Intelligent prediction method for thermal disturbance of enclosure structure of machine room

Abstract

The invention relates to the technical field of machine room thermal management, in particular to a machine room enclosure structure thermal disturbance intelligent prediction method, which comprises the following steps: data acquisition and data preprocessing; the theoretical value of the heat flow is calculated according to a physical heat transfer model, the model comprises a heat transfer attenuation function and a temperature delay function, the theoretical value is corrected based on a long-short term memory (LSTM) neural network, a heat transfer model parameter matrix is optimized, and an optimized heat flow prediction model is obtained; and utilizing the optimized heat flow prediction model to predict the thermal disturbance of the machine room enclosure structure in a future time period, generating a prediction result including temperature attenuation, temperature delay and heat flow, outputting the result, and performing deep learning and real-time optimization on historical data by the LSTM to obtain a prediction result of the machine room enclosure structure. And the adaptive LSTM model dynamically corrects a parameter matrix of a physical model according to historical data and real-time data, so that the problem of prediction deviation caused by data exception in a pure AI model is effectively solved.

Description

Technical Field

[0001] The invention relates to the technical field of thermal management of computer rooms, and in particular to an intelligent prediction method for thermal disturbance of a computer room enclosure structure. Background Art

[0002] Traditional computer room thermal management solutions mostly rely on the group control system of air-conditioning equipment, which adjusts the operating status of air-conditioning equipment by monitoring indoor temperature changes in real time. However, in the specific implementation process of such solutions, the impact of external environmental conditions on the thermal disturbance of the computer room enclosure is usually not fully considered. For example, some AI-based air-conditioning energy-saving research models predict heat changes by establishing temperature fields and long- and short-time prediction models, but these models usually only focus on the dynamic changes of internal heat sources, and do not conduct in-depth analysis of interference factors in the external environment and the heat transfer characteristics of the enclosure structure. This limitation leads to a decrease in prediction accuracy when facing drastic changes in the external environment, and the energy-saving effect of the system cannot meet expectations.

[0003] In addition, existing technologies are ambiguous in analyzing the source of heat. For example, when the outdoor temperature is low at night but the temperature inside the computer room still rises, traditional solutions cannot explain the specific cause of this phenomenon. In fact, this temperature change may be due to the heat storage effect of the enclosure structure of the computer room. However, existing technologies have not conducted in-depth research and modeling on the heat transfer attenuation and delay characteristics of the enclosure structure. This not only makes it difficult for the thermal management system to utilize the natural heat dissipation capacity of the enclosure structure, but also limits the efficient use of natural cold sources to a certain extent.

[0004] Furthermore, although existing pure AI models can achieve certain prediction functions through data fitting, due to the lack of correction based on physical laws, their prediction results often deviate from the actual operation situation, especially when the data is abnormal or the input features are incomplete, the model performance is particularly unstable. This approach that relies on black-box algorithms, in the absence of physical guidance, is prone to contradictions or error accumulation that the algorithm cannot explain. Summary of the invention

[0005] In order to solve the above technical problems, the present invention proposes an intelligent prediction method for thermal disturbance of the enclosure structure of the computer room, which combines the physical heat transfer model with the long short-term memory neural network (LSTM), and realizes a more accurate heat flow prediction through the complementarity and synergy of the algorithm and the physical model. The present invention aims to solve the following technical problems: First, how to accurately analyze and quantify the heat source of the computer room, especially the specific ratio of environmental heat transfer to internal heat generation; second, how to optimize the attenuation and delay problems in the heat transfer process based on the heat transfer characteristics of the enclosure structure; third, how to use the external natural cold source and the heat dissipation capacity of the enclosure structure to dynamically adjust the thermal management strategy of the computer room to achieve more efficient energy saving effects.

[0006] The present invention provides a method for intelligently predicting thermal disturbance of a machine room enclosure structure, comprising the following steps: Data collection: The external environment data of the computer room is collected through a micro-meteorological instrument, and the external environment data includes air temperature, air humidity, atmospheric pressure, wind speed, wind direction, radiation and light; the internal environment data of the computer room is collected through an indoor temperature and humidity sensor, and the internal environment data includes the temperature and humidity of each monitoring point in the computer room; Data preprocessing: Based on the external environment data and the internal environment data, combined with the geographic information of the computer room and the material characteristics of the enclosure structure, the data is formatted to generate an input feature matrix; Heat transfer modeling: Calculate theoretical values ​​of heat flow based on a physical heat transfer model, which includes a heat transfer decay function and a temperature delay function; Parameter optimization: Based on the long short-term memory neural network (LSTM), the theoretical value is corrected, and the heat transfer model parameter matrix is ​​further optimized to obtain the optimized heat flow prediction model; Heat flow prediction: using the optimized heat flow prediction model, predict the thermal disturbance of the enclosure structure of the computer room in the future time period, and generate prediction results including temperature attenuation, temperature delay and heat flow; Result output: Based on the prediction results, output thermal management recommendations for the computer room, including natural cooling source utilization strategy and equipment group control strategy.

[0007] Preferably, the data preprocessing step further comprises: Storing historical data; Based on the stored historical data and real-time collected data, the input feature matrix is ​​dynamically updated.

[0008] Preferably, the dynamic update of the historical data and the real-time collected data includes: Normalize historical data to eliminate differences in data dimensions; Anomaly detection algorithms are used to remove outliers from historical data to ensure the accuracy of model input data.

[0009] Preferably, the heat transfer modeling step specifically includes: The temperature attenuation when outdoor heat is transferred to indoor space is calculated by heat transfer attenuation function; The delay of outdoor temperature change being transmitted to indoor temperature is calculated through the temperature delay function; Based on the heat transfer attenuation function and the temperature delay function, a theoretical value of the preliminary heat flow is obtained.

[0010] Preferably, the preliminary heat flow theoretical value is further input into a long short-term memory neural network, which is used to correct the theoretical value and generate an optimized heat flow prediction model.

[0011] Preferably, the parameter optimization step specifically includes: Based on the optimized heat transfer attenuation function and temperature delay function, the model parameter matrix is ​​dynamically adjusted; The dynamic adjustment is accomplished through cyclic training of the long short-term memory neural network.

[0012] Preferably, the heat flow prediction step further comprises: Based on the current external and internal environmental data, predict the temperature decay, temperature delay and heat flow in the future time period; The prediction results are input into the computer room thermal management decision module to generate equipment group control strategies.

[0013] Preferably, the device group control strategy includes: Dynamically adjust the start and stop status of the air-conditioning unit; The fan operating speed is adjusted based on the heat flow prediction results to achieve energy-saving control.

[0014] Preferably, the thermal management decision module further determines whether to enable the natural cooling source based on the heat flow prediction result; when the predicted heat flow is negative and the heat dissipation of the enclosure structure is greater than the heat generation of the equipment inside the computer room, some air-conditioning equipment is turned off.

[0015] Preferably, the activation of the natural cooling source includes: Dynamically evaluate the use effect of the natural cooling source according to the heat dissipation capacity of the enclosure structure and the external ambient temperature; While meeting the heat dissipation requirements, natural cooling sources should be used first to reduce the energy consumption of the computer room.

[0016] The beneficial effects of the present invention are: (1) The present invention fully solves the problems of low model prediction accuracy, fuzzy analysis of heat source components, and insufficient utilization of natural cooling sources in the prior art by introducing the synergistic effect of the physical heat transfer model and the LSTM algorithm. The physical model provides a scientific theoretical basis for the algorithm, while LSTM further improves the prediction accuracy and adaptability of the model through deep learning and real-time optimization of historical data.

[0017] (2) In terms of heat prediction, the physical model makes a preliminary calculation of the theoretical value of heat flow by establishing a heat transfer attenuation function and a temperature delay function. The LSTM model dynamically corrects the parameter matrix of the physical model based on historical data and real-time data, effectively solving the problem of prediction deviation caused by data anomalies in pure AI models. This dual guarantee mechanism of algorithm and physical model ensures the accuracy and reliability of the prediction results.

[0018] (3) In terms of thermal management strategy optimization, the present invention proposes a mechanism for dynamically adjusting equipment group control. When the prediction results show that the heat flux is negative and the heat dissipation of the enclosure structure is greater than the heat generated by the equipment, the system can automatically adjust the start and stop status of the air-conditioning equipment to maximize energy efficiency. At the same time, through dynamic analysis of external environmental data, the system can give priority to natural cooling sources to further reduce the operating energy consumption of the computer room.

[0019] (4) In addition, the modular design and data processing flow of the present invention also provide guarantees for the scalability and robustness of the system. Through the collection and preprocessing of multi-source data, combined with the theoretical calculation of the physical model and the deep learning of LSTM, the system can flexibly respond to the thermal disturbance of the computer room under complex environmental conditions, significantly improving the intelligent level of thermal management.

[0020] In summary, the present invention has achieved a significant technical breakthrough in solving the problem of thermal management in computer rooms through the organic combination of algorithms and physical models, and has important engineering application value and promotion significance. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a schematic diagram of the arrangement of the micro-meteorological instrument and indoor temperature and humidity sensor required by the present invention.

[0022] Figure 2 It is a general schematic diagram of the data flow of the present invention.

[0023] Figure 3 It is a curve chart of indoor and outdoor temperature comparison in the present invention.

[0024] Figure 4 Schematic diagram of the model for heat transfer decay, temperature delay, and parameter optimization.

[0025] Figure 5 This is a detailed structural diagram of the LSTM neural network. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solutions 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. In addition, the technical features involved in each embodiment of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0027] like Figure 1-5 As shown, the present invention relates to an intelligent prediction method for thermal disturbance of a computer room enclosure structure, which provides a scientific basis for thermal management of a computer room by collecting, preprocessing, heat transfer modeling and optimizing algorithms of various data in combination with prediction output.

[0028] Figure 1The layout of the micro-meteorological instrument and indoor temperature and humidity sensors inside and outside the computer room is shown. The micro-meteorological instrument is placed outside the computer room to collect outdoor environmental data, including air temperature, humidity, atmospheric pressure, wind speed, wind direction, radiation and light, etc. These data can reflect the dynamic changes of the external environment. Indoor temperature and humidity sensors are placed at multiple key points in the computer room, such as near equipment cabinets and air-conditioning outlets, to accurately monitor the temperature and humidity in different areas. This layout ensures the comprehensiveness and accuracy of data collection, and provides a sufficient data basis for subsequent modeling.

[0029] pass Figure 1 The hardware layout shown in the figure can collect external meteorological data and internal temperature and humidity data in real time. Figure 2 The process shown in the figure first goes through the preprocessing module, normalization processing makes different data sources have consistent scales, and outliers are eliminated to ensure the reliability of the model input data. The preprocessed data is combined with the material properties of the enclosure structure (such as Figure 1 The feature matrix is ​​further constructed to provide input for subsequent model calculations.

[0030] Preferably, in one embodiment of the present invention, environmental data outside and inside the computer room are collected by a micro-meteorological instrument and an indoor temperature and humidity sensor. Specifically, the external environmental data include air temperature, air humidity, atmospheric pressure, wind speed, wind direction, radiation and light, which are collected by the micro-meteorological instrument and transmitted in real time. The internal environmental data includes the temperature and humidity of each monitoring point in the computer room, which are usually collected in real time by multiple distributed temperature and humidity sensors in the computer room. This type of data collection scheme ensures comprehensive perception of the internal and external environment of the computer room, and provides accurate data input for subsequent steps.

[0031] For example, preferably, when the air temperature is between -20°C and 45°C, the sensor is automatically calibrated to ensure data accuracy. The humidity range is usually 10% to 90% to meet the actual environmental requirements of the computer room. The preferred value of the atmospheric pressure parameter is 1000hPa to 1050hPa. These empirical values ​​have important reference value in the heat transfer model.

[0032] In the present invention, multi-source data collection is the basis of the entire thermal disturbance prediction process. The external environmental data collected by the micro-meteorological instrument include air temperature, air humidity, atmospheric pressure, wind speed, wind direction, radiation and light, which can fully reflect the dynamic environmental changes outside the computer room. For example, under extreme weather conditions, such as summer rainstorms or winter cold waves, the external environmental data will fluctuate significantly, which puts higher requirements on thermal disturbance prediction.

[0033] Figure 2The workflow from data collection to prediction output is shown. After the data is collected from the meteorological instrument and temperature and humidity sensor, it is first normalized and outliers are removed by the preprocessing module, and then input into the heat transfer model to calculate the preliminary heat flow value. Subsequently, the preliminary calculation results and historical data are jointly input into the LSTM model for optimization and correction, and finally the prediction results and optimization strategies are output. This flowchart intuitively shows the data flow and logical relationship between the modules of the system.

[0034] Preferably, to ensure the accuracy of data collection, the micro-meteorological instrument and indoor temperature and humidity sensor have an automatic calibration function. When the data is detected to deviate from the normal range, the system will trigger the automatic calibration mechanism to correct the sensor reading error. Especially for wind speed and radiation parameters, these data directly affect the heat transfer characteristics of the enclosure structure, so the accuracy of data collection is crucial.

[0035] Furthermore, the external and internal environmental data are formatted by the data preprocessing module, combined with the room's geographic information (such as longitude and latitude, altitude, wall elevation, azimuth inclination) and enclosure material properties (such as thermal conductivity, heat capacity, thickness, etc.) to generate an input feature matrix. The construction of the feature matrix is ​​the basis for subsequent model calculations, which comprehensively considers the dynamic characteristics of the environment and the static characteristics of the enclosure.

[0036] For example, when the wall material of the machine room is concrete, its thermal conductivity is generally between 1.2W / (m·K) and 1.8W / (m·K), and the heat capacity is preferably between 900J / (kg·K) and 1200J / (kg·K). With such parameter configuration, the actual heat transfer process can be better simulated.

[0037] Furthermore, the collected multi-source data needs to be formatted by the preprocessing module to generate a feature matrix. This feature matrix not only includes environmental data, but also integrates the room's geographic information and the characteristics of the enclosure material. For enclosure materials, such as concrete or glass, their thermal conductivity, thermal capacity, and thickness directly determine the rate and efficiency of heat transfer.

[0038] In order to enhance the adaptability of the feature matrix, the system adopts a dynamic update mechanism to integrate the latest data and historical data in real time. For example, the wall thickness of a certain computer room is 30cm and the thermal conductivity is 1.5W / (m·K). On this basis, combined with the external ambient temperature and the heat generated by the internal equipment, the feature matrix can be dynamically adjusted to ensure the comprehensiveness and accuracy of the model input.

[0039] In the present invention, the heat transfer characteristics of the computer room enclosure structure are calculated through a physical heat transfer model. Specifically, heat transfer attenuation and temperature delay are key model parameters. In some preferred cases, the heat transfer attenuation function can describe the energy loss when external heat passes through the wall, and its calculation formula is as follows:

[0040] in, represents the thermal conductivity of the enclosure structure, preferably 1.2W / (m·K) to 1.8W / (m·K), The wall area is usually 10㎡ to 30㎡. and are the temperature differences between the outside and inside, respectively. In practical applications, these parameters need to be adjusted according to the specific structure and environmental conditions of the computer room.

[0041] according to Figure 3 The indoor and outdoor temperature comparison curves shown in Figure 2 show that the enclosure structure has obvious attenuation and delay effects on heat transfer. Specifically, the physical model calculates the preliminary heat flow value through the heat transfer attenuation function and the temperature delay function. This method ensures that the system can accurately capture the transfer relationship between indoor and outdoor temperature dynamics. For example, when the outside temperature drops at night, the indoor temperature change will lag due to the heat storage characteristics of the wall, causing a temperature delay phenomenon.

[0042] Preferably, in certain high heat load scenarios, the heat storage characteristics of the wall have a particularly significant impact on the temperature delay. For example, the heat storage time of a concrete structure is often between 2 and 3 hours, which has a direct impact on the transmission of nighttime temperature changes. Through the above modeling, the trend of heat flow changes can be predicted at different time nodes, providing reliable preliminary data for subsequent algorithm corrections.

[0043] like Figure 4 As shown in the figure, the LSTM model is used to correct the preliminary heat flow value output by the physical model. By learning from historical data, the LSTM can dynamically adjust the weight matrix W to adapt it to real-time environmental changes. Combining the basic calculation of the physical model with the deep learning correction of the LSTM, a high-precision heat flow prediction value is finally obtained.

[0044] One of the innovations of the present invention is to combine the long short-term memory neural network (LSTM) with the physical model to correct the heat flow value calculated initially. Although the physical model can provide the theoretical value of the heat flow, due to the complex and changeable actual environmental conditions, relying solely on the physical model may not fully cover all dynamic factors. Therefore, through the LSTM model, these theoretical values ​​can be dynamically optimized and corrected.

[0045] The calculation of temperature delay is based on the heat storage characteristics of the building envelope. For example, for concrete structures, the delay time is usually between 1 and 3 hours. This delay parameter is calculated using the heat wave propagation formula and used for subsequent heat flow correction.

[0046] Figure 5 The internal structure of the LSTM model is shown in detail, including input, forget, update, output and other gating units. The model can maintain the long-term dependency of time series data in multi-step prediction, thus effectively solving the impact of short-term drastic changes in the environment on prediction accuracy.

[0047] Preferably, the present invention corrects the above preliminary heat flow prediction value through a long short-term memory neural network (LSTM). The input of the LSTM model includes a feature matrix and a preliminary calculated heat flow value, and an optimized heat flow prediction model is generated through cyclic training and parameter optimization. The formula is as follows:

[0048] in, is the theoretical heat flow value output by the physical model, is the weight matrix of the LSTM neural network, which is optimized by the deep learning algorithm.

[0049] Preferably, the training cycle of the LSTM model is 1 hour, and each training uses data from the past 24 hours to ensure that the model can maintain high prediction accuracy in both short-term and long-term changes.

[0050] For example, the time step of LSTM is preferably 5 to 10 minutes to ensure the capture of short-term thermal disturbances. The number of network layers is usually 2 to 3, which can be dynamically adjusted according to the actual data volume to balance computational efficiency and prediction accuracy.

[0051] The optimized model predicts the thermal disturbance in the future time period, and the system generates thermal management strategies based on the prediction results. These strategies include air conditioning start and stop, wind speed adjustment and natural cooling source utilization. For example, when the predicted heat flow is negative and the external temperature is below 15°C, the system will give priority to shutting down some air conditioning equipment and use natural cooling sources to dissipate heat to reduce energy consumption.

[0052] Finally, the optimized heat flux prediction model can predict future thermal disturbances. Based on the prediction results, a comprehensive report including temperature decay, temperature delay and heat flux trend is output to provide accurate decision support for the thermal management of the computer room. For example, when the prediction results show that the heat flux is negative and the heat dissipation of the enclosure structure exceeds the heat generated by the equipment in the computer room (usually 5kW to 10kW), the system will automatically recommend shutting down some air conditioning equipment and enabling natural cooling.

[0053] Preferably, the prediction function is suitable for 24-hour continuous operation scenarios, and optimal control of energy consumption is achieved by dynamically adjusting the thermal management strategy of the computer room.

[0054] Through the optimized heat flow prediction model, the present invention can predict the temperature decay, temperature delay and heat flow change trend in the future time period. This prediction function is of great value in practical applications. For example, when the system predicts that the external temperature will drop sharply within the next hour and the heat flow is negative, the system can shut down some air-conditioning equipment in advance and turn on the fan to assist in heat dissipation, thereby achieving efficient use of energy.

[0055] Preferably, in order to ensure the accuracy of the prediction results, the present invention adopts a multi-step verification mechanism to compare the predicted value with the real-time collected data. When the error between the two is less than 5%, the model prediction result is considered to be credible.

[0056] Through the above technical solution, the present invention provides a method for intelligently predicting thermal disturbances in the enclosure structure of a computer room. This method combines multi-source data, physical models and intelligent algorithms, which not only improves the accuracy of thermal management, but also enables dynamic optimization and forward-looking management, effectively reducing energy consumption in the computer room and ensuring stable operation of equipment.

[0057] Another important innovation of the present invention is the automatic generation and optimization of thermal management strategies. The system dynamically adjusts the equipment group control strategy based on the prediction results. This strategy includes equipment start and stop adjustment, wind speed control, and refrigerant flow optimization. For example, when the heat flow forecast for a certain area is too high, the system will give priority to increasing the wind speed in that area, or start the backup cooling equipment in advance to ensure that the ambient temperature is maintained within the range of 22°C to 26°C.

[0058] Preferably, the present invention further generates a specific thermal management strategy for the computer room based on the optimized heat flow prediction results. In one embodiment, the generation of the thermal management strategy is implemented by a decision module, which includes logical analysis and rule matching of the prediction results. In particular, when the heat flow is predicted to be a negative value, the system can automatically adjust the operating state of the HVAC equipment according to the comparison result of the heat dissipation and the heat generation of the equipment.

[0059] For example, when the prediction results show that the heat dissipation of the enclosure structure is -8kW and the heat generation of the equipment in the computer room is 5kW, the system will recommend turning off some air conditioning equipment and turning on the fans to make full use of the natural cooling source. This dynamic strategy adjustment can not only significantly reduce energy consumption, but also avoid the risk of equipment overheating.

[0060] Furthermore, the execution of the equipment group control strategy is completed through the collaboration of multiple modules. The strategy execution module is closely integrated with the LSTM model, and the equipment operation parameters are adjusted immediately after receiving new prediction results. Through this dynamic adjustment mechanism, the system can respond to environmental changes in a short time and improve the flexibility of thermal management in the computer room.

[0061] Furthermore, in order to ensure the accuracy of strategy generation, the system will generate strategies based on environmental conditions in different time periods, such as when the outdoor temperature is low at night or in the early morning, natural heat dissipation will be given priority. This flexible strategy generation mechanism enables the present invention to maintain good thermal management effects under a variety of complex environmental conditions.

[0062] In another preferred embodiment of the present invention, the prediction results are further used to dynamically adjust the equipment group control strategy. Through comprehensive analysis of the predicted heat flow value and historical data, the system can optimize the operating parameters of the air conditioning equipment and fans in the computer room in real time. This process includes multiple dimensions such as equipment start and stop control, wind speed adjustment, and refrigerant flow regulation.

[0063] For example, when the system detects that the temperature in a certain area is too high and the forecast shows that the heat flow will continue to increase in the next 5 minutes, the backup air conditioning equipment can be activated in advance to ensure that the temperature in the area remains within a safe range. Usually, in order to ensure efficient operation of the equipment, the temperature threshold of the start-stop control is preferably set between 22°C and 26°C.

[0064] Furthermore, in order to enhance the robustness of the system, before each strategy adjustment, the system will conduct a secondary check on the current environmental data to ensure the accuracy of the input data. The dynamic and accurate nature of this strategy can significantly improve the operational stability and energy efficiency of the computer room.

[0065] The present invention also proposes an optimization strategy for the use of natural cooling sources. Preferably, when the predicted heat flow is negative and the heat dissipation is greater than the heat generated by the equipment, the system will give priority to using the natural cooling source. This judgment process is based on the heat dissipation performance of the enclosure structure of the computer room and the external environmental conditions.

[0066] In a specific embodiment, the comparison of the heat dissipation and the heat generation of the device is calculated by the following formula:

[0067] When the calculation results show When the temperature drops below 15°C, the system will shut down some air conditioning equipment and use natural cooling sources to dissipate heat. To ensure the effectiveness of this strategy, threshold setting is crucial. For example, in spring and autumn, when the temperature difference is large, it is usually more suitable to enable natural cooling sources when the external temperature is below 15°C.

[0068] In terms of the use of natural cooling sources, the present invention sets a multi-level threshold judgment mechanism. When the system detects that the external ambient temperature is lower than 15°C and the predicted heat flow is negative, the natural cooling source will be enabled for heat dissipation first. This optimization strategy can not only reduce the power consumption of equipment operation, but also significantly improve the overall thermal management efficiency.

[0069] Through this strategy, the present invention can reduce energy consumption while ensuring the stability of the ambient temperature of the computer room. Especially in the scenario of long-term operation, this strategy optimization significantly extends the service life of the equipment.

[0070] The combination of the physical model and the LSTM model effectively improves the accuracy of thermal disturbance prediction and overcomes the defect of a single AI model that is slow to respond to environmental changes.

[0071] The thermal management strategy generation module based on the prediction results realizes the dynamic adjustment of the computer room environment and significantly improves the energy utilization efficiency of the computer room.

[0072] Multi-source data collection and outlier elimination mechanisms enhance the robustness of the system, ensuring stable operation in complex environments.

[0073] In summary, the present invention provides an efficient, accurate and intelligent thermal management solution for modern computer rooms by optimizing data collection, physical modeling and intelligent prediction.

[0074] In order to help operation and maintenance personnel better understand the overall status of thermal management in the computer room, the present invention can generate a detailed thermal management optimization report. The report content covers the equipment operation status, heat flow trend analysis and optimization suggestions, providing an important reference for subsequent strategy adjustments.

[0075] Preferably, the present invention can also generate a detailed thermal management optimization report based on the prediction results. The report comprehensively analyzes the thermal disturbance characteristics, equipment operating status and environmental conditions of the computer room, evaluates the execution effect of the thermal management strategy in the computer room, and provides suggestions for further optimization.

[0076] In a specific embodiment, the thermal management optimization report may include the following key parts: Analysis of thermal disturbance characteristics: Analysis of the heat transfer characteristics of the enclosure structure, including the distribution curves of temperature attenuation and temperature delay.

[0077] Equipment operation efficiency evaluation: statistics on performance parameters such as the start and stop frequency, average power consumption, etc. of air-conditioning equipment and fans.

[0078] Energy efficiency optimization suggestions: Provide adjustment suggestions for device group control strategies based on historical data and predicted trends.

[0079] For example, in a certain report, when the start and stop frequency of equipment in a certain area is significantly higher than that in other areas, the report will recommend optimizing the air conditioning settings in that area, such as lowering the wind speed or adjusting the startup temperature, to reduce equipment wear and power consumption.

[0080] Through the thermal management optimization report, operation and maintenance personnel can clearly understand the overall status of the thermal management of the computer room and take corresponding measures in a timely manner to improve management efficiency.

[0081] To ensure the high quality of input data, the present invention also designs a data anomaly detection and automatic correction module. Through the statistical outlier detection algorithm, the system can identify and correct abnormal data in real time to ensure that the input data of the model is always reliable. This module is particularly important in extreme environmental conditions, such as equipment failure or sensor failure.

[0082] Based on the distribution characteristics of historical data, this module can effectively identify outliers in environmental data and equipment data and perform automatic corrections.

[0083] Preferably, the anomaly detection algorithm uses the triple standard deviation method based on statistical distribution. When a data point collected at a certain time deviates from the mean by more than three standard deviations, the data point will be marked as abnormal. For example, when a temperature sensor is operating normally, the temperature variation range is usually 22°C to 26°C, but if the temperature value read suddenly jumps to 35°C, the system will determine it as abnormal data and automatically replace it with the average value of the time period.

[0084] This automatic correction mechanism can significantly improve the quality of the model's input data, thereby improving the accuracy and reliability of the prediction results.

[0085] The present invention combines innovative physical models with intelligent algorithms, which not only effectively solves many technical problems in the thermal management of computer rooms, but also provides reliable guarantees for improving energy efficiency and equipment operation stability. Its modular design and flexible application characteristics ensure the wide applicability of the present invention in various complex scenarios. Through the above detailed description, the present invention provides a comprehensive, accurate and flexible intelligent prediction method for thermal disturbances of computer room enclosures. Its core lies in the organic combination of multi-source data, physical models and intelligent algorithms to achieve intelligent, dynamic and forward-looking thermal management of computer rooms. The detailed design and parameter optimization of each module have been fully demonstrated to ensure efficient operation in different application scenarios.

[0086] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for intelligent prediction of thermal disturbance of a computer room enclosure structure, characterized in that: The following steps are involved: Data collection: The external environment data of the computer room is collected through a micro-meteorological instrument, and the external environment data includes air temperature, air humidity, atmospheric pressure, wind speed, wind direction, radiation and light; the internal environment data of the computer room is collected through an indoor temperature and humidity sensor, and the internal environment data includes the temperature and humidity of each monitoring point in the computer room; Data preprocessing: Based on the external environment data and the internal environment data, combined with the geographic information of the computer room and the material characteristics of the enclosure structure, the data is formatted to generate an input feature matrix; Heat transfer modeling: Calculate theoretical values ​​of heat flow based on a physical heat transfer model, which includes a heat transfer decay function and a temperature delay function; Parameter optimization: Based on the long short-term memory neural network (LSTM), the theoretical value is corrected, and the heat transfer model parameter matrix is ​​further optimized to obtain an optimized heat flow prediction model; Heat flow prediction: using the optimized heat flow prediction model, predict the thermal disturbance of the enclosure structure of the computer room in the future time period, and generate prediction results including temperature attenuation, temperature delay and heat flow; Result output: Based on the prediction results, output thermal management recommendations for the computer room, including natural cooling source utilization strategy and equipment group control strategy.

2. The intelligent prediction method for thermal disturbance of the enclosure structure of a computer room according to claim 1 is characterized in that: The data preprocessing step further comprises: Storing historical data; Based on the stored historical data and real-time collected data, the input feature matrix is ​​dynamically updated.

3. The intelligent prediction method for thermal disturbance of the enclosure structure of a computer room according to claim 2 is characterized in that: The dynamic update of the historical data and the real-time collected data includes: Normalize historical data to eliminate differences in data dimensions; Anomaly detection algorithms are used to remove outliers from historical data to ensure the accuracy of model input data.

4. The intelligent prediction method for thermal disturbance of the enclosure structure of a computer room according to claim 1 is characterized in that: The heat transfer modeling step specifically includes: The temperature attenuation when outdoor heat is transferred to indoor space is calculated by heat transfer attenuation function; The delay of outdoor temperature change being transmitted to indoor temperature is calculated through the temperature delay function; Based on the heat transfer attenuation function and the temperature delay function, a theoretical value of preliminary heat flow is obtained.

5. The intelligent prediction method for thermal disturbance of the enclosure structure of a computer room according to claim 4 is characterized in that: The preliminary heat flow theoretical value is further input into a long short-term memory neural network, which is used to correct the theoretical value and generate an optimized heat flow prediction model.

6. The intelligent prediction method for thermal disturbance of the enclosure structure of a computer room according to claim 5 is characterized in that: The parameter optimization step specifically includes: Based on the optimized heat transfer attenuation function and temperature delay function, the model parameter matrix is ​​dynamically adjusted; The dynamic adjustment is accomplished through cyclic training of the long short-term memory neural network.

7. The intelligent prediction method for thermal disturbance of a computer room enclosure structure according to claim 1 is characterized in that: The heat flow prediction step further comprises: Based on the current external and internal environmental data, predict the temperature decay, temperature delay and heat flow in the future time period; The prediction results are input into the computer room thermal management decision module to generate equipment group control strategies.

8. The intelligent prediction method for thermal disturbance of the enclosure structure of a computer room according to claim 7 is characterized in that: The device group control strategy includes: Dynamically adjust the start and stop status of the air-conditioning unit; The fan operating speed is adjusted based on the heat flow prediction results to achieve energy-saving control.

9. The intelligent prediction method for thermal disturbance of a computer room enclosure structure according to claim 7 is characterized in that: The thermal management decision module further determines whether to enable the natural cooling source based on the heat flow prediction result; when the predicted heat flow is negative and the heat dissipation of the enclosure structure is greater than the heat generated by the equipment inside the computer room, some air-conditioning equipment is turned off.

10. The intelligent prediction method for thermal disturbance of the enclosure structure of a computer room according to claim 9, characterized in that: The activation of the natural cooling source includes: Dynamically evaluate the use effect of the natural cooling source according to the heat dissipation capacity of the enclosure structure and the external ambient temperature; While meeting the heat dissipation requirements, natural cooling sources should be used first to reduce the energy consumption of the computer room.

Images