A seasonal thermal management device suitable for data centers

Abstract

The present application relates to the technical field of data center energy management and waste heat recovery, and particularly relates to a seasonal thermal management device suitable for a data center, comprising: data acquisition through a heat production information acquisition module and an environment information acquisition module; an analysis module determines a seasonal working condition thermal data adjustment range based on the environment data, and determines a plurality of comprehensive thermal indexes of the data center based on the seasonal working condition thermal data adjustment range, and divides thermal management areas; a control module determines each thermal management area based on the threshold of each thermal management area, and adjusts the load distribution of each cabinet based on the distribution state of each thermal management area; the analysis module further judges whether a corresponding chip protection strategy needs to be executed according to a hardware health degree model to optimize the equipment life. The present application realizes the collaborative and accurate control of the data center load and the liquid cooling system by dynamically identifying the thermal distribution and combining the seasonal working condition, and significantly improves the thermal management efficiency and the system energy efficiency.

Description

Technical Field

[0001] This invention relates to the field of data center energy management and waste heat recovery technology, specifically to a seasonal thermal management device suitable for data centers. Background Technology

[0002] With the rapid development of the digital economy, data centers, as computing infrastructure, are facing increasingly prominent energy consumption issues. Among these, energy consumption for cooling accounts for as much as 40%, while a large amount of low-grade waste heat generated by server operation is usually directly discharged into the environment, causing serious energy waste and thermal pollution.

[0003] Current heat management technologies remain at the level of heat dissipation, merely removing heat generated by IT equipment from the server room without recognizing it as a valuable, recyclable resource, resulting in a significant waste of low-grade heat energy. Secondly, existing systems lack fine-grained dynamic partitioning management capabilities, failing to accurately allocate cooling resources based on real-time changes in server load, instantaneous chip temperatures, and external environmental factors. This leads to insufficient cooling in high-load core areas and overcooling in low-load areas, resulting in low overall energy efficiency. Furthermore, when supplying waste heat from the data center to downstream users (such as residential heating and agricultural greenhouses), existing systems cannot effectively guarantee stable temperature and flow rates, failing to meet real-time heating demands. In addition, existing systems lack effective proactive management and optimization methods for addressing hardware lifespan degradation caused by long-term operation of chips in high-temperature fluctuating environments.

[0004] Therefore, there is an urgent need for an intelligent thermal management device that can achieve fine-grained zoning of data center heat sources, dynamically match downstream heat demand, and proactively manage hardware health. Summary of the Invention

[0005] Therefore, the present invention provides a seasonal thermal management device suitable for data centers to overcome the problems of existing technologies that cannot fully utilize the heat dissipation capacity of the computer room itself, lack refined dynamic zoning management capabilities, and have low overall energy utilization efficiency.

[0006] To achieve the above objectives, the present invention provides a seasonal thermal management device suitable for data centers, comprising: The heat generation information acquisition module is used to acquire temperature data of each rack inside the data center, wherein the temperature data includes coolant inlet temperature, coolant outlet temperature and chip junction temperature data; The environmental information acquisition module is used to acquire environmental data from the data center. The analysis module determines the seasonal operating condition thermal data adjustment range based on the environmental data, and determines several comprehensive thermal indices of the data center based on the seasonal operating condition thermal data adjustment range. The thermal management area includes a core thermal aggregation area, an auxiliary peak shaving area, and a fixed load area. The control module determines each thermal management zone based on its threshold values, and adjusts the load distribution of each rack based on the distribution status of each thermal management zone. In addition, the saturation rate is determined based on the distribution status of the thermal management area after the load distribution is adjusted, and the liquid cooling control parameters of each thermal management area are adjusted based on the saturation rate. The liquid cooling control parameters include the coolant flow distribution and the coolant inlet temperature setpoint.

[0007] Furthermore, the analysis module also includes determining a first thermal distribution map based on the location coordinates of each cabinet and the comprehensive thermal index.

[0008] Furthermore, the control module is also used to update several comprehensive thermal indices of the corresponding thermal management area based on the load distribution adjustment.

[0009] Furthermore, the analysis module also determines the adjustment range of seasonal operating condition heat data based on the predicted values ​​of the environmental data.

[0010] Furthermore, the control module adjusts the load distribution of each cabinet based on the distribution status of each thermal management area, including: The service load migration targets are determined based on the comparison between the chip junction temperature and the chip junction temperature threshold in the core thermal aggregation zone. The service load of the service load migration targets is then adjusted to adjust the load distribution of each cabinet.

[0011] Furthermore, a second thermal distribution map is determined based on several comprehensive thermal indices of the thermal management area that are reacquired after a preset time period.

[0012] Furthermore, the control module adjusts the liquid cooling control parameters of each thermal management zone based on the saturation rate of the core thermal aggregation zone of the second thermal distribution map, including: When the saturation rate of the core thermal polymerization zone is greater than the first saturation rate, the control module adjusts the liquid cooling control parameters of the auxiliary peak shaving zone and the fixed load zone. When the saturation rate of the core thermopolymerization zone is less than or equal to the first saturation rate and greater than the second saturation rate, the control module adjusts the liquid cooling control parameters of the auxiliary peak-shaving zone. When the saturation rate of the core thermal polymerization zone is less than or equal to the second saturation rate, the control module adjusts the liquid cooling control parameters of the core thermal polymerization zone.

[0013] Furthermore, the environmental data also includes: Dew point temperature data calculated based on relative humidity data from both the outside and inside of the data center; Based on medium- and long-term meteorological forecast data of the data center's location, a dew point temperature curve was plotted.

[0014] Furthermore, the control module also determines a condensation prevention and control threshold based on the judgment result that the dew point temperature data is lower than the coolant inlet temperature data, and the dew point temperature data and the coolant inlet temperature data.

[0015] Furthermore, the analysis module samples the junction temperature of each chip in the data center at preset time intervals, assigns weight coefficients based on the interval where the chip junction temperature is located, determines the cumulative lifespan consumption of each chip, and the control module executes the corresponding chip protection strategy.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention divides the heat source of a data center into a core heat aggregation zone, an auxiliary peak-shaving zone, and a fixed load zone, and constructs a threshold system for each zone. This enables on-demand allocation and precise supply of cooling resources, avoiding a uniform cooling method and reducing the energy consumption of the cooling system. Simultaneously, it utilizes waste heat resources, achieving stable output of waste heat temperature and flow rate through dynamic adjustment, thereby improving the overall energy utilization efficiency.

[0017] Furthermore, this invention provides a stable basic heat source through the core thermal aggregation zone, the auxiliary peak-shaving zone is responsible for dynamic adjustment to cope with the fluctuations in downstream heat demand, and the fixed load zone serves as a supplement or buffer for waste heat, so that the waste heat output of the data center as a whole can match the load changes of downstream heat users in real time, thus solving the problem of unstable heating in traditional systems.

[0018] Furthermore, this invention collects key data such as chip junction temperature at high frequency and calculates dew point temperature by combining it with external environmental data, constructing a condensation prevention threshold to effectively prevent condensation on the surface of cooling equipment and avoid safety accidents such as short circuits caused by condensation. Simultaneously, based on a hardware health management model, it proactively migrates loads and balances chip thermal stress, effectively extending the service life of critical hardware devices.

[0019] Furthermore, this invention can dynamically adjust the boundaries and attributes of each partition based on real-time load, chip temperature change rate, and medium- to long-term weather forecasts, enabling the system to flexibly cope with complex scenarios such as business peaks and seasonal changes, and achieving a high degree of intelligence in data center thermal management. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of a data center thermal management device according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the analysis module and control module in an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the division of a data center into three types of areas according to an embodiment of the present invention; Figure 4 This is a flowchart illustrating the control logic of an embodiment of the present invention. In the diagram, 1-core thermal polymerization zone; 2-auxiliary peak shaving zone; 3-fixed load zone. Detailed Implementation

[0021] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0022] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0023] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0024] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0025] With the rapid development of the digital economy, data centers, as computing infrastructure, are facing increasingly prominent energy consumption issues. Among these, energy consumption for cooling accounts for as much as 40%, while a large amount of low-grade waste heat generated by server operation is usually directly discharged into the environment, causing serious energy waste and thermal pollution.

[0026] Current technologies for managing heat generation in data centers remain at the level of heat dissipation, merely expelling heat generated by IT equipment from the server room without recognizing it as a valuable, recyclable resource, resulting in a significant waste of low-grade heat energy. Secondly, existing systems lack fine-grained dynamic partitioning management capabilities, failing to accurately allocate cooling resources based on real-time changes in server load, instantaneous chip temperatures, and external environmental factors. This leads to insufficient cooling in high-load core areas and overcooling in low-load areas, resulting in overall low energy efficiency. Furthermore, when supplying waste heat from the data center to downstream users (such as residential heating and agricultural greenhouses), existing systems cannot effectively guarantee stable temperature and flow rates, failing to meet real-time heating demands. In addition, existing systems lack effective proactive management and optimization methods for addressing hardware lifespan degradation caused by long-term operation of chips in high-temperature fluctuating environments.

[0027] Based on this, the present invention provides a seasonal thermal management device suitable for data centers, the system comprising: The heat generation information acquisition module is used to acquire temperature data of each cabinet inside the data center, including coolant inlet temperature, coolant outlet temperature and chip junction temperature data. The environmental information acquisition module is used to acquire environmental data from the data center. The analysis module determines the adjustment range of seasonal operating condition thermal data based on the environmental data. In addition, several comprehensive thermal indices of the data center are determined based on the adjustment range of seasonal operating condition thermal data, wherein the thermal management area includes a core thermal aggregation area, an auxiliary peak shaving area, and a fixed load area; The control module determines each thermal management zone based on its threshold values, and adjusts the load distribution of each rack based on the distribution status of each thermal management zone. In addition, the saturation rate is determined based on the distribution status of the thermal management area after the load distribution is adjusted, and the liquid cooling control parameters of each thermal management area are adjusted based on the saturation rate. The liquid cooling control parameters include the coolant flow distribution and the coolant inlet temperature setpoint.

[0028] Please see Figure 1 As shown, it is a schematic diagram of a data center thermal management device according to an embodiment of the present invention.

[0029] This embodiment provides a seasonal thermal management device suitable for data centers. After the system is started, the heat generation information acquisition module is used to acquire temperature data of each rack inside the data center, including coolant inlet temperature, coolant outlet temperature, and chip junction temperature data; The environmental information acquisition module is used to acquire environmental data from the data center. Specifically, the heat generation information acquisition module communicates with various sensors (such as temperature sensors and flow sensors) deployed inside the data center racks. It acquires temperature data from each rack within the server room, including the junction temperature of server chips, the inlet temperature of the coolant entering the rack's cold plate, and the outlet temperature of the coolant exiting the cold plate. To improve responsiveness to transient thermal shocks, the module sets the data acquisition frequency for chip junction temperatures, for example, once per second, and for temperatures higher than the coolant inlet and outlet temperatures, for example, once every 10 seconds. The module also calculates the coolant inlet and outlet temperature difference based on the acquired coolant inlet and outlet temperatures; this temperature difference directly reflects the real-time heat dissipation power of the current rack. All acquired raw data, calculated temperature difference data, and system partition data are synchronized to the historical database and sent together to the intelligent analysis and control module. It is understood that the chip junction temperature is the actual operating temperature of the PN junction of the transistor inside the semiconductor chip, and is a key indicator for evaluating chip performance, reliability, and lifespan.

[0030] The environmental information acquisition module collects environmental data from both inside and outside the data center. Internal environmental data includes air temperature and relative humidity within the server room; external environmental data includes outdoor air temperature, relative humidity, and medium- to long-term weather forecast data accessed via API, such as temperature and humidity forecast curves for the next 24 hours. Due to the rapid changes in the server room environment, its temperature data collection frequency, for example, once every 5 seconds, is higher than that of the external environmental data collection frequency. Based on the collected internal and external relative humidity data, combined with the corresponding temperature data, this module calculates the dew point temperature inside the server room to prevent condensation. Its core safety principle is to ensure that the coolant inlet temperature is always higher than the dew point temperature. Furthermore, this module also plots a dew point temperature curve for a future period based on medium- to long-term weather forecast data, providing a basis for proactive regulation. It can make predictions using any existing temperature forecasting technology, without limitation. All collected environmental data, seasonal operating condition data, and dew point temperature data are sent to the intelligent analysis and control module.

[0031] Please see Figure 2 As shown, it is a schematic diagram of the analysis module and control module in an embodiment of the present invention.

[0032] The analysis module determines the seasonal operating condition thermal data adjustment range based on the environmental data; it calculates the seasonal operating condition thermal data adjustment range based on the environmental forecast values ​​for the next 24 hours. The seasonal operating condition thermal data adjustment range is used to correct the thermal management area thresholds, which include the Tmax_th chip junction temperature threshold, the ΔTmax_th coolant inlet and outlet temperature difference threshold, and the Lmax load rate threshold.

[0033] In practice, the calculated seasonal operating condition thermal data adjustment range allows for higher coolant inlet temperatures under high-temperature conditions in summer.

[0034] Based on the adjustment range of seasonal operating condition thermal data, several comprehensive thermal indices of the data center are determined. The analysis module divides the thermal management area into core thermal aggregation area, auxiliary peak shaving area and fixed load area according to the calculated comprehensive thermal index Qindex. It can be understood that the high temperature aggregation area is set as the core thermal aggregation area, the area with the second highest temperature but obvious fluctuation is set as the auxiliary peak shaving area, and the stable area with lower temperature is set as the fixed load area.

[0035] Specifically, the construction of a heat map includes the following steps: 1. Data Acquisition and Coordinate Mapping First, the heat generation information acquisition module will establish physical coordinates for each rack node in the data center, such as row X, column Y, and layer Z.

[0036] The system simultaneously collects three core data points for this coordinate point: Real-time load rate: Reflects the current computing power and power consumption potential of the business; Chip junction temperature: reflects the temperature of the most critical heat-generating point; Coolant inlet and outlet temperature difference: reflects the actual heat removed by this node, i.e., the real-time heat dissipation.

[0037] 2. Calculation of thermal value The comprehensive thermal index Qindex is calculated using the following formula: ; Where: Tjunction is the chip junction temperature, ΔTcoolant is the temperature difference between the coolant inlet and outlet, and Lload is the load rate. Tmax_th is the chip junction temperature threshold, ΔTmax_th is the coolant inlet and outlet temperature difference threshold, and Lmax is the load rate threshold. α, β, and γ are the weighting coefficients of the comprehensive thermal index.

[0038] Specifically, Tmax_th, ΔTmax_th, and Lmax are determined based on three dimensions: hardware physical limits, system design boundaries, and partition management objectives.

[0039] In a specific implementation, Tmax_th is derived from the server chip's hardware specifications. For example, the rated operating junction temperature of a certain CPU or GPU model is typically 85°C, and the thermal protection trigger point is 95°C-100°C. To ensure hardware lifespan and operational safety, the threshold should not be set at the hardware's extreme limit; it is usually set to 90% of the thermal protection trigger point or the upper limit of the safe operating range. For example, if the chip's maximum allowable junction temperature is 95°C, then Tmax_th can be set to 85°C. When the junction temperature approaches Tmax_th, the system determines that the node is in a "high-stress" state and tends to classify it as part of the core thermal aggregation zone, i.e., a stable high-heat source requiring focused cooling.

[0040] ΔTmax_th depends on the specific heat capacity and flow rate of the coolant, as well as the design heat dissipation power of the cold plate or heat exchanger. The temperature difference directly reflects the real-time heat generation of the equipment. This threshold is typically set as the design temperature difference under rated operating conditions. For example, under standard design conditions, if the inlet and outlet temperature difference of a single cabinet at full load is 15°C, then this threshold can be set to 15°C. The closer the actual temperature difference is to this threshold, the higher the heat dissipation power of the node, and the greater its potential to serve as a stable heat source, i.e., a core thermal aggregation zone.

[0041] Lmax originates from the business scheduling strategy of IT equipment and typically refers to the percentage of CPU or GPU utilization. In this embodiment, Lmax is periodically read by the heat generation information acquisition module through the IPMI interface or virtualization platform API of the server baseboard management controller in each rack, with the collection frequency set to once every 10 seconds. It is used to distinguish whether the computing power of each rack in the data center is idle. For example, 70% can be set as the dividing line. If the load rate is consistently higher than this threshold, it means that the heat generated by that node has small fluctuations and belongs to the "base load" type heat source, which is suitable for classification into the core heat aggregation area or fixed load area, depending on the fluctuation range.

[0042] It is understood that the present invention does not fix the values ​​of Tmax_th, ΔTmax_th, and Lmax or set a range of values. Tmax_th, ΔTmax_th, and Lmax are set according to the actual situation.

[0043] α, β, and γ are used to define the importance of different indicators in the calculation of "heat value". The logic behind their determination should serve the purpose of this invention: to accurately distinguish between the stability and recyclability of a heat source.

[0044] α represents the junction temperature weight: the junction temperature is the source of heat generation. For this invention, junction temperature data is collected most frequently and best reflects transient thermal shock. It has the highest weight because it directly relates to chip safety and heat source quality. A higher junction temperature indicates a higher heat source temperature, making it more valuable for recycling. Preferably, α is set to 0.5.

[0045] β is the inlet and outlet temperature difference weight: the temperature difference represents the actual heat carried away by the coolant. It reflects the actual heat load of the node on the cooling system. The weight is moderate, used to quantify the actual amount of recoverable heat; preferably, β is 0.3.

[0046] γ is the load factor weight: the load factor represents the "sustainability" and "predictability" of heat generation. The weight is relatively low, but it determines the zoning attribute - only those with high load and small fluctuations can become "core heat aggregation zones"; those with high load but large fluctuations are suitable as "auxiliary peak-shaving zones". The preferred value for γ is 0.2.

[0047] The principle for setting the weights of the comprehensive thermal index is to prioritize stability and temperature quality. Please refer to Table 1, which is a table of preset thresholds and weight coefficients for this embodiment.

[0048] ; In this embodiment, Tmax_th, ΔTmax_th, and Lmax are adjusted based on the environmental forecast values ​​for the next 24 hours, using two dimensions: (1) Downstream heat demand Environmental prediction parameters: external temperature, especially the predicted minimum temperature.

[0049] Correction logic: The lower the external temperature, the greater the downstream heat demand, requiring the system to output more waste heat. Appropriately lower the threshold to allow more server racks to enter the "high heat" zone.

[0050] (2) System heat dissipation capacity Environmental prediction parameters: external temperature, humidity Correction logic: The higher the external temperature and humidity, the worse the heat dissipation conditions of the data center. Appropriately increase the safety threshold, reduce the load in advance, or enhance cooling.

[0051] Winter: Expand the heat source by reducing Tmax_th, increasing ΔTmax_th, and reducing Lmax.

[0052] Summer: Reduce heat sources and strengthen data center security by increasing Tmax_th, decreasing ΔTmax_th, and increasing Lmax.

[0053] Specifically, Tmax_th is set to 85℃, and is adjusted according to different correction values. When the hourly average temperature in the predicted 24-hour temperature is ≥32℃, the correction value of Tmax_th is set to -5℃, and Tmax_th is set to 80℃. When the hourly average temperature in the predicted 24-hour temperature is ≤0℃, the correction value of Tmax_th is set to 0℃, and Tmax_th is set to 85℃. When the hourly average temperature in the predicted 24-hour temperature is 0℃, the correction value of Tmax_th is set to 85℃.

[0054] ΔTmax_th is mainly affected by downstream heat demand. The heat demand coefficient is adjusted according to different temperatures. When the hourly average temperature in the predicted 24-hour temperature is ≤0℃, the heat demand coefficient is 1.3, and ΔTmax_th = 12 × 1.3 = 15.6℃. When the hourly average temperature in the predicted 24-hour temperature is between 0℃ and 15℃, the heat demand coefficient is 1.0, and ΔTmax_th = 12℃. When the hourly average temperature in the predicted 24-hour temperature is ≥30℃, the heat demand coefficient is 0.8, and ΔTmax_th = 9.6℃.

[0055] Lmax primarily adjusts the size of the heat source cabinet, adjusting it according to different load factors. When the predicted hourly average temperature for the next 24 hours is ≤0℃, the load factor is set to 0.7, at which point Lmax=49%; when the predicted hourly average temperature for the next 24 hours is between 0℃ and 15℃, the load factor is set to 1.0, at which point Lmax=70%; when the predicted hourly average temperature for the next 24 hours is ≥30℃, the load factor is set to 1.2, at which point Lmax=84%. When Lmax=49%, nodes with a load rate exceeding 49% can enter the core heat aggregation zone, expanding the heat source pool and outputting more waste heat. When Lmax=84%, only high-load nodes with a load rate exceeding 84% are considered core heat sources, avoiding over-reliance on equipment with deteriorating heat dissipation conditions.

[0056] Please see Figure 3 As shown, it is a schematic diagram of the data center being divided into three types of areas according to an embodiment of the present invention.

[0057] Based on several comprehensive thermal indices, the analysis module divides the physical space of the data center into three types of areas: Core Thermal Aggregation Zone 1: The cabinets in this area are under high load and high heat generation for a long time. The chip junction temperature is high and fluctuates little, which is a stable base load source for the waste heat supply of the data center.

[0058] Auxiliary peak shaving zone 2: The cabinet load and heat generation in this area have certain fluctuations and can be flexibly adjusted to respond to changes in demand from downstream heat users and play a peak shaving role.

[0059] Fixed load area 3: The rack load in this area is relatively stable but the heat generation is low. It can be used as a supplement to waste heat recovery or as a backup heat source when the core area needs maintenance.

[0060] The thermal management area is divided into three categories, with the following division principles and management strategies: The first preset comprehensive thermal index threshold is set to 0.55, and the second preset comprehensive thermal index threshold is set to 0.85.

[0061] The first thermal distribution map was determined using the location coordinates of each server rack and the overall thermal index. Specifically: Core thermal aggregation zone: Set Qindex≥0.85, stable base load heat source, Qindex is a continuously high value with small fluctuations, management strategy is to focus on cooling and prioritize its protection, as a stable output source for waste heat supply.

[0062] Auxiliary peak shaving zone: Set to 0.55≤Qindex<0.85, flexible peak shaving heat source, Qindex is medium to high value but fluctuates greatly, or the peak value is extremely high but not sustained, the management strategy is to dynamically adjust cooling, and flexibly adjust the load or flow according to downstream demand.

[0063] Fixed load area: Set Qindex < 0.55, supplementary or backup heat source, Qindex is continuously low value, minimum cooling, management strategy is to use as a backup area for waste heat supplementation or load migration.

[0064] It is understandable that the settings of the first and second preset comprehensive thermal indexes can be adjusted according to different scenarios, and can also be optimized through data iteration based on several historical operating conditions, which will not be elaborated here.

[0065] In this embodiment, the analysis module further combines historical data, seasonal operating condition data, and medium- and long-term meteorological forecasts to determine the adjustable range of thermal data for each region under different seasonal operating conditions, and constructs a specific threshold system for each region. This system includes, but is not limited to: a coolant inlet and outlet temperature difference threshold to prevent excessive temperature difference from causing thermal shock or insufficient temperature difference from causing insufficient heat dissipation; a chip junction temperature threshold to ensure that the chip operates within a safe temperature range; and a dew point temperature control threshold set according to the real-time dew point temperature to ensure that the coolant inlet temperature is higher than the dew point temperature and prevent condensation.

[0066] Specifically, based on the data types acquired by the environmental information module, the data is divided into two layers: a safety constraint layer and a dynamic adjustment layer. The environmental data types in the safety constraint layer are internal and external temperature and humidity within the data center. Its target is the dew point temperature control threshold. The dew point temperature is calculated based on the internal and external temperature and humidity to ensure that the coolant inlet temperature is always higher than the dew point temperature, preventing condensation. This is a mandatory safety baseline shared by all zones in the data center. The environmental data types in the dynamic adjustment layer are external temperature and medium- to long-term weather forecasts. Its target is the adjustable range of seasonal operating temperature data. Based on weather forecasts, the current seasonal operating conditions are determined, such as winter / summer / transitional season, and the leniency of the threshold system for each zone is dynamically adjusted to match the changing trends of downstream heating demand.

[0067] In one specific embodiment, the dew point temperature control threshold is set as the sum of the dew point temperature and the safety margin, with the safety margin set at 2℃~5℃. Changes in the dew point temperature are monitored and collected in real time, and the upper limit of the coolant inlet temperature is dynamically adjusted to ensure that no condensation risk occurs in any zone. A dew point temperature curve is plotted based on medium- and long-term meteorological forecast data of the data center's location.

[0068] Specifically, in the threshold system of the core thermal aggregation zone, the core thermal aggregation zone serves as a stable base load heat source. Its threshold setting needs to take into account both the stability of waste heat output and hardware safety. Environmental information affects its adjustment range through seasonal operating conditions.

[0069] The control module determines each thermal management zone based on the threshold of each thermal management zone, and adjusts the load distribution of each cabinet based on the distribution status of each thermal management zone.

[0070] In implementation, the service load migration targets are determined based on the comparison between the chip junction temperature and the chip junction temperature threshold in the core thermal aggregation zone. The service load of the migration targets is adjusted to adjust the load distribution of each rack. A second thermal distribution map is determined based on several comprehensive thermal indices of the thermal management area and the location coordinates of each rack, which are reacquired after a preset time. Then, the control module calculates the current saturation rate of the core thermal aggregation zone based on the second thermal distribution map after a preset time.

[0071] Specifically, the control module obtains the first thermal distribution map to get the current distribution status of each thermal management area (including the boundary range of each thermal management area and the system heat dissipation capacity), and generates a load distribution adjustment command accordingly.

[0072] When the junction temperature of a chip in a cabinet within the core thermal aggregation zone exceeds the chip junction temperature threshold (i.e., above 85°C), the control module determines that the chip is a target for service load migration. It then migrates the service load currently being processed by that chip (with less than 20% of its processing capacity) to a cabinet with remaining heat dissipation capacity in the auxiliary peak-shaving zone or fixed load zone. The selection of target cabinets for load migration follows this priority: first, chips in low-load areas within the same cabinet; second, chips in cabinets within the auxiliary peak-shaving zone; and finally, chips in cabinets within the fixed load zone. This ensures that the migrated service load is distributed tiered and matched to the processing capacity of the assigned chips, until the junction temperature of the chip currently being migrated is less than or equal to the chip junction temperature threshold. It is understood that if the migrated service load cannot bring the chip junction temperature below or equal to the chip junction temperature threshold, it is assumed that all cabinets are operating at full capacity, and the saturation rate for each area is calculated.

[0073] After the load distribution adjustment is completed and a preset time has elapsed, the heat generation information acquisition module re-collects the coolant inlet and outlet temperatures, chip junction temperatures, and flow rates of each cabinet. The analysis module recalculates the distribution status of each thermal management zone based on the updated data and determines the saturation rate of each zone. Preferably, the preset time is typically set to 20 to 60 minutes.

[0074] Based on several comprehensive thermal indices of the thermal management area and the coordinates of each cabinet location, which are reacquired after a preset time period, a second thermal distribution map is drawn according to the determined comprehensive thermal index threshold.

[0075] Specifically, for the saturation rate of a single area, the saturation rate = the actual total heat generation power of all heat sources (racks) in the current area / the maximum allowable heat dissipation power of the area under the design conditions.

[0076] The specific calculation steps are as follows: 1. Liquid-side calculation method: For a liquid-cooled circuit within a single area, measure the mass flow rate m, inlet temperature Tin, and outlet temperature Tout of the coolant, and calculate using the formula: ; Among them, P 实际 c is the actual total heat output of all heat sources (racks) in the current area; p This refers to the specific heat capacity of the coolant.

[0077] 2. The maximum allowable heat dissipation power of a single area under design conditions is calculated based on the minimum allowable inlet temperature of the coolant and the current actual outlet temperature. The coolant inlet temperature is limited by the dew point temperature.

[0078] In this embodiment, the first saturation rate is set to 90%, and the second saturation rate is set to 70%. It is understood that the first and second saturation rates are not fixed values ​​and can be set according to actual conditions.

[0079] Specifically, based on the second thermal distribution map, if the saturation rate of the core thermal aggregation zone is >90% (first saturation rate), it indicates that the zone has approached or reached its heat dissipation limit, and it is necessary to activate the cooling resources of the auxiliary peak shaving zone and the fixed load zone, and adjust the liquid cooling control parameters of the corresponding zone.

[0080] If 70% < saturation rate ≤ 90%, it indicates a high load but still controllable, and the liquid cooling control parameters of the auxiliary peak-shaving zone should be adjusted first.

[0081] If the saturation rate is ≤70% (second saturation rate), it indicates a low load, and only the liquid cooling control parameters of this area need to be adjusted.

[0082] Specifically, if the saturation rate of the core thermal aggregation zone is >90% (first saturation rate), it means that the zone has approached or reached its heat dissipation limit. The coolant flow rate of the auxiliary peak shaving zone will be increased by 20% to 30%, the flow rate of the fixed load zone will be increased by 10% to 15%, and the coolant inlet temperature of the auxiliary peak shaving zone will be reduced by 2℃ to 3℃ (with the real-time dew point temperature as a safety constraint). If 70% < saturation rate ≤ 90%, it indicates a high load but still controllable. The control module only adjusts the liquid cooling control parameters of the auxiliary peak-shaving zone: increases its coolant flow rate by 10% to 15% in a linear proportion, and reduces the coolant inlet temperature by 1°C to 2°C to share some of the heat dissipation pressure in the core area. If the saturation rate is ≤70% (second saturation rate), it indicates a low load. The control module only adjusts the liquid cooling control parameters of the core thermopolymerization zone itself: appropriately reducing its coolant flow rate by 5% to 15% and increasing the coolant inlet temperature by 1℃ to 2℃ to save energy consumption of the cooling system.

[0083] In some embodiments of the present invention, the analysis module samples the junction temperature of each chip in the data center at preset time intervals, assigns weight coefficients based on the interval where the chip junction temperature is located, determines the cumulative lifespan consumption of each chip, and the control module executes the corresponding chip protection strategy.

[0084] Specifically, the lifetime reduction threshold is set based on the weighted cumulative operating time of the chip junction temperature. The analysis module samples the chip junction temperature at preset time intervals (10 minutes in this embodiment), and assigns weight coefficients according to the interval in which the chip junction temperature falls. In this embodiment, the settings are as follows: Below 60℃, set the weight to 0.001 lifespan consumption / hour; 60℃~75℃, set weight 0.01 lifespan consumption / hour; 75℃~85℃, set weight 0.05 for lifespan consumption per hour; 85℃~95℃, set weight 0.20 for lifespan consumption per hour; For temperatures above 95℃, set a weight of 0.50 for lifespan consumption per hour.

[0085] The product of the cumulative running time (in hours) of each interval and the weighting coefficient is used as the cumulative lifetime consumption. When the cumulative consumption reaches the corresponding preset lifetime loss threshold, this embodiment sets: When the cumulative lifespan consumption reaches 0.5, the analysis module issues a warning signal. When the cumulative lifespan consumption reaches 0.7, the analysis module issues a migration action signal. When the cumulative lifespan consumption reaches 0.9, the analysis module issues a shutdown signal. The control module executes and triggers the corresponding protection action.

[0086] It is understood that the weighting coefficients and preset lifetime degradation thresholds can be adjusted based on chip manufacturer specifications and actual operating experience. Please refer to [link / reference]. Figure 4 As shown, it is a control logic flowchart of an embodiment of the present invention.

[0087] In some embodiments, the overall control logic of the device of the present invention is as follows: After the system starts up, it first collects data through the heat generation information acquisition module and the environmental information acquisition module; The analysis module determines the seasonal operating condition heat data adjustment range based on the environmental data, and determines several comprehensive thermal indices of the data center based on the seasonal operating condition heat data adjustment range. The thermal management area includes a core heat aggregation area, an auxiliary peak shaving area, and a fixed load area. The control module determines each thermal management zone based on the threshold of each thermal management zone, and adjusts the load distribution of each cabinet based on the distribution status of each thermal management zone; Finally, the system determines whether to execute the corresponding chip protection strategy to optimize equipment lifespan based on the hardware health model, and continues to cycle through the above steps to achieve intelligent management of data center thermal management.

[0088] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A seasonal thermal management device suitable for data centers, characterized in that, include: The heat generation information acquisition module is used to acquire temperature data of each rack inside the data center, wherein the temperature data includes coolant inlet temperature, coolant outlet temperature and chip junction temperature data; The environmental information acquisition module is used to acquire environmental data from the data center. The analysis module determines the seasonal operating condition thermal data adjustment range based on the environmental data, and determines several comprehensive thermal indices of the data center based on the seasonal operating condition thermal data adjustment range. The thermal management area includes a core thermal aggregation area, an auxiliary peak shaving area, and a fixed load area. The control module determines each thermal management zone based on its threshold values, and adjusts the load distribution of each rack based on the distribution status of each thermal management zone. In addition, the saturation rate is determined based on the distribution status of the thermal management area after the load distribution is adjusted, and the liquid cooling control parameters of each thermal management area are adjusted based on the saturation rate. The liquid cooling control parameters include the coolant flow distribution and the coolant inlet temperature setpoint.

2. The seasonal thermal management device for data centers according to claim 1, characterized in that, The analysis module also includes determining a first thermal distribution map based on the location coordinates of each cabinet and the comprehensive thermal index.

3. The seasonal thermal management device for data centers according to claim 1, characterized in that, The control module is also used to update several comprehensive thermal indices of the corresponding thermal management area based on the load distribution adjustment.

4. The seasonal thermal management device for data centers according to claim 2, characterized in that, The analysis module also determines the adjustment range of seasonal operating condition thermal data based on the predicted values ​​of environmental data.

5. The seasonal thermal management device for data centers according to claim 4, characterized in that, The control module adjusts the load distribution of each cabinet based on the distribution status of each thermal management zone, including: The service load migration targets are determined based on the comparison between the chip junction temperature and the chip junction temperature threshold in the core thermal aggregation zone. The service load of the service load migration targets is then adjusted to adjust the load distribution of each cabinet.

6. The seasonal thermal management device for data centers according to claim 5, characterized in that, A second thermal distribution map is determined based on several comprehensive thermal indices of the thermal management area that are reacquired after a preset time period.

7. The seasonal thermal management device for data centers according to claim 6, characterized in that, The control module adjusts the liquid cooling control parameters of each thermal management zone based on the saturation rate of the core thermal aggregation zone in the second thermal distribution map, including: When the saturation rate of the core thermal polymerization zone is greater than the first saturation rate, the control module adjusts the liquid cooling control parameters of the auxiliary peak shaving zone and the fixed load zone. When the saturation rate of the core thermopolymerization zone is less than or equal to the first saturation rate and greater than the second saturation rate, the control module adjusts the liquid cooling control parameters of the auxiliary peak-shaving zone. When the saturation rate of the core thermal polymerization zone is less than or equal to the second saturation rate, the control module adjusts the liquid cooling control parameters of the core thermal polymerization zone.

8. The seasonal thermal management device for data centers according to claim 1, characterized in that, The environmental data also includes: Dew point temperature data calculated based on relative humidity data from both the outside and inside of the data center; Based on medium- and long-term meteorological forecast data of the data center's location, a dew point temperature curve was plotted.

9. The seasonal thermal management device for data centers according to claim 8, characterized in that, The control module also determines a condensation prevention threshold based on the judgment result that the dew point temperature data is lower than the coolant inlet temperature data, and the dew point temperature data and the coolant inlet temperature data.

10. The seasonal thermal management device for data centers according to any one of claims 1 to 9, characterized in that, The analysis module samples the junction temperature of each chip in the data center at preset time intervals, assigns weight coefficients based on the interval of the chip junction temperature, determines the cumulative lifespan consumption of each chip, and the control module executes the corresponding chip protection strategy.

Images