Energy-saving intelligent control method for high-heat-density data center liquid cooling system

Abstract

The invention discloses an energy-saving intelligent control method for a liquid cooling system of a high-heat-density data center, which relates to the technical field of temperature control and mainly comprises the following steps: respectively establishing a predictive control model and a preset rule control model for a primary side cold source loop and a secondary side cooling liquid distribution loop; an hour-level optimization operation scheme of the primary side equipment is obtained and executed by executing the prediction model, and meanwhile, the outlet temperature of a secondary side cooling liquid distribution unit is monitored in real time; taking the outlet temperature as a triggering condition, and when the outlet temperature exceeds a preset range, triggering recalculation and updating of a primary side control scheme; by circulating the process, dynamic coordination linkage of primary side prediction optimization and secondary side real-time control is achieved, and finally the total energy consumption of the system is reduced on the premise that heat dissipation is guaranteed. According to the invention, high-efficiency and stable operation of the high-heat-density data center liquid cooling system under the dynamic load is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of temperature control, in particular to an energy-saving intelligent control method for a high-heat-density data center liquid cooling system. BACKGROUND

[0002] The liquid cooling technology of data center is a key solution to the heat dissipation challenge of high-heat-density servers. In current technical practice, the control strategy of the liquid cooling system has become a core factor affecting its energy efficiency and stability. The traditional control scheme generally adopts the classic PID control method, which focuses on maintaining the stability of a single control loop (such as the temperature or pressure loop). However, the liquid cooling system of data center is a complex coupled system composed of a primary-side cooling source and a secondary-side cooling liquid distribution network. Multiple independent PID loops often conflict when running. For example, to reduce the frequency of the secondary-side water pump for energy saving, it may lead to insufficient heat exchange temperature difference, and thus require the primary-side refrigeration machine to provide lower cold water temperature to compensate, ultimately causing significant increase in refrigeration machine energy consumption, forming a dilemma of "pushing down the gourd to float the dipper", and the overall energy efficiency of the system has not been optimized.

[0003] To break through the limitations of traditional control, the existing technical research mainly focuses on three directions: first, dynamic flow control based on IT load, which finely adjusts the end cooling liquid flow by real-time sensing of the power consumption and temperature of server CPU / GPU; second, multivariable decoupling control, aiming to solve the strong coupling relationship between system internal pressure, flow and other parameters, to realize independent precise control; the third is to introduce artificial intelligence and machine learning algorithms, through the analysis of historical load data to predict future heat load, so that the system changes from passive response to active optimization, thereby improving control quality and tapping energy-saving potential. Although these methods have made progress at the local or subsystem level, they mostly still focus on the independent optimization of the "secondary-side cooling liquid" loop or the "primary-side cooling source" loop, and have not effectively solved the problem of coordinated balance of the two subsystems at the heat exchange interface. This lack of consideration of the coupling relationship between systems makes it difficult for the control strategy to optimize the overall energy consumption under the premise of ensuring the safety of end heat dissipation, limiting the further improvement of the energy efficiency of the liquid cooling system. SUMMARY

[0004] To strengthen the construction of the coupling relationship between systems and ensure the overall energy consumption optimization under the premise of ensuring the safety of end heat dissipation, the present application proposes an energy-saving intelligent control method for a high-heat-density data center liquid cooling system, which coordinates the predictive optimization control of the primary-side cooling source loop and the real-time control of the secondary-side cooling liquid distribution loop by performing a cycle step, the cycle step comprising: S1: for the primary-side cooling source loop of the liquid cooling system, a predictive control model is established with the minimization of total equipment energy consumption in the future prediction interval as the optimization objective, while meeting the cooling capacity demand and various constraints of the system; S2: for the secondary side cooling liquid distribution circuit of the liquid cooling system, a preset rule control model is established to stabilize the temperature of the server device as the target, and the cooling liquid flow is dynamically distributed and controlled based on the real-time power consumption of the server device; S3: based on the preset rule control model, the devices in the secondary side cooling liquid distribution circuit are controlled in real time, and by running the prediction control model, the operation scheme sequence of each control variable of the primary side cooling source circuit in the future prediction interval is obtained, and the devices in the primary side cooling source circuit are controlled based on the operation scheme sequence; S4: during the execution of the operation scheme sequence, the outlet temperature of the cooling liquid distribution unit in the secondary side cooling liquid distribution circuit is monitored in real time; S5: taking the outlet temperature of the cooling liquid distribution unit as the trigger condition, when the outlet temperature exceeds the preset range, triggering and returning to S1 to update the operation scheme sequence of the primary side cooling source circuit.

[0005] The application combines the global and slow energy-saving optimization of the primary side circuit with the local and fast stable control of the secondary side circuit by constructing a double-layer multi-time scale coordinated control architecture, solves the collaborative balance problem of the two subsystems at the heat exchange interface, and realizes the efficient and stable operation of the liquid cooling system of the high-heat-density data center under dynamic load.

[0006] Further, in the S1 step, the data relied on by the prediction control model includes the data center computing task allocation plan and / or the predicted computing load.

[0007] Further, in the S2 step, the secondary side cooling liquid distribution circuit includes a cooling liquid distribution unit and an end circuit arranged in the data center, and the control variable of the cooling liquid flow in the preset rule control model includes at least one of the frequency conversion water pump frequency of the cooling liquid distribution unit, the frequency conversion water pump frequency of the end circuit, and the opening degree of the electric valve of the end circuit.

[0008] Further, when the cooling liquid flow is distributed by adjusting the valve opening degree, for the end circuit with stable heat load, a flow balance valve is used for flow control.

[0009] Further, when the cooling liquid flow is distributed by adjusting the valve opening degree, for the end circuit with large heat load variation, a pressure difference independent control valve is used for flow control.

[0010] Further, for the scene with high cooling liquid flow distribution control requirement, a frequency conversion water pump and electric valve cooperative scheme is adopted to actively distribute the flow of each end circuit in the cooling liquid distribution unit.

[0011] Further, in the S3 step, the control variable includes at least one of a chiller load rate, a variable frequency water pump frequency, and a cooling tower fan rotating speed.

[0012] Further, in the S5 step, the following is further included: When the outlet temperature does not exceed the preset range, then after a preset fixed time interval of executing the operation scheme sequence is reached, triggering and returning to the S1 step, and before a preset fixed time interval is reached, returning to the S5 step.

[0013] Further, the secondary side cooling liquid distribution circuit includes a plurality of cooling liquid distribution units, and the primary side cooling source circuit provides cooling capacity for the plurality of cooling liquid distribution units, thereby forming a one-to-many liquid cooling network.

[0014] Compared with the prior art, the present application has at least the following beneficial effects: (1) The energy-saving intelligent control method for the high-heat-density data center liquid cooling system is proposed, a double-layer multi-time scale coordinated control architecture is constructed, global slow energy-saving optimization of the primary side circuit and local fast stable control of the secondary side circuit are organically combined, and efficient and stable operation of the high-heat-density data center liquid cooling system under dynamic load is realized. (2) The model predictive control is used in the primary side, hour-level energy consumption optimization is carried out based on the computing task plan of the data center, the energy supply lag or waste problem caused by the large inertia of the traditional responsive control is fundamentally overcome, and the operation cost of main energy-consuming equipment such as the chiller and the water pump is reduced. (3) The rule control based on real-time power consumption is used in the secondary side, the cooling liquid flow is accurately and dynamically allocated at the second or minute level, the server equipment can always be in the best temperature interval when facing sudden computing tasks, and the performance and service life of the IT equipment are ensured. (4) Through the event-driven mechanism closely linked by the cooling liquid distribution unit outlet temperature, the temperature signal is used as the priority trigger condition for updating the primary side optimization scheme, the supply and demand relationship of the cooling capacity is intelligently coordinated, and energy dynamic balance is realized. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 It is a step diagram of the energy-saving intelligent control method for the high-heat-density data center liquid cooling system. Figure 2 It is a liquid cooling system schematic diagram under the condition of the single-edge controller of the secondary side cooling liquid distribution circuit. Figure 3 It is a liquid cooling system schematic diagram under the condition of the multi-edge controller of the secondary side cooling liquid distribution circuit. DETAILED DESCRIPTION

[0016] The following is a specific embodiment of the present application and further describes the technical solutions of the present application in conjunction with the drawings, but the present application is not limited to these embodiments.

[0017] With the rapid development of cloud computing, artificial intelligence and other technologies, the computing density and power consumption of data centers continue to rise, and the traditional air cooling cooling method has gradually failed to meet the cooling needs of high heat density cabinets. Liquid cooling technology has become an inevitable choice due to its high cooling capacity. In the control field of liquid cooling systems, existing technologies focus on local optimization within the system: or use traditional PID control methods to maintain the stability of single loop parameters, but multiple PID loops often conflict with each other and are difficult to achieve global energy efficiency optimization; or strive to deal with the strong coupling relationship between multiple variables in the system through complex decoupling algorithms, but face the challenges of complex models and high implementation costs; although some research has tried to introduce artificial intelligence technology for load prediction and active control, these solutions are still mostly limited to independent optimization of one of the "secondary side cooling liquid distribution" or "primary side cooling source", and have not effectively solved the dynamic balance problem of the two at the heat exchange interface. Due to the significant differences in physical characteristics and control objectives between the primary side system (cooling source) and the secondary side system (liquid cooling distribution) - the former has large inertia, slow response, and energy saving as the primary goal; the latter needs to respond to the rapid fluctuations of IT loads, with temperature stability as the core task - the traditional mode of splitting control of the two inevitably leads to system response lag, energy waste, and even affects the operation reliability of IT equipment. In order to overcome the above defects, the present application aims to provide an efficient and energy-saving control method for the primary and secondary sides, and for this purpose, as shown in Figure 1 the present application proposes an energy-saving intelligent control method for a high-heat-density data center liquid cooling system, mainly including the following steps: S1: For the primary side cooling source loop of the liquid cooling system, a prediction control model is established with the minimum total energy consumption of the equipment in the future prediction interval as the optimization goal, while meeting the cooling capacity demand and various constraints of the system; S2: For the secondary side cooling liquid distribution loop of the liquid cooling system, a preset rule control model is established for dynamic distribution control of the cooling liquid flow based on the real-time power consumption of the server equipment, with the goal of stabilizing the temperature of the server equipment; S3: Based on the preset rule control model, the equipment in the secondary side cooling liquid distribution loop is controlled in real time, and by running the prediction control model, the operation scheme sequence of each control variable of the primary side cooling source loop in the future prediction interval is obtained, and the equipment in the primary side cooling source loop is controlled based on the operation scheme sequence; S4: During the execution of the operation scheme sequence, the outlet temperature of the cooling liquid distribution unit in the secondary side cooling liquid distribution loop is monitored in real time; S5: Taking the outlet temperature of the cooling liquid distribution unit as a trigger condition, when the outlet temperature exceeds the preset range, triggering and returning to S1 step to update the operation scheme sequence of the primary side cold source circuit.

[0018] In practical applications, the liquid cooling system is usually physically composed of a cold source layer of the primary side circuit, a cooling liquid distribution layer of the secondary side circuit, and an upper computer or platform layer responsible for global monitoring and scheduling. The cold source layer of the primary side circuit mainly includes refrigeration units, primary side water pumps, cooling towers and their supporting pipelines, water supplementing and pressure regulating devices, water treatment equipment, etc., and its core function is to produce and provide sufficient cold energy through the reverse Carnot cycle or free cooling of refrigeration units, and store the cold energy in chilled water. The cooling liquid distribution layer of the secondary side circuit is a relatively independent internal circulation system, mainly including one or more cooling liquid distribution units (CDUs) and multiple end circuits connected to each CDU. The CDU usually contains plate heat exchangers, circulating water pumps, expansion tanks, precision filters, and monitoring instruments, etc., and its core function is to achieve efficient heat exchange between the primary side chilled water and the secondary side cooling liquid, while physically isolating the two water systems to improve system safety and reliability. Each CDU is connected to several end circuits, which deliver the cooled working medium (usually special cooling liquid such as deionized water, fluorinated liquid, etc.) to the liquid cooling plates in the server cabinet. The cooling liquid flows through the cooling plates close to the CPU, GPU, and other high-power chips, absorbing the huge heat generated by the chips through direct contact or micro-channel convection, completing the direct cooling of the heat-generating elements. The heated cooling liquid returns to the CDU, transferring the heat to the primary side chilled water, completing a cycle. The upper computer or platform layer, as the system core, usually deploys a data center infrastructure management system (DCIM) or a dedicated liquid cooling control platform, which communicates with the lower layer devices through industrial Ethernet, field bus or wireless sensor network, collects the running data of the whole system, and based on the business planning of the data center, historical load curve and artificial intelligence prediction algorithm, predicts the computing load in the next few hours to days, or formulates a fine computing task allocation plan, providing prospective decision basis for the front-end optimization control.

[0019] In order to realize fine and intelligent control, the system has been innovatively designed in the control architecture. Specifically, independent edge controllers are set up in the cold source layer of the primary side and the cooling liquid distribution layer of the secondary side. The edge controller of the cold source layer focuses on executing complex optimization algorithms, while the edge controller of the cooling liquid distribution layer is responsible for quickly and real-time controlling the devices of the CDU and end circuits. A direct data communication link is established between the two controllers, which can share key operating parameters, laying a foundation for coordinated control between subsystems. For example, Figure 2As shown, in a typical embodiment, a unified edge controller can be deployed at the secondary side to manage all CDUs and end devices thereunder, which is suitable for medium-sized data centers. For large-scale data center rooms, there can be dozens or even hundreds of CDU units in the cooling liquid distribution layer, serving different room modules or customer clusters, such as Figure 3 As shown, independent edge controllers are configured for each CDU or regional cluster thereof, which can achieve more precise and rapid local control, thereby better addressing the independent and dynamic heat dissipation requirements of high-heat-density server clusters in different regions, and avoiding the risk of insufficient processing capacity and single-point failure of a single controller.

[0020] Based on the above system architecture, the energy-saving intelligent control method of the present application proposes a double-layer multi-time scale coordinated control strategy. The core of the control logic of this method is that, in view of the different characteristics of the primary side system (large inertia and slow response) and the secondary side system (need to respond to rapid load fluctuations), different control strategies are adopted respectively, and the two are tightly coupled through key parameters.

[0021] For the control of the primary side cooling source circuit, the present application adopts an optimization method based on model predictive control (MPC).

[0022] Firstly, a state prediction control model needs to be established for the circuit. This model is obtained through mechanism modeling or system identification method, and can accurately represent the nonlinear energy consumption characteristics of devices such as chillers (whose COP curve changes with load rate and working conditions), water pumps (whose power consumption has a cubic relationship with flow rate and head), cooling towers (whose heat dissipation efficiency has a relationship with air volume and wet-bulb temperature), and the fluid dynamics and heat transfer processes within the system. Key forward-looking data required for establishing the model, such as future computing load predicted based on business logs and AI models provided by the upper computer or platform layer, provide external disturbance prediction input for the MPC model, enabling it to "anticipate" future cooling demand changes.

[0023] Subsequently, an MPC rolling optimization model is constructed, which takes a specific period (such as 4 hours or 8 hours, which can be configured according to the load cycle characteristics of the data center) as the prediction interval, and minimizes the total energy consumption of all devices in the primary side circuit within the interval as the optimization objective, while meeting the predicted cooling demand and various physical constraints of the system (such as upper and lower limits of chiller operating load rate, water pump frequency range, cooling tower approach limit, pipe network pressure capacity, etc.). By solving this constrained multivariate nonlinear optimization problem, the optimal operation scheme time series of chiller load rate, water pump frequency, cooling tower fan speed, etc. within the prediction interval can be obtained.

[0024] It is worth noting that the system does not rigidly execute the sequence, but introduces an intelligent update mechanism based on event-driven. During the execution of the scheme, the CDU coolant outlet temperature from the secondary side coolant distribution layer, which best reflects the balance of supply and demand, is continuously monitored. As long as the temperature stabilizes within the pre-set reasonable range (for example, set point ± 0.5°C), the system will perform MPC rolling optimization calculation as a backup trigger condition at pre-set fixed time intervals (for example, every hour) to periodically correct the control scheme to adapt to model mismatch and slow environmental changes. However, once the CDU outlet temperature is continuously monitored to exceed the pre-set range due to unpredictable load mutation, the emergency recalculation of MPC is triggered to update the control scheme. This "event-driven first, time-driven second" mechanism ensures that the primary side cooling can quickly respond to the emergency demand changes of the secondary side, maintains the dynamic thermal balance of the entire system, while avoiding unnecessary frequent optimization calculation, achieving the balance of optimization performance and calculation cost.

[0025] For the control of the secondary side coolant distribution circuit, a real-time control method based on pre-set rules is adopted to achieve second-level or minute-level fast response, and the core goal is to ensure that the temperature of each server's key components is always below the safety threshold.

[0026] First of all, it is necessary to establish the operation rule library or expert system of the circuit, which defines how to dynamically adjust the frequency of the CDU internal variable frequency water pump, the frequency of each terminal circuit variable frequency water pump and the opening of the electric regulating valve according to the real-time power consumption and CPU / GPU core temperature feedback read by the sensors deployed on the server motherboard. The combination adjustment of these control variables realizes the precise distribution scheme of coolant flow based on the real-time load change of the server. For example, a rule may be: "if the CPU temperature of a server exceeds T1, increase the valve opening of the terminal circuit where it is located linearly until the temperature falls to the target range; if all server temperatures are below T2, reduce the CDU water pump frequency in proportion to save energy".

[0027] In the specific implementation of flow distribution, the application also provides a hierarchical and on-demand matching solution. When a valve is used as the main adjustment means, the system intelligently matches the valve type according to the historical data and real-time characteristics of the server thermal load served by the terminal circuit. For terminal circuits with relatively stable thermal load, such as running databases and storage servers, which have a small fluctuation range, passive flow distribution type valves such as flow balance valves with lower cost and almost maintenance-free can meet the basic requirements, and fixed flow is distributed through pre-setting. For terminal circuits serving cloud computing instances, online transaction processing or scientific computing tasks, which have a large dynamic range of thermal load changes, active flow distribution type valves such as differential pressure independent control valves must be used. Such valves can automatically offset the pressure fluctuations in the system caused by the adjustment of other circuits through internal mechanical or electronic feedback mechanisms, ensuring that the flow to the circuit accurately follows the set value changes, providing accurate "on-demand cooling" for dynamic loads. Further, for special scenarios such as artificial intelligence training clusters and supercomputers, which have extremely high flow distribution control requirements and are extremely sensitive to temperature fluctuations, the application proposes an active distribution scheme using a variable frequency pump and an electric valve working together. By deploying small variable frequency pumps in the terminal circuit or the key branch of the CDU, an accurate and independent power source can be provided, and combined with the fine tuning of high-precision electric regulating valves, the flow and pressure of the CDU to each terminal circuit can be actively distributed accurately, quickly and independently, thereby providing the highest level of protection for the most demanding cooling requirements, ensuring that the chip can still maintain extremely low and stable temperatures under continuous high load.

[0028] During the entire system operation, continuous, comprehensive and high-precision data acquisition is the basis for achieving all the intelligent controls described above. Therefore, a large number of sensors need to be arranged in the primary and secondary circuits to monitor key parameters in real time, including but not limited to the temperature of the water cooling system and the coolant at key points (such as the inlet and outlet of the cold machine evaporator and condenser, the inlet and outlet of the primary and secondary sides of the CDU, the branch points of the main pipeline, the inlet of the server cabinet, and even the inlet and outlet of the cold plate of the key server), flow (measured by electromagnetic flow meters or ultrasonic flow meters), pressure (pressure and pressure difference at key nodes of the pipe network), liquid level (expansion tank, CDU water tank), and conductivity / purity of the coolant, etc. These data are uploaded to the respective edge controllers and upper computers / platform layers in real time, constituting the perception basis of the digital twin model, providing comprehensive data support for model calculation, rule judgment, system state monitoring, fault warning and energy efficiency evaluation.

[0029] In summary, the application uses event-driven model predictive control on the primary side for forward-looking system-level energy optimization, uses real-time rule-based control on the secondary side for fast and accurate device-level flow distribution, and uses the CDU outlet temperature as a key physical quantity to closely coordinate the two control layers at the heat exchange interface, thus forming a complete "perception-prediction-optimization-execution-feedback" intelligent closed loop. This method takes advantage of the dynamic characteristics of different parts of the system, effectively solves the inherent contradiction between the response lag of the primary side and the demand transient of the secondary side, and can fully meet the heat dissipation demand of high-heat-density server equipment at the end, maintain the long-term stability of its temperature (thus guaranteeing the completion quality of computing tasks and the service life of hardware), avoid "overcooling" through fine scheduling of the cold source equipment, eliminate energy waste caused by delayed response using load prediction information, optimize the average energy efficiency ratio of the entire liquid cooling system, and ultimately achieve the dual goals of reducing data center operating costs and improving overall sustainability.

[0030] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present application are only used to explain the relative positional relationship, motion condition, etc. between components in a certain posture (as shown in the drawings), and if the certain posture changes, the directional indications will also change accordingly.

[0031] In addition, the descriptions such as "first", "second", "one" and the like in the present application are only for the purpose of description, and cannot be understood as indicating or implying the relative importance of the indicated technical features or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" can explicitly or implicitly include at least one of the features. In the description of the present application, the meaning of "multiple" is at least two, such as two, three, etc., unless otherwise specifically limited.

[0032] In the present application, unless otherwise specifically defined and limited, the terms "connection", "fixation" and the like should be understood broadly, for example, "fixation" can be fixed connection, or detachable connection, or integral; can be mechanical connection, or electrical connection; can be direct connection, or indirect connection through intermediate medium; can be internal connection of two elements or interaction relationship between two elements, unless otherwise specifically limited. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0033] In addition, the technical solutions of each embodiment of the present application can be combined with each other, but it must be based on the fact that a person skilled in the art can realize it, and when the combination of technical solutions appears contradictory or unachievable, it should be considered that the combination of technical solutions does not exist, nor is it within the scope of protection required by the present application.

Claims

1. An energy-saving intelligent control method for a high-heat-density data center liquid cooling system, characterized in that, The coordination linkage between the predictive optimization control of the primary side cold source circuit and the real-time control of the secondary side coolant distribution circuit is achieved by performing a cycle step, which comprises: S1: for the primary side cold source circuit of the liquid cooling system, a predictive control model is established to minimize the total energy consumption of the equipment in the future prediction interval, while meeting the cooling demand and various constraints of the system; S2: for the secondary side coolant distribution circuit of the liquid cooling system, a preset rule control model is established to dynamically distribute the coolant flow based on the real-time power consumption of the server equipment, with the goal of stabilizing the temperature of the server equipment; S3: based on the preset rule control model, the equipment in the secondary side coolant distribution circuit is controlled in real time, and by running the predictive control model, the operation scheme sequence of each control variable of the primary side cold source circuit in the future prediction interval is obtained, and the equipment in the primary side cold source circuit is controlled based on the operation scheme sequence; S4: during the execution of the operation scheme sequence, the outlet temperature of the coolant distribution unit in the secondary side coolant distribution circuit is monitored in real time; S5: the outlet temperature of the coolant distribution unit is used as a trigger condition, and when the outlet temperature exceeds the preset range, the operation scheme sequence of the primary side cold source circuit is updated by triggering and returning to S1.

2. The energy-saving intelligent control method of a high-heat-density data center liquid cooling system according to claim 1, characterized in that, In the S1 step, the data relied on for establishing the predictive control model includes the data center computing task allocation plan and / or the predicted computing load, and the predictive control model is established by mechanism modeling or system identification method.

3. The energy-saving intelligent control method of a high-heat-density data center liquid cooling system according to claim 1, characterized in that, In the S2 step, the secondary side coolant distribution circuit includes a coolant distribution unit and an end circuit arranged in the data center, and the control variables of the preset rule control model for coolant flow include at least one of the frequency conversion water pump frequency of the coolant distribution unit, the frequency conversion water pump frequency of the end circuit, and the opening degree of the electric valve of the end circuit.

4. The energy-saving intelligent control method of a high-heat-density data center liquid cooling system according to claim 3, characterized in that, When the coolant flow is distributed by adjusting the valve opening degree, for the end circuit with stable thermal load, a flow balance valve is used for flow control.

5. The energy-saving intelligent control method of a high-heat-density data center liquid cooling system according to claim 4, characterized in that, When the coolant flow is distributed by adjusting the valve opening degree, for the end circuit with large thermal load variation, a pressure-independent control valve is used for flow control.

6. The energy-saving intelligent control method of a high-heat-density data center liquid cooling system according to claim 3, characterized in that, When it is necessary to actively distribute the coolant flow of each end circuit inside the coolant distribution unit, a frequency conversion water pump and an electric valve are used for flow control.

7. The energy-saving intelligent control method of a high-heat-density data center liquid cooling system according to claim 1, wherein, In the S3 step, the control variables include at least one of the chiller load rate, the frequency conversion water pump frequency, and the cooling tower fan speed.

8. The energy-saving intelligent control method of a high-heat-density data center liquid cooling system according to claim 1, wherein, In the S5 step, it also includes: When the outlet temperature does not exceed the preset range, then after a preset fixed time interval is reached during the execution of the operation scheme sequence, the S1 step is triggered and returned, and before a preset fixed time interval is reached, the S5 step is returned.

9. The energy-saving intelligent control method of a high-heat-density data center liquid cooling system according to claim 1, wherein, The secondary side coolant distribution circuit contains multiple coolant distribution units, and the primary side cold source circuit provides cooling capacity for the multiple coolant distribution units, forming a one-to-many liquid cooling network.

Images