AI-based data cooling system with self-adaptive regulation and control function

Abstract

The invention provides an AI-based data cooling system with a self-adaptive regulation and control function, and relates to the technical field of data cooling, and the AI-based data cooling system comprises the following steps: collecting equipment operation parameters and operation environment parameters of database equipment; calculating a multi-dimensional load value of the equipment according to the standard operation parameters, and generating a heat dissipation strategy based on a preset AI customized model, the standard environment parameters and the multi-dimensional load of the equipment in combination with a preset temperature safety threshold; and the heat dissipation effect is judged, and the heat dissipation strategy is adjusted. According to the method, rapid closed-loop regulation and control are achieved through dynamic weight optimization and the BP neural network, CPU, memory and I / O multi-dimensional loads are accurately quantified, the temperature trend is pre-judged, and PUE and energy consumption of a data center are remarkably reduced by means of a hierarchical priority strategy; through an online learning model and execution deviation alarm and feedback optimization, manual intervention is reduced, the adaptability is improved, and the deployment cost is reduced through lightweight design.

Description

Technical Field

[0001] This invention relates to the field of data cooling technology, and more particularly to an AI-based data cooling system with adaptive control capabilities. Background Technology

[0002] With the deep penetration of technologies such as cloud computing, artificial intelligence, big data analytics, and the Internet of Things, the global data volume is growing exponentially, and the density and performance of computing devices in scenarios such as data centers, supercomputing centers, and edge computing nodes are continuously improving. While the miniaturization of manufacturing processes and the leap in computing power of high-performance chips (such as GPUs, TPUs, and CPUs) have greatly improved data processing capabilities, they have also led to single-chip power consumption exceeding 300 watts or even higher, and local heat flux density reaching hundreds of watts per square centimeter, far exceeding the limits of traditional air cooling.

[0003] Against this backdrop, thermal management has become a core bottleneck restricting the release of computing power—excessive temperature directly reduces the reliability of electronic components (such as decreased semiconductor carrier mobility and accelerated solder joint fatigue), and may even cause equipment failure. Therefore, the core objective of a data cooling system is to maintain stable operation of the equipment within a safe temperature range through efficient heat transfer and dissipation mechanisms, while minimizing its own energy consumption as much as possible.

[0004] Data cooling systems integrate knowledge from multiple disciplines, including thermodynamics, fluid mechanics, materials science, and intelligent control. From early data center air conditioning and hot / cold aisle isolation to the gradual application of liquid cooling technology, and further to innovative explorations based on phase change materials and microchannel heat dissipation, its development has consistently revolved around two main themes: "improving heat exchange efficiency" and "optimizing energy utilization." Furthermore, the widespread adoption of the green data center concept has further promoted the integration of technologies such as natural cooling and waste heat recovery, making the adaptability of cooling systems to renewable energy sources and regional climate characteristics a crucial design dimension.

[0005] Existing data cooling systems often rely on a single dimension or fixed weights, resulting in insufficient assessment accuracy; control often adopts static strategies or simple algorithms, lacking temperature trend prediction and real-time closed-loop response, which can easily lead to overheating or excessive heat dissipation; parameters are configured based on human experience, lacking online learning and execution deviation feedback mechanisms, resulting in poor adaptability and frequent human intervention.

[0006] To address the shortcomings of the existing technologies, this technical solution proposes an AI-based data cooling system with adaptive control capabilities. Summary of the Invention

[0007] This invention provides an AI-based data cooling system with adaptive control function to solve the defects of existing data cooling systems that lack temperature trend prediction and real-time closed-loop response, which easily lead to overheating or excessive heat dissipation.

[0008] On one hand, this invention provides an AI-based data cooling system with adaptive control capabilities, comprising: a data acquisition module, a data processing module, a multi-dimensional load calculation module, a strategy optimization module, a heat dissipation execution module, and a status feedback module. The data acquisition module collects the operating parameters and environmental parameters of the database device. The data processing module preprocesses the operating parameters and environmental parameters to obtain standard operating parameters and standard environmental parameters. The multi-dimensional load calculation module calculates the multi-dimensional load values ​​of the device based on the standard operating parameters. The strategy optimization module generates a heat dissipation strategy based on a preset AI-customized model, standard environmental parameters, and the multi-dimensional load of the device, combined with a preset temperature safety threshold. The heat dissipation execution module dissipates heat from the database device based on the heat dissipation strategy and outputs a heat dissipation execution log. The status feedback module judges the heat dissipation effect based on the heat dissipation execution log and the real-time temperature of the database device, determines whether there is an execution deviation, and adjusts the heat dissipation strategy accordingly.

[0009] According to the AI-based data cooling system with adaptive control function provided by the present invention, the data processing module includes a high-efficiency data cleaning unit and a rapid standardization unit; the high-efficiency data cleaning unit is used to preprocess the equipment operating parameters and operating environment parameters to obtain processed comprehensive parameter data; the rapid standardization unit is used to map the processed comprehensive parameter data to the [0,1] interval according to the historical maximum / minimum value range of the pre-stored equipment operating parameters and operating environment parameters to obtain standard operating parameters and standard environmental parameters.

[0010] The AI-based data cooling system with adaptive control function provided by the present invention includes a multi-dimensional load calculation module comprising a weight optimization unit, an I / O load quantization unit, and a load level and type determination unit. The weight optimization unit is used to train an online learning model based on historical operating data of the database and output multi-parameter dynamic weight coefficients in real time. The I / O load quantization unit is used to calculate the multi-dimensional load value of the device using a weighted fusion algorithm. The load level and type determination unit is used to classify the load level according to the multi-dimensional load value of the device and identify the load type through an AI model.

[0011] According to the AI-based data cooling system with adaptive control function provided by the present invention, the step of the I / O load quantization unit calculating the multi-dimensional load value of the device includes: Extract standardized CPU utilization, memory utilization, IOPS, disk read / write bandwidth, and I / O latency from the standard operating parameters; Based on the standardized CPU utilization, memory utilization, IOPS, disk read / write bandwidth, and I / O latency, a weighted fusion algorithm is used to calculate the comprehensive I / O index. Based on the multi-parameter dynamic weighting coefficients, the load value is calculated through a multi-dimensional load quantification model to obtain the multi-dimensional load value of the device.

[0012] According to the AI-based data cooling system with adaptive control function provided by the present invention, the formula for calculating the comprehensive I / O index in the I / O load quantization unit is expressed as follows: Where L is the comprehensive load value of the device across multiple dimensions, with a value range of [0,1]; CPUutil is the standardized CPU utilization rate; Memutil is the standardized memory utilization rate; IOnorm is the standardized I / O comprehensive index; Wcpu is the dynamic weight coefficient of the CPU utilization dimension; Wmem is the dynamic weight coefficient of the memory utilization dimension; and Wio is the dynamic weight coefficient of the I / O operation dimension.

[0013] According to the AI-based data cooling system with adaptive control function provided by the present invention, the strategy optimization module includes a database customized feature engineering unit, a lightweight AI inference unit, and a threshold management and model optimization unit. The database customized feature engineering unit is used to construct a feature set adapted to the high-concurrency I / O and transactional load characteristics of the database, and to remove redundant features through a feature screening algorithm. The lightweight AI inference unit is used to obtain accurate heat dissipation adjustment parameters by using a BP neural network model, preloading inference weights and using a caching calculation mechanism, and to correct the accurate heat dissipation adjustment parameters by combining the instantaneous trend line fitted by the temperature data of the most recent second. The threshold management and model optimization unit is used to store and manage preset temperature safety thresholds, monitor the difference between the temperature of the monitoring point and the preset temperature safety threshold in real time, and dynamically adjust the priority of the heat dissipation strategy.

[0014] According to the AI-based data cooling system with adaptive control function provided by the present invention, the step of the lightweight AI inference unit correcting the precise heat dissipation adjustment parameters includes: Obtain temperature data from monitoring points in the standard operating parameters; A linear fitting algorithm is used to perform trend analysis on the temperature data at the monitoring points, generate a temperature change trend line, and calculate the slope of the temperature change. The precise heat dissipation adjustment parameters are corrected based on the slope of temperature change.

[0015] According to the AI-based data cooling system with adaptive control function provided by the present invention, the step of dynamically adjusting the priority of heat dissipation strategy by the threshold management and model optimization unit includes: Based on the standard environmental parameters of each monitoring point, the difference between the temperature data of each monitoring point and the preset temperature safety threshold is calculated to obtain the temperature difference. Priority levels are determined based on temperature difference; The priority of the heat dissipation strategy is dynamically adjusted based on the priority level.

[0016] According to the AI-based data cooling system with adaptive control function provided by the present invention, the status feedback module includes a heat dissipation effect judgment unit, a deviation processing and alarm unit, and a strategy correction feedback unit. The heat dissipation effect judgment unit is used to obtain the actual operating parameters of the heat dissipation components and the real-time temperature data of the database device in the heat dissipation execution log, calculate the deviation value between the actual heat dissipation effect and the expected effect, determine whether there is an execution deviation, and output the judgment result. The deviation processing and alarm unit is used to immediately trigger multi-channel alarms and start the backup heat dissipation strategy when the execution deviation is determined to exist. The strategy correction feedback unit is used to transmit the judgment result to the strategy optimization module in real time.

[0017] According to the AI-based data cooling system with adaptive control function provided by the present invention, the backup heat dissipation strategy in the deviation processing and alarm unit includes: when an execution deviation occurs in air cooling, the backup cooling fan is activated, and the speed of the backup fan is directly adjusted to the maximum heat dissipation power level under the corresponding load level, while the faulty fan is turned off; when an execution deviation occurs in water cooling, the system switches to air cooling auxiliary heat dissipation mode, increases the heat dissipation power of all air cooling executions, and issues a water cooling system fault alarm; when an execution deviation occurs in auxiliary heat dissipation, the local heat dissipation of all database devices is linked to increase the heat dissipation power level by one level, so that the local temperature of the device does not exceed the temperature safety threshold; after all backup heat dissipation strategies are activated, real-time temperature data of the device is continuously collected, and if the temperature does not drop to the safe range, an emergency alarm is triggered and fault details are recorded.

[0018] This invention provides an AI-based data cooling system with adaptive control capabilities. Through dynamic weight optimization and a BP neural network, it achieves rapid closed-loop control, accurately quantifying multi-dimensional loads of CPU, memory, and I / O and predicting temperature trends. Furthermore, it significantly reduces data center PUE and energy consumption based on a hierarchical priority strategy. By using online learning models and execution deviation alarms and feedback optimization, it reduces manual intervention and improves adaptability. Its lightweight design lowers deployment costs, ultimately achieving a comprehensive improvement in energy efficiency, security, response speed, and operational efficiency, significantly reducing TCO. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0020] Figure 1This is a schematic diagram of the structure of an AI-based data cooling system with adaptive control function provided in an embodiment of the present invention; Figure 2 This is a flowchart illustrating how the I / O load quantization unit calculates the multi-dimensional load value of the device, as provided in an embodiment of the present invention. Figure 3 This is a flowchart illustrating how the lightweight AI inference unit provided in this embodiment of the invention corrects the precise heat dissipation adjustment parameters. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0022] The following is combined Figures 1-3 The present invention describes an AI-based data cooling system with adaptive control functionality.

[0023] like Figures 1-3 As shown, the AI-based data cooling system with adaptive control function provided in this embodiment of the invention includes: The data acquisition module is used to collect the device operating parameters and operating environment parameters of the database device.

[0024] The data acquisition module includes a device operating parameter acquisition unit and an ambient temperature acquisition unit. The device operating parameter acquisition unit interfaces with the database device's motherboard BIOS and operating system monitoring interface, using a timed polling method to collect real-time CPU utilization. The specific steps are as follows: pre-configure the polling period; establish a stable data transmission channel through interface protocol negotiation; send parameter acquisition request frames to the BIOS and operating system monitoring interface according to the configured period, with each request frame carrying a unique device identifier and the type of parameter to be acquired; receive the raw data frames returned by the interface, parse and extract the raw CPU utilization data, perform CRC verification on the data frames to ensure data integrity, trigger a retransmission mechanism if verification fails (maximum of 3 retransmissions), and record an acquisition anomaly if it still fails; temporarily store the verified CPU utilization data in the acquisition buffer and mark the acquisition timestamp.

[0025] The system memory management interface is used to obtain the total physical memory capacity, used capacity, and cache usage capacity, and then calculate the actual memory utilization rate. Specifically, this involves: calling the system memory management API to obtain the three types of raw data: total physical memory capacity, used capacity, and cache usage capacity; calculating the memory utilization rate using the formula "Actual memory utilization rate = (Used capacity - Cache usage capacity) / Total physical memory capacity × 100%"; and validating the calculation results within the range of [0, 100%]. If the result exceeds this range, it is considered invalid data, triggering a re-collection process.

[0026] The system interfaces with the database storage controller and database management system to collect I / O operation parameters such as IOPS, disk read / write bandwidth, and I / O wait time. The steps include: establishing a dedicated I / O monitoring session with the storage controller and database management system, configuring the data collection granularity to millisecond level; for IOPS, counting the total number of disk read / write operations per unit time, and removing duplicate and invalid operation records; for disk read / write bandwidth, calculating the total amount of data read / written per unit time and converting it to MB / s; for I / O wait time, recording the time difference from initiation to completion of each I / O request, and taking the average value per unit time as the I / O wait time parameter.

[0027] The ambient temperature acquisition unit is used to deploy high-precision digital temperature sensors at key locations such as CPU heatsinks, memory slots, storage device surfaces, and server racks in the data center to collect temperature data from each monitoring point in real time. The acquisition steps include: address encoding of the temperature sensors at each deployment location to establish a mapping table between sensor addresses and monitoring point locations; configuring the sensor sampling frequency to 10Hz and communicating with the acquisition unit's main controller using the I2C bus protocol; the main controller sequentially reading the raw temperature data from each sensor according to address order and filtering the data; binding the filtered temperature data with the corresponding monitoring point location information and acquisition timestamp, and storing it in the acquisition buffer.

[0028] The data processing module is used to preprocess the equipment operating parameters and operating environment parameters to obtain standard operating parameters and standard environmental parameters.

[0029] The data processing module includes a high-efficiency data cleaning unit and a rapid standardization unit.

[0030] The high-efficiency data cleaning unit is used to remove outliers from the collected equipment operating parameters and operating environment parameters using the simplified 3σ criterion, and to supplement missing data using the nearest neighbor interpolation method, with a total data cleaning time of ≤10ms.

[0031] The fast standardization unit is used to map the device operating parameters and operating environment parameters of different dimensions to the [0,1] interval by using the min-max standardization algorithm based on the pre-stored historical maximum / minimum value range of each parameter, so as to obtain standard operating parameters and standard environment parameters. The standardization time for a single set of data is ≤5ms.

[0032] The multi-dimensional load calculation module is used to calculate the multi-dimensional load of the device based on standard operating parameters. It adopts an AI dynamic weight optimization mechanism, trains a weight optimization model based on historical operating data in the database, and takes into account input features such as CPU core usage distribution, memory page miss rate, I / O queue length, and transaction throughput. The dynamic weight coefficients are updated every 100ms.

[0033] The multi-dimensional load calculation module includes a weight optimization unit, an I / O load quantization unit, and a load level and type determination unit.

[0034] The weight optimization unit is used to train the online learning model based on historical runtime data from the database, and outputs dynamic weight coefficients for Wcpu, Wmem, and Wio in real time, satisfying Wcpu + Wmem + Wio = 1. The initial weight references are CPU utilization 0.35, memory utilization 0.25, and overall IO index 0.4, with a weight update cycle of 100ms.

[0035] The steps for training an online learning model include: Historical operational data from different database load scenarios were collected offline, and input features such as CPU core usage distribution, memory page miss rate, I / O queue length, and transaction throughput were extracted. The corresponding optimal weight combinations were labeled. An online gradient descent algorithm was used to train the weight optimization model and initialize the model parameters. Every 100ms, the latest input feature data is obtained from the normalization unit, the input features are input into the trained online learning model, and the initial weight coefficients Wcpu0, Wmem0, and Wio0 are output. The initial weight coefficients are normalized to ensure that Wcpu + Wmem + Wio = 1. The normalization formula is as follows: Wmem and Wio are calculated similarly. If the difference between the model output weight and the weight of the previous period is greater than 0.1, smoothing is performed using the formula "current weight = previous period weight × 0.7 + current model output weight × 0.3" to ensure a smooth weight transition.

[0036] The I / O load quantization unit is used to calculate the comprehensive I / O metric IOnorm using a weighted fusion algorithm. It integrates standardized data from three dimensions: IOPS, disk read / write bandwidth, and I / O latency, resulting in the load value, expressed by the formula: Where L is the comprehensive load value of the device across multiple dimensions, ranging from [0,1], used to classify load levels. CPUutil is the standardized CPU utilization rate, standardized by the min-max data processing module and mapped to the [0,1] interval. Memutil is the standardized memory utilization rate, mapped to the [0,1] interval. IOnorm is the standardized comprehensive I / O indicator, integrating multi-dimensional I / O parameters. IOPSnorm is the standardized disk read / write count per unit time. Bandwidthnorm is the standardized disk read / write bandwidth, and IOwaitnorm is the standardized I / O wait time. Wcpu is the dynamic weighting coefficient for the CPU utilization dimension, Wmem is the dynamic weighting coefficient for the memory utilization dimension, and Wio is the dynamic weighting coefficient for the I / O operation dimension.

[0037] The load level and type determination unit classifies load values ​​calculated by a multi-dimensional load quantification model into five levels: no load (L∈[0,0.2)), light load (L∈[0.2,0.4)), medium load (L∈[0.4,0.6)), heavy load (L∈[0.6,0.8)), and full load (L∈[0.8,1.0]). It identifies load types such as OLTP (transactional, short transaction I / O intensive) and OLAP (analytical, CPU / memory intensive) using an AI model. Specific steps include: Obtain the multi-dimensional comprehensive load value L; traverse the preset load level threshold range to determine the load level to which L belongs; record the load level, the corresponding L value, and the determination timestamp.

[0038] Extract feature sets, including transaction throughput, single transaction duration, CPU utilization percentage, and IOnorm percentage.

[0039] The feature set is input into a pre-trained load type recognition AI model, which outputs load type prediction results (OLTP / OLAP / mixed load) and confidence scores.

[0040] If the confidence level is less than 0.8, the historical load type data (last 5 minutes) is used for auxiliary judgment, and the type with the highest frequency is taken as the final result.

[0041] The strategy optimization module generates heat dissipation strategies based on a preset AI-customized model, standard environmental parameters, and multi-dimensional device load, combined with preset temperature safety thresholds. The AI-customized model is a lightweight BP neural network (3-layer network structure, 16 hidden layer nodes), supports INT8 quantization, and its input features include dynamic weight load values, load type labels, and temperatures at each monitoring point, with an inference latency of ≤50ms. Preset temperature safety thresholds include a normal CPU operating temperature ≤85℃ and a normal memory operating temperature ≤80℃. When the monitoring point temperature approaches the threshold, the priority of the heat dissipation strategy is automatically increased. The AI-customized model is periodically fine-tuned and optimized using the latest operational data.

[0042] The strategy optimization module includes a database-customized feature engineering unit, a lightweight AI inference unit, and a threshold management and model optimization unit.

[0043] The database customization feature engineering unit is used to build feature sets adapted to the high-concurrency I / O and transactional load characteristics of databases. These include OLTP-specific features (short transaction throughput, transaction response time, number of I / O operations per transaction), I / O concurrency features (I / O queue depth, percentage of parallel read / write operations across multiple disks), and resource contention features (CPU core switching frequency, memory lock wait time). Redundant features are then removed using feature filtering algorithms. OLTP is the domain where daily business transactions are processed in the database field. It focuses on data creation, deletion, modification, and querying, requiring the system to respond quickly to user requests and ensure data consistency and integrity. Specifically, this includes: Raw data is obtained from the multi-dimensional load calculation module and the data acquisition module, and eight basic features are extracted, including OLTP-specific features, I / O concurrency features, and resource contention features.

[0044] The basic features are processed by derivation, and the first difference and moving average of the features are calculated to expand the feature dimension to 16 dimensions.

[0045] The variance selection method was used to remove low-discrimination features with variance < 0.01. The Pearson correlation coefficient between each feature and the heat dissipation effect index was calculated, and features with an absolute value of correlation coefficient > 0.3 were retained.

[0046] The redundancy between features is calculated using the mutual information method. Redundant features with mutual information values ​​> 0.8 are removed, and finally, 8-10 core features are retained. The filtered core features are mapped to the [0,1] interval and encapsulated into the input format of the AI ​​model.

[0047] The lightweight AI inference unit employs a 3-layer BP neural network model quantized with INT8, pre-loads inference weights, and uses a caching mechanism to achieve an inference latency of ≤50ms, outputting precise thermal adjustment parameters. The adjustment parameters are then corrected by fitting an instantaneous trend line based on the temperature data from the most recent second.

[0048] The steps for preloading inference weights include: Load the INT8 quantized 3-layer BP neural network model and pre-trained weights, and initialize the input and output buffers; Hardware acceleration configuration is implemented for the model inference chain based on GPU / MCU hardware acceleration. The core feature set is obtained from the database customized feature engineering unit, and the preset temperature safety threshold is obtained from the threshold management unit; The feature set is input into the AI ​​model, forward inference calculation is performed, and the initial heat dissipation adjustment parameters are output. Record the inference time. If the time is greater than 40ms, trigger lightweight optimization.

[0049] The steps for the lightweight AI inference unit to correct precise thermal adjustment parameters include: Obtain the temperature data of the monitoring points within the most recent second from the standard operating parameters; A linear fitting algorithm is used to perform trend analysis on the temperature data at the monitoring points, generate a temperature change trend line, and calculate the slope of the temperature change. Based on the precise heat dissipation adjustment parameters, the parameters are corrected according to the slope of temperature change. The correction strategy includes: if the temperature data at the monitoring point shows an upward trend and the difference from the preset temperature safety threshold is ≤5℃, then the heat dissipation power is increased by 5%-10% based on the original adjustment parameters; if the temperature data at the monitoring point shows a downward trend and the difference from the preset temperature safety threshold is ≥10℃, then the heat dissipation power is decreased by 5%-10% based on the original adjustment parameters; if the temperature data at the monitoring point shows a stable trend, then the original precise heat dissipation adjustment parameters are kept unchanged.

[0050] The threshold management and model optimization unit stores and manages preset temperature safety thresholds, monitors the difference between the temperature at monitoring points and the thresholds in real time, and dynamically adjusts the priority of heat dissipation strategies. A regular optimization mechanism is established, using the latest operational data to fine-tune the AI-customized model weekly, improving heat dissipation efficiency and energy economy.

[0051] The threshold management method is as follows: preset temperature safety thresholds are stored according to the type of equipment component, a threshold configuration table is established, and users can manually modify it through the system interface; temperature data of each monitoring point is obtained from the data acquisition module in real time, and the difference between the temperature of each monitoring point and the corresponding threshold is calculated.

[0052] The steps involved in threshold management and model optimization units to dynamically adjust the priority of heat dissipation strategies include: Based on the standard environmental parameters of each monitoring point, the difference between the temperature data of each monitoring point and the preset temperature safety threshold is calculated to obtain the temperature difference. Priority levels are determined based on the temperature difference, including: high priority when the temperature difference is ≤3℃; medium priority when 3℃ < temperature difference ≤8℃; and low priority when the temperature difference is >8℃. The priority of the heat dissipation strategy is dynamically adjusted according to the priority level. When the priority is high, the current heat dissipation strategy is executed first, and unnecessary strategy optimization iterations are suspended to ensure timely heat dissipation response. When the priority is medium, the heat dissipation strategy is executed in the normal sequence, and strategy optimization analysis is carried out simultaneously. When the priority is low, the heat dissipation execution power can be reduced to save energy consumption, while the fine-tuning and optimization of the AI ​​customized model can be promoted normally.

[0053] The heat dissipation execution module is used to cool the database device based on a heat dissipation strategy and outputs a heat dissipation execution log. It includes an air-cooling execution unit, a water-cooling execution unit, and an auxiliary heat dissipation unit. The air-cooling execution unit adjusts the fan speed via PWM signals (adjustment range 30%~100%), the water-cooling execution unit adjusts the water pump speed and the exhaust fan speed, and the auxiliary heat dissipation unit connects to the data center air conditioning system for overall environmental optimization. The execution latency is ≤100ms, and the heat dissipation execution log records the actual operating parameters of the heat dissipation components.

[0054] The status feedback module is used to determine the cooling effect based on the cooling execution log and the real-time temperature of the database device, identify any execution deviations, and adjust the cooling strategy accordingly. It uses an SPI high-speed data transmission interface to collect the cooling execution status, with a feedback latency of ≤20ms. Execution deviations are determined by comparing the actual parameters in the cooling execution log with the strategy command parameters. When a deviation occurs, an audible and visual alarm, SMS notification, or system pop-up alarm is immediately issued, and a backup cooling strategy is triggered. Simultaneously, the feedback data is transmitted to the strategy optimization module for real-time correction of the cooling strategy.

[0055] The status feedback module includes a heat dissipation effect judgment unit, a deviation processing and alarm unit, and a strategy correction feedback unit.

[0056] The heat dissipation effect judgment unit is used to obtain the actual operating parameters of the heat dissipation components and the real-time temperature data of the database device in the heat dissipation execution log, calculate the deviation value between the actual heat dissipation effect and the expected effect, and determine whether there is an execution deviation.

[0057] The deviation handling and alarm unit is used to immediately trigger multi-channel alarms and activate backup cooling strategies when an execution deviation is detected, ensuring continuous heat dissipation. Specific strategies include: when an air-cooled execution unit experiences an execution deviation, activating the backup cooling fan and directly adjusting its speed to the maximum cooling power level for the corresponding load level, while simultaneously shutting down the faulty fan; when a water-cooled execution unit experiences an execution deviation, switching to air-cooled auxiliary cooling mode, increasing the cooling power of all air-cooled execution units to 90% of their rated power, and simultaneously issuing a water-cooling system fault alarm; when an auxiliary cooling unit experiences an execution deviation, coordinating with all database devices to increase their local cooling power by one level, ensuring that the local temperature of the devices does not exceed the safe temperature threshold; after all backup cooling strategies are activated, continuously collecting real-time temperature data of the devices, and if the temperature does not drop to a safe range within 1 minute, triggering an emergency alarm and recording fault details.

[0058] The strategy correction feedback unit is used to transmit the heat dissipation effect judgment results and execution deviation information to the strategy optimization module in real time, providing data support for the correction of inference parameters and fine-tuning of AI customized models, forming a closed-loop control of "decision-execution-feedback-optimization".

[0059] In summary, the AI-based adaptive data cooling system provided by this invention achieves rapid closed-loop control through dynamic weight optimization and a BP neural network. It accurately quantifies the multi-dimensional load of CPU, memory, and I / O and predicts temperature trends, while significantly reducing data center PUE and energy consumption through a hierarchical priority strategy. The system reduces manual intervention and improves adaptability through online learning models and execution deviation alarms and feedback optimization. Its lightweight design reduces deployment costs, ultimately achieving a comprehensive improvement in energy efficiency, security, response speed, and operational efficiency, significantly reducing TCO.

[0060] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.

[0061] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An AI-based data cooling system with adaptive regulation function, characterized in that, The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device.

2. The AI-based data cooling system with adaptive regulation function according to claim 1, characterized in that, The application relates to a database equipment intelligent cooling method and device. 3.The AI-based data cooling system with adaptive regulation function of claim 1, wherein, The application relates to a database equipment intelligent cooling method and device.

4. The AI-based data cooling system with adaptive regulation function according to claim 3, characterized in that, The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device.

5. The AI-based data cooling system with adaptive regulation function according to claim 4, characterized in that, The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment intelligent cooling method and device. The application relates to a database equipment 6.The AI-based data cooling system with adaptive regulation function of claim 1, wherein, The strategy optimization module comprises a database customization feature engineering unit, a lightweight AI inference unit and a threshold management and model optimization unit; the database customization feature engineering unit is configured to construct a feature set that is adapted to the high-concurrency I / O and transactional load characteristics of the database, and to remove redundant features through a feature screening algorithm; The lightweight AI inference unit is configured to use a BP neural network model, pre-load inference weights and use a cache computing mechanism to obtain accurate heat dissipation adjustment parameters, and correct the accurate heat dissipation adjustment parameters in combination with an instantaneous trend line fitted from temperature data in the last 1 second; the threshold management and model optimization unit is configured to store and manage preset temperature safety thresholds, monitor the difference between the temperature of a monitoring point and the preset temperature safety thresholds in real time, and dynamically adjust the priority of the heat dissipation strategy.

7. The AI-based data cooling system with adaptive regulation function according to claim 6, characterized in that, The step of correcting the accurate heat dissipation adjustment parameters by the lightweight AI inference unit comprises: obtaining monitoring point temperature data in the standard operating parameters; performing trend analysis on the monitoring point temperature data using a linear fitting algorithm to generate a temperature change trend line and calculate a temperature change slope; correcting the accurate heat dissipation adjustment parameters in combination with the temperature change slope.

8. The AI-based data cooling system with adaptive regulation function according to claim 7, characterized in that, The step of dynamically adjusting the priority of the heat dissipation strategy by the threshold management and model optimization unit comprises: calculating the difference between the temperature data of each monitoring point and the preset temperature safety thresholds to obtain temperature differences, according to the standard environmental parameters of each monitoring point; dividing priority levels according to the temperature differences; dynamically adjusting the priority of the heat dissipation strategy according to the priority levels. 9.The AI-based data cooling system with adaptive regulation function of claim 1, wherein, The state feedback module comprises a heat dissipation effect judgment unit, a deviation processing and alarm unit and a strategy correction feedback unit; the heat dissipation effect judgment unit is configured to obtain actual operating parameters of heat dissipation components and real-time temperature data of database equipment in the heat dissipation execution log, calculate a deviation value between actual and expected heat dissipation effects, determine whether there is an execution deviation, and output a determination result; The deviation processing and alarm unit is configured to trigger multi-channel alarms immediately when it is determined that there is an execution deviation, and simultaneously start a backup heat dissipation strategy; The strategy correction feedback unit is configured to transmit the determination result to the strategy optimization module in real time.

10. The AI-based data cooling system with adaptive regulation function according to claim 9, characterized in that, The backup heat dissipation strategy in the deviation processing and alarm unit comprises: when air cooling has an execution deviation, starting a backup heat dissipation fan, directly adjusting the speed of the backup fan to the maximum heat dissipation power level under the corresponding load level, and simultaneously shutting down the faulty fan; when water cooling has an execution deviation, switching to an air cooling auxiliary heat dissipation mode, increasing the heat dissipation power of all air cooling executions, and simultaneously issuing a water cooling system fault alarm; when auxiliary heat dissipation has an execution deviation, increasing the local heat dissipation of all database equipment by one heat dissipation power level so that the local temperature of the equipment does not exceed the temperature safety threshold; after all backup heat dissipation strategies are started, real-time temperature data of the equipment is continuously collected, and if the temperature has not decreased to a safe range, an emergency alarm is triggered and fault details are recorded.

Images