Liquid cooling system optimization control method, device and equipment, liquid cooling system, storage medium and product
By using a neural network optimization control method, objective functions for temperature difference response and control response are constructed. The liquid cooling system is then optimized using an Elman neural network and a non-dominated sorting genetic algorithm, which solves the problem of inaccurate control parameter tuning in traditional liquid cooling systems and achieves efficient and stable control of the liquid cooling system.
Patent Information
- Application Number
- CN202411027369.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-30
- Publication Date
- 2024-11-15
Smart Images

Figure CN118963510A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of infrastructure technology, and in particular to a liquid cooling system optimization control method, device, equipment, liquid cooling system, storage medium and product. Background Art
[0002] With the development of new infrastructure such as cloud computing and big data, the requirements for data computing speed are getting higher and higher, and the computing speed and amount of processors are also increasing, resulting in a continuous surge in the power consumption of components in data centers. In order to effectively solve the problem of excessive temperature of various electronic components, immersion liquid cooling has emerged. More and more data centers have begun to adopt immersion liquid cooling technology to meet the growing computing and storage needs, and the cooling effect has been significantly improved. However, with the widespread application of immersion liquid cooling systems, the adjustment, monitoring and control of the entire system have become very complicated. The control parameters of traditional data center liquid cooling cabinets rely on empirical rules or manually set parameters, lacking automation and intelligence, which makes the parameter setting process inefficient and inaccurate. Summary of the invention
[0003] The purpose of the embodiments of the present invention is to provide a liquid cooling system optimization control method, device, equipment, liquid cooling system, storage medium and product, which can effectively improve the efficiency of the liquid cooling system, accurately adjust the control parameters of the liquid cooling system, and improve the stability and reliability of the liquid cooling system.
[0004] To achieve the above objectives, an embodiment of the present invention provides a liquid cooling system optimization control method, comprising:
[0005] Acquiring real-time control parameters of the liquid cooling system, and inputting the real-time control parameters into a neural network so that the neural network outputs predicted characteristic parameters of the liquid cooling system;
[0006] Using the predicted characteristic parameters to construct a temperature difference response objective function and a control response objective function;
[0007] Solving the temperature difference response objective function and the control response objective function to obtain corresponding target control parameters;
[0008] The liquid cooling system is controlled according to the target control parameter.
[0009] As an improvement of the above solution, before obtaining the real-time control parameters of the liquid cooling system, the method further includes:
[0010] The neural network is optimized using a moving density estimation strategy and a non-dominated sorting genetic algorithm.
[0011] As an improvement of the above solution, after the real-time control parameter is input into the neural network so that the neural network outputs the predicted characteristic parameter of the liquid cooling system, the method further includes:
[0012] Constructing a power consumption function of the electronic device using the predicted characteristic parameters;
[0013] The neural network is optimized according to the power consumption function.
[0014] As an improvement of the above-mentioned scheme, the liquid cooling system includes an outdoor heat dissipation device, a primary side device, a secondary side device, a cold energy distribution unit and a cabinet for storing electronic equipment; wherein the primary side device is respectively connected to the outdoor heat dissipation device and the cold energy distribution unit, and the secondary side device is respectively connected to the cold energy distribution unit and the cabinet.
[0015] As an improvement of the above solution, the real-time control parameter includes at least one of the following parameters:
[0016] The fan speed, outdoor water pump speed and sprinkler pump speed of the outdoor cooling equipment;
[0017] Speed of primary circulation water pump of primary side equipment;
[0018] Speed of secondary circulation water pump of secondary side equipment;
[0019] Valve opening of the cooling distribution unit.
[0020] As an improvement of the above solution, the prediction characteristic parameters include temperature prediction parameters, flow rate prediction parameters, temperature difference prediction parameters, height prediction parameters and power consumption prediction parameters.
[0021] As an improvement of the above solution, the temperature prediction parameters include coolant temperature, cooling water temperature and cabinet inlet temperature;
[0022] The flow rate prediction parameters include the coolant flow rate, the cooling water flow rate and the cabinet inlet flow rate;
[0023] The temperature difference prediction parameter is the coolant temperature difference in the cabinet;
[0024] The height prediction parameter is the coolant height of the cabinet;
[0025] The power consumption prediction parameter is the actual power consumption of the electronic device;
[0026] Wherein, the cooling liquid circulates between the cabinet and the cold distribution unit, and the cooling water circulates between the cold distribution unit and the outdoor heat dissipation device.
[0027] As an improvement of the above solution, the use of the predicted characteristic parameters to construct a temperature difference response objective function includes:
[0028] Obtaining a preset temperature difference response weight coefficient, and obtaining a reference temperature difference corresponding to the temperature difference prediction parameter;
[0029] A temperature difference response objective function is constructed according to the temperature difference prediction parameter, the reference temperature difference and the temperature difference response weight coefficient.
[0030] As an improvement of the above solution, the use of the predicted characteristic parameters to construct a control response objective function includes:
[0031] Obtaining a preset control response weight coefficient, and obtaining a reference temperature corresponding to the temperature prediction parameter, a reference flow velocity corresponding to the flow velocity prediction parameter, and a reference height corresponding to the height prediction parameter;
[0032] A control response objective function is constructed according to the control response weight coefficient, the temperature prediction parameter, the reference temperature, the flow rate prediction parameter, the reference flow rate, the height prediction parameter and the reference height.
[0033] To achieve the above objectives, an embodiment of the present invention further provides a liquid cooling system optimization control device, comprising:
[0034] A real-time control parameter acquisition module is used to obtain real-time control parameters of the liquid cooling system;
[0035] A predicted characteristic parameter acquisition module, used for inputting the real-time control parameter into a neural network so that the neural network outputs the predicted characteristic parameter of the liquid cooling system;
[0036] A function construction module, used to construct a temperature difference response objective function and a control response objective function using the predicted characteristic parameters;
[0037] A function solving module, used for solving the temperature difference response objective function and the control response objective function to obtain corresponding target control parameters;
[0038] A control module is used to control the liquid cooling system according to the target control parameter.
[0039] To achieve the above-mentioned objectives, an embodiment of the present invention further provides a liquid cooling system, comprising an outdoor heat dissipation device, a primary side device, a secondary side device, a cold energy distribution unit, a cabinet for storing electronic equipment, and the liquid cooling system optimization control device described in the above-mentioned embodiment; wherein the primary side device is respectively connected to the outdoor heat dissipation device and the cold energy distribution unit, and the secondary side device is respectively connected to the cold energy distribution unit and the cabinet.
[0040] To achieve the above-mentioned objectives, an embodiment of the present invention also provides a liquid cooling system optimization control device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and when the processor executes the computer program, it implements the liquid cooling system optimization control method as described in any of the above embodiments.
[0041] To achieve the above objectives, an embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the liquid cooling system optimization control method as described in any of the above embodiments.
[0042] To achieve the above objectives, an embodiment of the present invention further provides a computer program product, including a computer program, which, when executed by a processor, implements the liquid cooling system optimization control method as described in any of the above embodiments.
[0043] Compared with the prior art, the liquid cooling system optimization control method, device, equipment, liquid cooling system, storage medium and product disclosed in the present invention establish a time series prediction model based on the liquid cooling system, determine the relationship between each control parameter of the liquid cooling system and the prediction parameter, and then establish and solve the temperature difference response objective function, control response objective function and power consumption function based on the prediction parameters output by the neural network. After solving, the target control parameters of the liquid cooling system are obtained. The target control parameters are used to control the operation of the liquid cooling system, which can effectively improve the efficiency of the liquid cooling system. In addition, since the objective function and power consumption function are constructed, the state of the liquid cooling system can be estimated more accurately, thereby more accurately adjusting the control parameters of the liquid cooling system and improving the stability and reliability of the liquid cooling system. BRIEF DESCRIPTION OF THE DRAWINGS
[0044] Figure 1 is a structural schematic diagram of a liquid cooling system provided by an embodiment of the present invention;
[0045] Figure 2 is a flow chart of a liquid cooling system optimization control method provided by an embodiment of the present invention;
[0046] Figure 3 is another flow chart of a liquid cooling system optimization control method provided by an embodiment of the present invention;
[0047] Figure 4 is a trend diagram between a control response objective function value and a temperature difference response objective function value provided by an embodiment of the present invention;
[0048] Figure 5 It is a structural block diagram of a liquid cooling system optimization control device provided by an embodiment of the present invention;
[0049] Figure 6 It is a structural block diagram of a liquid cooling system optimization control device provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0050] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only 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.
[0051] See also Figure 1 , Figure 1 It is a structural schematic diagram of an immersion liquid cooling system provided by an embodiment of the present invention, wherein the liquid cooling system comprises an outdoor heat dissipation device 10, a primary side device 20, a cold energy distribution unit 30, a secondary side device 40 and an immersion liquid cooling cabinet 50 for storing electronic equipment 60; wherein the primary side device is respectively connected to the outdoor heat dissipation device and the cold energy distribution unit, and the secondary side device is respectively connected to the cold energy distribution unit and the cabinet.
[0052] Exemplarily, the outdoor heat dissipation device 10 adopts a closed cooling tower, and the heat is dissipated by the unit of circulating water in the outdoor air cooling pipe. The specific accessories mainly include variable frequency fans, outdoor circulating water pumps, spray systems and valves. The primary side device 20 connects the outdoor heat dissipation device 10 to the cooling distribution unit (Coolant Distribution Unit, CDU) 30, and the primary side device 20 includes a primary pipeline and a primary circulating water pump (cooling water circulating water pump). The cooling distribution unit 30 can ensure the reasonable distribution, filtration and flow of the cooling liquid. The specific accessories of the cooling distribution unit 30 mainly include a plate heat exchanger, a filtration system and a plate electric valve. The secondary side device 40 connects the cooling distribution unit 30 to the full liquid cooling cabinet 50, and the secondary side device 40 includes a secondary pipeline and a secondary circulating water pump (cooling liquid circulating water pump). The cabinet 50 can provide the entry and exit of cooling liquid to cool the electronic equipment 60 stored therein. The cabinet is a fully enclosed structure, and all the heat is taken out of the cabinet 50 through the circulation of the cooling liquid.
[0053] It is worth noting that various sensors are deployed in the liquid cooling system to monitor the parameters such as coolant temperature, coolant flow rate, cooling water temperature, cooling water flow rate, cabinet inlet temperature, cabinet inlet flow rate, coolant temperature difference between the top and bottom of the cabinet, coolant height in the cabinet and actual power consumption of electronic equipment in real time, so as to provide data support for control decision. The control mechanism is responsible for receiving sensor data and controlling the operation status of the liquid cooling system, fully connecting with the actual situation, and realizing efficient, green and low-carbon refrigeration. The control mechanism mainly includes the variable frequency fan, spray pump, outdoor water pump of the outdoor heat dissipation equipment (cooling tower), the primary circulation water pump (cooling water circulation water pump) of the primary side equipment, the plate exchange electric valve opening of the cooling capacity distribution unit (plate heat exchanger) and the secondary circulation water pump (cooling liquid circulation water pump) of the secondary side equipment. Furthermore, intelligent algorithms are the core of intelligent optimization control of the liquid cooling system. According to the real-time collected data and the system operation status, the control strategy is optimized to realize the intelligent optimization control of the immersion liquid cooling system. The intelligent algorithms include Elman neural network algorithm, non-dominated sorting genetic algorithm (such as the second-generation non-dominated sorting genetic algorithm) and mobile density estimation strategy.
[0054] Exemplarily, in the embodiment of the present invention, the Elman neural network algorithm is used to find the relationship between the control parameters and the predicted characteristic parameters of the liquid cooling system, so as to achieve the purpose of changing the predicted characteristic parameters by adjusting the control parameters. Figure 3 , Figure 3 It is another flow chart of a liquid cooling system optimization control method provided by an embodiment of the present invention. In the embodiment of the present invention, a liquid cooling system rolling time series prediction model based on the Elman neural network is first constructed, and the historical liquid cooling system data is collected and processed to form training and test sets and the data is normalized (the common data normalization processing method in the prior art can be used), and then the number of neurons and the weight matrix in the Elman neural network are determined. Before this process, a moving density estimation strategy and a non-dominated sorting genetic algorithm can be introduced in advance, and the moving density estimation strategy and the non-dominated sorting genetic algorithm can be used to optimize the neural network, and then the prediction accuracy of the Elman neural network is evaluated, and finally the relationship between the predicted characteristic parameters and the control parameters of the liquid cooling system is output.
[0055] The calculation formula of Elman neural network is as follows:
[0056] x(k)=f{w 1 x c (k)+w 2 [u(k-1)]} (1);
[0058] x c (k) = αx c (k-1)+x(k-1) (2);
[0060] y(k)=g[w 3 x(k)] (3);
[0062] Among them, u, x, x c and y represent the input layer, hidden layer, receiving layer and output layer vectors in the Elman neural network respectively; w 1 、w 2 and w 3 They represent the weight matrices from the receiving layer to the hidden layer, from the input layer to the hidden layer, and from the hidden layer to the output layer respectively; f(x) and g(x) are the transfer functions of the hidden layer and the output layer, using the tansig function and the purelin function respectively; α represents the self-feedback factor, α∈[0,1].
[0063] The error index function E(k) of the Elman neural network is:
[0064]
[0065] Among them, y(k) and They represent the kth actual and predicted output values respectively, k=1, 2...n. After repeated iterations until the error accuracy meets the standard, the model is established.
[0066] The root mean square error (RMSE) is used to evaluate the prediction performance, and the calculation formula is:
[0067]
[0068] Among them, d pi is the predicted value; d ri is the actual value.
[0069] In order to increase the prediction accuracy of Elman neural network, the second generation non-dominated genetic algorithm is used to optimize the number of neurons and weight matrix in ELman neural network, and the neuron objective function and weight matrix objective function are established to perform multi-objective optimization solution.
[0070] Establish the neuron objective function:
[0071]
[0072] Among them, a, b, c, and d represent the number of neurons in the input layer, hidden layer, receiving layer, and output layer, respectively.
[0073] Establish the weight matrix objective function:
[0074]
[0075] Among them, w 1 ,w 2 ,w 3 They represent the weight matrices from the receiving layer to the hidden layer, from the input layer to the hidden layer, and from the hidden layer to the output layer. The non-dominated sorting genetic algorithm (NSGA-II) is used to solve the multi-objective optimization of the two objective functions.
[0076] Furthermore, in order to make the multi-objective optimization algorithm converge more easily and speed up the optimization, the moving density estimation strategy (SDE) is introduced. SDE contains both the distribution information and convergence information of individuals. In a population, when the density of individual p is estimated, the convergence of other individuals and p in each dimension of the target is compared to determine whether other individuals should be moved to the position of p in this dimension. If the performance of one or some individuals is better than that of individual p in this dimension of the target, they are moved to the position of individual p in this dimension of the target; otherwise, the positions of other individuals remain unchanged. For the minimization optimization problem, the density D'(p, P) of individual p in the population can be expressed as:
[0077] D'(p,P)=D(dist(p,q′ 1 ),dist(p,q′ 2 ),…,dist(p,q′ N-1 )) (10);
[0078] Among them, N represents the size of P, dist(p,q i ′)) represents individual p and individual q i ′. And q i ′ is the individual q i The individual after moving (p, q i ∈P and q i ≠p), which is defined as follows:
[0079]
[0080] Among them, p j ,q i(j) and q′ i(j) Represents p and q respectively i and q′ i The target value in the jth dimension, m is the target dimension.
[0081] Exemplarily, the SDE strategy can accelerate the convergence speed and distribution of the non-dominated sorting algorithm. Combined with the construction process of the above formulas (1) to (11), the specific steps of the NSGA-II+SDE algorithm include: 1) encoding, parameter setting and population initialization; 2) introducing the mobile density estimation strategy to accelerate the individual distribution diversity and convergence speed; 3) performing non-dominated sorting and crowding calculation; 4) performing the selection operation of the dominant individual; 5) performing crossover and mutation operations; 6) performing the elite strategy; 7) judging the evolutionary generation. After steps 1) to 7), the neural network optimization is completed.
[0082] See also Figure 2 , Figure 2 : is a flow chart of a liquid cooling system optimization control method provided by an embodiment of the present invention, the liquid cooling system optimization control method comprising:
[0083] S1. Acquire real-time control parameters of a liquid cooling system, and input the real-time control parameters into a neural network, so that the neural network outputs predicted characteristic parameters of the liquid cooling system;
[0084] S2, constructing a temperature difference response objective function and a control response objective function using the predicted characteristic parameters;
[0085] S3, solving the temperature difference response objective function and the control response objective function to obtain corresponding target control parameters;
[0086] S4. Control the liquid cooling system according to the target control parameter.
[0087] Specifically, in step S1, the real-time control parameter includes at least one of the following parameters: the fan speed V of the outdoor heat dissipation device; f , Outdoor water pump speed V s and spray pump speed V p ; Speed V of primary circulating water pump of primary equipment r ; Speed V of secondary circulation water pump of secondary side equipment k ; Valve opening degree K of the cooling distribution unit e .
[0088] Specifically, the prediction characteristic parameters include temperature prediction parameters, flow rate prediction parameters, temperature difference prediction parameters, height prediction parameters and power consumption prediction parameters based on time series. The temperature prediction parameters include the coolant temperature T A , Cooling water temperature T B and the cabinet inlet temperature T c ; The flow rate prediction parameters include the coolant flow rate V A , Cooling water flow rate V B and the cabinet inlet flow rate V C; The temperature difference prediction parameter is the coolant temperature difference T between the top and bottom of the cabinet D ; The height prediction parameter is the coolant height H of the cabinet L The power consumption prediction parameter is the actual power consumption W of the electronic device. R ; wherein the cooling liquid circulates between the cabinet and the cold distribution unit, and the cooling water circulates between the cold distribution unit and the outdoor heat dissipation device.
[0089] In the embodiment of the present invention, the Elman neural network algorithm is used to A , T B , T C 、V A 、V B 、V C , T D , H L and W R To make a prediction, the input data is V f 、V s 、V p 、V r 、V k and K e These control parameters, find the relationship between the control parameters and the prediction characteristic parameters, satisfy the following formulas (12) to (13), and achieve the purpose of changing the prediction characteristic parameters by adjusting the control parameters.
[0090] y(T AP ,V AP ,T BP ,V BP ,T CP ,V CP ,T DP ,H LP )=f(V f ,V p ,V s ,V r ,V k ,K e ) (12);
[0091] y(W RP )=f(T s ,T DP ,H LP ) (13);
[0092] Among them, T s is the time step, and the parameters with the subscript "P" are predicted values, such as T AP is the coolant temperature T A The predicted value of y(T AP ,V AP ,TBP ,V BP ,T CP ,V CP ,T DP ,H LP ) and y(W RP ) represents the output of the neural network, f(V f ,V p ,V s ,V r ,V k ,K e ) and f(T s ,T DP ,H LP ) represents the input of the neural network.
[0093] Specifically, in step S2, in order to adapt to the particularity of the immersion liquid cooling system, the present invention adopts a real number coding method, mainly considering the cascade control mode of immersion liquid cooling, using a cooling tower, a cooling water circulation pump, a plate heat exchanger, and a coolant circulation pump. For the optimization operator operation, only the first seven bits are valid, representing the time step T, respectively. s , Fan speed V of outdoor cooling equipment f , Spray pump speed V p , Outdoor water pump speed V s , primary circulation water pump speed V r , Secondary circulation water pump speed V k , Valve opening degree K of cooling distribution unit e , and the rest of the parameters are actual values of the immersion liquid cooling system. According to the refrigeration characteristics of the immersion liquid cooling system, the multi-objective optimization model of the two objective functions of temperature difference response and control response is constructed by using the relationship between various indicators.
[0094] Specifically, the use of the predicted characteristic parameters to construct a temperature difference response objective function includes:
[0095] S211, obtaining a preset temperature difference response weight coefficient, and obtaining a reference temperature difference corresponding to the temperature difference prediction parameter;
[0096] S212: constructing a temperature difference response objective function according to the temperature difference prediction parameter, the reference temperature difference and the temperature difference response weight coefficient.
[0097] For example, the temperature difference response objective function F T It is mainly based on the temperature difference trend of the coolant at the top and bottom of the full liquid cooling cabinet to timely feedback the operating status parameters of the immersion liquid cooling system to ensure that the equipment runs at the optimal temperature in the coolant. The temperature difference can also be flexibly adjusted according to the actual operating conditions to control the response objective function F TThe goal is to minimize the difference between the predicted value and the optimal value of the coolant temperature difference. Temperature difference response objective function F T Satisfies the following formula:
[0098]
[0099] in, is the temperature difference response weight coefficient; T DP , T DG They are respectively the predicted value and the optimal value of the coolant temperature difference between the top and the bottom of the cabinet. The optimal value here refers to the theoretical optimal value of the coolant temperature difference (which can be queried), which is generally determined according to the coolant characteristics and belongs to the characteristic parameters.
[0100] Specifically, in step S2, constructing a control response objective function using the predicted characteristic parameters includes:
[0101] S221, obtaining a preset control response weight coefficient, and obtaining a reference temperature corresponding to the temperature prediction parameter, a reference flow velocity corresponding to the flow velocity prediction parameter, and a reference height corresponding to the height prediction parameter;
[0102] S221. Construct a control response objective function according to the control response weight coefficient, the temperature prediction parameter, the reference temperature, the flow rate prediction parameter, the reference flow rate, the height prediction parameter and the reference height.
[0103] For example, the control response objective function F C Mainly based on the coolant temperature T A , Coolant flow rate V A , Cooling water temperature T B , Cooling water flow rate V B , Cabinet liquid inlet temperature T C , liquid flow rate V C and coolant height H L The immersion cooling system is controlled by the deviation of the operating data. The deviation is corrected to provide feedback, and the fan speed V f , Outdoor water pump speed V s , Spray pump speed V p , primary circulation water pump speed V r , Secondary circulation water pump speed V k , Valve opening degree K of cooling distribution unit e To control and improve the reliability and stability of the system. Control response objective function F C The control response objective function F is to minimize the difference between the predicted value and the historical average value of each predicted characteristic parameter. C Satisfies the following formula:
[0104]
[0105] Among them, β, γ and δ are control response weight coefficients respectively, among which β corresponds to temperature prediction parameter, γ corresponds to velocity prediction parameter, and δ corresponds to height prediction parameter; the subscript P is the prediction value of the corresponding indicator, and the subscript μ is the reference parameter corresponding to the corresponding indicator. These reference parameters are historical average values, such as T Aμ is the coolant temperature T A The corresponding reference parameter is the historical average value. The control response objective function F C It can reflect the degree of deviation between the predicted indicators and the actual indicators, and then determine whether adjustments need to be made through control mechanisms.
[0106] Further, see Figure 4 , Figure 4 It is a trend diagram between the control response objective function value and the temperature difference response objective function value provided by an embodiment of the present invention. The horizontal axis is the control response objective function value, and the vertical axis is the temperature difference response objective function value. The values between the two show a downward trend because there are coupled and opposing parameters between the two objectives. The larger one objective function is, the smaller the other objective function will be, so the overall trend is downward. Therefore, it can be shown that the two objectives are strongly correlated. Using these two objective functions, a flexible, reliable and optimal intelligent group control strategy for the immersion liquid cooling system can be implemented.
[0107] Specifically, after executing step S1, the method further includes:
[0108] S11, constructing a power consumption function of the electronic device using the predicted characteristic parameters;
[0109] S12. Optimizing the neural network according to the power consumption function.
[0110] Exemplarily, the temperature difference prediction parameter, the height prediction parameter and the preset time step are used as inputs, and the power consumption prediction parameter is used as output to construct a power consumption function. The mobile density estimation strategy (SDE) is introduced to optimize the second generation non-dominated sorting genetic algorithm, and the mobile density estimation strategy is improved according to the characteristics of the immersion liquid cooling system, and the power consumption function of the electronic device is introduced, and the power consumption function satisfies the following formula:
[0111] y(W RP )=f(T s ,T DP ,H LP ) (16);
[0112] definition:
[0113]
[0114] Among them, combined with the above formula (11), w RD w RP The value after the shift, that is, the same as W RP The objective function value of a certain dimension remains consistent, and then comparison and screening are performed. According to the objective function value, two objective functions are set in the present invention, so the dimension of j is 2. Each time the SDE strategy is used to compare and mark only one non-dominated individual, the multi-objective optimization algorithm is cyclical, and the SDE strategy is used again in each cycle.
[0115] Specifically, in step S3, the temperature difference response objective function and the control response objective function are solved to obtain corresponding target control parameters.
[0116] Exemplarily, the temperature difference response objective function and the control response objective function achieve their set goals by continuously iterating and predicting characteristic parameters, and then reversely obtain the corresponding control parameters as target control parameters, which are the optimal solution.
[0117] Furthermore, if Figure 3 As shown, after the neural network output predicts the relationship between the characteristic parameters and the control parameters, the initial parameters are integrated to obtain the above formulas (12) and (13), and then the objective functions of the temperature difference response and the control response are established, namely the above formulas (14) and (15), and the actual power consumption parameters of the electronic equipment (such as the server) are introduced to establish the power consumption function, namely the above formula (16), and the mobile density estimation strategy (SDE) is introduced to optimize the second generation non-dominated sorting genetic algorithm, and return to the model prediction of the neural network, and non-dominated sorting is performed on formula (17), and the optimal individual is selected according to the sorting registration and congestion, and finally the target control parameters corresponding to the temperature difference response objective function and the control response objective function are output. It is worth noting that since the present invention itself does not improve the second generation non-dominated sorting algorithm, the solution process of the algorithm is no different from the general second generation non-dominated sorting algorithm.
[0118] Specifically, in step S4, the liquid cooling system is controlled according to the target control parameter.
[0119] For example, the final intelligent group control optimization scheme of the immersion liquid cooling system is calculated by using the non-dominated sorting genetic algorithm with an improved mobile density estimation strategy, which can be optimized by adjusting the fan speed V of the outdoor cooling device within a certain time step. f , Outdoor water pump speed V s , Spray pump speed V p , the speed of the primary circulation water pump of the primary side equipment V r , the secondary circulation water pump speed V of the secondary side equipment k , Valve opening degree K of cooling distribution unit eThese parameters are used to adjust the coolant temperature T A , Coolant flow rate V A , Cooling water temperature T B , Cooling water flow rate V B , Cabinet liquid inlet temperature T C , liquid flow rate V C , top and bottom coolant temperature difference T D 、Coolant height H L and the actual power consumption of the electronic device W R Etc., to achieve flexible, reliable and optimal intelligent group control strategy for immersion liquid cooling system.
[0120] Compared with the prior art, the liquid cooling system optimization control method described in the embodiment of the present invention has the following beneficial effects:
[0121] The existing immersion liquid cooling method can not fundamentally change the problem of huge energy consumption and difficulty in point-to-point control of the immersion liquid cooling system by improving the hardware structure and coolant. The existing control method cannot realize large-scale intelligent optimization control of the immersion liquid cooling system. The present invention can solve the limitations of single-objective optimization algorithms and multi-objective optimization algorithms, and make full use of the improved algorithm to perform multi-objective optimization on the immersion liquid cooling system, so that the optimization results are more reliable. It can also be flexibly adjusted according to the actual situation of the data center, has great guiding significance, and improves work efficiency. It can effectively improve the efficiency of the liquid cooling system, accurately adjust the control parameters of the liquid cooling system, and improve the stability and reliability of the liquid cooling system.
[0122] See also Figure 5 , Figure 5 1 is a structural block diagram of a liquid cooling system optimization control device 100 provided in an embodiment of the present invention, wherein the liquid cooling system optimization control device 100 comprises:
[0123] A real-time control parameter acquisition module 11 is used to acquire real-time control parameters of the liquid cooling system;
[0124] A predicted characteristic parameter acquisition module 12, used for inputting the real-time control parameter into a neural network so that the neural network outputs the predicted characteristic parameter of the liquid cooling system;
[0125] A function construction module 13, used to construct a temperature difference response objective function and a control response objective function using the predicted characteristic parameters;
[0126] A function solving module 14 is used to solve the temperature difference response objective function and the control response objective function to obtain corresponding target control parameters;
[0127] The control module 15 is used to control the liquid cooling system according to the target control parameter.
[0128] Specifically, the liquid cooling system optimization control device 100 further includes:
[0129] The neural network optimization module is used to optimize the neural network by using a moving density estimation strategy and a non-dominated sorting genetic algorithm.
[0130] Specifically, the neural network optimization module is also used to: construct a power consumption function of the electronic device using the predicted characteristic parameters; and optimize the neural network according to the power consumption function.
[0131] Specifically, the liquid cooling system includes an outdoor heat dissipation device, a primary side device, a secondary side device, a cold energy distribution unit and a cabinet for storing electronic equipment; wherein the primary side device is respectively connected to the outdoor heat dissipation device and the cold energy distribution unit, and the secondary side device is respectively connected to the cold energy distribution unit and the cabinet.
[0132] Specifically, the real-time control parameters include at least one of the following parameters: fan speed, outdoor water pump speed and spray pump speed of the outdoor cooling equipment; primary circulation water pump speed of the primary side equipment; secondary circulation water pump speed of the secondary side equipment; valve opening of the cooling distribution unit.
[0133] Specifically, the prediction characteristic parameters include temperature prediction parameters, flow rate prediction parameters, temperature difference prediction parameters, height prediction parameters and power consumption prediction parameters.
[0134] Specifically, the temperature prediction parameters include the coolant temperature, the cooling water temperature and the cabinet inlet temperature; the flow rate prediction parameters include the coolant flow rate, the cooling water flow rate and the cabinet inlet flow rate; the temperature difference prediction parameter is the coolant temperature difference in the cabinet; the height prediction parameter is the coolant height of the cabinet; the power consumption prediction parameter is the actual power consumption of the electronic equipment; wherein the coolant circulates between the cabinet and the cold distribution unit, and the cooling water circulates between the cold distribution unit and the outdoor heat dissipation device.
[0135] Specifically, the function construction module 13 is used to: obtain a preset temperature difference response weight coefficient, and obtain a reference temperature difference corresponding to the temperature difference prediction parameter; and construct a temperature difference response objective function according to the temperature difference prediction parameter, the reference temperature difference and the temperature difference response weight coefficient.
[0136] Specifically, the function construction module 13 is used to: obtain a preset control response weight coefficient, and obtain a reference temperature corresponding to the temperature prediction parameter, a reference flow rate corresponding to the flow rate prediction parameter, and a reference height corresponding to the height prediction parameter; and construct a control response objective function according to the control response weight coefficient, the temperature prediction parameter, the reference temperature, the flow rate prediction parameter, the reference flow rate, the height prediction parameter and the reference height.
[0137] It is worth noting that the working process of each module in the liquid cooling system optimization control device 100 described in the embodiment of the present invention can refer to the working process of the liquid cooling system optimization control method described in the above embodiment, and will not be repeated here.
[0138] See also Figure 6 , Figure 6 1 is a structural block diagram of a liquid cooling system optimization control device 200 provided in an embodiment of the present invention, and the liquid cooling system optimization control device 200 includes a processor 21, a memory 22, and a computer program stored in the memory 22 and executable on the processor 21. When the processor 21 executes the computer program, the steps in the above-mentioned liquid cooling system optimization control method embodiments are implemented, such as steps S1 to S4.
[0139] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory 22 and executed by the processor 21 to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of implementing specific functions, which are used to describe the execution process of the computer program in the liquid cooling system optimization control device 200.
[0140] The liquid cooling system optimization control device 200 may include, but is not limited to, a processor 21 and a memory 22. Those skilled in the art may understand that the schematic diagram is merely an example of the liquid cooling system optimization control device 200 and does not constitute a limitation on the liquid cooling system optimization control device 200. The liquid cooling system optimization control device 200 may include more or fewer components than shown in the figure, or may combine certain components, or different components. For example, the liquid cooling system optimization control device 200 may also include input and output devices, network access devices, buses, etc.
[0141] The processor 21 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor 21 is the control center of the liquid cooling system optimization control device 200, and uses various interfaces and lines to connect various parts of the entire liquid cooling system optimization control device 200.
[0142] The memory 22 can be used to store the computer program and / or module. The processor 21 realizes various functions of the liquid cooling system optimization control device 200 by running or executing the computer program and / or module stored in the memory 22 and calling the data stored in the memory 22. The memory 22 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory 22 can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.
[0143] Wherein, if the module / unit integrated in the liquid cooling system optimization control device 200 is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor 21. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal and software distribution medium, etc.
[0144] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.
Claims
1. A liquid cooling system optimization control method, characterized in that: include: Acquiring real-time control parameters of the liquid cooling system, and inputting the real-time control parameters into a neural network so that the neural network outputs predicted characteristic parameters of the liquid cooling system; Using the predicted characteristic parameters to construct a temperature difference response objective function and a control response objective function; Solving the temperature difference response objective function and the control response objective function to obtain corresponding target control parameters; The liquid cooling system is controlled according to the target control parameter.
2. The liquid cooling system optimization control method according to claim 1, characterized in that: Before obtaining the real-time control parameters of the liquid cooling system, the method further includes: The neural network is optimized using a moving density estimation strategy and a non-dominated sorting genetic algorithm.
3. The liquid cooling system optimization control method according to claim 1, characterized in that: After inputting the real-time control parameters into the neural network so that the neural network outputs the predicted characteristic parameters of the liquid cooling system, the method further includes: Constructing a power consumption function of the electronic device using the predicted characteristic parameters; The neural network is optimized according to the power consumption function.
4. The liquid cooling system optimization control method according to claim 1, characterized in that: The liquid cooling system includes an outdoor heat dissipation device, a primary side device, a secondary side device, a cold energy distribution unit and a cabinet for storing electronic equipment; wherein the primary side device is respectively connected to the outdoor heat dissipation device and the cold energy distribution unit, and the secondary side device is respectively connected to the cold energy distribution unit and the cabinet.
5. The liquid cooling system optimization control method according to claim 4, characterized in that: The real-time control parameters include at least one of the following parameters: The fan speed, outdoor water pump speed and sprinkler pump speed of the outdoor cooling equipment; Speed of primary circulation water pump of primary side equipment; Speed of secondary circulation water pump of secondary side equipment; Valve opening of the cooling distribution unit.
6. The liquid cooling system optimization control method according to claim 4, characterized in that: The prediction characteristic parameters include temperature prediction parameters, flow rate prediction parameters, temperature difference prediction parameters, height prediction parameters and power consumption prediction parameters.
7. The liquid cooling system optimization control method according to claim 6, characterized in that: The temperature prediction parameters include coolant temperature, cooling water temperature and cabinet inlet temperature; The flow rate prediction parameters include the coolant flow rate, the cooling water flow rate and the cabinet inlet flow rate; The temperature difference prediction parameter is the coolant temperature difference in the cabinet; The height prediction parameter is the coolant height of the cabinet; The power consumption prediction parameter is the actual power consumption of the electronic device; Wherein, the cooling liquid circulates between the cabinet and the cold distribution unit, and the cooling water circulates between the cold distribution unit and the outdoor heat dissipation device.
8. The liquid cooling system optimization control method according to claim 6, characterized in that: The method of constructing a temperature difference response objective function using the predicted characteristic parameters includes: Obtaining a preset temperature difference response weight coefficient, and obtaining a reference temperature difference corresponding to the temperature difference prediction parameter; A temperature difference response objective function is constructed according to the temperature difference prediction parameter, the reference temperature difference and the temperature difference response weight coefficient.
9. The liquid cooling system optimization control method according to claim 6, characterized in that: The using the predicted characteristic parameters to construct a control response objective function comprises: Obtaining a preset control response weight coefficient, and obtaining a reference temperature corresponding to the temperature prediction parameter, a reference flow velocity corresponding to the flow velocity prediction parameter, and a reference height corresponding to the height prediction parameter; A control response objective function is constructed according to the control response weight coefficient, the temperature prediction parameter, the reference temperature, the flow rate prediction parameter, the reference flow rate, the height prediction parameter and the reference height.
10. A liquid cooling system optimization control device, characterized in that: include: A real-time control parameter acquisition module is used to obtain real-time control parameters of the liquid cooling system; A predicted characteristic parameter acquisition module, used for inputting the real-time control parameter into a neural network so that the neural network outputs the predicted characteristic parameter of the liquid cooling system; A function construction module, used to construct a temperature difference response objective function and a control response objective function using the predicted characteristic parameters; A function solving module, used for solving the temperature difference response objective function and the control response objective function to obtain corresponding target control parameters; A control module is used to control the liquid cooling system according to the target control parameter.
11. A liquid cooling system, characterized in that: It includes an outdoor heat dissipation device, a primary-side device, a secondary-side device, a cold distribution unit, a cabinet for storing electronic equipment, and the liquid cooling system optimization control device as described in claim 10; wherein the primary-side device is respectively connected to the outdoor heat dissipation device and the cold distribution unit, and the secondary-side device is respectively connected to the cold distribution unit and the cabinet.
12. A liquid cooling system optimization control device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the liquid cooling system optimization control method according to any one of claims 1 to 9 is implemented.
13. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the liquid cooling system optimization control method according to any one of claims 1 to 9.
14. A computer program product, characterized in that It comprises a computer program, which, when executed by a processor, implements the liquid cooling system optimization control method as claimed in any one of claims 1 to 9.
Citation Information
Cited By
Liquid cooling server flow dynamic adjusting system based on genetic algorithm
CN120872115A
Liquid-cooled server flow dynamic adjustment system based on genetic algorithm
CN120872115B
Flow rate control method and apparatus for CDU group, electronic device, and storage medium
WO2026223020A1
