A data center computer room liquid cooling cycle energy-saving control method and system

Abstract

The present invention provides a data center computer room liquid cooling cycle energy-saving control method and system. First, a computer room cellular network is established, where each cellular node includes multiple IT cabinets. Time-of-use electricity prices are obtained. Then, a real-time data matrix of each cellular node is obtained and DT-CWT decomposition is performed. A transient temperature curve and a steady-state temperature curve are obtained, and a heat flux density distribution diagram is established. Real-time waste heat marking is performed. Based on the real-time waste heat marking and the time-of-use electricity price, hierarchical and time-of-use recovery is performed on the heat generated by each cellular node. A cold storage strategy is implemented. Then, PID control or MPC control is selectively used to obtain a liquid cooling demand matrix. PID control or SMC control is performed according to the real-time temperature deviation. A step adjustment instruction or a steady-state adjustment instruction is generated to adjust the liquid cooling demand matrix. Finally, a cold storage module or a liquid cooling cycle module is activated to perform real-time liquid cooling and heat dissipation. This method can improve energy saving while ensuring liquid cooling and heat dissipation effects, thereby reducing the total electricity price.

Description

Technical Field

[0001] The present invention relates to the field of liquid cooling and heat dissipation technology, and in particular to a liquid cooling cycle energy-saving control method and system for a data center computer room. Background Art

[0002] At a time when liquid cooling technology in data center rooms is developing rapidly, with the large-scale deployment of high-density computing equipment and the increasingly urgent need for green energy conservation, liquid cooling control faces many technical challenges that need to be addressed. Traditional liquid cooling control mainly relies on fixed parameter control or manual adjustment of cooling parameters, and consumes a lot of electricity.

[0003] On the one hand, relying on fixed parameter control or adjusting cooling parameters through manual experience is often one-sided. Manual experience cannot perceive the transient changes in the thermal load of IT cabinets in the computer room in real time, such as the instantaneous temperature rise caused by the sudden load of the GPU cluster, resulting in delayed cooling response and frequent local hotspots. Secondly, relying on fixed parameter control often focuses on local nodes and uses single-channel control, resulting in poor liquid cooling effect.

[0004] On the other hand, when liquid cooling units are used to cool IT cabinets in a computer room, there is often a lack of a liquid cooling linkage mechanism and a waste heat utilization mechanism for time-of-use electricity prices, resulting in high electricity consumption and energy waste.

[0005] Therefore, existing liquid cooling control technology is unable to cope with energy saving and liquid cooling heat dissipation effects at the same time. Summary of the Invention

[0006] In view of the above-mentioned problems, in combination with the first aspect of the present invention, an embodiment of the present invention provides a method for controlling energy saving of a liquid cooling cycle in a data center computer room, the method comprising:

[0007] Establishing a cellular network in the computer room, where each cellular node includes multiple IT cabinets, assigning node numbers to each cellular node, and obtaining time-of-use electricity prices;

[0008] Obtain the real-time data matrix of each honeycomb node, decompose the real-time data matrix based on DT-CWT, obtain the transient temperature curve and steady-state temperature curve of each honeycomb node, and establish a heat flux density distribution map based on the steady-state temperature curve and honeycomb node;

[0009] Based on the heat flux density distribution map, each cell node is marked for waste heat in real time. Based on the real-time waste heat marking and time-of-use electricity prices, heat generated by multiple IT cabinets in each cell node during operation is recovered in a hierarchical and time-of-use manner. A cooling storage strategy is implemented based on the steady-state temperature curve and time-of-use electricity prices.

[0010] Selecting to use PID control or MPC control based on a transient temperature curve and a steady-state temperature curve to obtain a liquid cooling demand matrix, wherein the liquid cooling demand matrix includes a priority sequence;

[0011] Acquire real-time temperature deviation based on real-time data matrix, perform PID control or SMC control according to the real-time temperature deviation, generate step adjustment instruction or steady-state adjustment instruction, and adjust the liquid cooling demand matrix according to the step adjustment instruction or steady-state adjustment instruction;

[0012] Based on the adjusted liquid cooling demand matrix and the time-of-use electricity price, the cold storage module or the liquid cooling circulation module is started to perform real-time liquid cooling on each cellular node.

[0013] In another aspect, an embodiment of the present invention further provides a data center computer room liquid cooling cycle energy-saving control system, comprising:

[0014] A cellular module, configured to establish a cellular network comprising a plurality of cellular nodes;

[0015] An acquisition module, comprising a sensor module and an electricity price module;

[0016] A sensor module is used to obtain a real-time data matrix, and the sensor module is correspondingly deployed on an IT cabinet;

[0017] Electricity price module, used to obtain time-of-use electricity prices;

[0018] A storage module, comprising a cold storage module and a heat storage module;

[0019] The cold storage module is used to implement the cold storage strategy, store the cold energy, and release the cold energy, thereby providing real-time liquid cooling for each cellular node;

[0020] The heat storage module is used to perform hierarchical and time-sharing recovery of heat generated by multiple IT cabinets in each cellular node during operation, and can provide auxiliary heat to the external preheating module;

[0021] Liquid cooling circulation module, used for liquid cooling and heat dissipation of cellular nodes;

[0022] a processing module, configured to obtain a liquid cooling demand matrix and adjust the liquid cooling demand matrix based on a step adjustment instruction or a steady-state adjustment instruction;

[0023] The control module is used to start the cold storage module or the liquid cooling circulation module to perform real-time liquid cooling and heat dissipation on each cellular node according to the adjusted liquid cooling demand matrix and the time-of-use electricity price.

[0024] Based on the above aspects, the embodiment of the present application realizes that liquid cooling technology can simultaneously meet the needs of energy saving and liquid cooling effect. By constructing a cellular network node control architecture, the method is not only focused on the local area, but also on the global area, which can prevent local overcooling. Secondly, based on the liquid cooling linkage mechanism and waste heat utilization mechanism of time-of-use electricity prices, cold storage can be carried out during the period of low electricity prices, and liquid cooling and heat dissipation can be carried out through the stored cold energy during the period of peak electricity prices. The heat generated by the IT cabinet when it is working can be recovered in a graded and time-sharing manner, and the waste heat can be used in the external preheating module to provide heat for the outside, thereby saving electricity and reducing the ineffective waste of energy. It effectively solves the problems of high power consumption and energy waste in traditional liquid cooling control technology. In addition, through the dual-loop control of steady state and transient, transient temperature rise can be timely discovered, and sudden temperature rise and steady-state temperature can be accurately responded to, thereby improving the effect of liquid cooling and heat dissipation. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 This is a schematic diagram of the execution flow of a data center computer room liquid cooling cycle energy-saving control method provided by an embodiment of the present invention;

[0026] Figure 2 This is a schematic diagram of the execution flow of establishing a data center computer room cellular network in a liquid cooling cycle energy-saving control method provided by an embodiment of the present invention;

[0027] Figure 3 This is a schematic diagram of a liquid cooling cycle energy-saving control system for a data center computer room provided by an embodiment of the present invention;

[0028] Figure 4 This is a schematic diagram of a storage module in a liquid cooling cycle energy-saving control system for a data center computer room provided by an embodiment of the present invention;

[0029] Figure 5 The present invention provides a schematic diagram of an acquisition module in a liquid cooling cycle energy-saving control system for a data center computer room. DETAILED DESCRIPTION

[0030] The present invention will be described in detail below with reference to the accompanying drawings. Figure 1 This is a schematic diagram of the execution flow of a data center computer room liquid cooling cycle energy-saving control method provided by an embodiment of the present invention. The data center computer room liquid cooling cycle energy-saving control method is introduced in detail below.

[0031] Step S1: Establish a cellular network in a computer room, wherein each cellular node in the cellular network includes a plurality of IT cabinets, label each cellular node, and obtain a time-of-use electricity price.

[0032] It can be understood that by establishing a cellular network, the global temperature control can be decomposed into the coordinated response of local independent nodes by dynamically dividing the computer room into multiple hexagonal nodes and matching the hybrid control strategy of MPC control and PID control. This not only achieves the rapid positioning and suppression of sudden heat loads, but also ensures the balance of the overall thermal environment through modeling of the thermal coupling relationship between nodes. Each cellular node adaptively switches between MPC control or PID control according to the transient temperature rise rate, and generates a liquid cooling demand matrix. It can then choose to switch between SMC control or PID control according to the disturbance value to correct the disturbance. While eliminating local hot spots, it dynamically coordinates the cooling resource allocation of adjacent nodes to avoid the chain temperature fluctuations caused by thermal interference in traditional global unified control.

[0033] Furthermore, the topology of the cellular network can also provide spatial priority mapping for cooling scheduling driven by time-of-use electricity prices, so that the release of stored cooling capacity can match the needs of high-load areas, thereby achieving energy efficiency optimization at the global level. Based on this multi-scale control mechanism from local to global, the adaptability to complex thermal environments can be improved.

[0034] Specifically, time-of-use electricity price is an electricity price mechanism that is dynamically adjusted based on the relationship between electricity supply and demand. It divides the 24 hours of the day into peak period, off-peak period and average period. The peak period is usually from 18:00 to 22:00, when electricity demand is high and electricity prices are also high. Reducing the power consumption of liquid cooling control during peak period can greatly save electricity prices. The off-peak period is usually from 00:00 to 6:00, when electricity demand is low and electricity prices are the lowest. At this time, the liquid cooling circulation module can be started at full power to dissipate heat for the IT cabinet and store cold energy based on the cold storage module, which can save electricity consumption. The average period is the rest of the time except the peak period and off-peak period, and the electricity price is between the peak period and off-peak period.

[0035] Figure 2 It is a schematic diagram of the execution flow of establishing a data center computer room cellular network in a liquid cooling cycle energy-saving control method provided by an embodiment of the present invention.

[0036] In this embodiment, establishing a cellular network in a computer room specifically includes the following steps:

[0037] Step S11 , obtaining the physical locations of all IT cabinets and sensor modules deployed on the IT cabinets in the computer room, wherein the physical locations include the three-dimensional coordinates of each IT cabinet and each sensor module.

[0038] Step S12: synchronize the timestamps of all sensor modules through a unified protocol, and use cubic spline interpolation to fill in the fault data appearing in all sensor modules.

[0039] In step S13, the physical positions of all IT cabinets and sensor modules are mapped to a unified coordinate system based on the ICP algorithm. The physical position of the sensor module is used as the reference point to generate a minimum circumscribed triangle. The angle of the internal angle of the minimum circumscribed triangle is greater than or equal to 30 degrees, and a triangular network is formed based on the minimum circumscribed triangle.

[0040] Step S14: based on the duality relationship of the triangular network, convert the vertex of each triangle in the triangular network into a hexagonal honeycomb node, wherein the honeycomb node shares at least one edge with an adjacent node.

[0041] Step S15 , calculating the thermal coupling coefficient between the honeycomb nodes, and performing a grid optimization operation based on the thermal coupling coefficient.

[0042] In this embodiment, it is assumed that a computer room in a data center is 20 meters long and 10 meters wide, with 30 cabinets deployed. The physical locations of the 30 cabinets are obtained, including three-dimensional coordinates. For example, the location of cabinet A1 is x=2.3m, y=5.1m, and z=0m. A Delaunay triangulation network is constructed with the sensor module locations as nodes, ensuring that the internal angles of all triangles are greater than or equal to 30 degrees to avoid the generation of sharp and narrow units. For high-density areas, virtual nodes can be forcibly inserted to improve local mesh resolution. Then, based on the duality relationship of Delaunay triangulation, each triangle vertex is converted into a hexagonal honeycomb node. For example, the side length of a honeycomb node is 1.2m, covering cabinets A1-A3. Each honeycomb node shares at least one edge with an adjacent honeycomb node. The common edges of the honeycomb nodes can then be smoothed to eliminate jagged edges and reduce disturbances in the fluid boundary layer. Finally, the thermal coupling coefficient between the honeycomb nodes is calculated, and mesh optimization operations are performed based on the thermal coupling coefficient.

[0043] Furthermore, Delaunay triangulation generates a triangular mesh with the maximum minimum angle by using the sensor module positions in the computer room as nodes, ensuring mesh uniformity and providing a geometrically stable topological basis for subsequent hexagonal transformation. The triangular network after triangulation is strictly dual to the honeycomb hexagon, ensuring that the distance from each hexagonal node to each edge is equal, providing a uniform spatial benchmark for subsequent dynamic liquid cooling distribution.

[0044] In this embodiment, the grid optimization operation includes:

[0045] Step S151 , traverse all adjacent cellular nodes, and perform a strong coupling threshold judgment based on the thermal coupling coefficient between the cellular nodes, wherein the strong coupling threshold is 5%.

[0046] Step S151-1: If the thermal coupling coefficient between the two groups of cellular nodes is less than or equal to 5%, no action is taken.

[0047] In step S151-2, if the thermal coupling coefficient between the two groups of cellular nodes is greater than 5%, the two groups of cellular nodes are deemed to be strongly coupled nodes, and the strongly coupled nodes are split into three cellular nodes. The process is repeated until the thermal coupling coefficient between the two groups of cellular nodes is less than or equal to 5%.

[0048] Step S16: label the cellular nodes that have completed the grid optimization operation and contain at least one IT cabinet. If the node does not contain an IT cabinet, it is marked as invalid and will no longer participate in subsequent processing.

[0049] It can be understood that by quantifying the heat transfer ratio between adjacent honeycomb nodes, it can be ensured that local temperature control will not cause a chain reaction across nodes, preventing global temperature imbalance caused by thermal interference. For example, when the thermal coupling coefficient of an adjacent honeycomb node is detected to exceed the limit of 6%, it is identified as a strongly coupled node and the strongly coupled node is split to prevent the local temperature rise from spreading to regional overheating.

[0050] Specifically, the node label can be expressed as "G01", "G02", "...", etc., and each cellular node is given a unique identity through the node label to achieve accurate mapping and rapid positioning of the cellular network. The node label can be directly associated with the physical location of the cabinet. For example, the cellular node of "G01" covers cabinets A1-A3, and the specific location of cabinets A1-A3 can be directly found based on the node label of "G01". In this way, the hot spot area can be quickly located and the cellular node can be given priority during liquid cooling. As for the failure label, when the cellular network is constructed, the cellular nodes in the cellular network do not contain IT cabinets. The nodes are marked as failed, so that the nodes with the failure label no longer participate in subsequent processing, thereby saving unnecessary computing resource loss.

[0051] Step S2: obtain the real-time data matrix of each honeycomb node, decompose the real-time data matrix based on DT-CWT, obtain the transient temperature curve and steady-state temperature curve of each honeycomb node, and establish a heat flux density distribution map based on the steady-state temperature curve and the honeycomb node.

[0052] Specifically, the real-time data matrix contains temperature signal data, and the temperature signal data is denoised. In addition, the real-time data matrix can also include real-time flow data, coolant parameters, etc., so that users can expand other additional functions based on the real-time data matrix.

[0053] In this embodiment, step S2 includes:

[0054] Step S21, decompose the temperature signal data based on DT-CWT into 5 layers, the first to second layers correspond to high-frequency sub-bands, the high-frequency sub-bands represent temperature signal data with a frequency domain range of 0.1 Hz to 10 Hz, and the third to fifth layers correspond to low-frequency sub-bands, the low-frequency sub-bands represent temperature signal data with a frequency domain range of less than 0.1 Hz.

[0055] In this embodiment, the high-frequency sub-band is represented by the capture of temperature mutations at the second level, and the high-frequency sub-band retains the details of the temperature mutation, while the low-frequency sub-band is represented by the capture of long-term thermal trends at the minute level or above, and the low-frequency sub-band retains the slowly changing trend of the thermal environment. After defining the frequency domain coverage range, the signal characteristics are separated. Among them, the transient signal is represented by a high-frequency, short-term, high-energy pulse signal, such as the instantaneous temperature rise caused by the sudden task load of the IT cabinet, and the steady-state signal is represented by a low-frequency, long-term, and evenly distributed energy signal, such as the basic heat dissipation requirements of the IT cabinet.

[0056] It can be understood that by performing multi-scale decomposition of temperature signal data through DT-CWT, the global frequency domain averaging of traditional Fourier transform can be avoided and the extraction of local time-frequency features can be achieved.

[0057] Step S22 , performing inverse transformation based on the high frequency subband coefficients of the first and second layers to obtain a temperature transient component, and obtaining a transient temperature curve based on the temperature transient component, wherein the transient temperature curve represents a sudden temperature rise of the cellular node based on time.

[0058] Step S23 , performing inverse transformation based on the low-frequency subband coefficients of layers 3 to 5 to obtain a temperature steady-state component, and obtaining a steady-state temperature curve based on the temperature steady-state component, wherein the steady-state temperature curve represents the operating temperature of the cellular node based on time.

[0059] In this embodiment, an inverse transformation is performed based on the high-frequency subband coefficients of the 1st to 2nd layers to reconstruct the short-term temperature rise signal, and an inverse transformation is performed based on the low-frequency subband coefficients of the 3rd to 5th layers to reconstruct the long-term thermal equilibrium signal. The horizontal axis of the transient temperature curve and the steady-state temperature curve is time, and the vertical axis is temperature.

[0060] Step S24: Calculate the dynamic temperature gradient group between adjacent honeycomb nodes based on the steady-state temperature curve, and calculate the dynamic heat flux density group of each honeycomb node based on the dynamic temperature gradient group. Obtain the average heat flux density according to the dynamic heat flux density group, and map the average heat flux density to each honeycomb node in the computer room honeycomb network to generate a heat flux density distribution map.

[0061] As can be understood, DT-CWT (dual-tree complex wavelet transform) uses two parallel real wavelet trees, one of which processes the real part of the signal and the other processes the imaginary part, to generate complex-valued wavelet coefficients, overcoming the directional ambiguity and spectral leakage problems of traditional wavelet transform, and has approximate translation invariance, making it suitable for processing non-stationary signals.

[0062] Furthermore, DT-CWT (Dual-Tree Complex Wavelet Transform) uses multi-scale frequency domain decomposition to separate temperature signal data into transient components (high-frequency subband, 0.1Hz-10Hz) and steady-state components (low-frequency subband, less than 0.1Hz). The transient component can capture rapid temperature rises caused by sudden tasks in cellular nodes, such as those caused by them, and provide a response basis for MPC control. The steady-state component represents the long-term stable basic heat load of cellular nodes, which is used to support subsequent waste heat time-sharing and graded recovery and cold storage strategies. In addition, the dual-tree structure of the complex filter bank in DT-CWT (Dual-Tree Complex Wavelet Transform) can eliminate the frequency band aliasing problem of traditional wavelet transform.

[0063] Step S3: Mark the waste heat of each honeycomb node in real time based on the heat flux density distribution map, perform hierarchical and time-sharing recovery of the heat generated by multiple IT cabinets in each honeycomb node during operation based on the real-time waste heat marking and time-sharing electricity prices, and implement a cold storage strategy based on the steady-state temperature curve and time-sharing electricity prices.

[0064] Specifically, the heat flux density distribution diagram includes the heat flux color intensity of each honeycomb node. The heat flux density distribution diagram is displayed in the hexagonal shape of the honeycomb node, and the corresponding node number and average heat flux density are mapped in each honeycomb node. Each honeycomb node is filled with the corresponding heat flux color intensity.

[0065] In this embodiment, the hierarchical and time-sharing recycling includes:

[0066] Step S31 : Mark each honeycomb node for waste heat in real time based on the heat flow color intensity. The waste heat mark is updated every 1 minute. The real-time waste heat mark includes a high heat mark, a medium heat mark, and a low heat mark.

[0067] Specifically, the heat generated by multiple IT cabinets in each cellular area during operation is graded and labeled based on high heat marks, medium heat marks, and low heat marks, so as to better recover and utilize waste heat in a graded manner.

[0068] Step S31-1: If the honeycomb node is marked as high heat, the high heat generated by the honeycomb node is recovered to the heat storage module, and the heat storage module can provide auxiliary heat to the external preheating module.

[0069] In this embodiment, high-calorie high-temperature coolant can be introduced into the heat storage module through connecting components such as a three-way valve, and then auxiliary heating can be performed through an external preheating module. For example, heating can be provided for office areas and living areas. Compared with traditional gas boilers, the electricity cost of heating can be reduced. In actual applications, waste heat recovery can be selectively enabled or disabled according to the season. For example, in summer, the temperature is high and no heating is required, so the waste heat recovery can be turned off. In winter, the temperature is low and heating is required, so the waste heat recovery can be started to perform auxiliary heating and save electricity.

[0070] Step S31 - 2 : If the honeycomb node is marked as medium heat, the medium heat energy generated by the honeycomb node is recovered to the heat storage module.

[0071] Step S31-2-1: If the time-of-use electricity price is at a low point, the heat storage module heats the medium heat until the heat reaches a high point and then stops.

[0072] It is understandable that when the time-of-use electricity price is at a low point, the electricity price is lower. By heating the medium-temperature coolant with medium heat during the period of lower electricity price, the electricity consumption can be reduced, while in winter, heat can be released during peak hours to reduce electricity costs.

[0073] Step S31-2-2: If the time-of-use electricity price is in the peak period or the average period, the medium heat is transferred to the cold storage module.

[0074] Step S31 - 3 : If the honeycomb node is marked as low heat, the low heat generated by the honeycomb node is transferred to the cold storage module.

[0075] It is understandable that when the time-of-use electricity price is at peak or average times, the electricity cost is higher, so the medium-temperature coolant with medium heat is not heated. Instead, the medium-temperature coolant with medium heat is heated and transported to the cold storage module. The medium-temperature coolant with medium heat is heated and cooled through the cold storage strategy, and then stored as cold energy.

[0076] In this embodiment, the cold storage strategy includes:

[0077] The average temperature of IT cabinets in the total cellular network is obtained based on the steady-state temperature curve. The cold storage threshold for cooling and storing cold in the cold storage module is proportional to the average temperature of the IT cabinet. The cold storage module implements a cold storage strategy based on the average temperature. When the time-of-use electricity price is at a low period, the cold storage module cools and stores the medium and low heat transmitted by the thermal storage module. If the cold storage in the cold storage module reaches the cold storage threshold, the cold storage module stops working.

[0078] In this embodiment, the cold storage threshold required to be stored by the cold storage module is obtained based on the average temperature of the IT cabinets in the total cellular network, and the cold storage threshold is proportional to the average temperature of the IT cabinet, so that the cold storage module can store redundant cold to prevent sudden temperature rise abnormalities. For example, when the AI ​​training task suddenly increases and the heat load of the cellular node exceeds expectations, the redundant cold can be quickly released to avoid local overheating. Secondly, if the power grid is interrupted, the redundant cold can still be temporarily called for liquid cooling to ensure that users have sufficient time to start the backup power supply.

[0079] Step S4: Select PID control or MPC control based on the transient temperature curve and the steady-state temperature curve to obtain a liquid cooling demand matrix, wherein the liquid cooling demand matrix includes a priority sequence.

[0080] In this embodiment, step S4 includes:

[0081] Step S41, calculate the transient temperature rise rate of each honeycomb node based on the transient temperature curve, calculate the steady-state temperature rise rate of each honeycomb node based on the steady-state temperature curve, and obtain the temperature rise limit according to the transient temperature rise rate and the steady-state temperature rise rate. When performing the calculation, the weight of the transient temperature rise rate is greater than the steady-state temperature rise rate.

[0082] In step S41-1, if the transient temperature rise rate is greater than or equal to the temperature rise limit, MPC control is selected, and the liquid cooling demand of the cellular node in the next 30 seconds is actively predicted based on the MPC control to obtain MPC adjustment data.

[0083] Step S41 - 2 : If the transient temperature rise rate is less than the temperature rise limit, PID control is selected to passively obtain PID adjustment data of the current liquid cooling demand of the cellular node based on the PID control.

[0084] Step S42 : obtaining a liquid cooling demand matrix based on the PID adjustment data and the MPC adjustment data, wherein the priority sequence in the liquid cooling demand matrix is ​​sorted based on the liquid cooling demand of the cellular node.

[0085] It should be noted that for high-load cellular nodes, such as those with a temperature rise rate greater than or equal to 2 degrees Celsius, MPC (model predictive control) can achieve forward-looking control by constructing a thermodynamic dynamic model to predict the temperature trend in the next 30 seconds. After predicting the temperature trend in the next 30 seconds, the coolant flow rate can be actively adjusted to complete intervention before the temperature rise gets out of control. The response time is faster than PID control and is suitable for cellular nodes with a transient temperature rise rate greater than or equal to the temperature rise limit.

[0086] It should be noted that for low-load cellular nodes, such as those with a temperature rise rate of less than 2 degrees Celsius, PID control does not require the prediction of complex models. It only needs to calculate and output the control quantity that needs to adjust the coolant flow, which is suitable for the real-time requirements of low-load scenarios.

[0087] Specifically, the PID control adopts an improved differential-first PID structure. The selected parameters are a proportional band of 5-50%, an integral time of 10-300 seconds, and a dead zone setting of plus or minus 0.3 degrees Celsius to avoid frequent adjustments. When the deviation continues to exceed the dead zone, the integral action is suspended to prevent excessive adjustment of the coolant flow.

[0088] Furthermore, through the dynamic switching strategy of MPC control and PID control, both cellular nodes with sudden loads and cellular nodes with low loads can be taken into account. MPC control targets the critical temperature rise problem of high-load cellular nodes, while PID control maintains a large steady-state area with low energy consumption, thereby saving energy while improving the efficiency of liquid cooling.

[0089] Step S5, obtaining the real-time temperature deviation based on the real-time data matrix, performing PID control or SMC control according to the real-time temperature deviation, generating a step adjustment instruction or a steady-state adjustment instruction, and adjusting the liquid cooling demand matrix according to the step adjustment instruction or the steady-state adjustment instruction.

[0090] Specifically, step S5 includes:

[0091] Step S51 : obtaining the real-time temperature deviation of each honeycomb node, importing the real-time temperature deviation of each honeycomb node into a step-transient switching function, and obtaining a disturbance value based on the step-transient switching function.

[0092] Specifically, the step-transient switching function is expressed as:

[0093] ;

[0094] in, Expressed as the perturbation value, Expressed as real-time temperature deviation change rate, It is expressed as real-time temperature deviation. For example, if the temperature of a cell node rises from +0.2 degrees Celsius to +0.8 degrees Celsius within 1 second, , and the disturbance value is expressed as .

[0095] It should be noted that the step-transient switching function is used to dynamically determine the disturbance intensity and trigger the switching of the control mode. Its essence is to integrate the real-time temperature deviation and its rate of change into a scalar indicator through a mathematical function. When the indicator exceeds the preset threshold (preset threshold = 0.3), it is determined to be a transient disturbance, which in turn triggers a step-type gain adjustment, thereby eliminating or minimizing the transient disturbance, making the MPC control more precise and allowing the high-load cellular node to quickly return to stability. When the indicator is less than or equal to the preset threshold, it is determined to be a steady-state disturbance, and then the PID control is fine-tuned to eliminate or minimize the steady-state disturbance, making the PID control more precise.

[0096] In step S51 - 1 , if the disturbance value is greater than 0.3, it is considered as a transient disturbance, and SMC control is performed. A step adjustment instruction is generated based on the SMC control. The step adjustment instruction is expressed as a gain adjustment of the liquid cooling demand matrix.

[0097] In step S51-2, if the disturbance value is less than or equal to 0.3, it is regarded as a steady-state disturbance, and PID control is performed. A steady-state adjustment instruction is generated based on the PID control. The steady-state adjustment instruction is expressed as a fine-tuning of the liquid cooling demand matrix.

[0098] Step S6: Based on the adjusted liquid cooling demand matrix and the time-of-use electricity price, the cold storage module or the liquid cooling circulation module is started to perform real-time liquid cooling and heat dissipation on each cellular node.

[0099] When the time-of-use electricity price is at a peak period, the cold storage module is started to release cold energy. The cold storage module transmits cold energy to each honeycomb node for liquid cooling and heat dissipation based on the liquid cooling demand matrix. If the maximum cold energy in the cold storage module drops to 35%, the power of the cold storage module is reduced to 50%, and the liquid cooling circulation module is started to assist in liquid cooling and heat dissipation for each honeycomb node. If the maximum cold energy in the cold storage module drops to 20%, the cold storage module is shut down, and each honeycomb node is liquid cooled and heat dissipated separately through the liquid cooling circulation module. When the time-of-use electricity price is at a low period, the cold storage strategy is restarted to replenish cold energy for the cold storage module.

[0100] When the time-of-use electricity price is at the average period, the cold storage module and the liquid cooling circulation module are started at the same time to perform liquid cooling on each cellular node;

[0101] When the time-of-use electricity price is at a low period, the liquid cooling circulation module is started, and the liquid cooling circulation module performs liquid cooling on each cellular node based on the liquid cooling demand matrix.

[0102] Furthermore, the method also includes: constructing a reinforcement learning model, importing the liquid cooling demand matrix in the past 24 hours and the real-time data matrix of each cellular node as historical training data into the reinforcement learning model, the reinforcement learning model obtains the flow compensation coefficient based on the TD3 algorithm and historical training data, and optimizes PID control and MPC control based on the flow compensation coefficient.

[0103] Figure 3 A schematic diagram of a liquid cooling cycle energy-saving control system for a data center computer room provided by some embodiments of the present application that can implement the concept of the present application is shown. Figure 4 A schematic diagram of a storage module in a liquid cooling cycle energy-saving control system for a data center computer room, provided by some embodiments of the present application and capable of realizing the concept of the present application, is shown. Figure 5 A schematic diagram of an acquisition module in a liquid cooling cycle energy-saving control system for a data center computer room provided by some embodiments of the present application, which can realize the concept of the present application, is shown.

[0104] Specifically, a data center computer room liquid cooling cycle energy-saving control system includes:

[0105] A cellular module, configured to establish a cellular network comprising a plurality of cellular nodes;

[0106] An acquisition module, comprising a sensor module and an electricity price module;

[0107] A sensor module is used to obtain a real-time data matrix, and the sensor module is correspondingly deployed on an IT cabinet;

[0108] Electricity price module, used to obtain time-of-use electricity prices;

[0109] A storage module, comprising a cold storage module and a heat storage module;

[0110] The cold storage module is used to implement the cold storage strategy, store the cold energy, and release the cold energy, thereby providing real-time liquid cooling for each cellular node;

[0111] The heat storage module is used to perform hierarchical and time-sharing recovery of heat generated by multiple IT cabinets in each cellular node during operation, and can provide auxiliary heat to the external preheating module;

[0112] Liquid cooling circulation module, used for liquid cooling and heat dissipation of cellular nodes;

[0113] a processing module, configured to obtain a liquid cooling demand matrix and adjust the liquid cooling demand matrix based on a step adjustment instruction or a steady-state adjustment instruction;

[0114] The control module is used to start the cold storage module or the liquid cooling circulation module to perform real-time liquid cooling and heat dissipation on each cellular node according to the adjusted liquid cooling demand matrix and the time-of-use electricity price.

[0115] The specific usage and function of this embodiment are described below:

[0116] First, a computer room cellular network is established. Each cellular node in the computer room cellular network includes multiple IT cabinets. Each cellular node is labeled and the time-of-use electricity price is obtained. Then, the real-time data matrix of each cellular node is obtained and decomposed based on DT-CWT. The transient temperature curve and steady-state temperature curve of each cellular node are obtained, and a heat flux density distribution map is established. Then, based on the heat flux density distribution map, real-time waste heat marking is performed on each cellular node. Based on the real-time waste heat marking and time-of-use electricity price, the heat generated during the operation of multiple IT cabinets in each cellular node is recycled in a hierarchical and time-of-use manner, thereby reusing the waste heat energy. A cooling storage strategy is implemented according to the steady-state temperature curve and the time-of-use electricity price. Then, the use of the heat storage strategy is selected based on the transient temperature curve and the steady-state temperature curve. Use PID control or MPC control to obtain the liquid cooling demand matrix. Through PID control or SMC control, step adjustment instructions or steady-state adjustment instructions can be generated to adjust the liquid cooling demand matrix. Based on the adjusted liquid cooling demand matrix and the time-of-use electricity price, the cold storage module or the liquid cooling circulation module is started to perform real-time liquid cooling and heat dissipation on each cellular node. This method can regulate the liquid cooling control based on the time-of-use electricity price in different electricity price periods. It can also recover the waste heat through graded and time-sharing recovery, and recover the waste heat and use it for external heating, so that the waste heat energy can be reused, saving electricity. In addition, through steady-state and transient dual-loop control, transient temperature rise can be timely discovered, and sudden temperature rise and steady-state temperature can be accurately responded to, thereby improving the effect of liquid cooling and heat dissipation.

[0117] In addition, an embodiment of the present invention further provides an electronic device, including:

[0118] At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method proposed in the first embodiment of the present invention.

[0119] The following is a detailed introduction to the various components of electronic equipment:

[0120] The term "processor" is the control center of an electronic device and can be a single processor or a collective term for multiple processing elements. For example, the processor can be one or more central processing units (CPUs), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the first embodiment of the present invention, such as one or more digital signal processors (DSPs) or one or more field programmable gate arrays (FPGAs).

[0121] The processor can execute various functions of the electronic device by running or executing software programs stored in the memory and calling data stored in the memory.

[0122] The memory is used to store the software program for executing the solution of the present invention, and the execution is controlled by the processor. The specific implementation method can refer to the above method embodiment and will not be repeated here.

[0123] The memory may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, a random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only (CD-ROM), or other optical disc storage, optical disc storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto. The memory may be integrated with the processor or exist independently and be coupled to the processor via an interface circuit of the electronic device, and this is not specifically limited in the embodiments of the present invention.

[0124] The above embodiments can be implemented in whole or in part via software, hardware (e.g., circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product comprises one or more computer instructions or computer programs. When loaded or executed on a computer, the processes or functions described in accordance with the embodiments of the present invention are fully or partially performed. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another. For example, the computer instructions can be transferred from one website, computer, server, or data center to another website, computer, server, or data center via wireless communication (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium accessible by a computer, or a data storage device such as a server or data center that contains a collection of one or more available media. The available medium can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. The semiconductor media can be a solid-state drive.

[0125] It should be understood that the term "and / or" as used herein simply describes an association between related objects, indicating that three possible relationships exist. For example, "A and / or B" can represent the existence of A alone, the existence of both A and B, or the existence of B alone. A and B can be singular or plural. Furthermore, the character " / " as used herein generally indicates an "or" relationship between the related objects, but it may also indicate an "and / or" relationship. For specific understanding, please refer to the context.

[0126] It should be understood that in the embodiments of the present invention, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0127] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should all be included in the scope of protection of the present invention.

Claims

1. A data center room liquid cooling cycle energy saving control method, characterized in that: The method comprises: Establishing a cellular network in the computer room, where each cellular node includes multiple IT cabinets, assigning node numbers to each cellular node, and obtaining time-of-use electricity prices; Obtain the real-time data matrix of each honeycomb node, decompose the real-time data matrix based on DT-CWT, obtain the transient temperature curve and steady-state temperature curve of each honeycomb node, and establish a heat flux density distribution map based on the steady-state temperature curve and honeycomb node; Based on the heat flux density distribution map, each cell node is marked for waste heat in real time. Based on the real-time waste heat marking and time-of-use electricity prices, heat generated by multiple IT cabinets in each cell node during operation is recovered in a hierarchical and time-of-use manner. A cooling storage strategy is implemented based on the steady-state temperature curve and time-of-use electricity prices. Selecting to use PID control or MPC control based on a transient temperature curve and a steady-state temperature curve to obtain a liquid cooling demand matrix, wherein the liquid cooling demand matrix includes a priority sequence; Acquire real-time temperature deviation based on real-time data matrix, perform PID control or SMC control according to the real-time temperature deviation, generate step adjustment instruction or steady-state adjustment instruction, and adjust the liquid cooling demand matrix according to the step adjustment instruction or steady-state adjustment instruction; Based on the adjusted liquid cooling demand matrix and the time-of-use electricity price, the cold storage module or the liquid cooling circulation module is started to perform real-time liquid cooling on each cellular node.

2. A data center computer room liquid cooling cycle energy saving control method according to claim 1, characterized in that: The method of separating the real-time data matrix based on DT-CWT, obtaining the transient temperature curve and steady-state temperature curve of each honeycomb node, and establishing a heat flux density distribution map based on the steady-state temperature curve and the honeycomb node includes: The real-time data matrix includes temperature signal data, and denoising is performed on the temperature signal data; The temperature signal data is decomposed into 5 layers based on DT-CWT, wherein the first and second layers correspond to high-frequency sub-bands, which represent temperature signal data with a frequency range of 0.1 Hz to 10 Hz, and the third to fifth layers correspond to low-frequency sub-bands, which represent temperature signal data with a frequency range of less than 0.1 Hz. Performing an inverse transformation based on the high frequency subband coefficients of the first and second layers to obtain a temperature transient component, and obtaining a transient temperature curve based on the temperature transient component, wherein the transient temperature curve represents a sudden temperature rise of the cellular node based on time; Performing an inverse transformation based on the low-frequency subband coefficients of layers 3 to 5 to obtain a temperature steady-state component, and obtaining a steady-state temperature curve based on the temperature steady-state component, wherein the steady-state temperature curve represents a temperature of a cellular node operating based on time; Based on the steady-state temperature curve, the dynamic temperature gradient group between adjacent honeycomb nodes is calculated, and based on the dynamic temperature gradient group, the dynamic heat flux density group of each honeycomb node is calculated. The average heat flux density is obtained according to the dynamic heat flux density group, and the average heat flux density is mapped to each honeycomb node in the cellular network of the computer room to generate a heat flux density distribution map.

3. A data center computer room liquid cooling cycle energy saving control method according to claim 1, characterized in that: The PID control or MPC control is selected based on the transient temperature curve and the steady-state temperature curve to obtain a liquid cooling demand matrix, wherein the liquid cooling demand matrix includes a priority sequence, including: The transient temperature rise rate of each honeycomb node is calculated based on the transient temperature curve, and the steady-state temperature rise rate of each honeycomb node is calculated based on the steady-state temperature curve. The temperature rise limit is obtained based on the transient temperature rise rate and the steady-state temperature rise rate. When performing the calculation, the weight of the transient temperature rise rate is greater than the steady-state temperature rise rate. If the transient temperature rise rate is greater than or equal to the temperature rise limit, MPC control is selected. Based on MPC control, the liquid cooling demand of the cellular node in the next 30 seconds is proactively predicted to obtain MPC adjustment data. If the transient temperature rise rate is less than the temperature rise limit, PID control is selected to passively obtain PID adjustment data of the current liquid cooling demand of the cellular node based on PID control; A liquid cooling demand matrix is ​​obtained based on the PID adjustment data or the MPC adjustment data, and a priority sequence in the liquid cooling demand matrix is ​​sorted based on the liquid cooling demand of the cellular node.

4. A data center computer room liquid cooling cycle energy saving control method according to claim 1, characterized in that: The method of acquiring a real-time temperature deviation based on a real-time data matrix, performing PID control or SMC control according to the real-time temperature deviation, and generating a step adjustment instruction or a steady-state adjustment instruction includes: Obtaining a real-time temperature deviation of each honeycomb node, importing the real-time temperature deviation of each honeycomb node into a step-transient switching function, and obtaining a disturbance value based on the step-transient switching function; If the disturbance value is greater than 0.3, it is considered a transient disturbance and SMC control is performed. A step adjustment instruction is generated based on the SMC control. The step adjustment instruction is expressed as a gain adjustment of the liquid cooling demand matrix. If the disturbance value is less than or equal to 0.3, it is regarded as a steady-state disturbance and PID control is performed. A steady-state adjustment instruction is generated based on the PID control. The steady-state adjustment instruction is expressed as a fine-tuning of the liquid cooling demand matrix.

5. The method for controlling energy saving of liquid cooling cycle in a data center computer room according to claim 1, characterized in that: The method of establishing a computer room cellular network, wherein each cellular node in the computer room cellular network includes multiple IT cabinets, labeling each cellular node, and obtaining a time-of-use electricity price includes: Obtain the physical locations of all IT cabinets and sensor modules deployed on the IT cabinets in the computer room, including the three-dimensional coordinates of each IT cabinet and each sensor module; Synchronize the timestamps of all sensor modules through a unified protocol, and use cubic spline interpolation to fill in the fault data in all sensor modules; Based on the ICP algorithm, the physical locations of all IT cabinets and sensor modules are mapped to a unified coordinate system. The physical location of the sensor module is used as the reference point to generate a minimum circumscribed triangle. The angle of the minimum circumscribed triangle is greater than or equal to 30 degrees. A triangular network is formed based on the minimum circumscribed triangle. Based on the duality relationship of the triangular network, the vertex of each triangle in the triangular network is converted into a hexagonal honeycomb node, and the honeycomb node shares at least one edge with the adjacent node; Calculate the thermal coupling coefficient between cellular nodes and perform grid optimization based on the thermal coupling coefficient; The cellular nodes that have completed the grid optimization operation and contain at least one IT cabinet are labeled. If the node does not contain an IT cabinet, it is marked as invalid and no longer participates in subsequent processing.

6. A data center computer room liquid cooling cycle energy saving control method according to claim 5, characterized in that: The grid optimization operation based on the thermal coupling coefficient includes: Traverse all adjacent cellular nodes and determine a strong coupling threshold based on the thermal coupling coefficient between the cellular nodes. The strong coupling threshold is 5%. If the thermal coupling coefficient between two groups of cellular nodes is less than or equal to 5%, no action will be taken; If the thermal coupling coefficient between two groups of honeycomb nodes is greater than 5%, the two groups of honeycomb nodes are considered to be strongly coupled nodes. The strongly coupled nodes are split into three honeycomb nodes and the process is repeated until the thermal coupling coefficient between the two groups of honeycomb nodes is less than or equal to 5%.

7. The method for controlling energy saving of liquid cooling cycle in a data center computer room according to claim 1, characterized in that: The method of starting the cold storage module or the liquid cooling circulation module based on the adjusted liquid cooling demand matrix and the time-of-use electricity price to perform real-time liquid cooling and heat dissipation on each cellular node includes: When the time-of-use electricity price is at a peak period, the cold storage module is started to release cold energy. The cold storage module transmits cold energy to each honeycomb node for liquid cooling and heat dissipation based on the liquid cooling demand matrix. If the maximum cold energy in the cold storage module drops to 35%, the power of the cold storage module is reduced to 50%, and the liquid cooling circulation module is started to assist in liquid cooling and heat dissipation for each honeycomb node. If the maximum cold energy in the cold storage module drops to 20%, the cold storage module is shut down, and each honeycomb node is liquid cooled and heat dissipated separately through the liquid cooling circulation module. When the time-of-use electricity price is at a low period, the cold storage strategy is restarted to replenish cold energy for the cold storage module. When the time-of-use electricity price is at the average period, the cold storage module and the liquid cooling circulation module are started at the same time to perform liquid cooling on each cellular node; When the time-of-use electricity price is at a low period, the liquid cooling circulation module is started, and the liquid cooling circulation module performs liquid cooling on each cellular node based on the liquid cooling demand matrix.

8. The method for controlling energy saving of liquid cooling cycle in a data center computer room according to claim 1, characterized in that: The method performs real-time waste heat marking on each honeycomb node based on the heat flux density distribution map, performs hierarchical and time-sharing recovery of heat generated during the operation of multiple IT cabinets in each honeycomb node based on the real-time waste heat marking and time-sharing electricity prices, and implements a cooling strategy based on the steady-state temperature curve and time-sharing electricity prices, including: The heat flux density distribution diagram includes the heat flux color intensity of each honeycomb node, and each honeycomb node is marked with real-time waste heat based on the heat flux color intensity. The real-time waste heat mark is updated every 1 minute and includes a high heat mark, a medium heat mark, and a low heat mark. If the honeycomb node is marked as high heat, the high heat generated by the honeycomb node is recovered to the heat storage module, and the heat storage module can provide auxiliary heat to the external preheating module; If the honeycomb node is marked as medium heat, the medium heat generated by the honeycomb node is recovered to the heat storage module. If the time-of-use electricity price is in the off-peak period, the heat storage module heats the medium heat until it reaches a high heat and stops. If the time-of-use electricity price is in the peak period or the average period, the medium heat is transferred to the cold storage module. If the honeycomb node is marked as low heat, the low heat generated by the honeycomb node is transferred to the cold storage module; The average temperature of IT cabinets in the total cellular network is obtained based on the steady-state temperature curve. The cold storage threshold for cooling and storing cold in the cold storage module is proportional to the average temperature of the IT cabinet. The cold storage module implements a cold storage strategy based on the average temperature. When the time-of-use electricity price is at a low period, the cold storage module cools and stores the medium and low heat transmitted by the thermal storage module. If the cold storage in the cold storage module reaches the cold storage threshold, the cold storage module stops working.

9. The method for controlling energy saving of liquid cooling cycle in a data center computer room according to claim 1, characterized in that: The method further comprises: A reinforcement learning model was constructed, and the liquid cooling demand matrix over the past 24 hours and the real-time data matrix of each cellular node were imported into the reinforcement learning model as historical training data. The reinforcement learning model obtained the flow compensation coefficient based on the TD3 algorithm and historical training data, and optimized PID control and MPC control based on the flow compensation coefficient.

10. A data center computer room liquid cooling cycle energy-saving control system, characterized in that: include: A cellular module, configured to establish a cellular network comprising a plurality of cellular nodes; An acquisition module, comprising a sensor module and an electricity price module; A sensor module is used to obtain a real-time data matrix, and the sensor module is correspondingly deployed on an IT cabinet; Electricity price module, used to obtain time-of-use electricity prices; A storage module, comprising a cold storage module and a heat storage module; The cold storage module is used to implement the cold storage strategy, store the cold energy, and release the cold energy, thereby providing real-time liquid cooling for each cellular node; The heat storage module is used to perform hierarchical and time-sharing recovery of heat generated by multiple IT cabinets in each cellular node during operation, and can provide auxiliary heat to the external preheating module; Liquid cooling circulation module, used for liquid cooling and heat dissipation of cellular nodes; a processing module, configured to obtain a liquid cooling demand matrix and adjust the liquid cooling demand matrix based on a step adjustment instruction or a steady-state adjustment instruction; The control module is used to start the cold storage module or the liquid cooling circulation module to perform real-time liquid cooling and heat dissipation on each cellular node according to the adjusted liquid cooling demand matrix and the time-of-use electricity price.

Images