AI tuning of liquid cooling system and optimal PUE implementation method based on multi-parameter fusion

Abstract

The present invention relates to the field of liquid cooling tuning technology, specifically to a method for implementing AI tuning and optimal PUE of a liquid cooling system with multi-parameter fusion, comprising the following steps: collecting temperature, flow rate and pump speed data of the liquid cooling system through sensors, calculating current performance indicators, and generating real-time performance monitoring values; according to the real-time performance monitoring values, adjusting the dynamic inertia weight and acceleration coefficient in the particle swarm algorithm, optimizing the global search strategy, and generating optimized search parameters. In the present invention, a particle swarm optimization algorithm is used to adjust the dynamic inertia weight and acceleration coefficient, which can quickly find parameter configurations under various operating conditions, thereby improving adaptive capabilities and operating efficiency. The introduction of long short-term memory networks makes it possible not only to predict future cooling needs, but also to make real-time adjustments to pump speed and valve opening based on these predictions, thereby optimizing energy use and enhancing cooling effects.

Description

Technical Field

[0001] The present invention relates to the field of liquid cooling tuning technology, and in particular to a method for realizing multi-parameter fusion liquid cooling system AI tuning and optimal PUE. Background Art

[0002] The single-phase immersion liquid cooling system places the server in a liquid cooling box and uses the flow of coolant to remove the heat generated by the server. The heated coolant flows into the plate heat exchanger, exchanges heat with the cooling water, and then flows back into the liquid cooling box.

[0003] Traditional single-phase immersion liquid cooling systems mainly rely on fixed heat exchange mechanisms, which show obvious limitations when dealing with high heat loads or drastic changes in ambient temperature. Due to the lack of real-time data monitoring and automatic adjustment mechanisms, they are not able to respond promptly when encountering emergencies, making it difficult to effectively adapt to load changes, which in turn leads to overheating and increases the risk of failure. Therefore, improvements are needed. Summary of the invention

[0004] The purpose of the present invention is to solve the shortcomings existing in the prior art and propose a multi-parameter fusion liquid cooling system AI tuning and optimal PUE implementation method.

[0005] In order to achieve the above objectives, the present invention adopts the following technical solution, a multi-parameter fusion liquid cooling system AI tuning and optimal PUE implementation method, comprising the following steps:

[0006] The temperature, flow rate and pump speed data of the liquid cooling system are collected through sensors, the current performance index is calculated, and a real-time performance monitoring value is generated; according to the real-time performance monitoring value, the dynamic inertia weight and acceleration coefficient in the particle swarm algorithm are adjusted, the global search strategy is optimized, and the optimized search parameters are generated;

[0007] Based on the optimized search parameters, a particle swarm optimization algorithm is run to obtain a performance evaluation under each parameter configuration and generate a preliminary operation result; by comparing each operation result, a configuration is selected to generate a system configuration;

[0008] Collecting temperature information of the data center, analyzing the data change trend using a long short-term memory network, and obtaining a preliminary cooling demand forecast value; adjusting the pump speed and valve opening according to the preliminary cooling demand forecast value to generate an adjusted output value;

[0009] Monitor the deviation between the adjusted output value and the real-time data, determine whether the deviation exceeds a set safety threshold, and generate a deviation analysis result.

[0010] Preferably, the steps for obtaining the real-time performance monitoring value are:

[0011] The temperature, flow rate and pump speed of the liquid cooling system are collected through sensors, and three types of real-time data are recorded to generate a data set;

[0012] Based on the data set, the real-time performance monitoring value is calculated, and the calculation formula is:

[0013]

[0014] Among them, k T is the square weight of temperature data, k F is the square root weight of the velocity data, k S is the inverse weight of the pump speed, M is the real-time performance monitoring value, T is the temperature, S is the pump speed, and F is the flow rate.

[0015] Preferably, the step of obtaining the optimized search parameters is:

[0016] receiving the real-time performance monitoring value as input to adjust parameters in a particle swarm algorithm;

[0017] Calculate the dynamic inertia weight and acceleration coefficient using the following formula:

[0018]

[0019] and

[0020]

[0021] Where, e is the natural logarithm base, M is the real-time performance monitoring value, w is the dynamic inertia weight, and c is the acceleration coefficient;

[0022] Based on the dynamic inertia weight and acceleration coefficient, the global search strategy is optimized to generate optimized search parameters.

[0023] Preferably, the steps for obtaining the preliminary operation results are:

[0024] The optimized search parameters are loaded, and parameters are configured in the particle swarm optimization algorithm, including adjusting the initial position and speed settings of the particles, starting the algorithm operation process, and obtaining preliminary operation results of the algorithm under each parameter configuration.

[0025] Preferably, the steps of obtaining the system configuration are:

[0026] Based on the preliminary operation results, analyze the fitness value of each particle, evaluate the effectiveness of the parameter configuration corresponding to each particle according to the fitness value, identify the parameter configuration by comparing the fitness of each particle, and form a performance evaluation report;

[0027] Based on the performance evaluation report, according to the convergence speed of the algorithm and the quality of the solution, the parameter configuration is screened to obtain the system configuration.

[0028] Preferably, the steps for obtaining the preliminary cooling demand forecast value are:

[0029] Deploy temperature sensors in the data center to collect temperature data in real time, and send the temperature data to the database through the network for storage to form a comprehensive temperature data set;

[0030] Based on the comprehensive temperature data set, using a long short-term memory network to analyze the temperature change trend and generate a temperature trend analysis report;

[0031] Based on the temperature trend analysis report, a cooling demand forecast value within a certain time range in the future is obtained to obtain a preliminary cooling demand forecast value.

[0032] Preferably, the step of obtaining the adjusted output value is:

[0033] Evaluating whether the current pump speed and valve settings meet the predicted cooling demand based on the preliminary cooling demand prediction value to obtain an adjustment demand analysis;

[0034] Based on the adjustment demand analysis, the required pump speed and valve opening are recalculated using the following formula:

[0035]

[0036] and

[0037]

[0038] Among them, Q c is the cooling demand forecast value, Q max is the maximum cooling capacity, S max and O max are the maximum setting values ​​of pump speed and valve opening respectively, S is the pump speed, and O is the valve opening;

[0039] Based on the recalculated pump speed and valve opening, the pump and valve are adjusted to the new set points to obtain the adjusted output value.

[0040] Preferably, the steps of obtaining the deviation analysis result are:

[0041] Track the adjusted pump speed and valve opening output values ​​to obtain real-time monitoring data sets;

[0042] Based on the real-time monitoring data set, calculating the deviation between the adjusted output value and the real-time monitoring data to form a preliminary deviation analysis result;

[0043] Based on the preliminary deviation analysis results, the deviation situation is evaluated to check whether any deviation exceeds the safety threshold. If so, an alarm is triggered and a deviation analysis result is generated.

[0044] Compared with the prior art, the advantages and positive effects of the present invention are:

[0045] In the present invention, the particle swarm optimization algorithm is used to adjust the dynamic inertia weight and acceleration coefficient, which can quickly find parameter configurations under various operating conditions, thereby improving the adaptive ability and operating efficiency. The introduction of the long short-term memory network not only makes it possible to predict future cooling needs, but also to make real-time adjustments to the pump speed and valve opening based on these predictions, thereby optimizing energy use and enhancing the cooling effect. In addition, through real-time deviation monitoring and analysis, problems that may cause performance degradation or equipment damage can be discovered and adjusted in a timely manner, thereby enhancing reliability and safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a schematic diagram of the steps of the present invention. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0048] See also Figure 1 The present invention provides a technical solution, a multi-parameter fusion liquid cooling system AI tuning and optimal PUE implementation method, including the following steps:

[0049] The temperature, flow rate and pump speed data of the liquid cooling system are collected through sensors, the current performance indicators are calculated, and real-time performance monitoring values ​​are generated; according to the real-time performance monitoring values, the dynamic inertia weight and acceleration coefficient in the particle swarm algorithm are adjusted, the global search strategy is optimized, and the optimized search parameters are generated;

[0050] Based on the optimized search parameters, run the particle swarm optimization algorithm to obtain the performance evaluation under each parameter configuration and generate preliminary operation results; select the configuration by comparing each operation result and generate the system configuration;

[0051] Collect temperature information from the data center, use a long short-term memory network to analyze data change trends, and obtain a preliminary cooling demand forecast value; adjust the pump speed and valve opening according to the preliminary cooling demand forecast value to generate an adjusted output value;

[0052] Monitor the deviation between the adjusted output value and the real-time data, determine whether the deviation exceeds the set safety threshold, and generate deviation analysis results.

[0053] The steps to obtain real-time performance monitoring values ​​are as follows:

[0054] The temperature, flow rate and pump speed of the liquid cooling system are collected through sensors, and three types of real-time data are recorded to generate a data set;

[0055] Based on the data set, the real-time performance monitoring value is calculated using the following formula:

[0056]

[0057] Among them, k T is the square weight of temperature data, k F is the square root weight of the velocity data, k S is the inverse weight of the pump speed, M is the real-time performance monitoring value, T is the temperature, S is the pump speed, and F is the flow rate.

[0058] Specifically, the temperature, flow rate and pump speed of the liquid cooling system are collected respectively through sensors. This process is realized through temperature sensors, flow meters and pump speed sensors. The three work together for data acquisition. Data acquisition is responsible for recording and transmitting data to ensure that the collected data is accurate and real-time. The temperature, flow rate and pump speed data collected by the sensors are transmitted to the processing unit wirelessly or wiredly. The processing unit updates these data to the monitoring interface in real time and stores them in the database for subsequent analysis and historical data comparison.

[0059] The benefit of the formula is that it takes into account the different influences of temperature, flow rate and pump speed, reflects the actual influence of different parameters through weight adjustment, and enhances the accuracy and flexibility of performance indicators;

[0060] Calculation process: The real-time collected temperature T is 30°C, the flow rate F is 0.5m / s, the pump speed S is 1500rpm, and the weights are set as k T =0.8, k F =0.05, k S =0.1, the calculation process is as follows:

[0061] Calculate the temperature squared and multiply it by the weight: k T ·T 2 =0.8·30 2 =720;

[0062] Calculate the square root of the velocity multiplied by the weight:

[0063] Calculate the inverse of the pump speed multiplied by the weight:

[0064] Substitute the above results into the formula to calculate M:

[0065]

[0066] The result shows that the real-time performance monitoring value M is 240.053, which is a high value, indicating that the current system performance is in good condition and the system does not need to be adjusted; if the M value is lower than a certain threshold, it may indicate unstable performance or a fault, requiring further inspection and maintenance; the threshold is calculated by analyzing historical performance indicator data, and its mean and standard deviation can be calculated. If the historical data shows that the mean value of M is 200 and the standard deviation is 20; based on these statistical data, the threshold can be set to the mean value minus twice the standard deviation, that is, Threshold = 200-2×20 = 160; this means that any M value below 160 may indicate that the system performance is not up to standard or there is a potential mechanical failure.

[0067] The steps to obtain the optimized search parameters are:

[0068] Receive real-time performance monitoring values ​​as input to adjust parameters in the particle swarm algorithm;

[0069] Calculate the dynamic inertia weight and acceleration coefficient using the following formula:

[0070]

[0071] and

[0072]

[0073] Where, e is the natural logarithm base, M is the real-time performance monitoring value, w is the dynamic inertia weight, and c is the acceleration coefficient;

[0074] Based on the dynamic inertia weight and acceleration coefficient, the global search strategy is optimized to generate optimized search parameters.

[0075] Specifically, the formula is beneficial in that by applying the natural logarithm base and the conversion of the real-time performance monitoring value, flexibility is increased and the ability to adapt to the state is enhanced; e represents the natural logarithm base, which is used to adjust the shape of the curve so that the changes of the weight w and the acceleration coefficient c are smoother, and M is the real-time performance monitoring value;

[0076] Calculation process: The real-time performance monitoring value M is 5, and the process of calculating the values ​​of w and c is as follows:

[0077]

[0078] The global search strategy is optimized based on the dynamic inertia weight and acceleration coefficient. During the optimization process, the recalculated dynamic inertia weight and acceleration coefficient are first applied to the particle swarm algorithm to adjust the particle speed and position update formula. Through these adjustments, the particles can explore the search space more effectively, avoiding local optimality and failing to reach the global optimality. In addition, the optimized search parameters also include adjustments to the search range and search speed to ensure that the algorithm can find a better solution within a given number of iterations, thereby generating optimized search parameters, ensuring the global search capability and the probability of finding the optimal solution.

[0079] The steps to obtain the preliminary running results are:

[0080] Load the optimized search parameters, configure the parameters in the particle swarm optimization algorithm, including adjusting the initial position and velocity settings of the particles, start the algorithm running process, and obtain the preliminary running results of the algorithm under each parameter configuration.

[0081] The steps to obtain the system configuration are:

[0082] Based on the preliminary running results, analyze the fitness value of each particle, evaluate the effectiveness of the corresponding parameter configuration of each particle according to the fitness value, identify the parameter configuration by comparing the fitness of each particle, and form a performance evaluation report;

[0083] Based on the performance evaluation report, the parameter configuration is screened according to the convergence speed of the algorithm and the quality of the solution to obtain the system configuration.

[0084] Specifically, load the optimized search parameters, set the initial position and speed of the particles, which are set according to the previous performance test results to ensure that each particle can start searching from the most advantageous position; start the particle swarm optimization process, during which the particles dynamically adjust their positions and speeds according to the historical best positions of individuals and groups, and each particle calculates its fitness based on the current position evaluation function. Such iterative updates help identify effective exploration paths and ensure that the global search is not limited to the local optimal area explored in the early stage. Continue to iterate until the predetermined number of iterations or improvement threshold is reached, thereby forming preliminary operation results.

[0085] Based on the preliminary running results, the fitness of each particle is analyzed. This analysis uses the performance evaluation function of the particle at the target position to calculate its fitness. During the calculation process, the fitness of each particle at the current position is compared with the fitness of all particles to determine which particles have better parameter configurations than other particles. This method can identify parameter configurations with excellent performance. Subsequently, these data are integrated to form a detailed performance evaluation report.

[0086] Using the performance evaluation report, the best performing configuration is screened out by comparing the algorithm convergence speed and solution quality under different configurations. The convergence speed is determined by the number of iterations required to reach the preset fitness threshold, while the quality of the solution is evaluated by the stability and fitness value of the solution. The selected parameter configuration is the configuration that consistently shows fast convergence speed and high solution quality in multiple iterations. These configurations are confirmed through the final validation test to confirm their actual effect and reliability, ensuring that the selected configuration can be applied to actual operations to obtain the system configuration.

[0087] The steps to obtain the preliminary cooling demand forecast value are:

[0088] Deploy temperature sensors in the data center to collect temperature data in real time, and send the temperature data to the database through the network for storage to form a comprehensive temperature data set;

[0089] Based on the comprehensive temperature data set, the temperature change trend is analyzed using the long short-term memory network to generate a temperature trend analysis report;

[0090] Based on the temperature trend analysis report, the cooling demand forecast value within a certain time range in the future is obtained to obtain the preliminary cooling demand forecast value.

[0091] Specifically, temperature sensors are installed in the data center and deployed according to a preset layout plan to ensure that the temperature of each area can be accurately monitored. The sensor collection frequency is adjusted to adapt to the rapid changes in ambient temperature, thereby capturing real-time temperature fluctuations. The data collected by each sensor is transmitted to the central database in real time through an encrypted network protocol. Data encryption and security verification measures are used during the transmission process to prevent data tampering or theft. After arriving at the database, data cleaning is performed to eliminate abnormal temperature records caused by sensor failure or environmental interference, forming a comprehensive and accurate temperature data set.

[0092] After receiving the cleaned temperature data set, the data is analyzed to identify the patterns and trends of temperature changes. The time series data is processed using a long short-term memory network. LSTM is able to identify time dependencies in the data, and feature extraction is performed to extract key change points from continuous temperature readings. The trends behind these change points are analyzed, the possible impact of these trends is quantified, and a temperature trend analysis report is generated. The report details the expected temperature rise or fall trend and provides a quantitative prediction of future temperature changes.

[0093] According to the temperature trend analysis report, the highest and lowest temperatures in the future are obtained and compared with the design parameters of the data center's cooling system to determine the required cooling intensity. The fan speed and coolant flow rate are adjusted to adapt to the predicted demand. These adjustments are based on the set temperature safety threshold, which is determined based on the thermal tolerance of the equipment and historical operating data. These operations generate a preliminary forecast of cooling demand, which ensures that the data center operates within a safe and efficient range, thereby maintaining the working condition of the equipment.

[0094] The steps to obtain the adjusted output value are:

[0095] Based on the preliminary cooling demand forecast value, evaluate whether the current pump speed and valve settings meet the forecasted cooling demand, and obtain the adjustment demand analysis;

[0096] Based on the adjustment demand analysis, recalculate the required pump speed and valve opening, the calculation formula is:

[0097]

[0098] and

[0099]

[0100] Among them, Q c is the cooling demand forecast value, Q max is the maximum cooling capacity, S max and O max are the maximum setting values ​​of pump speed and valve opening respectively, S is the pump speed, and O is the valve opening;

[0101] Based on the recalculated pump speed and valve opening, the pump and valve are adjusted to the new set points to obtain the adjusted output value.

[0102] Specifically, the preliminary cooling demand forecast is used to evaluate whether the current pump speed and valve settings are sufficient to meet the predicted cooling demand. By analyzing the current load and comparing it with historical data, it is evaluated whether the existing equipment configuration meets future operating requirements. At the same time, it is determined whether the current pump speed and valve opening can meet future cooling needs. If the forecast data shows that the existing configuration cannot meet future needs, equipment adjustment is recommended.

[0103] The benefit of the formula is that by applying exponential decay and square root functions, the two formulas can more accurately adjust the pump speed and valve opening to match the actual cooling demand, thereby optimizing energy efficiency; where Q c represents the current cooling demand forecast value, Q max The maximum cooling capacity that can be provided is usually determined based on equipment specifications and historical operating data. max and O maxThey represent the maximum possible values ​​of pump speed and valve opening, respectively, which are set according to the equipment design limits;

[0104] Calculation process: Set Q c =100kW is the current cooling demand forecast value, Q max =150kW is the maximum cooling capacity of the system, S max =1000rpm and O max =100% is the maximum setting value, substitute it into the formula and calculate:

[0105]

[0106] The results show that under the current cooling demand forecast, the pump speed should be adjusted to about 433rpm and the valve opening should be adjusted to about 82%. Such adjustments help to optimize the energy efficiency of the cooling system while ensuring that the equipment operates within a safe and effective working range.

[0107] Based on the recalculated pump speed and valve opening, the pump and valve are adjusted to the new set point, including sending control signals to the pump and valve drivers, and monitoring the feedback data in real time to ensure the implementation of the set value. The adjustment process involves PID control and real-time feedback adjustment. The PID controller compares the set pump speed and valve opening with the actual operating parameters, and adjusts the control signal to reduce the error between the target value and the actual value. Real-time feedback continuously monitors the equipment operating status through sensors, such as the actual speed of the pump and the actual position of the valve, to ensure that each adjustment is accurate, thereby ensuring the operation of the cooling system.

[0108] The steps to obtain the deviation analysis results are:

[0109] Track the adjusted pump speed and valve opening output values ​​to obtain real-time monitoring data sets;

[0110] Based on the real-time monitoring data set, the deviation between the adjusted output value and the real-time monitoring data is calculated to form a preliminary deviation analysis result;

[0111] Based on the preliminary deviation analysis results, the deviation situation is evaluated to check whether any deviation exceeds the safety threshold. If so, an alarm is triggered and the deviation analysis results are generated.

[0112] Specifically, the temperature, flow rate and pump speed data are collected through sensors, data aggregation and preliminary deviation analysis are performed, data deviation and loss are corrected to ensure data reliability, and the collected data are preliminarily screened to remove outliers and obvious erroneous data to ensure the accuracy of the data set for subsequent analysis. Through these accurate real-time monitoring data, a comprehensive data monitoring set is formed;

[0113] Quantitative analysis is used to process monitoring data sets, and each data point is compared one by one. The deviation between the current value and the historical average is calculated, and the safety threshold is set to ±10% of the historical data to define fluctuations within the normal operating range. This safety threshold is preset based on past equipment performance and typical operating conditions. Data points that exceed this range are identified as potential risk points. This method ensures immediate identification of anomalies in operation and generates preliminary deviation analysis results, which will be used to determine whether the system is within the safe operating range;

[0114] Continue to conduct in-depth assessment based on the deviation analysis results, classify and prioritize the deviations that exceed the safety threshold, determine the causes and possible impacts of the deviations, and immediately trigger an alarm if the deviation value exceeds the preset safety threshold by ±10%, notifying maintenance personnel to conduct inspections and generating deviation analysis results at the same time.

[0115] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. The multi-parameter fusion liquid cooling system AI tuning and optimal PUE implementation method is characterized by: The following steps are involved: The temperature, flow rate and pump speed data of the liquid cooling system are collected through sensors, the current performance indicators are calculated, and real-time performance monitoring values ​​are generated; According to the real-time performance monitoring value, the dynamic inertia weight and acceleration coefficient in the particle swarm algorithm are adjusted to optimize the global search strategy and generate optimized search parameters; Based on the optimized search parameters, a particle swarm optimization algorithm is run to obtain a performance evaluation under each parameter configuration and generate a preliminary operation result; by comparing each operation result, a configuration is selected to generate a system configuration; Collecting temperature information of the data center, analyzing the data change trend using a long short-term memory network, and obtaining a preliminary cooling demand forecast value; adjusting the pump speed and valve opening according to the preliminary cooling demand forecast value to generate an adjusted output value; Monitor the deviation between the adjusted output value and the real-time data, determine whether the deviation exceeds a set safety threshold, and generate a deviation analysis result.

2. According to the multi-parameter fusion liquid cooling system AI tuning and optimal PUE implementation method of claim 1, it is characterized in that: The steps for obtaining the real-time performance monitoring value are: The temperature, flow rate and pump speed of the liquid cooling system are collected through sensors, and three types of real-time data are recorded to generate a data set; Based on the data set, the real-time performance monitoring value is calculated using the following formula: Among them, k T is the square weight of temperature data, k F is the square root weight of the velocity data, k S is the inverse weight of the pump speed, M is the real-time performance monitoring value, T is the temperature, S is the pump speed, and F is the flow rate.

3. According to the multi-parameter fusion liquid cooling system AI tuning and optimal PUE implementation method of claim 1, it is characterized in that: The steps for obtaining the optimized search parameters are: receiving the real-time performance monitoring value as input to adjust parameters in a particle swarm algorithm; Calculate the dynamic inertia weight and acceleration coefficient using the following formula: and Where, e is the natural logarithm base, M is the real-time performance monitoring value, w is the dynamic inertia weight, and c is the acceleration coefficient; Based on the dynamic inertia weight and acceleration coefficient, the global search strategy is optimized to generate optimized search parameters.

4. The method for realizing AI tuning and optimal PUE of a liquid cooling system with multi-parameter fusion according to claim 1 is characterized in that: The steps for obtaining the preliminary operation results are: The optimized search parameters are loaded, parameters are configured in the particle swarm optimization algorithm, including adjusting the initial position and speed settings of the particles, the algorithm operation process is started, and preliminary operation results of the algorithm under each parameter configuration are obtained.

5. According to the multi-parameter fusion liquid cooling system AI tuning and optimal PUE implementation method of claim 1, it is characterized in that: The steps for obtaining the system configuration are: Based on the preliminary operation results, analyze the fitness value of each particle, evaluate the effectiveness of the parameter configuration corresponding to each particle according to the fitness value, identify the parameter configuration by comparing the fitness of each particle, and form a performance evaluation report; Based on the performance evaluation report, according to the convergence speed of the algorithm and the quality of the solution, the parameter configuration is screened to obtain the system configuration.

6. The method for realizing AI tuning and optimal PUE of a liquid cooling system with multi-parameter fusion according to claim 1 is characterized in that: The steps for obtaining the preliminary cooling demand forecast value are: Deploy temperature sensors in the data center to collect temperature data in real time, and send the temperature data to the database through the network for storage to form a comprehensive temperature data set; Based on the comprehensive temperature data set, using a long short-term memory network to analyze the temperature change trend and generate a temperature trend analysis report; Based on the temperature trend analysis report, a cooling demand forecast value within a certain time range in the future is obtained to obtain a preliminary cooling demand forecast value.

7. The method for realizing AI tuning and optimal PUE of a liquid cooling system with multi-parameter fusion according to claim 1 is characterized in that: The steps for obtaining the adjusted output value are: Evaluating whether the current pump speed and valve settings meet the predicted cooling demand based on the preliminary cooling demand prediction value to obtain an adjustment demand analysis; Based on the adjustment demand analysis, the required pump speed and valve opening are recalculated using the following formula: and Among them, Q c is the cooling demand forecast value, Q max is the maximum cooling capacity, S max and O max are the maximum setting values ​​of pump speed and valve opening respectively, S is the pump speed, and O is the valve opening; Based on the recalculated pump speed and valve opening, the pump and valve are adjusted to the new set points to obtain the adjusted output value.

8. The method for realizing AI tuning and optimal PUE of a liquid cooling system with multi-parameter fusion according to claim 1 is characterized in that: The steps for obtaining the deviation analysis results are: Track the adjusted pump speed and valve opening output values ​​to obtain real-time monitoring data sets; Based on the real-time monitoring data set, calculating the deviation between the adjusted output value and the real-time monitoring data to form a preliminary deviation analysis result; Based on the preliminary deviation analysis results, the deviation situation is evaluated to check whether any deviation exceeds the safety threshold. If so, an alarm is triggered and a deviation analysis result is generated.

Images