Server cabinet liquid immersion cooling system

Abstract

The invention discloses a server cabinet liquid immersion cooling system, and relates to the technical field of big data and artificial intelligence cabinet cooling, and the system comprises an information monitoring module which is used for collecting the operation state data of a server cabinet and monitoring the data change condition; the data processing module is used for cleaning and integrating the running state data acquired by the information monitoring module; the artificial intelligence prediction module is used for predicting the future cooling demand of the server cabinet based on the historical operation state data and the operation state data output by the current data processing module; the optimization control module is used for solving the multi-objective optimization function according to the cooling requirement and formulating a system operation strategy; the execution module is used for performing control optimization on the server cabinet according to the system operation strategy; the monitoring and alarming module is used for monitoring the running state of the system in real time and sending out an alarming signal when the system is abnormal; the performance evaluation module is used for evaluating the energy efficiency of the system. According to the system, accurate prediction and optimal control of future demands of the server cabinet can be realized.

Description

Technical Field

[0001] The present invention relates to the technical field of big data and artificial intelligence cabinet cooling, and in particular to a server cabinet liquid immersion cooling system. Background Art

[0002] With the rapid growth of computing power in data centers, traditional air cooling has been unable to meet the needs of high-density servers. Liquid immersion cooling technology uses special liquids (such as hydrocarbons or fluorides) as refrigerants to immerse all heat-generating components in coolant, and removes heat through liquid circulation. Compared with air cooling, it has better heat exchange effect and heat dissipation efficiency. With the continuous advancement of technology, traditional cooling water systems have developed from simple mechanical control in the early days to today's complex automated control systems.

[0003] In recent years, through real-time monitoring and data analysis, the system can adjust operating parameters in time and optimize energy consumption. However, the existing technology still has many shortcomings in practical applications, especially in intelligent prediction and optimization control, and has not yet achieved the desired effect.

[0004] Existing server cabinet cooling water systems usually rely on preset cooling control parameters and rules, which are difficult to cope with complex and changing working conditions. This fixed control strategy is prone to problems such as energy waste and increased equipment wear in actual operation. In order to solve this problem, some systems have tried to introduce data analysis and prediction technologies, but most of them are still at the basic data monitoring and recording stage, lacking in-depth data mining and intelligent analysis capabilities, and unable to achieve accurate prediction and optimization control of future needs. Summary of the invention

[0005] In view of the above-mentioned problems existing in the existing circulating cooling water system, the present invention provides a server cabinet liquid immersion cooling system, which can achieve accurate prediction and optimized control of future demand.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: In a first aspect, an embodiment of the present invention provides a server cabinet liquid immersion cooling system, the system comprising: an information monitoring module, a data processing module, an artificial intelligence prediction module, an optimization control module, an execution module, a monitoring and alarm module, and a performance evaluation module; the output end of the information monitoring module is connected to the input end of the data processing module, the output end of the data processing module is connected to the input end of the artificial intelligence prediction module, the output end of the artificial intelligence prediction module is connected to the input end of the optimization control module, the output end of the optimization control module is connected to the input end of the execution module, the output end of the execution module is connected to the input of the monitoring and alarm module, and the output end of the monitoring and alarm module is connected to the input end of the performance evaluation module; The information monitoring module is used to collect the operating status data of the server cabinet and monitor the data changes during the operation of the server cabinet; The data processing module is used to clean the running status data collected by the information monitoring module, and after cleaning the running status data, process the collection time difference of different running status data by timestamp merging to obtain the running status data under the same timestamp; An artificial intelligence prediction module is used to predict the future cooling needs of the server cabinet based on the historical operating status data and the operating status data output by the current data processing module; The optimization control module is used to establish a multi-objective optimization function with energy consumption, water quality maintenance and temperature control accuracy as the goals, and solve the multi-objective optimization function according to the cooling demand to obtain the system operation strategy of the system; An execution module, used to control and optimize the server cabinet according to the system operation strategy obtained by the optimization control module; The monitoring and alarm module is used to monitor the system operation status in real time and send out an alarm signal if there is any abnormality; The performance evaluation module is used to evaluate the energy efficiency of the system and provide a basis for long-term optimization.

[0007] As a preferred solution of the server cabinet liquid immersion cooling system described in the present invention, the information monitoring module is specifically used to set the sampling frequency, based on the set sampling frequency, collect analog signals through a variety of water quality sensors, temperature sensors, flow meters and energy consumption sensors, and convert the collected analog signals into digital signals to obtain operating status data; and set a normal operating range for each operating status parameter, and trigger an alarm when the operating status data exceeds the normal range.

[0008] As a preferred solution of the server cabinet liquid immersion cooling system of the present invention, the data processing module specifically includes: Predefine normal data ranges for different types of sensor data, and check whether the operating status data collected by each sensor is within the corresponding predefined normal data range. If not, the operating status data that is not within the corresponding normal data range will be marked as abnormal; if it is, the interquartile range IQR method is used to perform a secondary inspection on the operating status data. If the operating status data falls below Q1-1.5IQR or above Q3+1.5IQR, the operating status data is marked as abnormal, where Q3 is the upper quartile, Q1 is the lower quartile, and IQR is the inner moment; After removing the abnormal operating status data, the operating status data from different sensors and devices are merged into a unified data set. During the merging process, the timestamp is used as the keyword, and the approximate merging method is adopted to deal with the collection time difference of the operating status data from different sensors.

[0009] As a preferred solution of the server cabinet liquid immersion cooling system of the present invention, the artificial intelligence prediction module predicts the future cooling demand of the server cabinet based on the historical operating status data and the operating status data output by the current data processing module, specifically including: According to the historical operation status data, the operation status data output by the current data processing module is normalized; the normalization method is: ; in, is the normalized running status data, X is the running status data output by the data processing module, and They are the minimum and maximum values ​​of the historical operation status data respectively; The change rate feature is extracted from the normalized operating status data; the calculation formula of the change rate feature is: ; in, is the rate of change characteristic, for t 1. Operation status data at a moment in time, yes( t 1-1) Operation status data at the moment; Set multiple prediction time ranges and output parameters, each prediction time range corresponds to an output parameter, and train a separate extreme gradient boosting model for each prediction time range and each output parameter. The change rate feature is input into the extreme gradient boosting model to obtain the future cooling demand of the server cabinet.

[0010] As a preferred solution of the server cabinet liquid immersion cooling system of the present invention, the construction process of the extreme gradient boosting model specifically includes: Construct an extreme gradient boosting model by using Bayesian optimization to find the best hyperparameters, using the gradient boosting decision tree algorithm, using mean square error as the objective function, and using early stopping. Early stopping methods include: Set the initial optimal validation loss is positive infinity, sets the initial patience count to 0, and sets the patience value Patience value Set as the optimal validation threshold; after each training round, calculate the current validation loss ;if , then update the best validation loss , and reset the patience count; otherwise, reduce the patience count; the patience value represents the best validation loss The maximum number of times training is allowed to continue when there is no improvement. The patience count is the number of times the model is trained. If the patience count exceeds the patience value , stop training; The validation loss is calculated as: ; in, is the number of samples in the validation set, L is the loss function, It is i The true value of the validation set samples, It is i The predicted values ​​of the validation set samples.

[0011] As a preferred solution of the server cabinet liquid immersion cooling system of the present invention, the optimization control module is specifically used for: Define multi-objective optimization functions, including energy consumption function, water quality maintenance function and temperature control accuracy function; The energy consumption function is: ; in, For energy consumption, is the fan power, is the water pump power, is the running time; The water quality maintenance function is: ; in, For water resources value, for The target value of the value, is the conductivity, is the target conductivity, and is the weight coefficient, for( ), for( )’s absolute value; The temperature control accuracy function is: ; in, For temperature control accuracy, is the actual water temperature, To set the temperature, for( )’s absolute value; The multi-objective optimization function is: ; in, α , β , γ are the weight coefficients of energy consumption function, water quality maintenance function and temperature control accuracy function respectively; The set constraints include: cooling tower fan speed range, circulating water pump flow range, water temperature set point range and system cooling capacity constraints.

[0012] As a preferred solution of the server cabinet liquid immersion cooling system of the present invention, the optimization control module is specifically used to adopt the model predictive control MPC combined with the particle swarm optimization algorithm PSO optimization control strategy to determine the system operation strategy; Among them, the MPC prediction model in the model predictive control MPC is: ; ; PSO update formula: ; ; in, for The system status at the moment, for k The control input at the moment, for k The system output at the moment, A , B and C are system matrices, for t +1 moment i The particle velocity of a particle, for t +1 moment i The particle position of each particle, for t Moment i The particle velocity of a particle, for t Moment i The particle position of each particle, is the individual optimal solution, is the global optimal solution, is the inertia weight, and is the acceleration constant, and is a random number; The control strategy optimization process is as follows: The particle swarm is initialized, and the position of each particle represents a set of cooling control parameters, which include fan speed, water pump flow and temperature set point; The server cabinet is controlled through cooling control parameters to determine the operation status data of the server cabinet; According to the operation status data, the MPC prediction model is used to predict the system response for each time step in the prediction time domain to obtain the predicted operation status of the server cabinet; Calculate the objective function value of the multi-objective optimization function according to the predicted operating status and cooling demand; According to the objective function value, update the individual optimal solution and the global optimal solution of the particle; The PSO update formula is used to update the particle position and velocity. The individual optimal solution and the global optimal solution are continuously updated through the updated particles and velocity until the maximum number of iterations is reached or the convergence condition is met. The particle position corresponding to the global optimal solution is determined as the system operation strategy.

[0013] As a preferred solution of the server cabinet liquid immersion cooling system of the present invention, the optimization control module is also specifically used for: During the control strategy optimization process, check whether the system operation strategy meets all constraints. If not, use the penalty function to correct the objective function value: The penalty function is: ; in, is the corrected objective function value, F is the original objective function value, is the penalty parameter, For the i constraints.

[0014] As a preferred solution of the server cabinet liquid immersion cooling system described in the present invention, the performance evaluation module is used to calculate the energy utilization efficiency index and generate an optimization effect report; the energy utilization efficiency index includes the energy efficiency index and the temperature control accuracy.

[0015] In a second aspect, an embodiment of the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the server cabinet liquid immersion cooling system as described in the first aspect of the present invention is implemented.

[0016] In a third aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the server cabinet liquid immersion cooling system as described in the first aspect of the present invention is implemented.

[0017] The beneficial effects of the present invention are: The server cabinet liquid immersion cooling system of the present invention is provided with an information monitoring module, a data processing module, an artificial intelligence prediction module, an optimization control module, an execution module, a monitoring and alarm module, and a performance evaluation module. Through the coordinated cooperation of these modules, accurate prediction and optimization control of future needs can be achieved. In addition, the server cabinet liquid immersion cooling system of the present invention can realize comprehensive monitoring, diagnosis, prediction and optimization functions, greatly improving the operating efficiency and reliability of the system. The intelligent and adaptive characteristics of the system enable it to maintain optimal performance under various working conditions, while providing powerful decision-making support for operators.

[0018] Furthermore, the information monitoring module collects the operating status data of the server cabinet and monitors the data changes, providing detailed and accurate basic data for subsequent analysis and decision-making; the data processing module cleans the collected data, and after cleaning the operating status data, processes the collection time difference of different operating status data by timestamp merging to obtain the operating status data under the same timestamp, effectively removing the "impurities" such as errors, duplications, and incompleteness that may exist in the original data, and after removing the impurities, the operating status data is integrated under the same timestamp, making the data used by subsequent modules more standardized and high-quality, improving the reliability of the entire system's analysis and decision-making based on data, and avoiding erroneous judgments due to "dirty data"; the artificial intelligence prediction module combines historical operating status data and currently processed data to predict The future cooling needs of server cabinets make full use of past experience and current real-time conditions to make the prediction results more scientific and forward-looking. Knowing the cooling needs in advance can make preparations in advance, effectively avoiding the impact of insufficient cooling or excessive cooling on the performance and life of the server cabinet, and achieving accurate resource allocation expectations; the optimization control module takes energy consumption, water quality maintenance and temperature control accuracy as goals, establishes a multi-objective optimization function, and solves the multi-objective optimization function according to the cooling needs to obtain the system operation strategy of the system, which can optimize resource utilization to the greatest extent while meeting the normal operation of the server cabinet, so as to achieve the purpose of energy saving and consumption reduction, and improve overall operating efficiency, ensure that the system always runs in the best state, avoid resource waste and unnecessary losses, thereby realizing accurate prediction and optimization control of future needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them: Figure 1 A structural diagram of a server cabinet liquid immersion cooling system provided by the present invention.

[0020] Figure 2 A schematic diagram of the structure of a terminal provided by the present invention. DETAILED DESCRIPTION

[0021] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0022] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0023] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0024] Example 1 Reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, and which provides a server cabinet liquid immersion cooling system, such as Figure 1 As shown, Figure 1 The structure diagram of the liquid immersion cooling system of the server cabinet is shown in FIG. The system includes: an information monitoring module, a data processing module, an artificial intelligence prediction module, an optimization control module, an execution module, a monitoring and alarm module, and a performance evaluation module; the output end of the information monitoring module is connected to the input end of the data processing module, the output end of the data processing module is connected to the input end of the artificial intelligence prediction module, the output end of the artificial intelligence prediction module is connected to the input end of the optimization control module, the output end of the optimization control module is connected to the input end of the execution module, the output end of the execution module is connected to the input end of the monitoring and alarm module, and the output end of the monitoring and alarm module is connected to the input end of the performance evaluation module. Figure 2 As shown, Figure 2 The terminal includes an input and output device, a transmission device, a processor and a memory. The system is applied to Figure 2 The terminal in is used as an example to illustrate the following contents:

[0025] The information monitoring module is used to collect the operating status data of the server cabinet and monitor the data changes during the operation of the server cabinet; the data processing module is used to clean the operating status data collected by the information monitoring module, and after cleaning the operating status data, process the collection time difference of different operating status data by timestamp merging to obtain the operating status data under the same timestamp; the artificial intelligence prediction module is used to predict the future cooling requirements of the server cabinet based on the historical operating status data and the operating status data output by the current data processing module; the optimization control module is used to establish a multi-objective optimization function with energy consumption, water quality maintenance and temperature control accuracy as the goals, and solve the multi-objective optimization function according to the cooling demand to obtain the system operation strategy of the system; the execution module is used to control and optimize the server cabinet according to the system operation strategy formulated by the optimization control module; the monitoring and alarm module is used to monitor the system operation status in real time, and if there is an abnormality, an alarm signal is issued; the performance evaluation module is used to evaluate the energy efficiency of the system and provide a basis for long-term optimization.

[0026] Among them, the server cabinet includes multiple servers, and obtaining the operating status data of the server cabinet is essentially the operating status data of the servers in the server cabinet; controlling the server cabinet is essentially optimizing the control of the servers in the server cabinet.

[0027] Specifically, the information monitoring module is used to monitor the changes in operating parameters such as water quality, temperature, flow rate and energy consumption when the server cabinet is running; the data processing module is used to clean and integrate the original data in preparation for subsequent analysis; the artificial intelligence prediction module is used to predict future cooling needs based on historical data and current conditions; the optimization control module is used to formulate the optimal system operation strategy according to the prediction results; the execution module is used to control and optimize the server cabinet according to the system operation strategy obtained by the optimization control module; the monitoring and alarm module is used to monitor the system operation status in real time and promptly alarm when an abnormality is found; the performance evaluation module is used to evaluate the energy efficiency of the system and provide a basis for long-term optimization.

[0028] Furthermore, the information monitoring module is specifically used to set the sampling frequency. Based on the set sampling frequency, analog signals are collected through a variety of water quality sensors, temperature sensors, flow meters and energy consumption sensors, and the collected analog signals are converted into digital signals to obtain operating status data; and a normal operating range is set for each operating status parameter, and an alarm is triggered when the operating status data exceeds the normal range.

[0029] Specifically, the information monitoring module includes: installing a variety of water quality sensors at key locations of the system, including pH sensors for monitoring the acidity and alkalinity of water; conductivity sensors for detecting the ion content in water; dissolved oxygen sensors for measuring the dissolved oxygen content in water; deploying temperature sensors at multiple key locations of the system, such as: cooling water inlet, cooling water outlet and cooling tower; installing flow meters in the main pipeline and bypass; installing energy consumption sensors to monitor the energy consumption of water pumps and cooling towers.

[0030] Set the sampling frequency (e.g. once every 5 minutes), use a high-precision analog-to-digital converter to convert the analog signal into a digital signal, develop a real-time monitoring dashboard to display the real-time values ​​and trend charts of key parameters, set the normal operating range for each parameter, and trigger an alarm when the parameter exceeds the range.

[0031] Further, as a preferred solution of the server cabinet liquid immersion cooling system described in the present invention, the data processing module specifically includes: pre-defining normal data ranges for different types of sensor data, and checking whether the operating status data collected by each sensor is within the corresponding predefined normal data range. If not, the operating status data that is not within the corresponding normal data range is marked as abnormal; if it is, the interquartile range IQR method is used to perform a secondary detection on the operating status data. If the operating status data falls below Q1-1.5IQR or above Q3+1.5IQR, the operating status data is marked as abnormal, wherein Q3 is the upper quartile, Q1 is the lower quartile, and IQR is the inner moment; after the abnormal operating status data is eliminated, the operating status data from different sensors and devices are merged into a unified data set. During the merging process, the timestamp is used as a keyword, and the approximate merging method is used to process the collection time difference of the operating status data from different sensors.

[0032] Specifically, the data processing module includes the following contents: processing data points beyond the normal range, detecting outliers, and deciding whether to delete or replace outliers according to specific circumstances.

[0033] Furthermore, the process of detecting outliers includes: first, defining normal ranges for different types of sensor data, such as pH (6.0-9.0), conductivity (100-1000 μS / cm), temperature (0-50°C), flow rate (0-1000 L / min), etc., and then, using multiple judgment criteria to identify outliers.

[0034] First, check whether it is within the predefined normal range. If not, it is marked as an abnormality. If it is, the interquartile range (IQR) method is further used. If the data point falls below Q1-1.5IQR or above Q3+1.5IQR, it is considered an abnormality, where Q3 is the upper quartile, Q1 is the lower quartile, and IQR is the inner moment.

[0035] The data processing module also includes: unifying the timestamp format to ensure the consistency of time series data, normalizing field names so that data from different sources can be seamlessly integrated; converting categorical variables into numerical form, and normalizing or standardizing numerical variables to make their distribution more suitable for subsequent analysis; Merge data from different sensors and devices into a unified dataset. First, merge data from different sensors (pH value, conductivity, temperature, flow, etc.) into a unified dataset. During the merging process, use the timestamp as the keyword and adopt the approximate merge_asof method to handle the situation where the data collection time of different sensors may be slightly different.

[0036] In an exemplary embodiment, the artificial intelligence prediction module predicts the future cooling demand of the server cabinet based on the historical operating status data and the operating status data output by the current data processing module, specifically including: According to the historical operation status data, the operation status data output by the current data processing module is normalized to ensure the quality and consistency of the data: the normalization method is: (1); in, is the normalized running status data, X is the running status data output by the data processing module, and They are the minimum and maximum values ​​of the historical running status data respectively.

[0037] The change rate feature is extracted from the normalized operating status data. Specifically, key features such as temperature change rate, humidity change rate, equipment load rate, etc. are extracted from the preprocessed data and the periodic features of the time series are extracted using Fourier transform. The calculation formula of the change rate feature is as follows: (2); in, is the rate of change characteristic, for t 1. Operation status data at a moment in time, yes( t 1-1) Operation status data at the moment.

[0038] Set multiple prediction time ranges and output parameters, each prediction time range corresponds to an output parameter, and train a separate extreme gradient boosting model for each prediction time range and each output parameter. The change rate feature is input into the extreme gradient boosting model to obtain the future cooling demand of the server cabinet.

[0039] Specifically, a machine learning algorithm is used to train historical data to generate a prediction model. The algorithm of the present invention uses an extreme gradient boosting (XGBoost) model to train a separate XGBoost model for each prediction time range (24 hours, 48 ​​hours, 7 days) and each output parameter (cooling water flow, required temperature).

[0040] Furthermore, the construction process of the extreme gradient boosting model specifically includes: constructing the extreme gradient boosting model by using the Bayesian optimization method to find the best hyperparameters, using the gradient boosting decision tree algorithm, using the mean square error as the objective function, and using the early stopping method.

[0041] Specifically, the data set is divided into 70% training set, 15% validation set and 15% test set. The Bayesian optimization method is used to find the best hyperparameters, including the maximum depth of the tree, the minimum child node weight, the feature sampling ratio, etc. The gradient boosting decision tree algorithm is used, and the objective function is the mean square error. The early stopping method is used to prevent overfitting. The early stopping method includes: setting the initial best validation loss is positive infinity, sets the initial patience count to 0, and sets the patience value Patience value Set as the optimal validation threshold; after each training round, calculate the current validation loss ;if , then update the best validation loss , and reset the patience count; otherwise, reduce the patience count; the patience value represents the best validation loss The maximum number of times training is allowed to continue when there is no improvement. The patience count is the number of times the model is trained. If the patience count exceeds the patience value , stop training.

[0042] Furthermore, the calculation formula of the validation loss is: (3); in, is the number of samples in the validation set, L is the loss function, It is i The true value of the validation set samples, It is i The predicted values ​​of the validation set samples.

[0043] It should be noted that by introducing the early stopping method, the model can be effectively prevented from overfitting during the training process, and the generalization ability and prediction accuracy of the model can be improved. The root mean square error (RMSE) and mean absolute percentage error (MAPE) are used to evaluate the model performance.

[0044] The optimization control module includes the following contents: obtaining the cooling water flow forecast for the next 24 hours, 48 ​​hours and 7 days and the required temperature forecast for the next 24 hours, 48 ​​hours and 7 days from the artificial intelligence prediction module; obtaining the current system status, including the cooling tower fan speed, circulating water pump flow, current water temperature and system load; obtaining external environmental data, including ambient temperature, relative humidity and atmospheric pressure.

[0045] Define multi-objective optimization functions, including energy consumption function, water quality maintenance function, and temperature control accuracy function.

[0046] Specifically, the energy consumption function is: (4); in, For energy consumption, is the fan power, is the water pump power, For the running time.

[0047] The water quality maintenance function is: (5); in, For water resources value, for The target value of the value, is the conductivity, is the target conductivity, and is the weight coefficient, for( ), for( )’s absolute value.

[0048] The temperature control accuracy function is: (6); in, For temperature control accuracy, is the actual water temperature, To set the temperature, for( )’s absolute value.

[0049] The comprehensive objective function is: (7); in, α , β , γ They are the weight coefficients of energy consumption function, water quality maintenance function and temperature control accuracy function respectively.

[0050] The set constraints include: cooling tower fan speed range, circulating water pump flow range, water temperature set point range and system cooling capacity constraints.

[0051] Furthermore, the optimization control module specifically includes: using Model Predictive Control (MPC) combined with Particle Swarm Optimization (PSO) to optimize the control strategy and determine the system operation strategy; including the following contents: The MPC prediction model in the model predictive control MPC is: (8); (9).

[0052] PSO update formula: (10); (11); in, for The system status at the moment, for k The control input at the moment, for k The system output at the moment, A , B and C are system matrices, for t +1 moment i The particle velocity of a particle, for t +1 moment i The particle position of each particle, for t Moment i The particle velocity of a particle, for t Moment i The particle position of each particle, is the individual optimal solution, is the global optimal solution, is the inertia weight, and is the acceleration constant, and Is a random number.

[0053] Preferably, the control strategy optimization process is as follows: Set the initialization particle group, and the position of each particle represents a set of cooling control parameters, which include fan speed, water pump flow, and temperature set point; Control the server cabinet through cooling control parameters to determine the operating status data of the server cabinet; According to the operation status data, the MPC prediction model is used to predict the system response for each time step in the prediction time domain to obtain the predicted operation status of the server cabinet; Calculate the objective function value of the multi-objective optimization function according to the predicted operating status and cooling demand; According to the objective function value, update the individual optimal solution and the global optimal solution of the particle; The PSO update formula is used to update the particle position and velocity. The individual optimal solution and the global optimal solution are continuously updated through the updated particles and velocity until the maximum number of iterations is reached or the convergence condition is met. The particle position corresponding to the global optimal solution is determined as the system operation strategy.

[0054] The Kalman filter algorithm can be used to update the MPC model parameters. Calculating the objective function value of the multi-objective optimization function can specifically include: defining the predicted operating state and cooling demand as two objective functions, and setting the two objective functions as f 1( x )and f 2( x ), two objective functions f 1( x )and f 2( x ) respectively set the weights as w 1 and w 2, then the calculation formula of the objective function value min J(x) of the multi-objective optimization function is: min J ( x )= w 1 f 1( x )+ w 2 f 2( x ); It should be noted that the objective function value of the multi-objective optimization function is the value of the minimized multi-objective optimization function F.

[0055] During the control strategy optimization process, check whether the system operation strategy meets all constraints. If not, use the penalty function method to correct the result.

[0056] (12); in, is the corrected objective function value, F is the original objective function value, is the penalty parameter, For the i constraints.

[0057] The actual operating status is the fan power , Pump power , water quality resources value and actual water temperature, the cooling demand is Target value of value , Target conductivity and set temperature .

[0058] Generate control instructions according to the system operation strategy, including fan inverter set value, water pump inverter set value and PLC temperature set point, and send the control instructions to the execution module.

[0059] It should be noted that the prediction results provided by XGBoost, such as the cooling water flow and temperature requirements for the next 24 hours, 48 ​​hours and 7 days, are used as the input of MPC; MPC uses these predictions to plan future control strategies, while PSO is used to find the optimal cooling control parameters within the MPC framework, that is, to obtain the optimal solution in combination with cooling.

[0060] Execution module, including the following: The execution module receives control instructions from the optimization control module and parses the control instructions into specific execution instructions, such as pump start instructions, fan speed settings, etc. The control instructions are sent to each actuator through the network or bus system. Each actuator will receive a specific control instruction, and the actuator will perform corresponding operations according to the received instructions, such as pump start, fan speed adjustment, etc.

[0061] The monitoring and alarm module, including the main function of the monitoring and alarm module, is to monitor the key parameters of the system in real time to ensure that these parameters operate within the normal range. If any abnormal situation is found, the module will immediately notify the relevant personnel via SMS or email to ensure that the problem can be dealt with quickly, including: Monitor key parameters in the system, such as temperature, pressure, flow, current, etc., and set the normal range value of each parameter; compare the monitored parameters with their normal range values ​​in real time, and if the parameters are out of the normal range, mark them as abnormal; when an abnormality is detected, trigger the alarm mechanism and generate alarm information, including abnormal parameters, current values, timestamps, etc., and notify relevant personnel through preset communication methods (such as SMS, email).

[0062] Record the system's operating status, including real-time values ​​of all key parameters, and record alarm information for subsequent analysis and troubleshooting.

[0063] The performance evaluation module is used to calculate energy utilization efficiency indicators and generate optimization effect reports; the energy utilization efficiency indicators include energy efficiency indicators and temperature control accuracy.

[0064] The calculation formula of energy efficiency index is: (13); in, is the optimized energy efficiency index; is the cooling capacity provided, in kilowatts (kW); is the part load ratio, is the power consumption of the fan, in kilowatts (kW); is the power consumption of the pump in kilowatts (kW); The power consumption of the cooling tower in kilowatts (kW).

[0065] The calculation formula for temperature control accuracy is: (14); in, To optimize the temperature control accuracy, For the i The actual temperature at a time point, For the i The set temperature at a time point, N is the total number of measurement points, For the i The weight of a time point.

[0066] Furthermore, the optimization effect reports generated include: Daily report: 24-hour energy consumption curve, daily average energy efficiency ratio (EER) and specific energy consumption (SEC), comparison with the previous day, and abnormal event marking; Weekly report: daily energy consumption and EER trend, weekly average EER and SEC, comparison with last week, and identification of the best and worst performance days; Monthly report: monthly total energy consumption and distribution, monthly average EER and SEC, comparison with the previous month and the same period last year, performance trend analysis, and energy saving potential estimation.

[0067] This embodiment also provides a computer device, which is suitable for the liquid immersion cooling system of a server cabinet, including: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the server cabinet liquid immersion cooling system proposed in the above embodiment.

[0068] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected through a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, near field communication (NFC) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covered on the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0069] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the server cabinet liquid immersion cooling system is implemented as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, EPROM), programmable read-only memory (Programmable Red-Only Memory, PROM), read-only memory (Read-Only Memory, ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0070] In summary, the server cabinet liquid immersion cooling system of the present invention can realize comprehensive monitoring, diagnosis, prediction and optimization functions, greatly improve the operating efficiency and reliability of the system, and the intelligent and adaptive characteristics of the system enable it to maintain optimal performance under various working conditions, while providing powerful decision-making support for operators.

[0071] Example 2 Reference Figure 1 and Figure 2 , which is the second embodiment of the present invention, in order to further verify the beneficial effects of the present invention, experimental simulation comparison data of the server cabinet liquid immersion cooling system and the traditional method are provided.

[0072] The experimental scenario of the present invention is selected in the cooling water system of a large industrial park. The system includes multiple key components, including a cooling tower, a circulating water pump, a sensor network, etc.

[0073] The experiment is divided into two phases: the traditional system operation phase and the invented system operation phase. In order to ensure the accuracy and consistency of the data, the experimental preparation work includes installing sensors, setting monitoring parameters, and configuring the data acquisition and processing system.

[0074] First, a variety of water quality sensors are installed at key locations in the cooling system, including pH sensors, conductivity sensors, dissolved oxygen sensors, etc. In addition, temperature sensors are installed at the inlet, outlet and cooling tower of the cooling water, flow meters are installed in the main pipeline and bypass, and energy consumption sensors are installed on the water pump and cooling tower. The sampling frequency of all sensors is set to once every 5 minutes to ensure high-frequency data collection and real-time performance. A high-precision analog-to-digital converter is used to convert the analog signal of the sensor into a digital signal. The real-time monitoring dashboard displays the real-time values ​​and trend graphs of key parameters, and sets the normal operating range and alarm mechanism.

[0075] After collecting and preprocessing the data, the data from different sensors are merged into a unified dataset, and the differences in data collection time of different sensors are handled by the timestamp approximate merging method.

[0076] After the data processing is completed, the artificial intelligence prediction module is started. In order to improve the prediction accuracy of the model, the Bayesian optimization method is used to find the optimal hyperparameters, and the gradient boosting decision tree algorithm is used for optimization to prevent overfitting.

[0077] Finally, according to the future cooling demand forecast provided by the artificial intelligence prediction module, a multi-objective optimization function is defined, including minimization of energy consumption, water quality maintenance and temperature control accuracy. The control strategy is optimized by combining model predictive control (MPC) with particle swarm optimization (PSO); the optimized control strategy is parsed into specific execution instructions, and the control instructions are sent to each actuator through the network or bus system. The actuator performs corresponding operations according to the received instructions. Some experimental data of the present invention are shown in Table 1.

[0078] Table 1

[0079] Based on the above data analysis, it can be concluded that the present invention has shown significant advantages in water quality maintenance, temperature control, flow stability and energy consumption reduction; compared with traditional systems, the invented system improves the intelligence level and operation efficiency of the system by integrating big data and artificial intelligence technology, and significantly improves the performance and reliability of the system.

[0080] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

[0081] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of the present invention.

Claims

1. A server cabinet liquid immersion cooling system, characterized in that: The system comprises: an information monitoring module, a data processing module, an artificial intelligence prediction module, an optimization control module, an execution module, a monitoring and alarm module, and a performance evaluation module; the output end of the information monitoring module is connected to the input end of the data processing module, the output end of the data processing module is connected to the input end of the artificial intelligence prediction module, the output end of the artificial intelligence prediction module is connected to the input end of the optimization control module, the output end of the optimization control module is connected to the input end of the execution module, the output end of the execution module is connected to the input of the monitoring and alarm module, and the output end of the monitoring and alarm module is connected to the input end of the performance evaluation module; The information monitoring module is used to collect the operating status data of the server cabinet and monitor the data changes during the operation of the server cabinet; The data processing module is used to clean the running status data collected by the information monitoring module, and after cleaning the running status data, process the collection time difference of different running status data by timestamp merging to obtain the running status data under the same timestamp; An artificial intelligence prediction module is used to predict the future cooling needs of the server cabinet based on the historical operating status data and the operating status data output by the current data processing module; The optimization control module is used to establish a multi-objective optimization function with energy consumption, water quality maintenance and temperature control accuracy as the goals, and solve the multi-objective optimization function according to the cooling demand to obtain the system operation strategy of the system; An execution module, used for controlling and optimizing the server cabinet according to the system operation strategy obtained by the optimization control module; The monitoring and alarm module is used to monitor the system operation status in real time and send out an alarm signal if there is any abnormality; The performance evaluation module is used to evaluate the energy efficiency of the system and provide a basis for long-term optimization.

2. The server cabinet liquid immersion cooling system according to claim 1, characterized in that: The information monitoring module is specifically used to set the sampling frequency. Based on the set sampling frequency, analog signals are collected through a variety of water quality sensors, temperature sensors, flow meters and energy consumption sensors, and the collected analog signals are converted into digital signals to obtain operating status data; and a normal operating range is set for each operating status parameter, and an alarm is triggered when the operating status data exceeds the normal range.

3. The server cabinet liquid immersion cooling system according to claim 1, characterized in that: The data processing module specifically includes: Predefine normal data ranges for different types of sensor data, and check whether the operating status data collected by each sensor is within the corresponding predefined normal data range. If not, the operating status data that is not within the corresponding normal data range will be marked as abnormal; if it is, the interquartile range IQR method is used to perform a secondary inspection on the operating status data. If the operating status data falls below Q1-1.5IQR or above Q3+1.5IQR, the operating status data is marked as abnormal, where Q3 is the upper quartile, Q1 is the lower quartile, and IQR is the inner moment; After removing the abnormal operating status data, the operating status data from different sensors and devices are merged into a unified data set. During the merging process, the timestamp is used as the keyword, and the approximate merging method is adopted to deal with the collection time difference of the operating status data from different sensors.

4. The server cabinet liquid immersion cooling system according to claim 1, characterized in that: The implementation method of predicting the future cooling demand of the server cabinet in the artificial intelligence prediction module based on the historical operating status data and the operating status data output by the current data processing module specifically includes: According to the historical operation status data, the operation status data output by the current data processing module is normalized; the normalization method is: ； in, is the normalized running status data, X is the running status data output by the data processing module, and They are the minimum and maximum values ​​of the historical operation status data respectively; The change rate feature is extracted from the normalized operating status data; the calculation formula of the change rate feature is: ； in, is the rate of change characteristic, for t 1. Operation status data at a moment in time, yes( t 1-1) Operation status data at the moment; Set multiple prediction time ranges and output parameters, each prediction time range corresponds to an output parameter, and train a separate extreme gradient boosting model for each prediction time range and each output parameter. The change rate feature is input into the extreme gradient boosting model to obtain the future cooling demand of the server cabinet.

5. The server cabinet liquid immersion cooling system according to claim 4, characterized in that: The construction process of the extreme gradient boosting model specifically includes: Construct an extreme gradient boosting model by using Bayesian optimization to find the best hyperparameters, using the gradient boosting decision tree algorithm, using mean square error as the objective function, and using early stopping. Early stopping methods include: Set the initial optimal validation loss is positive infinity, sets the initial patience count to 0, and sets the patience value Patience value Set as the optimal validation threshold; after each training round, calculate the current validation loss ;if , then update the best validation loss , and reset the patience count; otherwise, reduce the patience count; the patience value represents the best validation loss The maximum number of times training is allowed to continue when there is no improvement. The patience count is the number of times the model is trained. If the patience count exceeds the patience value , stop training; The validation loss is calculated as: ； in, is the number of samples in the validation set, L is the loss function, It is i The true value of the validation set samples, It is i The predicted values ​​of the validation set samples.

6. The server cabinet liquid immersion cooling system according to claim 4, characterized in that: The optimization control module is specifically used for: Define multi-objective optimization functions, including energy consumption function, water quality maintenance function and temperature control accuracy function; The energy consumption function is: ； in, For energy consumption, is the fan power, is the water pump power, is the running time; The water quality maintenance function is: ； in, For water resources value, for The target value of the value, is the conductivity, is the target conductivity, and is the weight coefficient, for( ), for( )’s absolute value; The temperature control accuracy function is: ； in, For temperature control accuracy, is the actual water temperature, To set the temperature, for( )’s absolute value; The multi-objective optimization function is: ； in, α , β , γ are the weight coefficients of energy consumption function, water quality maintenance function and temperature control accuracy function respectively; The set constraints include: cooling tower fan speed range, circulating water pump flow range, water temperature set point range and system cooling capacity constraints.

7. The server cabinet liquid immersion cooling system according to claim 6, characterized in that: The optimization control module is specifically used to use model predictive control MPC combined with particle swarm optimization algorithm PSO to optimize the control strategy and determine the system operation strategy; Among them, the MPC prediction model in the model predictive control MPC is: ； ； PSO update formula: ； ； in, for The system status at the moment, for k The control input at the moment, for k The system output at the moment, A , B and C are system matrices, for t +1 moment i The particle velocity of a particle, for t +1 moment i The particle position of each particle, for t Moment i The particle velocity of a particle, for t Moment i The particle position of each particle, is the individual optimal solution, is the global optimal solution, is the inertia weight, and is the acceleration constant, and is a random number; The control strategy optimization process is as follows: The particle swarm is initialized, and the position of each particle represents a set of cooling control parameters, which include fan speed, water pump flow and temperature set point; Control the server cabinet through cooling control parameters to determine the operating status data of the server cabinet; According to the operation status data, the MPC prediction model is used to predict the system response for each time step in the prediction time domain to obtain the predicted operation status of the server cabinet; Calculate the objective function value of the multi-objective optimization function according to the predicted operating status and cooling demand; According to the objective function value, update the individual optimal solution and the global optimal solution of the particle; The PSO update formula is used to update the particle position and velocity. The individual optimal solution and the global optimal solution are continuously updated through the updated particles and velocity until the maximum number of iterations is reached or the convergence condition is met. The particle position corresponding to the global optimal solution is determined as the system operation strategy.

8. The server cabinet liquid immersion cooling system according to claim 7, characterized in that: The optimization control module is also specifically used for: During the control strategy optimization process, check whether the system operation strategy meets all constraints. If not, use the penalty function to correct the objective function value: The penalty function is: ； in, is the corrected objective function value, F is the original objective function value, is the penalty parameter, For the i constraints.

9. The server cabinet liquid immersion cooling system according to claim 1, characterized in that: The performance evaluation module is used to calculate energy utilization efficiency indicators and generate optimization effect reports; the energy utilization efficiency indicators include energy efficiency indicators and temperature control accuracy.

Images