A Liquid Cooling Energy Consumption Regulation and Optimization Method Based on Digital Twin
By co-optimizing digital twin models and deep reinforcement learning models, and combining energy consumption assessment and scenario information of the heat dissipation control unit, the problem of accurate energy consumption regulation of liquid cooling systems was solved, achieving high efficiency, energy saving and stable operation of liquid cooling systems.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Filing Date
- 2026-06-10
- Publication Date
- 2026-07-17
Smart Images
Figure CN122421318A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to data center liquid cooling optimization, specifically to a liquid cooling energy consumption regulation and optimization method based on digital twins. Background Technology
[0002] With the large-scale deployment of high-density computing clusters, liquid cooling is gradually becoming the mainstream heat dissipation solution for data centers. Dynamic fluctuations in server room load and real-time changes in ambient temperature mean that relying on manual experience or fixed thresholds to adjust liquid cooling units can easily lead to issues such as redundant operation of cold pumps and over-cooling of the chillers, resulting in high energy consumption of the liquid cooling system and significant challenges in managing the server room's Power Usage Effectiveness (PUE). Currently, most mainstream liquid cooling system optimization solutions rely on repeated debugging of control parameters on physical equipment. Frequent on-site debugging consumes equipment operating time and incurs high debugging costs. Furthermore, physical testing carries operational risks such as server overheating and liquid cooling unit overload.
[0003] Most existing digital twin liquid cooling control solutions rely solely on the digital twin model's output control strategy, lacking intelligent algorithms to assist in optimization. The model's output control strategy is susceptible to modeling biases, making it difficult to achieve optimal global energy consumption. Furthermore, deep reinforcement learning, when used for optimization alone, lacks constraints from real-world equipment operating conditions, resulting in control strategies with insufficient adaptability for practical applications.
[0004] Meanwhile, existing liquid cooling system optimization solutions mostly start from the unified control of the whole machine, lacking refined energy efficiency identification of each heat dissipation control unit, unable to accurately locate inefficient heat dissipation control units, difficult to define targeted unit optimization heat dissipation control targets, and do not rely on scenario information to build a dedicated control optimization space, resulting in a lack of constraints in the optimization process and limited degree of alignment between optimization results and actual working conditions. Summary of the Invention
[0005] (a) Technical problems to be solved
[0006] In view of the above-mentioned shortcomings of the existing technology, the present invention provides a liquid cooling energy consumption regulation and optimization method based on digital twin, which can effectively overcome the shortcomings of the existing technology in that it is difficult to accurately regulate and optimize the energy consumption of liquid cooling systems.
[0007] (II) Technical Solution
[0008] To achieve the above objectives, the present invention provides the following technical solution:
[0009] A liquid cooling energy consumption regulation and optimization method based on digital twins includes the following steps:
[0010] S1. A digital twin model of the liquid cooling system is used to output the first control strategy and the corresponding first PUE value;
[0011] S2. Input the virtual running data of the digital twin model into the pre-trained deep reinforcement learning model, and output the second control policy and the corresponding second PUE value;
[0012] S3. Compare the first PUE value and the second PUE value to obtain the control strategy of the liquid cooling system;
[0013] S4. Collect the operation monitoring data of all heat dissipation control units, combine them with the data center heat dissipation energy saving control rules, evaluate the energy consumption efficiency of all heat dissipation control units, and obtain the optimized heat dissipation control units and optimized heat dissipation control targets.
[0014] S5. Based on the optimized heat dissipation control unit, load the data center heat dissipation related scenario information, and perform heat dissipation control feature mining and classification on the optimized heat dissipation control unit to establish an optimized heat dissipation control space;
[0015] S6. Combining the optimized heat dissipation control objectives, data center heat dissipation-related scenario information, and optimized heat dissipation control space, optimize the heat dissipation and energy-saving control of the optimized heat dissipation control unit to obtain a liquid cooling optimized control scheme.
[0016] S7. Combine the liquid cooling system control strategy with the liquid cooling optimization control scheme to generate the optimal liquid cooling system control scheme and execute the corresponding control.
[0017] Preferably, before the digital twin model of the liquid cooling system in S1 outputs the first control strategy and the corresponding first PUE value, it includes:
[0018] Collect equipment parameters and historical operating data of the liquid cooling system, and construct a digital twin model of the liquid cooling system based on the equipment parameters and historical operating data.
[0019] Preferably, the digital twin model of the liquid cooling system running in S1 outputs a first control strategy and a corresponding first PUE value, including:
[0020] Determine if the data acquisition sensors of the liquid cooling system are functioning properly. When the data acquisition sensors are functioning properly, collect the real-time operating data of the liquid cooling system.
[0021] The digital twin model of the liquid cooling system is optimized using real-time operating data, and the optimized digital twin model is run to output the first control strategy and the corresponding first PUE value.
[0022] Preferably, after collecting the equipment parameters, historical operating data and real-time operating data of the liquid cooling system, the data needs to undergo preprocessing, including handling missing values, removing outliers, normalizing data, and filtering related features.
[0023] Preferably, in step S2, the virtual running data of the digital twin model is input into a pre-trained deep reinforcement learning model, and the second control policy and the corresponding second PUE value are output, including:
[0024] The historical virtual operation data of the digital twin model is input into the deep reinforcement learning model, and the coolant flow rate is used as the action variable. The optimization objective is to obtain the minimum PUE value. The deep reinforcement learning model is trained to obtain a pre-trained deep reinforcement learning model.
[0025] The real-time virtual operation data of the digital twin model is input into the pre-trained deep reinforcement learning model, which outputs a second control policy and a corresponding second PUE value.
[0026] Preferably, in step S3, the first PUE value and the second PUE value are compared to obtain the liquid cooling system control strategy, including:
[0027] If the first PUE value is greater than the second PUE value, then the second control strategy will be used as the control strategy for the liquid cooling system, and the digital twin model of the liquid cooling system will be optimized using the real-time operating data of the regulated liquid cooling system.
[0028] Otherwise, the first control strategy is used as the control strategy for the liquid cooling system, and the deep reinforcement learning model is optimized using the real-time virtual operation data of the regulated digital twin model.
[0029] Preferably, in S4, operational monitoring data of all heat dissipation control units are collected, and combined with the data center heat dissipation energy-saving control rules, energy efficiency assessments are performed on all heat dissipation control units to obtain optimized heat dissipation control units and optimized heat dissipation control targets, including:
[0030] Energy efficiency is evaluated based on the operational monitoring data of each heat dissipation control unit to obtain the corresponding energy efficiency index;
[0031] The energy efficiency index of each heat dissipation control unit is compared with the energy efficiency threshold of each heat dissipation control unit in the data center heat dissipation energy saving control rules.
[0032] Based on the comparison results, the heat dissipation control unit with an energy efficiency index less than the energy efficiency threshold is selected as the optimized heat dissipation control unit, and the energy efficiency threshold of the optimized heat dissipation control unit is taken as the optimized heat dissipation control target.
[0033] Preferably, the step of evaluating the energy efficiency of the operational monitoring data of each heat dissipation control unit to obtain the corresponding energy efficiency index includes:
[0034] Energy efficiency evaluation records were collected for each heat dissipation control unit to obtain an energy efficiency evaluation record set.
[0035] The energy efficiency assessment model was trained under supervision using the energy efficiency assessment record set to obtain the energy efficiency assessment error coefficient.
[0036] When the energy efficiency assessment error coefficient is less than the energy efficiency assessment error threshold, model training is stopped, and a pre-trained energy efficiency assessment model is obtained.
[0037] The operation monitoring data of each heat dissipation control unit is input into the pre-trained energy efficiency evaluation model to evaluate energy efficiency and output the corresponding energy efficiency index.
[0038] Preferably, in S5, the optimized heat dissipation control unit loads the computer room heat dissipation-related scenario information, and performs heat dissipation control feature mining and classification on the optimized heat dissipation control unit to establish an optimized heat dissipation control space, including:
[0039] Based on the data center heat dissipation related scenario information, the heat dissipation control records of the optimized heat dissipation control unit are correlated and mined to obtain the data center heat dissipation control correlation domain.
[0040] Based on the associated domain of computer room heat dissipation control, heat dissipation control features are mined and classified to obtain multiple heat dissipation control feature areas, and an optimized heat dissipation control space is established.
[0041] Preferably, the step of mining and classifying heat dissipation control features based on the data center heat dissipation control association domain to obtain multiple heat dissipation control feature regions and establishing an optimized heat dissipation control space includes:
[0042] A concentrated interval calculation is performed on multiple heat dissipation control feature regions to obtain the optimized heat dissipation control domain;
[0043] Based on the optimized heat dissipation control domain, heat dissipation control decisions are made to obtain multiple optimized heat dissipation control decisions, and these multiple optimized heat dissipation control decisions are added to the optimized heat dissipation control space.
[0044] (III) Beneficial Effects
[0045] Compared with existing technologies, the liquid cooling energy consumption regulation and optimization method based on digital twins provided by this invention has the following beneficial effects:
[0046] 1) Optimize energy consumption configuration and improve energy utilization efficiency.
[0047] A collaborative optimization mechanism combining digital twin and deep reinforcement learning models is constructed, breaking the limitations of the traditional single control mode of liquid cooling systems. Relying on the digital twin model, which is accurately modeled and iteratively optimized in real time, it can realistically replicate the full operating state of the liquid cooling system and output a first control strategy that fits the actual hardware. The pre-trained deep reinforcement learning model takes the lowest PUE as its core objective and iteratively optimizes the output of the second control strategy based on the virtual operating data of the digital twin model. At the same time, by comparing and selecting the PUE values of the two strategies, the control strategy of the liquid cooling system is dynamically selected, avoiding energy waste problems such as overcooling liquid supply and idle redundant operation of liquid cooling units caused by traditional fixed parameter control. It accurately matches the heat dissipation needs of the computer room, maximizes the energy-saving potential of the liquid cooling system, effectively reduces the PUE of the computer room, and achieves refined energy utilization.
[0048] 2) Dynamic iterative optimization to improve the control efficiency of the liquid cooling system.
[0049] An innovative bidirectional model iteration and update mechanism is set up. While selecting the optimal control strategy for the liquid cooling system, it drives the continuous optimization of the digital twin model and the deep reinforcement learning model in reverse. When the deep reinforcement learning strategy is selected, the digital twin model is updated using the real-time operating data of the liquid cooling system. When the digital twin model strategy is selected, the deep reinforcement learning model is optimized based on the real-time virtual operating data of the digital twin model, realizing dynamic adaptive upgrading of the two models. At the same time, the energy efficiency assessment model accurately identifies inefficient heat dissipation control units, and targets optimization objects and objectives. It abandons the inefficient mode of traditional indiscriminate control and manual parameter adjustment, and constructs a dedicated control space by combining scene feature mining to complete precise optimization. This significantly shortens the control parameter adjustment cycle and significantly improves the response speed and operating efficiency of the liquid cooling system's dynamic control.
[0050] 3) Layered and precise optimization improves the accuracy of energy-saving control optimization.
[0051] This invention abandons the traditional extensive optimization model of liquid cooling systems and establishes a two-layer optimization system of "global strategy optimization + precise unit optimization". First, it obtains the global liquid cooling system control strategy through dual-model comparison. Then, it performs energy efficiency quantitative evaluation on each heat dissipation control unit, accurately locates the units with substandard energy efficiency and determines their specific optimization targets. By mining the heat dissipation-related scenario information and heat dissipation control characteristics of the computer room, it classifies and constructs a dedicated optimized heat dissipation control space. Under the constraints of matching actual working conditions, it completes refined energy-saving optimization, effectively solving the problems of poor scenario adaptability, insufficient targeting and easy getting trapped in local optima in traditional optimization methods. Multi-level linkage optimization allows the final integrated control scheme to take into account both global energy consumption optimization and efficient unit operation, greatly improving the accuracy and stability of energy-saving regulation of the liquid cooling system. Attached Figure Description
[0052] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.
[0053] Figure 1 This is a schematic diagram of the process of the present invention;
[0054] Figure 2 This is a schematic diagram of the process for obtaining the control strategy of the liquid cooling system in this invention. Detailed Implementation
[0055] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.
[0056] The following describes the specific process of the liquid cooling energy consumption regulation and optimization method based on digital twins provided by this invention, using specific examples (e.g.) Figure 1 (as shown) and technical effects.
[0057] S1. Run the digital twin model of the liquid cooling system and output the first control strategy and the corresponding first PUE value.
[0058] Before the digital twin model of the liquid cooling system running in S1 outputs the first control strategy and the corresponding first PUE value, as follows: Figure 2 As shown, it includes:
[0059] Collect equipment parameters and historical operating data of the liquid cooling system, and construct a digital twin model of the liquid cooling system based on the equipment parameters and historical operating data.
[0060] The digital twin model of the liquid cooling system running in S1 outputs the first control strategy and the corresponding first PUE value, such as Figure 2 As shown, it includes:
[0061] Determine if the data acquisition sensors of the liquid cooling system are functioning properly. When the data acquisition sensors are functioning properly, collect the real-time operating data of the liquid cooling system.
[0062] The digital twin model of the liquid cooling system is optimized using real-time operating data, and the optimized digital twin model is run to output the first control strategy and the corresponding first PUE value.
[0063] The above technical solution requires preprocessing of the data after collecting equipment parameters, historical operating data and real-time operating data of the liquid cooling system, including missing value handling, outlier removal, data normalization and correlation feature filtering.
[0064] S2. Input the virtual operation data of the digital twin model into the pre-trained deep reinforcement learning model, and output the second control policy and the corresponding second PUE value, such as... Figure 2 As shown, it includes:
[0065] The historical virtual operation data of the digital twin model is input into the deep reinforcement learning model, and the coolant flow rate is used as the action variable. The optimization objective is to obtain the minimum PUE value. The deep reinforcement learning model is trained to obtain a pre-trained deep reinforcement learning model.
[0066] The real-time virtual operation data of the digital twin model is input into the pre-trained deep reinforcement learning model, which outputs a second control policy and a corresponding second PUE value.
[0067] S3. Compare the first PUE value and the second PUE value to obtain the control strategy of the liquid cooling system, such as... Figure 2 As shown, it includes:
[0068] If the first PUE value is greater than the second PUE value, then the second control strategy will be used as the control strategy for the liquid cooling system, and the digital twin model of the liquid cooling system will be optimized using the real-time operating data of the regulated liquid cooling system.
[0069] Otherwise, the first control strategy is used as the control strategy for the liquid cooling system, and the deep reinforcement learning model is optimized using the real-time virtual operation data of the regulated digital twin model.
[0070] The above technical solution constructs a collaborative optimization mechanism of digital twin and deep reinforcement learning dual models, breaking the limitations of the traditional single control mode of liquid cooling systems. Relying on the digital twin model with accurate modeling and real-time iterative optimization, it can realistically replicate the full operating state of the liquid cooling system and output a first control strategy that fits the actual hardware. The pre-trained deep reinforcement learning model takes the lowest PUE as the core objective and iteratively optimizes the output of the second control strategy based on the virtual operating data of the digital twin model. At the same time, by comparing and selecting the best PUE value of the two strategies, the control strategy of the liquid cooling system is dynamically selected, avoiding the energy waste problems caused by traditional fixed parameter control, such as overcooling liquid supply and idle redundant operation of liquid cooling units. It accurately matches the heat dissipation needs of the computer room, maximizes the energy-saving potential of the liquid cooling system, effectively reduces the PUE of the computer room, and achieves refined energy utilization.
[0071] In addition, the above technical solution innovatively sets up a two-way model iterative update mechanism. While selecting the best control strategy for the liquid cooling system, it drives the continuous optimization of the digital twin model and the deep reinforcement learning model. When the deep reinforcement learning strategy is selected, the digital twin model is updated using the real-time operating data of the liquid cooling system. When the digital twin model strategy is selected, the deep reinforcement learning model is optimized based on the real-time virtual operating data of the digital twin model, so as to realize the dynamic adaptive upgrade of the two models.
[0072] S4. Collect operational monitoring data from all thermal management control units, and in conjunction with the data center's energy-saving thermal management control rules, evaluate the energy efficiency of all thermal management control units to obtain optimized thermal management control units and optimized thermal management control objectives, including:
[0073] Energy efficiency is evaluated based on the operational monitoring data of each heat dissipation control unit to obtain the corresponding energy efficiency index;
[0074] The energy efficiency index of each heat dissipation control unit is compared with the energy efficiency threshold of each heat dissipation control unit in the data center heat dissipation energy saving control rules.
[0075] Based on the comparison results, the heat dissipation control unit with an energy efficiency index less than the energy efficiency threshold is selected as the optimized heat dissipation control unit, and the energy efficiency threshold of the optimized heat dissipation control unit is taken as the optimized heat dissipation control target.
[0076] Specifically, the energy efficiency of each heat dissipation control unit's operational monitoring data is evaluated to obtain the corresponding energy efficiency index, including:
[0077] Energy efficiency evaluation records were collected for each heat dissipation control unit to obtain an energy efficiency evaluation record set.
[0078] The energy efficiency assessment model was trained under supervision using the energy efficiency assessment record set to obtain the energy efficiency assessment error coefficient.
[0079] When the energy efficiency assessment error coefficient is less than the energy efficiency assessment error threshold, model training is stopped, and a pre-trained energy efficiency assessment model is obtained.
[0080] The operation monitoring data of each heat dissipation control unit is input into the pre-trained energy efficiency evaluation model to evaluate energy efficiency and output the corresponding energy efficiency index.
[0081] The above technical solution accurately identifies inefficient heat dissipation control units through an energy efficiency assessment model, specifically targets optimization objects and goals, abandons the inefficient mode of traditional all-domain indiscriminate control and manual parameter adjustment, and combines scenario feature mining to construct a dedicated control space to achieve precise optimization, significantly shortens the control parameter adjustment cycle, and significantly improves the response speed and operating efficiency of the liquid cooling system's dynamic control.
[0082] S5. Based on the optimized heat dissipation control unit, load the data center heat dissipation-related scenario information, and perform heat dissipation control feature mining and classification on the optimized heat dissipation control unit to establish an optimized heat dissipation control space, including:
[0083] Based on the data center heat dissipation related scenario information, the heat dissipation control records of the optimized heat dissipation control unit are correlated and mined to obtain the data center heat dissipation control correlation domain.
[0084] Based on the associated domain of computer room heat dissipation control, heat dissipation control features are mined and classified to obtain multiple heat dissipation control feature areas, and an optimized heat dissipation control space is established.
[0085] Specifically, based on the data center heat dissipation control association domain, heat dissipation control features are mined and classified to obtain multiple heat dissipation control feature regions, and an optimized heat dissipation control space is established, including:
[0086] A concentrated interval calculation is performed on multiple heat dissipation control feature regions to obtain the optimized heat dissipation control domain;
[0087] Based on the optimized heat dissipation control domain, heat dissipation control decisions are made to obtain multiple optimized heat dissipation control decisions, and these multiple optimized heat dissipation control decisions are added to the optimized heat dissipation control space.
[0088] S6. Combining the optimized heat dissipation control objectives, the data center heat dissipation related scenario information, and the optimized heat dissipation control space, the optimized heat dissipation control unit is optimized for heat dissipation and energy saving control to obtain a liquid cooling optimized control scheme.
[0089] S7. Combine the liquid cooling system control strategy with the liquid cooling optimization control scheme to generate the optimal liquid cooling system control scheme and execute the corresponding control.
[0090] This application's technical solution abandons the traditional extensive optimization model of liquid cooling systems and establishes a two-layer optimization system of "global strategy optimization + precise unit optimization". First, it obtains the global liquid cooling system control strategy through dual-model comparison, and then performs energy efficiency quantitative evaluation on each heat dissipation control unit to accurately locate units with substandard energy efficiency and determine their specific optimization targets. By mining the heat dissipation-related scenario information and heat dissipation control characteristics of the computer room, it classifies and constructs a dedicated optimized heat dissipation control space. Under the constraints of matching actual working conditions, it completes refined energy-saving optimization, effectively solving the problems of poor scenario adaptability, insufficient targeting, and easy getting trapped in local optima in traditional optimization methods. Multi-level linkage optimization allows the final integrated control scheme to take into account both global energy consumption optimization and efficient unit operation, greatly improving the accuracy and stability of energy-saving regulation of the liquid cooling system.
[0091] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A liquid cooling energy consumption regulation and optimization method based on digital twins, characterized in that: Includes the following steps: S1. A digital twin model of the liquid cooling system is used to output the first control strategy and the corresponding first PUE value; S2. Input the virtual running data of the digital twin model into the pre-trained deep reinforcement learning model, and output the second control policy and the corresponding second PUE value; S3. Compare the first PUE value and the second PUE value to obtain the control strategy of the liquid cooling system; S4. Collect the operation monitoring data of all heat dissipation control units, combine them with the data center heat dissipation energy saving control rules, evaluate the energy consumption efficiency of all heat dissipation control units, and obtain the optimized heat dissipation control units and optimized heat dissipation control targets. S5. Based on the optimized heat dissipation control unit, load the data center heat dissipation related scenario information, and perform heat dissipation control feature mining and classification on the optimized heat dissipation control unit to establish an optimized heat dissipation control space; S6. Combining the optimized heat dissipation control objectives, data center heat dissipation-related scenario information, and optimized heat dissipation control space, optimize the heat dissipation and energy-saving control of the optimized heat dissipation control unit to obtain a liquid cooling optimized control scheme. S7. Combine the liquid cooling system control strategy with the liquid cooling optimization control scheme to generate the optimal liquid cooling system control scheme and execute the corresponding control.
2. The liquid cooling energy consumption regulation and optimization method based on digital twin according to claim 1, characterized in that: Before the digital twin model of the liquid cooling system running in S1 outputs the first control strategy and the corresponding first PUE value, it includes: Collect equipment parameters and historical operating data of the liquid cooling system, and construct a digital twin model of the liquid cooling system based on the equipment parameters and historical operating data.
3. The liquid cooling energy consumption regulation and optimization method based on digital twin according to claim 2, characterized in that: The digital twin model of the liquid cooling system running in S1 outputs the first control strategy and the corresponding first PUE value, including: Determine if the data acquisition sensors of the liquid cooling system are functioning properly. When the data acquisition sensors are functioning properly, collect the real-time operating data of the liquid cooling system. The digital twin model of the liquid cooling system is optimized using real-time operating data, and the optimized digital twin model is run to output the first control strategy and the corresponding first PUE value.
4. The liquid cooling energy consumption regulation and optimization method based on digital twin according to claim 3, characterized in that: After collecting the equipment parameters, historical operating data, and real-time operating data of the liquid cooling system, preprocessing is required, including handling missing values, removing outliers, normalizing data, and filtering related features.
5. The liquid cooling energy consumption regulation and optimization method based on digital twin according to claim 3, characterized in that: In S2, the virtual running data of the digital twin model is input into a pre-trained deep reinforcement learning model, which outputs a second control policy and a corresponding second PUE value, including: The historical virtual operation data of the digital twin model is input into the deep reinforcement learning model, and the coolant flow rate is used as the action variable. The optimization objective is to obtain the minimum PUE value. The deep reinforcement learning model is trained to obtain a pre-trained deep reinforcement learning model. The real-time virtual operation data of the digital twin model is input into the pre-trained deep reinforcement learning model, which outputs a second control policy and a corresponding second PUE value.
6. The liquid cooling energy consumption regulation and optimization method based on digital twin according to claim 5, characterized in that: In S3, the first PUE value and the second PUE value are compared to obtain the liquid cooling system control strategy, including: If the first PUE value is greater than the second PUE value, then the second control strategy will be used as the control strategy for the liquid cooling system, and the digital twin model of the liquid cooling system will be optimized using the real-time operating data of the regulated liquid cooling system. Otherwise, the first control strategy is used as the control strategy for the liquid cooling system, and the deep reinforcement learning model is optimized using the real-time virtual operation data of the regulated digital twin model.
7. The liquid cooling energy consumption regulation and optimization method based on digital twin according to claim 1, characterized in that: S4 collects operational monitoring data from all thermal control units, combines it with data center energy-saving control rules, and evaluates the energy efficiency of all thermal control units to obtain optimized thermal control units and optimized thermal control objectives, including: Energy efficiency is evaluated based on the operational monitoring data of each heat dissipation control unit to obtain the corresponding energy efficiency index; The energy efficiency index of each heat dissipation control unit is compared with the energy efficiency threshold of each heat dissipation control unit in the data center heat dissipation energy saving control rules. Based on the comparison results, the heat dissipation control unit with an energy efficiency index less than the energy efficiency threshold is selected as the optimized heat dissipation control unit, and the energy efficiency threshold of the optimized heat dissipation control unit is taken as the optimized heat dissipation control target.
8. The liquid cooling energy consumption regulation and optimization method based on digital twin according to claim 7, characterized in that: The energy efficiency evaluation of the operational monitoring data of each heat dissipation control unit to obtain the corresponding energy efficiency index includes: Energy efficiency evaluation records were collected for each heat dissipation control unit to obtain an energy efficiency evaluation record set. The energy efficiency assessment model was trained under supervision using the energy efficiency assessment record set to obtain the energy efficiency assessment error coefficient. When the energy efficiency assessment error coefficient is less than the energy efficiency assessment error threshold, model training is stopped, and a pre-trained energy efficiency assessment model is obtained. The operation monitoring data of each heat dissipation control unit is input into the pre-trained energy efficiency evaluation model to evaluate energy efficiency and output the corresponding energy efficiency index.
9. The liquid cooling energy consumption regulation and optimization method based on digital twin according to claim 7, characterized in that: In S5, the optimized heat dissipation control unit loads data center heat dissipation-related scenario information, and performs heat dissipation control feature mining and classification on the optimized heat dissipation control unit to establish an optimized heat dissipation control space, including: Based on the data center heat dissipation related scenario information, the heat dissipation control records of the optimized heat dissipation control unit are correlated and mined to obtain the data center heat dissipation control correlation domain. Based on the associated domain of computer room heat dissipation control, heat dissipation control features are mined and classified to obtain multiple heat dissipation control feature areas, and an optimized heat dissipation control space is established.
10. The liquid cooling energy consumption regulation and optimization method based on digital twin according to claim 9, characterized in that: The process involves mining and classifying heat dissipation control features based on the associated domain of the computer room heat dissipation control, obtaining multiple heat dissipation control feature regions, and establishing an optimized heat dissipation control space, including: A concentrated interval calculation is performed on multiple heat dissipation control feature regions to obtain the optimized heat dissipation control domain; Based on the optimized heat dissipation control domain, heat dissipation control decisions are made to obtain multiple optimized heat dissipation control decisions, and these multiple optimized heat dissipation control decisions are added to the optimized heat dissipation control space.