CN-121586244-B - Intelligent liquid feeding control method and system for single-phase immersed liquid cooling data center cabinet

CN121586244BCN 121586244 BCN121586244 BCN 121586244BCN-121586244-B

Abstract

The invention discloses an intelligent liquid feeding control method and system for a single-phase immersed liquid cooling data center cabinet, and relates to the technical field of cabinet liquid cooling. The intelligent liquid feeding control method and system for the single-phase immersed liquid cooling data center cabinet comprise the steps of S1, collecting cabinet liquid cooling telemetry data, preprocessing the cabinet liquid cooling telemetry data, S2, evaluating cabinet liquid cooling supply and demand matching degree, judging cooling supply and demand states and generating flow adjusting instructions, triggering redistribution of cooling resources when insufficient cooling states and unbalanced heat conditions are met, S3, analyzing the flow adjusting instructions, executing pump valve adjustment of a liquid feeding branch, collecting feedback data after execution to carry out deviation evaluation, and carrying out compensation adjustment when an evaluation result does not meet expectations, S4, evaluating total liquid cooling operation risks after failure event triggering. The problem that liquid cooling supply cannot be matched in real time due to severe heat load change among cabinets, and therefore temperature rise overrun of a high-load cabinet and waste of liquid flow in a low-load area are caused is solved.

Inventors

SUN HAIWANG
QIN GUOQIANG
LI XUEQIANG
LIU SHENGCHUN
ZHANG ZHIQIANG
WANG XINGHAO
ZHAO LE

Assignees

天津提尔科技有限公司

Dates

Publication Date: 20260508
Application Date: 20260127

Claims (9)

1. The intelligent liquid feeding control method for the single-phase immersed liquid cooling data center cabinet is characterized by comprising the following steps of: S1, collecting cabinet liquid cooling telemetering data, performing validity check, abnormality rejection and smoothing treatment on the cabinet liquid cooling telemetering data, and completing normalization and standardization; S2, evaluating the matching degree of the liquid cooling supply and demand of the cabinet based on the preprocessed liquid cooling telemetering data of the cabinet, judging the cooling supply and demand state and generating a flow regulating instruction, and triggering the redistribution of cooling resources when the insufficient cooling state and the unbalanced heat condition are met; the specific steps for evaluating the cabinet liquid cooling supply and demand matching degree based on the preprocessed cabinet liquid cooling telemetering data are as follows: Multiplying the inlet flow, the cooling liquid density and the specific heat capacity of the cooling liquid based on the preprocessed liquid cooling telemetry data of the cabinet, and multiplying the difference between the temperature of the liquid outlet and the temperature of the liquid inlet to obtain the heat absorption power of the cooling liquid; dividing the heat absorption power of the cooling liquid by the working condition comprehensive heat demand, and finally subtracting a constant one to obtain a cabinet cold and hot adaptation evaluation value; S3, analyzing a flow regulation instruction, executing pump valve regulation of the liquid feeding branch, collecting feedback data for deviation evaluation after execution, performing compensation regulation when an evaluation result does not meet expectations, and recording as a fault event when the evaluation result does not meet expectations; and S4, after the fault event is triggered, evaluating the total cold running risk, executing liquid feeding branch switching when the evaluation result is in an acceptable range, and triggering emergency stop and uploading fault information when the evaluation result exceeds a risk threshold.
2. The intelligent liquid feeding control method for a single-phase immersed liquid cooling data center cabinet according to claim 1, wherein the specific steps of collecting the liquid cooling telemetry data of the cabinet, performing validity check, anomaly rejection and smoothing on the liquid cooling telemetry data of the cabinet, and completing normalization and normalization are as follows: Collecting liquid cooling telemetry data of the cabinet in real time, wherein the liquid cooling telemetry data of the cabinet comprise liquid inlet flow, liquid inlet temperature, liquid outlet temperature, single machine heat load, liquid inlet pressure, liquid outlet pressure, specific heat capacity of cooling liquid, cooling liquid density, pump energy consumption power and pump frequency; the method comprises the steps of carrying out validity verification on cabinet liquid cooling telemetry data through a boundary constraint identification algorithm, removing abnormal jump values, timestamp disordered data and records exceeding a physical working range of equipment, carrying out local fluctuation suppression on the cabinet liquid cooling telemetry data through a sliding window smoothing algorithm, relieving transient anomalies caused by coolant disturbance, pump starting impact and flow fluctuation, carrying out distribution normalization on the cabinet liquid cooling telemetry data through a mean standard deviation normalization method, maintaining an original characteristic structure, improving the contrast availability among different parameters, carrying out normalization processing on the cabinet liquid cooling telemetry data through a interval linear remapping algorithm, and mapping measurement values of heterogeneous dimensions to a unified scale interval.
3. The intelligent liquid feeding control method for a single-phase immersed liquid cooling data center cabinet according to claim 1, wherein the specific steps of judging a cooling supply and demand state and generating a flow regulating instruction, and triggering the redistribution of cooling resources when the cooling deficiency state and the thermal unbalance condition are satisfied are as follows: the cooling supply and demand state is judged by comparing the cabinet cold and hot adaptation evaluation value and the cold and hot adaptation threshold value in real time, namely, when the cabinet cold and hot adaptation evaluation value is smaller than or equal to the primary cold and hot adaptation threshold value, the cooling shortage state is judged, a flow increasing control instruction for increasing the pump frequency and the valve opening is generated, when the cabinet cold and hot adaptation evaluation value is larger than the primary cold and hot adaptation threshold value and smaller than the secondary cold and hot adaptation threshold value, the heat balance state is judged, the current flow is maintained unchanged, and when the cabinet cold and hot adaptation evaluation value is larger than or equal to the secondary cold and hot adaptation threshold value, the cooling surplus state is judged, and a flow decreasing control instruction for decreasing the pump frequency and the valve opening is generated; summarizing the cold and hot adaptation evaluation values of all the cabinets at a cluster level, outputting a thermal risk early warning signal and triggering a cooling resource reallocation operation when any condition that a fixed number of cabinets are in an insufficient cooling state and the same cabinet is not in a thermal balance state in a continuous fixed sampling period is met.
4. The intelligent liquid feeding control method for a single-phase immersed liquid cooling data center cabinet according to claim 3, wherein the specific steps of analyzing the flow adjustment command and executing the pump valve adjustment of the liquid feeding branch are as follows: after triggering the cooling resource redistribution operation, issuing a flow control instruction, establishing an execution path between a target cabinet and a corresponding flow control device, and positioning an electric control flow valve, a variable frequency driving pump and a branch pipeline valve (1) in a target liquid delivery branch; Analyzing the operation type and the parameter content in the flow control instruction, extracting the flow increase and decrease amplitude, the adjustment duration and the execution precision requirement, performing linear mapping based on the difference degree between the cabinet cold and hot adaptation evaluation value and the corresponding cold and hot adaptation threshold value to obtain an initial set value of a target quantity, further dynamically correcting the initial set value by combining with a reinforcement learning strategy optimization algorithm to generate a target liquid inlet flow, a target pump frequency and a target inlet and outlet pressure difference, and adjusting the pump frequency and the valve opening of the target liquid feeding branch.
5. The intelligent liquid feeding control method for the single-phase immersed liquid cooling data center cabinet according to claim 1, wherein the specific steps of collecting feedback data for deviation evaluation after execution are as follows: After the flow control instruction is executed, the actual liquid inlet flow, the actual pump frequency and the actual inlet-outlet pressure difference are acquired in real time, and whether the current instruction accords with the control quantity expectation is evaluated, wherein the difference value between the actual liquid inlet flow and the target liquid inlet flow is divided by the sum of the target liquid inlet flow and a minimum term, the minimum term is a positive real number with extremely small value but non-zero, and the minimum term is used for avoiding the problem of unstable numerical value caused by zero divisor in the operation process, and the value range is To the point of Obtaining a flow rate deviation ratio of the inlet fluid, dividing the difference between the actual pump frequency and the target pump frequency by the sum of the target pump frequency and the minimum term to obtain a pump frequency deviation ratio, dividing the difference between the actual inlet and outlet pressure differences and the target inlet and outlet pressure differences by the sum of the target inlet and outlet pressure differences and the minimum term to obtain a pressure difference deviation ratio, sequentially squaring the flow rate deviation ratio of the inlet fluid, the pump frequency deviation ratio and the pressure difference deviation ratio, adding, taking the square root, and obtaining an execution deviation evaluation value.
6. The intelligent liquid feeding control method for a single-phase immersed liquid cooling data center cabinet according to claim 5, wherein the specific steps of performing compensation adjustment when the evaluation result does not meet expectations, and recording as a fault event when the evaluation result does not meet expectations after compensation are as follows: The method comprises the steps of comparing an execution deviation evaluation value with an execution deviation threshold in real time, when the execution deviation evaluation value is smaller than or equal to the execution deviation threshold, not processing, when the execution deviation evaluation value is larger than the execution deviation threshold, judging a main abnormal source according to the values of a feed liquid flow deviation ratio, a pump frequency deviation ratio and a pressure difference deviation ratio, compensating by correspondingly adjusting the pump frequency and the valve opening, recalculating the execution deviation evaluation value after compensation, and recording as a fault event and outputting an alarm signal if the execution deviation evaluation value is still larger than the tolerance threshold.
7. The intelligent liquid feeding control method for a single-phase immersed liquid cooling data center cabinet according to claim 6, wherein the specific steps of evaluating the overall liquid cooling operation risk after the fault event is triggered are as follows: After the fault event is triggered, the cabinet cold and hot adaptation evaluation value, the execution deviation evaluation value, the inlet and outlet pressure difference and the pump energy consumption power of each cabinet are adjusted, the difference value between the cabinet cold and hot adaptation evaluation value of the ith cabinet and the primary cold and hot adaptation threshold value is taken as the maximum value, the difference value between the cabinet cold and hot adaptation evaluation value of the ith cabinet and the secondary cold and hot adaptation threshold value is taken as the maximum value and added, the sum of the difference value between the secondary cold and hot adaptation threshold value and the primary cold and hot adaptation threshold value and the minimum item is divided, the execution deviation correction item is obtained by dividing the execution deviation evaluation value of the ith cabinet by the sum of the execution deviation threshold value and the minimum item, the inlet and outlet pressure difference of the ith cabinet is divided by the sum of the pump energy consumption power of the ith cabinet and the minimum item, the cold and hot adaptation correction item and the hydraulic load correction item are sequentially added, the single machine risk amount of all cabinets is obtained, and the single machine risk amount of liquid cooling operation risk is obtained by adding and arithmetic average.
8. The intelligent liquid feeding control method for a single-phase immersed liquid cooling data center cabinet according to claim 7, wherein the specific steps of executing liquid feeding branch switching when the evaluation result is in an acceptable range, triggering emergency stop and uploading fault information when the evaluation result exceeds a risk threshold are as follows: The liquid cooling operation risk assessment value and the liquid cooling risk threshold value are compared in real time, when the liquid cooling operation risk assessment value is smaller than or equal to the liquid cooling risk threshold value, the liquid feeding branch switching operation is executed, namely a standby branch execution path is established, the frequency of a current liquid feeding branch pump is gradually reduced, a branch pipeline valve (1) is closed, and meanwhile the frequency of the standby branch pump and the opening of the valve are improved; when the liquid cooling operation risk assessment value is larger than the liquid cooling risk threshold value, triggering an emergency protection mechanism, executing forced shutdown, and sending fault information to an upper manager.
9. The intelligent liquid feeding control system for the single-phase immersed liquid cooling data center cabinet is characterized by comprising a cabinet liquid cooling telemetering data acquisition preprocessing module, a heat load evaluation and control instruction issuing module, an intelligent liquid feeding execution and compensation feedback regulation module and a liquid cooling operation risk evaluation and fault handling module, wherein the intelligent liquid feeding control system is used for the single-phase immersed liquid cooling data center cabinet according to any one of claims 1-8, and the intelligent liquid feeding control system is characterized in that the intelligent liquid feeding control system comprises the cabinet liquid cooling telemetering data acquisition preprocessing module, the heat load evaluation and control instruction issuing module, the intelligent liquid feeding execution and compensation feedback regulation module and the liquid cooling operation risk evaluation and fault handling module, wherein: The cabinet liquid cooling telemetering data acquisition preprocessing module is used for acquiring cabinet liquid cooling telemetering data, performing validity check, abnormal rejection and smoothing processing on the cabinet liquid cooling telemetering data, and completing normalization and standardization; The heat load evaluation and control instruction issuing module is used for evaluating the matching degree of the liquid cooling supply and demand of the cabinet based on the preprocessed liquid cooling telemetering data of the cabinet, judging the cooling supply and demand state and generating a flow adjustment instruction, and triggering the redistribution of cooling resources when the insufficient cooling state and the heat imbalance condition are met; The intelligent liquid feeding execution and compensation feedback adjustment module is used for analyzing the flow adjustment instruction, executing pump valve adjustment of a liquid feeding branch, collecting feedback data after execution to carry out deviation evaluation, carrying out compensation adjustment when the evaluation result does not meet the expectation, and recording as a fault event if the evaluation result does not meet the expectation; And the liquid cooling operation risk assessment and fault handling module is used for assessing the whole liquid cooling operation risk after the fault event is triggered, executing liquid feeding branch switching when the assessment result is in an acceptable range, triggering emergency stop when the assessment result exceeds a risk threshold value and uploading fault information.

Description

Intelligent liquid feeding control method and system for single-phase immersed liquid cooling data center cabinet Technical Field The invention relates to the technical field of cabinet liquid cooling, in particular to an intelligent liquid feeding control method and system for a single-phase immersed liquid cooling data center cabinet. Background Under the background of continuous expansion of the scale of the data center, the power consumption and the heat load of the single cabinet continuously rise, and the traditional air cooling mode gradually exposes bottlenecks in the aspects of heat dissipation efficiency, energy consumption control and space occupation. The liquid cooling technology gradually becomes a main stream cooling means of high-performance computing, cloud computing and large-scale data centers due to the high heat transfer capability and good temperature control performance. The immersed liquid cooling can directly exchange heat with the heating device in a large area, and has obvious advantages in the aspects of energy efficiency ratio, noise control and system stability. For example, the invention with publication number CN116017951a provides a liquid-cooled cabinet and a liquid cooling system, which relate to the technical field of computers, in particular to the technical fields of cloud computing, data centers and the like. The liquid cooling cabinet comprises a cabinet body, a circulation system and a heating device, wherein the circulation system and the heating device are arranged in the cabinet body, the circulation system is used for transmitting cooling liquid, the heating device is immersed in the cooling liquid, the heating device comprises a whole cabinet server, a centralized power supply module and a centralized power supply bus, and the whole cabinet server is connected with the centralized power supply module through the centralized power supply bus. The invention can improve the effect of the liquid-cooled cabinet. For example, the invention of publication number CN117222171a provides a cabinet liquid cooling system comprising a plurality of servers, a liquid cooling device, a first manifold, a second manifold, an overflow pipe, and a control valve. The server, the liquid cooling device and at least one overflow pipe are connected in parallel through the first branch pipe and the second branch pipe. The control valve comprises a first valve clack, a second valve clack, a pushed piece and an elastic piece. The first valve clack and the second valve clack are rotatably arranged in the overflow pipe. Opposite ends of the elastic piece are respectively fixed on the overflow pipe and the pushed piece. The elastic piece is used for pushing the pushing piece to enable the pushing piece to be abutted between the first valve clack and the second valve clack so as to block the overflow pipe from communicating with the server and the liquid cooling device. When the control valve bears a critical pressure, the pushed part overcomes the elasticity of the elastic part and is far away from the first valve clack and the second valve clack, so that the overflow pipe is communicated with the server and the liquid cooling device. However, the above-mentioned prior art solutions mainly focus on structural optimization and flow path design of the liquid cooling cabinet, and do not carry out intensive research on real-time matching and intelligent regulation of the liquid cooling supply and demand states, and especially lack dynamic response mechanisms in situations of insufficient cooling, excessive cooling and hot air risk. Once the liquid cooling supply and demand deviation or the execution deviation occurs, the system is easy to generate insufficient heat dissipation, low energy efficiency and even thermal safety risk. Therefore, in view of the above problems, there is a need for intelligent liquid delivery control methods and systems for single-phase immersion liquid cooled data center cabinets. Disclosure of Invention Technical problem to be solved Aiming at the defects of the prior art, the invention provides an intelligent liquid feeding control method and system for single-phase immersed liquid cooling data center cabinets, which solve the problems that liquid cooling supply cannot be matched in real time due to severe heat load change among cabinets, and further high-load cabinet temperature rise overrun and low-load area liquid flow waste are caused. Technical proposal The intelligent liquid feeding control method and system for the single-phase immersed liquid cooling data center cabinet comprise the following steps of S1, collecting cabinet liquid cooling telemetry data, performing validity check, abnormal rejection and smoothing treatment on the cabinet liquid cooling telemetry data, and completing normalization and standardization, S2, evaluating cabinet liquid cooling supply and demand matching degree based on the preprocessed cabinet liquid cooling telemetry data, judging co