolant flow rate, cooling water temperature, cooling water flow rate, cooling tower fan speed, cooling tower spray pump speed, circulating water pump flow rate, plate heat exchanger flow rate, cooling capacity distribution unit (CDU) temperature controller parameters, cooling capacity distribution unit (CDU) flow rate controller parameters, cold plate inlet temperature, cold plate inlet flow, server shell temperature and server power consumption. After data cleaning, missing values and outliers are processed to ensure the quality of time series data. Substitute into the Informer time series prediction model, use the cross-validation method to determine the number of layers of the encoder and decoder, the number of multi-head attention layers, time step and learning rate parameters. The first column of the data set is time. , the step length is 1 hour, and the rest are data center ambient temperatures , Coolant temperature , Coolant flow rate , Cooling water temperature , Cooling water flow rate , Cooling tower fan speed , Cooling tower spray pump speed , Circulating water pump flow rate , Plate heat exchanger flow rate , Cooling Distribution Unit (CDU) temperature controller parameters , Cooling Distribution Unit (CDU) flow rate controller parameters , Cold plate inlet temperature , Cold plate inlet flow rate , Server shell temperature and server power consumption The main parameters predicted are coolant temperature, coolant flow rate, cooling water temperature, cooling water flow rate, cold plate inlet temperature, cold plate inlet flow rate, server shell temperature and server power consumption.
[0030] The data is divided into training set, test set and validation set. The training set data is standardized and substituted into the time series prediction model to calculate the prediction results. Parameters are adjusted and improved in the validation set. After the model is determined, the test set data is used for error analysis. The accuracy of the model is measured by the root mean square error (RMSE) and mean absolute error (MAE). Finally, the predicted data of coolant temperature, coolant flow rate, cooling water temperature, cooling water flow rate, cold plate inlet temperature, cold plate inlet flow rate, server shell temperature and server power consumption are obtained, as well as the relationship between them and other parameters.
[0031] In order to adapt to the particularity of the cold plate liquid cooling system, the real number coding method is adopted, mainly considering the cascade control mode of the cold plate liquid cooling, using cooling towers, circulating water pumps-plate heat exchangers, and cold distribution units (CDU) for flexible cooling. For the optimization operator operation, only the first seven bits are valid, representing the time step. , cooling tower fan speed , Cooling tower spray pump speed , Circulating water pump flow rate , Plate heat exchanger flow rate , Cooling Distribution Unit (CDU) temperature controller parameters , Cooling Distribution Unit (CDU) flow rate controller parameters , the rest are real-time parameters. According to the actual situation, the value range of all optimization operators is defined. That is to say, the parameters in the control scheme include the time step , cooling tower fan speed , Cooling tower spray pump speed , Circulating water pump flow rate , Plate heat exchanger flow rate , Cooling Distribution Unit (CDU) temperature controller parameters , Cooling Distribution Unit (CDU) flow rate controller parameters The prediction process of the operating status prediction model can be expressed as: ; .
[0032] The subscript P represents the value in the running status prediction data, that is, They respectively represent the predicted value of coolant temperature, the predicted value of coolant flow rate, the predicted value of cooling water temperature, the predicted value of cooling water flow rate, the predicted value of cold plate inlet temperature, the predicted value of cold plate inlet flow rate, the predicted value of cooling server shell temperature and the predicted value of cooling server power consumption.
[0033] Based on the operation status prediction data corresponding to each control scheme, multiple control schemes are optimized to obtain the target control scheme of the cold plate liquid cooling system to be controlled, including: Based on the running state prediction data corresponding to the control scheme, the objective function value corresponding to the control scheme is obtained, and the objective function value reflects the temperature control effect, fault information and energy consumption state of the cold plate liquid cooling system to be controlled corresponding to the running state prediction data; Based on the objective function values corresponding to the control schemes, multiple control schemes are optimized to obtain a target control scheme.
[0034] The method provided by the present invention, according to the refrigeration characteristics of the cold plate liquid cooling system, uses the relationship between various indicators to construct a multi-objective optimization control model of temperature control, fault diagnosis and energy consumption optimization. Since the research object has a strong coupling, the indicators can appear repeatedly in the objective function. The control scheme is multi-objective optimized for the three objectives of temperature control effect, fault diagnosis and energy consumption state, fully considering the influencing factors between the objectives, overcoming the limitations and instability of single-objective solution, and making the optimization solution more authentic and advanced. The cold plate liquid cooling modular products currently available on the market can meet the refrigeration needs, but there is no limit on the power consumption, and cannot comply with the development requirements of green, low-carbon, energy conservation and emission reduction. Through the method provided by the present invention, the heat dissipation capacity of the cold plate liquid cooling system can be controlled separately according to the power consumption and heat dissipation requirements of the data center server, so as to achieve precise heat dissipation, reduce the loss of heat dissipation capacity, improve the overall energy efficiency level of the data center, and make the entire liquid cooling system have a great improvement in temperature control, fault diagnosis and energy consumption optimization. It has a very high guiding significance, fills the gap in the prior art, and opens up new ideas for the development of cold plate liquid cooling group control systems.
[0035] Specifically, based on the running state prediction data corresponding to the control scheme, the objective function value corresponding to the control scheme is obtained, including: Determine a first function value based on the server shell temperature and the cold plate inlet temperature in the operation state prediction data, wherein the first function value reflects the temperature control effect of the cold plate liquid cooling system to be controlled corresponding to the operation state prediction data; Determine a second function value based on each operating parameter in the operating state prediction data and the actual statistical value of each operating parameter, the second function value reflects the fault information of the cold plate type liquid cooling system to be controlled corresponding to the operating state prediction data; A third function value is determined based on the difference between the server shell temperature and the ambient temperature in the operating status prediction data and the server power consumption in the operating status prediction data. The third function value reflects the energy consumption status of the cold plate liquid cooling system to be controlled and the cooling server corresponding to the operating status prediction data.
[0036] The first function value is the temperature control objective function, the second function value is the fault diagnosis objective function, and the third function value is the energy consumption optimization objective function. The specific calculation methods of these three objective functions are described below.
[0037] Temperature control objective function It is mainly based on the changing trends of server shell temperature and ambient temperature to timely adjust the operating status of the liquid cooling system to keep the data center equipment running within the most appropriate temperature range. The server shell temperature can also be flexibly adjusted according to actual conditions.
[0038] ; In the formula, , is the temperature control weight coefficient; , They are the cold plate inlet temperature and server shell temperature prediction values, , They are the optimal values of cold plate inlet temperature and server shell temperature respectively.
[0039] Fault diagnosis objective function The diagnosis and prediction of potential faults are mainly based on the discrete degree of operating data such as coolant temperature, coolant flow rate, cooling water temperature, cooling water flow rate, cold plate inlet temperature, cold plate inlet flow rate, server shell temperature, etc. By monitoring abnormal data patterns, equipment failures can be discovered in a timely manner and corresponding maintenance measures can be taken to improve the reliability and stability of the system.
[0040] ; In the formula, , is the fault diagnosis weight coefficient; the subscript P is the corresponding index prediction value, the subscript The standard deviation of the actual data of the corresponding indicator can reflect the degree of dispersion between the predicted indicator and the actual indicator, and then reflect whether there is an abnormal state in the data, and remove the outliers during the optimization process.
[0041] Energy consumption optimization objective function Mainly based on the server shell temperature, ambient temperature change trend and server power consumption data, the cold plate liquid cooling system can reduce the energy consumption of the liquid cooling system and server as much as possible while ensuring the stable operation of the data center equipment. By continuously optimizing the control strategy, the energy efficiency of the cold plate liquid cooling system in the data center can be optimized.
[0042] ; In the formula, , Optimize weight coefficients for energy consumption; , are the actual maximum and minimum values of the ambient temperature, They are the maximum and minimum actual power consumption of the server respectively.
[0043] The first function value, the second function value and the third function value can be fused to obtain the objective function value corresponding to the control scheme, and multiple control schemes can be optimized based on the objective function value.
[0044] Optimizing multiple control schemes may be achieved by using a genetic algorithm. Specifically, optimizing multiple control schemes based on the objective function values corresponding to the control schemes to obtain a target control scheme for the cold plate liquid cooling system to be controlled includes: Based on the dung beetle algorithm, each individual in the first population is updated to obtain the second population, and the corresponding control scheme of the individuals in the first population; The individuals in the second population are non-dominatedly sorted based on the objective function values corresponding to the individuals in the second population, and the individuals in the second population are selected based on the sorting results to obtain a third population; The fourth population is obtained by selecting individuals from the third population based on the reference point; Clear the first population and add individuals from the fourth population to the first population; Repeat the step of updating each individual in the first population based on the dung beetle algorithm to obtain the second population until a preset end condition is met, where the total number of individuals in all fourth populations reaches a preset number or the number of times the fourth population is generated reaches a preset number; A targeted control scheme is determined based on the individuals in each fourth population.
[0045] In the method provided by the present invention, multiple iterations of optimization are performed, each iteration of optimization generates a fourth population, and finally all individuals in the fourth population are combined into a set, which is an optimal solution set, from which the target control scheme is selected. In the first round of iterations, the first population can be composed of a plurality of randomly generated individuals, and in subsequent iterations, the first population is the fourth population generated in the previous iteration.
[0046] In each iteration, the dung beetle algorithm (DBO) is first used to update the individuals. Four division methods, namely, rolling dung beetles, breeding dung beetles, foraging dung beetles, and stealing dung beetles, are used for updating. Several candidate solutions in the search space are randomly generated as individuals. The proportions of the four types of dung beetles in the population are 20%, 20%, 25%, and 35%, respectively.
[0047] The formula for updating the position of the rolling dung beetle is as follows: ; ; .
[0048] In the formula, is a random number between (0,1), is the probability of a dung beetle encountering an obstacle, ranging from (0,1); is the deviation coefficient between (0,0.02], is a constant between (0,1), is the light intensity, The worst position in the population; is the deviation coefficient between [0,π], when =0, the position is not updated.
[0049] The formula for updating the position of breeding dung beetles is as follows: Define the security zone: ; ; In the formula, is the local optimal position of the iterative group, Lb and Ub represent the lower and upper bounds of the feasible region respectively. are the lower and upper bounds of the safe area.
[0050] After the dung beetle finds a safe area, it breeds chicks. The position update formula is as follows: ; ; In the formula, , is 1 D is an independent random vector of D, where D is the optimization dimension.
[0051] The formula for updating the position of a foraging dung beetle is as follows: ; ; ; In the formula, is the global optimal position, , are the lower and upper bounds of the optimal foraging area, are the lower and upper bounds of the safe area, respectively. is 1 D, is a random number vector of normal distribution, is a random vector of (0,1).
[0052] The formula for updating the position of the stealing dung beetle is as follows: ; In the formula, The size is 1 D is a random vector that follows a normal distribution, and S is a constant.
[0053] In order to increase the optimization efficiency, the method provided by the present invention adds mutation probability to the dung beetle algorithm to increase the optimization efficiency, and adds non-dominated sorting so that the algorithm can optimize and solve multiple objectives. Furthermore, in order to increase the degree of population convergence and update the population in a targeted manner, a penalty function is constructed based on the thermal performance and flow resistance performance of the cold plate in the cold plate liquid cooling system, which can dynamically adjust the individual position and accelerate the convergence efficiency of the algorithm. Specifically, based on the dung beetle algorithm, each individual in the first population is updated to obtain a second population, including: Based on the dung beetle algorithm, each individual in the first population is updated to obtain the intermediate population; Based on the operating state prediction data corresponding to the individuals in the intermediate population, the penalty function value corresponding to each individual in the intermediate population is determined, and the individuals in the intermediate population are updated based on the penalty function value to obtain a second population. The penalty function value reflects the thermal performance and flow resistance performance of the cold plate in the cold plate liquid cooling system to be controlled corresponding to the state prediction data.
[0054] The individuals in the intermediate population are updated based on the penalty function value. The penalty function value may be added to the objective function value corresponding to the individuals in the intermediate population, and the objective function value of the individuals in the intermediate population is updated, so that a better individual can be approached more quickly.
[0055] Determining the penalty function value corresponding to each individual in the intermediate population based on the running state prediction data corresponding to the individuals in the intermediate population includes: When the cold plate inlet flow rate in the running state prediction data corresponding to the individual in the intermediate population is not within the first preset range, determining the penalty function value based on the actual value of the cold plate outlet flow rate and the cold plate inlet flow rate in the running state prediction data; When the server power consumption in the operating status prediction data corresponding to the individual in the intermediate population is not within the second preset range, the penalty function value is determined based on the cold plate thermal resistance of the cold plate liquid cooling system to be controlled, the service shell temperature in the operating status prediction data, the cold plate inlet temperature, and the server power consumption.
[0056] The penalty function value can be expressed as: ; in, is the cold plate thermal resistance, is the actual value of the cold plate outlet velocity, is the second preset range, It is the first preset range.
[0057] After the second population is generated, the individuals in the second population are non-dominatedly sorted based on the objective function values corresponding to the individuals in the second population. Figure 3 As shown, a reference point-based approach is used to maintain the diversity of the population when selecting the critical layer Pareto solution set.
[0058] Setting of reference points: In the process of environmental selection, the mapping relationship between reference points and individuals in the population is determined by setting reference points to ensure the diversity of the population. The method of constructing weights by boundary crossing is usually used to make the reference points distributed on a (M-1)-dimensional hyperplane. The expression of the number of reference points P generated is: ; Wherein, M is the dimension of the target space, that is, the number of optimization targets, H is the number of evenly divided parts of each target, H is determined according to the actual object and selected according to the degree of detail of the division. The larger H is, the finer the division is and the more reference points there are. In the present invention, the optimization is performed for three-dimensional targets, H is set to 4, and there are 15 reference points in total.
[0059] Standardization of population: Selecting population The minimum value of each target range of individuals in is taken as the ideal point of the current population ,In order to simplify the algorithm, after translation operation , the origin can be set to an ideal point.
[0060] The intercept is usually determined by the extreme point. In multi-objective problems, the extreme point is a point with a large value in only one objective function direction and a small value in the other directions. The calculation formula is: ; Where: is the unit direction vector of the coordinate axis.
[0061] According to the relationship between the hyperplane and the intercept, the relationship between the extreme point and the intercept can be calculated, so the standardization of the population can be expressed as: ; Where: , is the intercept of each dimension, is the objective function value represented by the individual in the population.
[0062] Association operation: After the reference point and ideal point are constructed, the association operation between the population individuals and the reference point is performed. Figure 1 As shown, the line connecting the ideal point and the reference point is defined as the reference line, the distance from the individual to the reference line is calculated, the individual is connected to the reference line with the nearest distance, and is correlated with the corresponding reference point.
[0063] Individual retention operation: After individuals are associated with reference points, individuals corresponding to those reference points with fewer connections should be retained to maintain diversity. According to this principle, individual retention operations are performed until the population size reaches the requirement.
[0064] In all retained individuals can be Figure 4 As shown, the target control scheme obtained by selecting therein can adjust the cooling tower fan speed, cooling tower spray pump speed, circulating water pump flow rate, plate heat exchanger flow rate, cooling capacity distribution unit (CDU) temperature controller parameters, and cooling capacity distribution unit (CDU) flow rate controller parameters at different time steps to adjust the coolant temperature, coolant flow rate, cooling water temperature, cooling water flow rate, cold plate inlet temperature, cold plate inlet flow, server shell temperature, and server power consumption, etc., to achieve a flexible, reliable and better cold plate liquid cooling system intelligent group control strategy.
[0065] The cold plate type liquid cooling system control device provided by the present invention is described below. The cold plate type liquid cooling system control device described below and the cold plate type liquid cooling system control method described above can be referred to each other. Figure 5 As shown, the cold plate type liquid cooling system control device provided by the present invention comprises: A scheme generating module 510 is used to obtain a plurality of control schemes for the cold plate type liquid cooling system to be controlled, each control scheme including values of a plurality of controllable parameters of the cold plate type liquid cooling system to be controlled; The state prediction module 520 is used to input input data into the trained operation state prediction model to obtain operation state prediction data output by the operation state prediction model, wherein the input data includes the control scheme, the cold plate liquid cooling system to be controlled, and the historical operation state data of the cooling server corresponding to the cold plate liquid cooling system to be controlled, and the operation state prediction data reflects the operation state of the cold plate liquid cooling system to be controlled and the cooling server at a future time; An optimization module 530 is used to optimize multiple control schemes based on the operation state prediction data corresponding to the control schemes, and obtain a target control scheme for the cold plate liquid cooling system to be controlled; The running status prediction model is trained based on multiple sets of training data, and each set of training data includes sample input data and running status prediction data labels corresponding to the sample input data.
[0066] Figure 6 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 6As shown, the electronic device may include: a processor (processor) 610, a communication interface (Communications Interface) 620, a memory (memory) 630 and a communication bus 640, wherein the processor 610, the communication interface 620, and the memory 630 communicate with each other through the communication bus 640. The processor 610 can call the logic instructions in the memory 630 to execute the cold plate liquid cooling system control method, which includes: obtaining multiple control schemes for the cold plate liquid cooling system to be controlled, each control scheme including the values of multiple controllable parameters of the cold plate liquid cooling system to be controlled; inputting input data into a trained operation status prediction model to obtain operation status prediction data output by the operation status prediction model, the input data including the control scheme, the cold plate liquid cooling system to be controlled, and the historical operation status data of the cooling server corresponding to the cold plate liquid cooling system to be controlled, and the operation status prediction data reflects the operation status of the cold plate liquid cooling system to be controlled and the cooling server at a future moment; based on the operation status prediction data corresponding to the control schemes, optimizing multiple control schemes to obtain a target control scheme for the cold plate liquid cooling system to be controlled; wherein the operation status prediction model is trained based on multiple groups of training data, and each group of training data includes sample input data and operation status prediction data labels corresponding to the sample input data.
[0067] In addition, the logic instructions in the above-mentioned memory 630 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.
[0068] On the other hand, the present invention also provides a computer program product, which includes a computer program, which can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the cold plate liquid cooling system control method provided above, and the cold plate liquid cooling system control method includes: obtaining multiple control schemes for the cold plate liquid cooling system to be controlled, each control scheme including values of multiple controllable parameters of the cold plate liquid cooling system to be controlled; inputting input data into a trained operation state prediction model to obtain operation state prediction data output by the operation state prediction model, the input data including the control scheme, the cold plate liquid cooling system to be controlled, and the historical operation state data of the cooling server corresponding to the cold plate liquid cooling system to be controlled, and the operation state prediction data reflects the operation state of the cold plate liquid cooling system to be controlled and the cooling server at a future moment; based on the operation state prediction data corresponding to the control schemes, multiple control schemes are optimized to obtain a target control scheme for the cold plate liquid cooling system to be controlled; wherein the operation state prediction model is obtained by training based on multiple groups of training data, and each group of training data includes sample input data and operation state prediction data labels corresponding to the sample input data.
[0069] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which is implemented when the processor executes the cold plate liquid cooling system control method provided above, the cold plate liquid cooling system control method comprising: obtaining multiple control schemes for the cold plate liquid cooling system to be controlled, each control scheme comprising values of multiple controllable parameters of the cold plate liquid cooling system to be controlled; inputting input data into a trained operation status prediction model to obtain operation status prediction data output by the operation status prediction model, the input data comprising the control scheme, the cold plate liquid cooling system to be controlled, and historical operation status data of a cooling server corresponding to the cold plate liquid cooling system to be controlled, the operation status prediction data reflecting the operation status of the cold plate liquid cooling system to be controlled and the cooling server at a future moment; based on the operation status prediction data corresponding to the control schemes respectively, optimizing multiple control schemes to obtain a target control scheme for the cold plate liquid cooling system to be controlled; wherein the operation status prediction model is obtained through training based on multiple groups of training data, each group of training data comprising sample input data and operation status prediction data labels corresponding to the sample input data.
[0070] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.
[0071] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.
[0072] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A cold plate liquid cooling system control method, It is characterized in that include: Acquire multiple control schemes for the cold plate type liquid cooling system to be controlled, each of the control schemes including values of multiple controllable parameters of the cold plate type liquid cooling system to be controlled; Inputting input data into a trained operation status prediction model to obtain operation status prediction data output by the operation status prediction model, wherein the input data includes the control scheme, the cold plate type liquid cooling system to be controlled, and historical operation status data of a cooling server corresponding to the cold plate type liquid cooling system to be controlled, and the operation status prediction data reflects the operation status of the cold plate type liquid cooling system to be controlled and the cooling server at a future moment; Based on the operation state prediction data respectively corresponding to the control schemes, multiple control schemes are optimized to obtain a target control scheme for the cold plate liquid cooling system to be controlled; The running status prediction model is obtained by training based on multiple sets of training data, and each set of training data includes sample input data and running status prediction data labels corresponding to the sample input data.
2. The cold plate liquid cooling system control method according to claim 1, It is characterized in that The optimizing of the plurality of control schemes based on the operation state prediction data respectively corresponding to the control schemes to obtain the target control scheme of the cold plate type liquid cooling system to be controlled includes: Based on the running state prediction data corresponding to the control scheme, an objective function value corresponding to the control scheme is obtained, wherein the objective function value reflects the temperature control effect, fault information and energy consumption state of the cold plate type liquid cooling system to be controlled corresponding to the running state prediction data; Based on the objective function values corresponding to the control schemes, multiple control schemes are optimized to obtain the target control scheme.
3. The cold plate type liquid cooling system control method according to claim 2, It is characterized in that The obtaining of the objective function value corresponding to the control scheme based on the running state prediction data corresponding to the control scheme includes: Determine a first function value based on the server shell temperature and the cold plate inlet temperature in the operation state prediction data, wherein the first function value reflects the temperature control effect of the cold plate type liquid cooling system to be controlled corresponding to the operation state prediction data; Determine a second function value based on each operating parameter in the operating state prediction data and actual statistical values of each operating parameter, wherein the second function value reflects fault information of the cold plate type liquid cooling system to be controlled corresponding to the operating state prediction data; A third function value is determined based on the difference between the server shell temperature and the ambient temperature in the operating status prediction data and the server power consumption in the operating status prediction data, and the third function value reflects the energy consumption status of the cold plate liquid cooling system to be controlled and the cooling server corresponding to the operating status prediction data.
4. The cold plate liquid cooling system control method according to claim 2, It is characterized in that The optimizing of the multiple control schemes based on the objective function value corresponding to the control scheme to obtain the target control scheme of the cold plate type liquid cooling system to be controlled includes: updating each individual in the first population based on the dung beetle algorithm to obtain a second population, wherein the individuals in the first population correspond to the control scheme; performing non-dominated sorting on the individuals in the second population based on the objective function value corresponding to each individual in the second population, and selecting the individuals in the second population based on the sorting result to obtain a third population; Selecting individuals from the third population based on the reference point to obtain a fourth population; Clearing the first population, and adding individuals in the fourth population to the first population; Repeat the step of updating each individual in the first population based on the dung beetle algorithm to obtain the second population until a preset end condition is met, wherein the total number of individuals in the fourth population reaches a preset number or the number of times the fourth population is generated reaches a preset number; The target control scheme is determined based on each individual in the fourth population.
5. The cold plate type liquid cooling system control method according to claim 4, It is characterized in that The method of updating each individual in the first population based on the dung beetle algorithm to obtain the second population includes: Based on the dung beetle algorithm, each individual in the first population is updated to obtain the intermediate population; Based on the operating status prediction data corresponding to the individuals in the intermediate population, the penalty function value corresponding to each individual in the intermediate population is determined, and the individuals in the intermediate population are updated based on the penalty function value to obtain the second population, wherein the penalty function value reflects the thermal performance and flow resistance performance of the cold plate in the cold plate liquid cooling system to be controlled corresponding to the state prediction data.
6. The cold plate type liquid cooling system control method according to claim 5, It is characterized in that The step of determining the penalty function value corresponding to each individual in the intermediate population based on the running state prediction data corresponding to the individuals in the intermediate population includes: When the cold plate inlet flow rate in the operating state prediction data corresponding to the individual in the intermediate population is not within a first preset range, determining the penalty function value based on the actual value of the cold plate outlet flow rate and the cold plate inlet flow rate in the operating state prediction data; When the server power consumption in the operating status prediction data corresponding to the individual in the intermediate population is not within the second preset range, the penalty function value is determined based on the cold plate thermal resistance of the cold plate liquid cooling system to be controlled, the service shell temperature in the operating status prediction data, the cold plate inlet temperature, and the server power consumption.
7. A cold plate liquid cooling system control device, It is characterized in that include: A scheme generating module, used for acquiring a plurality of control schemes for the cold plate type liquid cooling system to be controlled, each of the control schemes comprising values of a plurality of controllable parameters of the cold plate type liquid cooling system to be controlled; A state prediction module, used for inputting input data into a trained operation state prediction model, and obtaining operation state prediction data output by the operation state prediction model, wherein the input data includes the control scheme, the cold plate type liquid cooling system to be controlled, and historical operation state data of a cooling server corresponding to the cold plate type liquid cooling system to be controlled, and the operation state prediction data reflects the operation state of the cold plate type liquid cooling system to be controlled and the cooling server at a future moment; An optimization module, used for optimizing a plurality of control schemes based on the operation state prediction data respectively corresponding to the control schemes, to obtain a target control scheme for the cold plate type liquid cooling system to be controlled; The running status prediction model is obtained by training based on multiple sets of training data, and each set of training data includes sample input data and running status prediction data labels corresponding to the sample input data.
8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, It is characterized in that When the processor executes the computer program, the cold plate liquid cooling system control method according to any one of claims 1 to 6 is implemented.
9. A non-transitory computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the cold plate liquid cooling system control method according to any one of claims 1 to 6 is implemented.
10. A computer program product comprising a computer program, It is characterized in that When the computer program is executed by a processor, the cold plate liquid cooling system control method according to any one of claims 1 to 6 is implemented.
