WO2023010556A1 - 精密空调的动态预测控制方法、装置和系统 - Google Patents

精密空调的动态预测控制方法、装置和系统 Download PDF

Info

Publication number
WO2023010556A1
WO2023010556A1 PCT/CN2021/111276 CN2021111276W WO2023010556A1 WO 2023010556 A1 WO2023010556 A1 WO 2023010556A1 CN 2021111276 W CN2021111276 W CN 2021111276W WO 2023010556 A1 WO2023010556 A1 WO 2023010556A1
Authority
WO
WIPO (PCT)
Prior art keywords
precision air
air conditioner
dynamic
model
simulation
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2021/111276
Other languages
English (en)
French (fr)
Inventor
白新
周晓舟
孙天瑞
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Siemens Schweiz AG
Siemens Ltd China
Original Assignee
Siemens Schweiz AG
Siemens Ltd China
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Siemens Schweiz AG, Siemens Ltd China filed Critical Siemens Schweiz AG
Priority to US18/681,136 priority Critical patent/US20240280286A1/en
Priority to CN202180098830.0A priority patent/CN117413147A/zh
Priority to PCT/CN2021/111276 priority patent/WO2023010556A1/zh
Priority to EP21952430.3A priority patent/EP4375585A4/en
Publication of WO2023010556A1 publication Critical patent/WO2023010556A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F24—HEATING; RANGES; VENTILATING
    • F24F—AIR-CONDITIONING; AIR-HUMIDIFICATION; VENTILATION; USE OF AIR CURRENTS FOR SCREENING
    • F24F11/00—Control or safety arrangements
    • F24F11/62—Control or safety arrangements characterised by the type of control or by internal processing, e.g. using fuzzy logic, adaptive control or estimation of values
    • F24F11/63—Electronic processing
    • F24F11/64—Electronic processing using pre-stored data
    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F24—HEATING; RANGES; VENTILATING
    • F24F—AIR-CONDITIONING; AIR-HUMIDIFICATION; VENTILATION; USE OF AIR CURRENTS FOR SCREENING
    • F24F11/00—Control or safety arrangements
    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F24—HEATING; RANGES; VENTILATING
    • F24F—AIR-CONDITIONING; AIR-HUMIDIFICATION; VENTILATION; USE OF AIR CURRENTS FOR SCREENING
    • F24F11/00—Control or safety arrangements
    • F24F11/70—Control systems characterised by their outputs; Constructional details thereof
    • F24F11/72—Control systems characterised by their outputs; Constructional details thereof for controlling the supply of treated air, e.g. its pressure
    • F24F11/74—Control systems characterised by their outputs; Constructional details thereof for controlling the supply of treated air, e.g. its pressure for controlling air flow rate or air velocity
    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F24—HEATING; RANGES; VENTILATING
    • F24F—AIR-CONDITIONING; AIR-HUMIDIFICATION; VENTILATION; USE OF AIR CURRENTS FOR SCREENING
    • F24F11/00—Control or safety arrangements
    • F24F11/70—Control systems characterised by their outputs; Constructional details thereof
    • F24F11/80—Control systems characterised by their outputs; Constructional details thereof for controlling the temperature of the supplied air
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/28—Design optimisation, verification or simulation using fluid dynamics, e.g. using Navier-Stokes equations or computational fluid dynamics [CFD]
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/08—Thermal analysis or thermal optimisation

Definitions

  • the invention relates to the field of precision air conditioners, in particular to a dynamic predictive control method, device and system for precision air conditioners.
  • thermal and energy management is critical in reducing performance usage effectiveness (PUE, performance usage effectiveness) and protecting IT equipment from running smoothly.
  • IT equipment includes server racks, blade servers, and the like.
  • the HAVC system, especially the CRAC (Computer Room Air Conditioner) unit accounts for most of the energy loss except for IT equipment.
  • An average 1°C increase in temperature inside a data center can lead to energy savings of 4% to 5% due to reduced cooling requirements. Therefore, dynamic analysis of thermal conditions has the potential to improve the energy performance of energy in data centers and other smart infrastructures and reduce operating costs.
  • the first aspect of the present invention provides a dynamic predictive control method for a precision air conditioner, which includes the following steps: S1, generating a three-dimensional simulation model based on a simulation template based on the static information of the hardware configuration and layout of the application scene where the precision air conditioner is located, wherein , the three-dimensional simulation model and the static model of the hardware configuration and layout of the application scene correspond to each other; S2, perform simulation based on the three-dimensional simulation model corresponding to the dynamic parameters of the precision air conditioner, and use machine learning to establish dynamic parameters and simulation Correspondence of the results to generate a proxy model; S3, based on the proxy model and the real-time dynamic data of the precision air conditioner, predict the temperature distribution and change of the application scene.
  • the dynamic predictive control method of the precision air conditioner further includes the following steps: iteratively optimizing at least one control parameter of the precision air conditioner based on the three-dimensional simulation model and/or the proxy model, wherein the control parameter Including the number of air conditioners turned on, supply air temperature and flow.
  • the dynamic predictive control method of the precision air conditioner also includes the following steps: comparing the space temperature demand of the application scenario with a first predetermined threshold according to the load plan of the hardware configuration of the application scenario and its power curve, such as If it is judged that the precision air conditioner is undercooled or overcooled, the operating parameters of the precision air conditioner are adjusted or the load plan of the application scenario is adjusted.
  • the dynamic predictive control method of the precision air conditioner also includes the following steps: comparing the actual value of the physical sensor of the hardware configuration of the application scenario with the predicted value of the simulation model, if the actual value and the predicted value If the deviation exceeds a second predetermined threshold, the static parameter is adjusted and a parameter model is generated based on the change of the static parameter and the prediction is performed using a computational fluid dynamics simulator.
  • the three-dimensional simulation model is based on computational fluid dynamics, which numerically simulates and calculates mass, energy, and momentum transfer generated during fluid flow in three-dimensional space, and displays the The calculation results of the temperature, pressure and velocity of the fluid in the application scene space are presented graphically.
  • the second aspect of the present invention provides a dynamic predictive control system for precision air-conditioning, including: a processor; and a memory coupled to the processor, the memory having instructions stored therein, and when the instructions are executed by the processor, the The dynamic predictive control system of the precision air conditioner performs an action, and the action includes: S1, generating a three-dimensional simulation model based on a simulation template based on the hardware configuration and layout of the application scene where the precision air conditioner is located, and the static information of the layout, wherein the three-dimensional The simulation model and the static model of the hardware configuration and layout of the application scene correspond to each other; S2, perform simulation based on the three-dimensional simulation model corresponding to the dynamic parameters of the precision air conditioner, and use machine learning to establish a correspondence between the dynamic parameters and the simulation results to generate a proxy model; S3, based on the proxy model and the real-time dynamic data of the precision air conditioner, predict the temperature distribution and change of the application scene.
  • the dynamic predictive control system of the precision air conditioner also includes the following steps: iteratively optimizing at least one control parameter of the precision air conditioner based on the three-dimensional simulation model and/or the proxy model, wherein the control parameter Including the number of air conditioners turned on, supply air temperature and flow.
  • the action includes comparing the space temperature requirement of the application scenario with a first predetermined threshold according to the load plan of the hardware configuration of the application scenario and its power curve, such as judging that the precision air conditioner is undercooled or overcooled Then adjust the operating parameters of the precision air conditioner or adjust the load plan of the application scenario.
  • the action includes: comparing the actual value of the physical sensor of the hardware configuration of the application scenario with the predicted value of the simulation model, and if the deviation between the actual value and the predicted value exceeds a second predetermined threshold, then Static parameters are adjusted and parametric models are generated based on changes in the static parameters and predictions are performed using a computational fluid dynamics simulator.
  • the three-dimensional simulation model is based on computational fluid dynamics, which numerically simulates and calculates mass, energy, and momentum transfer generated during fluid flow in three-dimensional space, and displays the The calculation results of the temperature, pressure and velocity of the fluid in the application scene space are presented graphically.
  • the third aspect of the present invention provides a dynamic predictive control device for a precision air conditioner, which includes: a generating device that generates a three-dimensional simulation model based on a simulation template based on the hardware configuration and layout of the application scene where the precision air conditioner is located and the static information of the layout, wherein, The three-dimensional simulation model and the static model of the hardware configuration and layout of the application scene correspond to each other; the simulation device performs simulation based on the three-dimensional simulation model corresponding to the dynamic parameters of the precision air conditioner, and uses machine learning to establish dynamic parameters and simulation The corresponding relationship of the results is used to generate a proxy model; the prediction device predicts the temperature distribution and change of the application scene based on the proxy model and the real-time dynamic data of the precision air conditioner.
  • a fourth aspect of the present invention provides a computer program product tangibly stored on a computer-readable medium and comprising computer-executable instructions which, when executed, cause at least one processor to perform the The method described in the first aspect of the present invention.
  • a fifth aspect of the present invention provides a computer-readable medium on which are stored computer-executable instructions that, when executed, cause at least one processor to perform the method according to any one of the first aspects of the present invention. .
  • the dynamic predictive control mechanism of the precision air conditioner fills the gap in the dynamic monitoring and optimization system between the hardware and the data center management software, such as energy management and task scheduling, so that the operator can understand and control the operation of the precision air conditioner more clearly thermal performance.
  • the optimization feature based on the physical simulator helps to reduce unnecessary energy loss on the power supply side, and does not affect the daily working conditions of IT equipment or require an excessive number of physical sensors.
  • the irregularity detection module provides early warning of potential system failures to avoid unnecessary downtime.
  • Fig. 1 is a structural schematic diagram of a dynamic predictive control system of a precision air conditioner according to a specific embodiment of the present invention
  • Fig. 2 is the principle schematic diagram of the emulator of the dynamic predictive control system of precision air conditioner according to a specific embodiment of the present invention
  • Fig. 3 shows the influence of different data center cabinet air supply temperature on the cabinet outlet air temperature under different load conditions of the simulator of the dynamic predictive control system of the precision air conditioner according to a specific embodiment of the present invention.
  • the present invention provides a dynamic predictive control mechanism for precision air-conditioning.
  • the present invention provides a dynamic prediction and optimization system for thermal management of precision air-conditioning application scenarios based on digital twins. performance modeling.
  • the dynamic predictive control system of precision air conditioner CRAC includes a monitoring and predicting device 100 , an optimizing device 200 , an abnormality detecting device 300 , a simulator 400 , and a database 500 .
  • the simulator 400 can be called by any one of the above-mentioned monitoring and predicting device 100 , optimizing device 200 , and abnormality detecting device 300 and return predicted performance data for further analysis.
  • the southbound interface 600 interfaces with field data and other external systems, and simultaneously provides status of current operations and recommendations for adjustments or execution of external applications.
  • the on-site data includes sensors and DDC control from precision air-conditioning application scenarios
  • the other external systems include, for example, management systems of application scenarios
  • other external applications include job schedulers
  • the dynamic predictive control system of precision air conditioner also includes external interface P 2 , sensor S, data center management tool IT and software interface P 1 .
  • the database 500 saves the system configuration, operating status and parameters required by the monitoring and forecasting device 100 , the optimization device 200 , and the anomaly detection device 300 , as well as the output of predetermined prediction or special analysis.
  • the system configuration includes data center server and ventilation system layout parameters, IT equipment ID and power capacity, precision air-conditioning unit ID and power capacity, energy billing, etc.
  • Operational status includes sensor readings, individual server load percentages, primary positioning points for precision air conditioning units, and more.
  • the output data includes virtual temperature distribution and predicted energy demand etc.
  • the monitoring and predicting device 100 executes data extraction from the database 500 at a specific interval, for example, 15 to 30 minutes. Every two sets of data should be considered to perform a prediction, and the dynamic data describe the collective information of the precision air conditioner target application scenario, that is, the dynamic data of the operating status.
  • the simulator 400 stores the collected data in the form of XML files.
  • the simulation engine uses any CFD model, such as the Flotherm model or the surrogate model, to predict the temperature distribution taking into account the current operating status and temperature information at key locations in the application scenario, which will be displayed with virtual sensors, enabling the operator to identify potential hot spots (hot spots) and evaluate existing run settings.
  • the choice of simulation strategy is based on the accuracy and availability of the empirical surrogate model.
  • the above-mentioned modules can also provide forecasts for upcoming events and provide early warning of demand-side operations if current operating settings are not able to meet expected temperature demands.
  • GUI is the user interface for the user and the dynamic predictive control of the precision air conditioner provided by the present invention.
  • the optimization device 200 provides recommendations from providers for running or external applications, such as defining optimization setpoints. Considering energy cost/loss minimization goals and constraints, such as maximum server inlet data center server temperature, the above module invokes the optimizer to generate a set of operating parameters within a specific range, the simulator tests and executes the corresponding predictions, optimizes The processor uses these predictions to further refine and recommend optimal combinations. For example, if there are multiple precision air-conditioning units in the application scenario, the optimization module 200 can help determine an optimized on/off schedule while keeping the hardware configuration and layout temperature of the application scenario within a safe range. In other cases, the optimizer may recommend an increased or decreased setpoint to take into account specific temperature requirements or actual task loads. If the next next task is clear, such as the next task from the job scheduler system, the optimizer will help find the optimal task path and scheduling, taking into account both cost and time limit.
  • providers for running or external applications such as defining optimization setpoints. Considering energy cost/loss minimization goals and constraints, such as maximum server inlet data center server temperature
  • the anomaly detection device 300 compares the stored data of the physical sensor and the read data of the virtual sensor in the database 500 . Consistent deviations may indicate a sensor or IT device failure, in which case operations and maintenance should be noted to ensure the safety and smooth operation of the application's hardware configuration and layout.
  • the first aspect of the present invention provides a dynamic predictive control method for precision air conditioners, which includes the following steps:
  • First execute step S1 generate a 3D simulation model based on a simulation template based on the static information of the hardware configuration and layout of the application scene where the precision air conditioner is located, wherein the 3D simulation model and the static model of the hardware configuration and layout of the application scene interact with each other correspond.
  • the application scenario of the precision air conditioner is a data center
  • the target data center has multiple cabinets, and multiple servers are placed on each cabinet, and the data center also has a separate data center management system (DCIM) .
  • DCIM data center management system
  • CRAC three precision air conditioners
  • PUE power usage efficiency
  • the 3D simulation model of CITIC Data Center can be quickly generated.
  • the simulation model needs to correspond to the physical model, for example, the number of cabinets and air conditioners in the data center and the The location of the hardware should be in one-to-one correspondence with the 3D simulation model.
  • the three-dimensional simulation model is a three-dimensional simulation model based on computational fluid dynamics, which uses a computer to numerically simulate and calculate the mass, energy, and momentum transfer generated during the fluid flow process in three-dimensional space, and presents it in a cloud or line diagram Graphical presentation of calculation results such as temperature, pressure, velocity, etc.
  • the image presentation is optionally in the monitoring and prediction device 100 .
  • the three-dimensional simulation model is a geometric model, which in this embodiment includes a data center computer room model and layout, a cabinet model and layout, a precision air-conditioning equipment model, etc., and the above-mentioned models and layouts can be imported from third-party software.
  • the simulation and calculation of the 3D simulation model need to introduce boundary conditions, which include ambient temperature, surface material heat transfer coefficient, server heating power, inlet air temperature and speed, etc., to correspond to the working conditions to be predicted or simulated.
  • the solver may use general-purpose software, which has mature support in the prior art, and will not be repeated for the sake of brevity.
  • the static information is system configuration information.
  • the static information includes information such as the number, location, and orientation of cabinets in the data center, each server number and corresponding power curve, ventilation layout, perforated floor layout, and Opening degree, air conditioner serial number, power curve, maximum cooling capacity, etc.
  • step S2 the simulator 400 performs simulation based on the three-dimensional simulation model corresponding to the dynamic parameters of the precision air conditioner, and uses machine learning to establish a correspondence between the dynamic parameters and the simulation results to generate a proxy model.
  • step S3 is executed, based on the agent model and the real-time dynamic data of the precision air conditioner, the temperature distribution and change of the application scene are predicted.
  • the data input by the monitoring and forecasting device 100 includes running status A and job progress B collected at specific time intervals, wherein the running status A and job progress B are dynamic data.
  • the monitoring and predicting device 100 outputs dynamic data and real-time data to the simulator 400 .
  • the dynamic parameters of the simulator 400 include application scene dynamic parameters E and precision air conditioner dynamic parameters F, wherein the scene dynamic parameters E include space temperature, air conditioner position, space layout, thermal conductivity coefficient and other thermal insulation materials performance.
  • the spatial layout is the number of seats, air-conditioning locations, ventilation system locations, etc. in the data center.
  • the dynamic parameters F of the precision air conditioner include the key setting parameters and air volume of the air conditioner, the readings of physical sensors such as temperature and humidity, and the heat generation and load of the hardware.
  • the temperature sensor and the humidity sensor are exemplarily placed on the cabinet of the data center, and the heat generation and load of the hardware refer to the heat generation and load of the server in the data center.
  • the emulator 400 stores the collected data in the form of an XML file, and then executes the emulation based on the emulation strategy selected by the selector 420,
  • Empirical models or high-fidelity CFD tools can be applied as simulation engines to adapt to actual operational needs or different accuracy and speed of front-end applications, such as FloTherm.
  • the CFD tool is a simulation software for computational fluid dynamics.
  • the above parameters are input into the three-dimensional model in the form of parameter model M1, and the output is the three-dimensional temperature distribution space flow field in space.
  • the spatial flow field is the directional velocity of the air flow
  • the spatial streamline is the flow trajectory of the air.
  • the calculator 410 performs calculations based on the three-dimensional temperature distribution spatial flow field and the proxy model M2 output by the CFD tool, and outputs prediction information to the monitoring and prediction device 100 .
  • the monitoring and predicting device 100 sends the reading C of the virtual sensor and the alarm information D to the client.
  • the proxy model M2 is a fast response model regressed from a high-precision simulation model, for example, based on fluid mechanics modeling. After the simulation model has a certain amount of data, there will be a fast proxy model to respond to real-time changes in a timely manner.
  • the parameter model M1 is based on the rapid modeling of the above parameters based on the simulation template.
  • the dynamic predictive control method of the precision air conditioner further includes the following steps: the optimization device 200 conducts an optimization process for at least one control parameter of the precision air conditioner based on the three-dimensional simulation model and/or the proxy model. Iterative optimization, wherein the control parameters include the number of air conditioners turned on, air supply temperature and flow.
  • key control points are adjusted according to the prediction results, such as air supply temperature, air-conditioning inlet and outlet, air-conditioning start and stop, etc., and then according to the detection and prediction comparison, if there is a deviation between the predicted value and the actual value, an alarm will be issued if the predicted temperature is too high.
  • the optimization module 100 can first implement the degree and configuration of precision air conditioners, user goals and restrictions, such as the maximum power of air conditioners and the maximum flow of exhaust equipment in order to ensure the temperature range of data center cabinets, and then predict according to the load plan of the future data center, Perform simulation verification through adjustable parameters to determine whether the optimization requirements are met, and further adjust the energy consumption control strategy of precision air conditioners, etc. For example, in practical applications, the heat generated by a cabinet such as 90% load and 50% is different.
  • the dynamic predictive control method of the precision air conditioner also includes the following steps: comparing the space temperature demand of the application scenario with a first predetermined threshold according to the load plan of the hardware configuration of the application scenario and its power curve, such as If it is judged that the precision air conditioner is undercooled or overcooled, the operating parameters of the precision air conditioner are adjusted or the load plan of the application scenario is adjusted.
  • the selector 420 selects different simulation strategies, and when the agent model is stable, the temperature distribution and changes of the application scenario are performed based on the agent model and the real-time dynamic data of the precision air conditioner. predict. If the agent model is not stable, then compare the actual value of the physical sensor of the hardware configuration of the application scenario with the predicted value of the simulation model, and if the deviation between the actual value and the predicted value exceeds a second predetermined threshold, then The selector 420 selects different simulation strategies, adjusts static parameters and generates a parametric model based on changes in the static parameters and performs prediction using a computational fluid dynamics simulator.
  • the abscissa is the cabinet air inlet temperature of the data center
  • the vertical coordinate is the cabinet outlet air temperature of the data center.
  • Different histograms correspond to the cabinet heating power of different data centers.
  • Figure 3 illustrates the influence of the air supply temperature of different channels on the outlet air temperature of the cabinet under different load conditions.
  • the heating power of the cabinet corresponds to the load, and the heating power of the cabinet can have a positive correspondence with the load, that is, the greater the load The greater the heating power of the cabinet.
  • FIG. 3 For example, different histograms correspond to cabinet heating powers of 1000W, 2000W, 3000W, 4000W and 5000W respectively. From Figure 3, it can be roughly concluded that under a certain load, in order to control the outlet air temperature of the cabinet, the maximum range of the cabinet inlet air temperature can be set. As shown in Figure 3, the range of the outlet air temperature of the cabinet is also rising when the inlet air temperature of the cabinet is around 17 degrees, 22 degrees and 27 degrees. For example, in the case of a heating power of 1000W, different cabinet inlet air temperatures of 17°C, 22°C, and 27°C correspond to different outlet air temperatures of 17.5°C, 22.5°C, and 27°C, respectively. Therefore, if there is a threshold for the cabinet outlet air temperature, you can see the approximate range that the cabinet inlet air temperature can meet. The corresponding maximum cabinet inlet air temperature is the recommended channel supply air temperature.
  • the second aspect of the present invention provides a dynamic predictive control system for precision air-conditioning, including: a processor; and a memory coupled to the processor, the memory having instructions stored therein, and when the instructions are executed by the processor, the The dynamic predictive control system of the precision air conditioner performs an action, and the action includes: S1, generating a three-dimensional simulation model based on a simulation template based on the hardware configuration and layout of the application scene where the precision air conditioner is located, and the static information of the layout, wherein the three-dimensional The simulation model and the static model of the hardware configuration and layout of the application scene correspond to each other; S2, perform simulation based on the three-dimensional simulation model corresponding to the dynamic parameters of the precision air conditioner, and use machine learning to establish a correspondence between the dynamic parameters and the simulation results to generate a proxy model; S3, based on the proxy model and the real-time dynamic data of the precision air conditioner, predict the temperature distribution and change of the application scene.
  • the dynamic predictive control system of the precision air conditioner also includes the following steps: iteratively optimizing at least one control parameter of the precision air conditioner based on the three-dimensional simulation model and/or the proxy model, wherein the control parameter Including the number of air conditioners turned on, supply air temperature and flow.
  • the action includes comparing the space temperature requirement of the application scenario with a first predetermined threshold according to the load plan of the hardware configuration of the application scenario and its power curve, such as judging that the precision air conditioner is undercooled or overcooled Then adjust the operating parameters of the precision air conditioner or adjust the load plan of the application scenario.
  • the action includes: comparing the actual value of the physical sensor of the hardware configuration of the application scenario with the predicted value of the simulation model, and if the deviation between the actual value and the predicted value exceeds a second predetermined threshold, then Static parameters are adjusted and parametric models are generated based on changes in the static parameters and predictions are performed using a computational fluid dynamics simulator.
  • the three-dimensional simulation model is based on computational fluid dynamics, which numerically simulates and calculates mass, energy, and momentum transfer generated during fluid flow in three-dimensional space, and displays the The calculation results of the temperature, pressure and velocity of the fluid in the application scene space are presented graphically.
  • the third aspect of the present invention provides a dynamic predictive control device for a precision air conditioner, which includes: a generating device that generates a three-dimensional simulation model based on a simulation template based on the hardware configuration and layout of the application scene where the precision air conditioner is located and the static information of the layout, wherein, The three-dimensional simulation model and the static model of the hardware configuration and layout of the application scene correspond to each other; the simulation device performs simulation based on the three-dimensional simulation model corresponding to the dynamic parameters of the precision air conditioner, and uses machine learning to establish dynamic parameters and simulation The corresponding relationship of the results is used to generate a proxy model; the prediction device predicts the temperature distribution and change of the application scene based on the proxy model and the real-time dynamic data of the precision air conditioner.
  • a fourth aspect of the present invention provides a computer program product tangibly stored on a computer-readable medium and comprising computer-executable instructions which, when executed, cause at least one processor to perform the The method described in the first aspect of the present invention.
  • a fifth aspect of the present invention provides a computer-readable medium on which are stored computer-executable instructions that, when executed, cause at least one processor to perform the method according to any one of the first aspects of the present invention. .
  • the dynamic predictive control mechanism of the precision air conditioner fills the gap in the dynamic monitoring and optimization system between hardware and data center management software, such as energy management and task scheduling, so that the operator can more clearly understand and control the operation of the precision air conditioner to Meet the dynamic thermal management needs of data centers.
  • the optimization feature based on the physical simulator helps to reduce unnecessary energy loss on the power supply side, and does not affect the daily working conditions of IT equipment or require an excessive number of physical sensors.
  • the irregularity detection module provides early warning of potential system failures to avoid unnecessary downtime.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Combustion & Propulsion (AREA)
  • Mechanical Engineering (AREA)
  • Chemical & Material Sciences (AREA)
  • Signal Processing (AREA)
  • Theoretical Computer Science (AREA)
  • Mathematical Physics (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Fluid Mechanics (AREA)
  • Fuzzy Systems (AREA)
  • Computer Hardware Design (AREA)
  • Geometry (AREA)
  • Mathematical Analysis (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Pure & Applied Mathematics (AREA)
  • Mathematical Optimization (AREA)
  • Computing Systems (AREA)
  • Algebra (AREA)
  • Air Conditioning Control Device (AREA)

Abstract

本发明提供了精密空调的动态预测控制方法、装置和系统，其中，包括如下步骤：S1，根据所述精密空调所在应用场景的硬件配置和布局的静态信息基于仿真模板生成三维仿真模型；S2，基于所述三维仿真模型对应所述精密空调的动态参数执行仿真，并采用机器学习建立动态参数和仿真结果的对应关系以生成代理模型；S3，基于所述代理模型和所述精密空调的实时动态数据对所述应用场景的温度分布和变化进行预测。本发明提供的精密空调的动态预测控制机制填补了硬件和数据中心管理软件之间动态监控和优化系统的空白，例如能量管理和任务调度，使得操作员可以更清楚地理解和控制精密空调运行以满足数据中心的动态热管理需求。

Description

精密空调的动态预测控制方法、装置和系统 技术领域
本发明涉及精密空调领域，尤其涉及一种精密空调的动态预测控制方法、装置和系统。
背景技术
在数据中心中，热量和能量的管理在减少性能使用效率(PUE，performance usage effectiveness)方面和保护IT设备顺利运行方面至关重要。其中，IT设备包括服务器机架、刀片式服务器等。HAVC系统特别是计算机房空调设备(CRAC，Computer Room Air Conditioner)单元占用了除了IT设备以外的大部分能量损耗。由于冷却需求降低，数据中心内部平均1摄氏度的温度升高能导致4％～5％的能量节省。因此，动态分析热力工况对于改善数据中心以及其他智能基础设施的能量的能量性能来说很有潜力，并且能够降低运行成本。
如今，市场上大部分解决方案来自硬件端，例如，通过升级CRAR单元的升级，通风系统的重新设计，服务器机架的物理附件或者用其他介质来更换空气冷却系统。从软件方案的方面，一些数据中心管理系统提供了工作分配功能来调整服务器负载，以减少热能产生和节约能量。然而，这些方法缺少整体热量分布的配合，并仅仅依赖于经验计算。最近，利用CFD软件的仿真获取其动能来提供数据中心热能表现得细节，但其应用主要限制在设计阶段。例如，在任何实际变化发生之前，一些公司尝试利用CFD软件6SigmaDC为其自身数据中心测试和验证充当虚拟测试平台。一些公司也为其自己的软件提供了数据中心基础设施管理(DCIM，Data Center Infrastructure Management)开发工具包，其能为复杂能量和热量管理挑战带来额外能力。在这种情况下，一个热量和能量管理得动态预测和优化工具/系统能帮助在硬件和软件方案之间提供桥梁。
发明内容
本发明第一方面提供了精密空调的动态预测控制方法，其中，包括如下步骤：S1，根据所述精密空调所在应用场景的硬件配置和布局和布局的静态信息基于仿真模板生成三维仿真模型，其中，所述三维仿真模型和所述应用场景的硬件配置和布局的静态模型相互对应；S2，基于所述三维仿真模型对应所述精密空调的动态参数执行仿真，并采用机器学习建立动态参数和仿真结果的对应关系以生成代理模型；S3，基于所述代理模型和所述精密空调的实时动态数据对所述应用场景的温度分布和变化进行预测。
进一步地，所述精密空调的动态预测控制方法还包括如下步骤：基于所述三维仿真模型和/或所述代理模型针对至少一个所述精密空调的控制参数进行迭代优化，其中，所述控制参数包括空调开启数量、送风温度和流量。
进一步地，所述精密空调的动态预测控制方法还包括如下步骤：根据所述应用场景的硬件配置的负载计划及其功率曲线对该应用场景的空间温度需求和一个第一预定阈值进行比较，如判断所述精密空调制冷不足或者制冷过度则调整所述精密空调的运行参数或者调整所述应用场景的负载计划。
进一步地，所述精密空调的动态预测控制方法还包括如下步骤：将所述应用场景的硬件配置的物理传感器的实际值和所述仿真模型的预测值进行对比，如果所述实际值和预测值的偏差超过一个第二预定阈值则调整静态参数并基于所述静态参数的变化生成参数模型并利用计算流体力学的仿真器执行预测。
进一步地，所述三维仿真模型是基于计算流体力学的，其对三维空间内的流体流动过程中产生的质量、能量、动量传递进行数值模拟和计算，并以云图或线图的方式将所述应用场景空间中流体的温度、压力、速度的计算结果进行图像化呈现。
本发明第二方面提供了精密空调的动态预测控制系统，包括：处理器；以及与所述处理器耦合的存储器，所述存储器具有存储于其中的指令，所述指令在被处理器执行时使所述精密空调的动态预测控制系统执行动作，所述动作包括：S1，根据所述精密空调所在应用场景的硬件配置和布局和布局的静态信息基于仿真模板生成三维仿真模型，其中，所述三维仿真模型和所述应用场景的硬件配置和布局的静态模型相互对应；S2，基于所述三维仿真模型对应所述精密空调的动态参数执行仿真，并采用机器学习建立动态参数和仿真结果的对应关系以生成代理模型；S3，基于所述代理模型和所述精密空 调的实时动态数据对所述应用场景的温度分布和变化进行预测。
进一步地，所述精密空调的动态预测控制系统还包括如下步骤：基于所述三维仿真模型和/或所述代理模型针对至少一个所述精密空调的控制参数进行迭代优化，其中，所述控制参数包括空调开启数量、送风温度和流量。
进一步地，所述动作包括根据所述应用场景的硬件配置的负载计划及其功率曲线对该应用场景的空间温度需求和一个第一预定阈值进行比较，如判断所述精密空调制冷不足或者制冷过度则调整所述精密空调的运行参数或者调整所述应用场景的负载计划。
进一步地，所述动作包括：将所述应用场景的硬件配置的物理传感器的实际值和所述仿真模型的预测值进行对比，如果所述实际值和预测值的偏差超过一个第二预定阈值则调整静态参数并基于所述静态参数的变化生成参数模型并利用计算流体力学的仿真器执行预测。
进一步地，所述三维仿真模型是基于计算流体力学的，其对三维空间内的流体流动过程中产生的质量、能量、动量传递进行数值模拟和计算，并以云图或线图的方式将所述应用场景空间中流体的温度、压力、速度的计算结果进行图像化呈现。
本发明第三方面提供了精密空调的动态预测控制装置，其中，包括：生成装置，根据所述精密空调所在应用场景的硬件配置和布局和布局的静态信息基于仿真模板生成三维仿真模型，其中，所述三维仿真模型和所述应用场景的硬件配置和布局的静态模型相互对应；仿真装置，基于所述三维仿真模型对应所述精密空调的动态参数执行仿真，并采用机器学习建立动态参数和仿真结果的对应关系以生成代理模型；预测装置，基于所述代理模型和所述精密空调的实时动态数据对所述应用场景的温度分布和变化进行预测。
本发明第四方面提供了计算机程序产品，所述计算机程序产品被有形地存储在计算机可读介质上并且包括计算机可执行指令，所述计算机可执行指令在被执行时使至少一个处理器执行根据本发明第一方面所述的方法。
本发明第五方面提供了计算机可读介质，其上存储有计算机可执行指令，所述计算机可执行指令在被执行时使至少一个处理器执行根据本发明第一方面任一项所述的方法。
本发明提供的精密空调的动态预测控制机制填补了硬件和数据中心管理软件之间动态监控和优化系统的空白，例如能量管理和任务调度，使得操作 员可以更清楚地理解和控制精密空调运行的热量表现。其中，基于物理仿真器得优化特征有助于减少供电侧的非必要能量损耗，并且不会影响到IT设备的日常工作条件或者要求过多数量的物理传感器。此外，不规则探测模块能够提供潜在系统故障的早期预警，以避免不必要的停机时间。
附图说明
图1是根据本发明一个具体实施例的精密空调的动态预测控制系统的结构示意图；
图2是根据本发明一个具体实施例的精密空调的动态预测控制系统的仿真器的原理示意图；
图3是根据本发明一个具体实施例的精密空调的动态预测控制系统的仿真器的不同负载情况下不同数据中心机柜送风温度对机柜出风温度的影响。
具体实施方式
以下结合附图，对本发明的具体实施方式进行说明。
本发明提供了精密空调的动态预测控制机制，本发明基于数字双胞胎提供了精密空调应用场景的热管理的动态预测和优化系统，其利用几个数据源来对应用场景的硬件配置和布局的热力性能建模。如图1所示，精密空调CRAC的动态预测控制系统包括监控预测装置100、优化装置200、异常探测装置300、仿真器400、数据库500。其中，仿真器400能被上述监控预测装置100、优化装置200、异常探测装置300的其中任一个调用并为进一步分析返回预测性能数据。南向接口600连接有现场数据和其他外部系统，并且同时提供当前操作的状态和推荐以进行调整或者执行外部应用。其中，示例性地，所述现场数据包括来自精密空调应用场景的传感器和DDC控制，所述其他外部系统包括例如应用场景的管理系统，其他外部应用包括作业调度程序。此外，精密空调的动态预测控制系统还包括外部接口P 2、传感器S、数据中心管理工具IT和软件接口P 1。
其中，数据库500保存了监控预测装置100、优化装置200、异常探测装置300所需的系统配置、运行状态和参数以及预定预测或者特别分析的输出。其中，系统配置包括数据中心服务器和通风系统布局参数、IT设备ID和功率容量、精密空调单元ID和功率容量、能量计费等。运行状态包括传感器度数、 单个服务器负载百分比、精密空调单元的主要定位点等。输出数据包括虚拟温度分布和预测能量需求等。
其中，监控预测装置100执行一个特定间隔时间从数据库500来的数据提取，例如15到30分钟。应当考虑每两组数据以执行一个预测，动态数据描述了精密空调目标应用场景的集合信息，即运行状态的动态数据。仿真器400以XML文件的形式来存储收集到的数据。仿真引擎利用任一CFD模型，例如Flotherm模型或者代理模型(surrogate model)会考虑到应用场景关键位置目前的运行状态和温度信息预测温度分布，其会用虚拟传感器展示，促成运行器来确定潜在热点(hot spots)和评估现有运行设置。仿真策略的选择是基于经验代理模型的准确度和可用性的。当连接于一个作业调度系统时，上述模块也能为即将到来的事件预测，以及如果当前运行设置不能够满足预期温度需求时对需求端的操作发出早期预警。
此外图形用户界面GUI是用户和本发明提供的精密空调的动态预测控制的用户界面。
优化装置200提供运行或者外部应用从供应端来的推荐，例如定义优化设置点。考虑到能量成本/损耗的最小化目标和约束，例如最大服务器入口数据中心服务器温度，上述模块会调用优化器来在一个特定范围内产生一系列操作参数，仿真器会测试和执行相应预测，优化器利用这些预测进一步提炼和推荐优化组合。例如，如果应用场景中具有多个精密空调单元，则优化模块200能有助于决定优化开启/关断计划表并同时保持应用场景的硬件配置和布局的温度于一个安全范围。在另外的情况下，考虑到特定温度需求或者实际任务负载，优化器会推荐一个增加或者减少的设置点。如果接下来的下一个任务是明确的，例如从作业调度器系统来的下一个任务，优化器会帮助寻找优化任务路径和调度，并且同时考虑到成本和时限。
异常探测装置300会比较数据库500中的物理传感器的存入数据和虚拟传感器的读取数据。一致的偏差会指示一个传感器或者IT设备的故障，在这种情况下运作和维持应该被注意到以确保应用场景硬件配置和布局的安全性和顺利运行。
本发明第一方面提供了精密空调的动态预测控制方法，其中，包括如下步骤：
首先执行步骤S1，根据所述精密空调所在应用场景的硬件配置和布局的 静态信息基于仿真模板生成三维仿真模型，其中，所述三维仿真模型和所述应用场景的硬件配置和布局的静态模型相互对应。
在本实施例中，假设所述精密空调的应用场景为数据中心，目标数据中心有多台机柜，每台机柜上放置有多个服务器，数据中信还具有有单独的数据中心管理系统(DCIM)。为保障数据中心服务器的正常运行，配备精密空调(CRAC)三个，其采用穿孔地板送风的形式进行空间降温，送风温度为15-22度，如果依靠完全物理温度传感器的反馈执行精密空调的控制，则会存在机房温控的及时性和精确性较低，电源使用效率(PUE)和空调能耗过高等问题。因此，根据数据库500中的静态数据，结合软件预设参数化仿真模板，快速生成改数据中信的三维仿真模型，该仿真模型需和物理模型相互对应，例如，数据中心的机柜、空调数量以及上述硬件位置都应当和三维仿真模型中一一对应。
优选地，所述三维仿真模型是基于计算流体力学的三维仿真模型，其采用计算机对三维空间内的流体流动过程中产生的质量、能量、动量传递进行数值模拟和计算，并以云图或线图的方式将如温度、压力、速度等计算结果进行图像化呈现。其中，所述图像化呈现可选地是在监控预测装置100中。
具体地，所述三维仿真模型是几何模型，其在本实施例中包括数据中心机房模型及布局，机柜模型及布局，精密空调设备模型等，上述模型以及布局可以从第三方软件导入。三维仿真模型的模拟和计算需要引入边界条件，所述边界条件包括环境温度、表面材料传热系数、服务器发热功率、进风温度速度等，以对应要预测或仿真的工况。其中，所述求解器可以采用通用软件，其在现有技术中有成熟的支持，为简明起见不再赘述。
其中，所述静态信息是系统配置信息，例如，在本实施例中所述静态信息包括数据中心的机柜数量、位置、朝向等信息，各服务器编号及对应功率曲线，通风布局、穿孔地板布局及开度、空调设备编号，功率曲线，最大制冷量等。
然后执行步骤S2，仿真器400基于所述三维仿真模型对应所述精密空调的动态参数执行仿真，并采用机器学习建立动态参数和仿真结果的对应关系以生成代理模型。
接着执行步骤S3，基于所述代理模型和所述精密空调的实时动态数据对所述应用场景的温度分布和变化进行预测。
具体地，基于建立的三维仿真模型，引入不同动态参数对应的不同的精密空调及其应用场景的工况执行多次仿真，仿真结果包括温度、压力、速度等存储在数据库中，然后采用机器学习的方式建立动态参数和仿真结果的对应关系，生成代理模型(surrogate model)。其中，仿真结果就是虚拟传感器的值。代理模型是基于查表或者对应关系。
具体地，如图2所示，监控预测装置100输入的数据包括每隔特定时间采集的运行状态A和作业进度B，其中，运行状态A和作业进度B都是动态数据。监控预测装置100输出动态数据和实时数据给仿真器400。具体地，在本实施例中，仿真器400的动态参数包括应用场景动态参数E和精密空调动态参数F，其中，场景动态参数E包括空间温度、空调位置、空间布局，热传导系数等隔热材料的性能。在本实施例中，空间布局是数据中心的座位、空调位置、通风系统位置数量等。精密空调动态参数F包括空调的关键设定参数和风量，温度、湿度等物理传感器读书，硬件的发热量和负载。其中，在本实施例中温度传感器和湿度传感器示例性地放在数据中心的机柜上，硬件的发热量和负载是指数据中心的服务器发热量和负载。
仿真器400以XML文件的形式来存储收集到的上述数据，然后基于选择器420所选择的仿真策略执行仿真，
经验模型(empirical models)或者高保真CFD工具(high-fidelity CFD tools)都能够被应用为仿真引擎来适应实际操作需求或者前端应用的不同的准确和速度，例如FloTherm。CFD工具是计算流体力学的仿真软件，将上述参数以参数模型M1的方式输入三维模型，输出是空间的三维温度分布空间流场。其中，空间流场是空气流动的方向速度，空间流线是空气的流动轨迹。然后计算器410基于CFD工具输出的空间的三维温度分布空间流场以及代理模型M2执行计算，并输出预测信息给监控预测装置100。然后监控预测装置100将虚拟传感器的读数C和报警信息D发送给客户。
其中，代理模型M2是根据高精度仿真模型回归出来一个快速响应模型，例如基于流体力学建模。仿真模型有一定数据量以后，会有一个快速代理模型，对实时变化进行及时响应。参数模型M1基于仿真模板根据上述参数进行的快速建模。
进一步地，如图1所示，所述精密空调的动态预测控制方法还包括如下步骤：优化装置200基于所述三维仿真模型和/或所述代理模型针对至少一个 所述精密空调的控制参数进行迭代优化，其中，所述控制参数包括空调开启数量、送风温度和流量。根据本发明一个优选实施例，根据预测结果调整关键的控制点，例如送风温度，空调出入口，空调启停等，然后根据检测和预测对比，如果预测值和实际值有偏离则报警预测温度过高，提示在特定时间内有温度过高建议操作。优化模块100可以先执行精密空调的度数和配置，用户目标和限制，比如为了保证数据中心机柜的温度范围空调设备的最大功率和排风设备最大流量，然后根据未来数据中心的负载计划进行预测，通过可以调的参数执行仿真验证，以确定是否满足优化需求，进一步调整精密空调能耗控制的策略等。比如，在实际应用中，机柜例如90％负载和50％的发热量是不一样的。
进一步地，所述精密空调的动态预测控制方法还包括如下步骤：根据所述应用场景的硬件配置的负载计划及其功率曲线对该应用场景的空间温度需求和一个第一预定阈值进行比较，如判断所述精密空调制冷不足或者制冷过度则调整所述精密空调的运行参数或者调整所述应用场景的负载计划。
进一步地，如图2所示，选择器420选择不同的仿真策略，当代理模型稳定时，则基于所述代理模型和所述精密空调的实时动态数据对所述应用场景的温度分布和变化进行预测。如果代理模型并不稳定，则将所述应用场景的硬件配置的物理传感器的实际值和所述仿真模型的预测值进行对比，如果所述实际值和预测值的偏差超过一个第二预定阈值则选择器420选择不同的仿真策略，调整静态参数并基于所述静态参数的变化生成参数模型并利用计算流体力学的仿真器执行预测。
如图3所示，横坐标是数据中心的机柜进风温度，竖坐标是数据中心的机柜出风温度，不同的柱状图对应着不同的数据中心的机柜发热功率。图3说明了在不同负载情况下不同通道送风温度对机柜出风温度的影响，其中，机柜的发热功率对应着负载不同，其中机柜的发热功率可以和负载有正对应关系，即负载越大机柜的发热功率越大。数据中心的机柜进风温度和机柜送风温度也是正相关关系，即机柜进风温度越高机柜送风温度也越大。其中，示例性地，不同柱状图对应机柜发热功率分别为1000W、2000W、3000W、4000W和5000W。图3可以大概得出在一定的负载下为了控制机柜出风温度，能够设定机柜进风温度的最大范围。如图3所示，在机柜进风温度在17度、22度和27度左右机柜的出风温度的范围也在上升。比如，对比发热功率 1000W的情况下，不同的机柜进风温度17度、22度和27度对应不同的出风温度大概分别在17.5度、22.5度和27度。因此如果有机柜出风温度的阈值，就可以看到大概机柜进风温度可以满足的大概范围，相应的最大的机柜进风温度就是推荐的通道送风温度。
本发明第二方面提供了精密空调的动态预测控制系统，包括：处理器；以及与所述处理器耦合的存储器，所述存储器具有存储于其中的指令，所述指令在被处理器执行时使所述精密空调的动态预测控制系统执行动作，所述动作包括：S1，根据所述精密空调所在应用场景的硬件配置和布局和布局的静态信息基于仿真模板生成三维仿真模型，其中，所述三维仿真模型和所述应用场景的硬件配置和布局的静态模型相互对应；S2，基于所述三维仿真模型对应所述精密空调的动态参数执行仿真，并采用机器学习建立动态参数和仿真结果的对应关系以生成代理模型；S3，基于所述代理模型和所述精密空调的实时动态数据对所述应用场景的温度分布和变化进行预测。
进一步地，所述精密空调的动态预测控制系统还包括如下步骤：基于所述三维仿真模型和/或所述代理模型针对至少一个所述精密空调的控制参数进行迭代优化，其中，所述控制参数包括空调开启数量、送风温度和流量。
进一步地，所述动作包括根据所述应用场景的硬件配置的负载计划及其功率曲线对该应用场景的空间温度需求和一个第一预定阈值进行比较，如判断所述精密空调制冷不足或者制冷过度则调整所述精密空调的运行参数或者调整所述应用场景的负载计划。
进一步地，所述动作包括：将所述应用场景的硬件配置的物理传感器的实际值和所述仿真模型的预测值进行对比，如果所述实际值和预测值的偏差超过一个第二预定阈值则调整静态参数并基于所述静态参数的变化生成参数模型并利用计算流体力学的仿真器执行预测。
进一步地，所述三维仿真模型是基于计算流体力学的，其对三维空间内的流体流动过程中产生的质量、能量、动量传递进行数值模拟和计算，并以云图或线图的方式将所述应用场景空间中流体的温度、压力、速度的计算结果进行图像化呈现。
本发明第三方面提供了精密空调的动态预测控制装置，其中，包括：生成装置，根据所述精密空调所在应用场景的硬件配置和布局和布局的静态信息基于仿真模板生成三维仿真模型，其中，所述三维仿真模型和所述应用场 景的硬件配置和布局的静态模型相互对应；仿真装置，基于所述三维仿真模型对应所述精密空调的动态参数执行仿真，并采用机器学习建立动态参数和仿真结果的对应关系以生成代理模型；预测装置，基于所述代理模型和所述精密空调的实时动态数据对所述应用场景的温度分布和变化进行预测。
本发明第四方面提供了计算机程序产品，所述计算机程序产品被有形地存储在计算机可读介质上并且包括计算机可执行指令，所述计算机可执行指令在被执行时使至少一个处理器执行根据本发明第一方面所述的方法。
本发明第五方面提供了计算机可读介质，其上存储有计算机可执行指令，所述计算机可执行指令在被执行时使至少一个处理器执行根据本发明第一方面任一项所述的方法。
本发明提供的精密空调的动态预测控制机制填补了硬件和数据中心管理软件之间动态监控和优化系统的空白，例如能量管理和任务调度，使得操作员可以更清楚地理解和控制精密空调运行以满足数据中心的动态热管理需求。其中，基于物理仿真器得优化特征有助于减少供电侧的非必要能量损耗，并且不会影响到IT设备的日常工作条件或者要求过多数量的物理传感器。此外，不规则探测模块能够提供潜在系统故障的早期预警，以避免不必要的停机时间。
尽管本发明的内容已经通过上述优选实施例作了详细介绍，但应当认识到上述的描述不应被认为是对本发明的限制。在本领域技术人员阅读了上述内容后，对于本发明的多种修改和替代都将是显而易见的。因此，本发明的保护范围应由所附的权利要求来限定。此外，不应将权利要求中的任何附图标记视为限制所涉及的权利要求；“包括”一词不排除其它权利要求或说明书中未列出的装置或步骤；“第一”、“第二”等词语仅用来表示名称，而并不表示任何特定的顺序。

Claims (13)

  1. 精密空调的动态预测控制方法，其中，包括如下步骤：
    S1，根据所述精密空调所在应用场景的硬件配置和布局的静态信息基于仿真模板生成三维仿真模型，其中，所述三维仿真模型和所述应用场景的硬件配置和布局的静态模型相互对应；
    S2，基于所述三维仿真模型对应所述精密空调的动态参数执行仿真，并采用机器学习建立动态参数和仿真结果的对应关系以生成代理模型；
    S3，基于所述代理模型和所述精密空调的实时动态数据对所述应用场景的温度分布和变化进行预测。
  2. 根据权利要求1所述的精密空调的动态预测控制方法，其特征在于，所述精密空调的动态预测控制方法还包括如下步骤：
    基于所述三维仿真模型和/或所述代理模型针对至少一个所述精密空调的控制参数进行迭代优化，其中，所述控制参数包括空调开启数量、送风温度和流量。
  3. 根据权利要求1所述的精密空调的动态预测控制方法，其特征在于，所述精密空调的动态预测控制方法还包括如下步骤：根据所述应用场景的硬件配置的负载计划及其功率曲线对该应用场景的空间温度需求和一个第一预定阈值进行比较，如判断所述精密空调制冷不足或者制冷过度则调整所述精密空调的运行参数或者调整所述应用场景的负载计划。
  4. 根据权利要求1所述的精密空调的动态预测控制方法，其特征在于，所述精密空调的动态预测控制方法还包括如下步骤：
    将所述应用场景的硬件配置的物理传感器的实际值和所述仿真模型的预测值进行对比，如果所述实际值和预测值的偏差超过一个第二预定阈值则调整静态参数并基于所述静态参数的变化生成参数模型并利用计算流体力学的仿真器执行预测。
  5. 根据权利要求1所述的精密空调的动态预测控制方法，其特征在于，所述三维仿真模型是基于计算流体力学的，其对三维空间内的流体流动过程中产生的质量、能量、动量传递进行数值模拟和计算，并以云图或线图的方式将所述应用场景空间中流体的温度、压力、速度的计算结果进行图像化呈现。
  6. 精密空调的动态预测控制系统，包括：
    处理器；以及
    与所述处理器耦合的存储器，所述存储器具有存储于其中的指令，所述指令在被处理器执行时使所述精密空调的动态预测控制系统执行动作，所述动作包括：
    S1，根据所述精密空调所在应用场景的硬件配置和布局和布局的静态信息基于仿真模板生成三维仿真模型，其中，所述三维仿真模型和所述应用场景的硬件配置和布局的静态模型相互对应；
    S2，基于所述三维仿真模型对应所述精密空调的动态参数执行仿真，并采用机器学习建立动态参数和仿真结果的对应关系以生成代理模型；
    S3，基于所述代理模型和所述精密空调的实时动态数据对所述应用场景的温度分布和变化进行预测。
  7. 根据权利要求1所述的精密空调的动态预测控制系统，其特征在于，所述精密空调的动态预测控制系统还包括如下步骤：
    基于所述三维仿真模型和/或所述代理模型针对至少一个所述精密空调的控制参数进行迭代优化，其中，所述控制参数包括空调开启数量、送风温度和流量。
  8. 根据权利要求1所述的精密空调的动态预测控制系统，其特征在于，所述动作包括根据所述应用场景的硬件配置的负载计划及其功率曲线对该应用场景的空间温度需求和一个第一预定阈值进行比较，如判断所述精密空调制冷不足或者制冷过度则调整所述精密空调的运行参数或者调整所述应用场景的负载计划。
  9. 根据权利要求1所述的精密空调的动态预测控制系统，其特征在于，所述动作包括：
    将所述应用场景的硬件配置的物理传感器的实际值和所述仿真模型的预测值进行对比，如果所述实际值和预测值的偏差超过一个第二预定阈值则调整静态参数并基于所述静态参数的变化生成参数模型并利用计算流体力学的仿真器执行预测。
  10. 根据权利要求1所述的精密空调的动态预测控制系统，其特征在于，所述三维仿真模型是基于计算流体力学的，其对三维空间内的流体流动过程中产生的质量、能量、动量传递进行数值模拟和计算，并以云图或线图的方 式将所述应用场景空间中流体的温度、压力、速度的计算结果进行图像化呈现。
  11. 精密空调的动态预测控制装置，其中，包括：
    生成装置，根据所述精密空调所在应用场景的硬件配置和布局和布局的静态信息基于仿真模板生成三维仿真模型，其中，所述三维仿真模型和所述应用场景的硬件配置和布局的静态模型相互对应；
    仿真装置，基于所述三维仿真模型对应所述精密空调的动态参数执行仿真，并采用机器学习建立动态参数和仿真结果的对应关系以生成代理模型；
    预测装置，基于所述代理模型和所述精密空调的实时动态数据对所述应用场景的温度分布和变化进行预测。
  12. 计算机程序产品，所述计算机程序产品被有形地存储在计算机可读介质上并且包括计算机可执行指令，所述计算机可执行指令在被执行时使至少一个处理器执行根据权利要求1至5中任一项所述的方法。
  13. 计算机可读介质，其上存储有计算机可执行指令，所述计算机可执行指令在被执行时使至少一个处理器执行根据权利要求1至5中任一项所述的方法。
PCT/CN2021/111276 2021-08-06 2021-08-06 精密空调的动态预测控制方法、装置和系统 Ceased WO2023010556A1 (zh)

Priority Applications (4)

Application Number Priority Date Filing Date Title
US18/681,136 US20240280286A1 (en) 2021-08-06 2021-08-06 Dynamic Prediction Control Method, Apparatus and System for Precision Air Conditioner
CN202180098830.0A CN117413147A (zh) 2021-08-06 2021-08-06 精密空调的动态预测控制方法、装置和系统
PCT/CN2021/111276 WO2023010556A1 (zh) 2021-08-06 2021-08-06 精密空调的动态预测控制方法、装置和系统
EP21952430.3A EP4375585A4 (en) 2021-08-06 2021-08-06 Method, apparatus and system for dynamic predictive control of precision air conditioning systems

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2021/111276 WO2023010556A1 (zh) 2021-08-06 2021-08-06 精密空调的动态预测控制方法、装置和系统

Publications (1)

Publication Number Publication Date
WO2023010556A1 true WO2023010556A1 (zh) 2023-02-09

Family

ID=85153999

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/111276 Ceased WO2023010556A1 (zh) 2021-08-06 2021-08-06 精密空调的动态预测控制方法、装置和系统

Country Status (4)

Country Link
US (1) US20240280286A1 (zh)
EP (1) EP4375585A4 (zh)
CN (1) CN117413147A (zh)
WO (1) WO2023010556A1 (zh)

Cited By (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116382377A (zh) * 2023-05-12 2023-07-04 安徽中科新源半导体科技有限公司 一种用于基站机柜温控的多联机控制方法及系统
CN116989432A (zh) * 2023-09-27 2023-11-03 深圳市美兆环境股份有限公司 一种空调系统节能运行优化调度方法及系统
CN117154301A (zh) * 2023-10-27 2023-12-01 内蒙古中电储能技术有限公司 一种用于储能电池组的风冷散热控制方法及系统
CN117171914A (zh) * 2023-09-05 2023-12-05 三河市皓智精密机械制造有限公司 一种高精密主轴性能测试方法及系统
CN117241561A (zh) * 2023-10-13 2023-12-15 北京科技大学 一种数据中心分布式精密空调机组优化运行参数估计方法
CN117412555A (zh) * 2023-10-13 2024-01-16 晋永源科技(深圳)有限公司 一种数据中心机房的空调控温系统
CN117725636A (zh) * 2023-12-22 2024-03-19 西安交通大学 一种数据中心的内部物理场预测方法、系统及设备
CN118274417A (zh) * 2024-05-31 2024-07-02 南通宝雪冷冻设备有限公司 用于工业制冷的全局性调控方法及装置
CN118499977A (zh) * 2024-07-17 2024-08-16 国鲸科技(广东横琴粤澳深度合作区)有限公司 一种温度快速响应的电子制冷系统
CN118765100A (zh) * 2024-09-05 2024-10-11 国网浙江省电力有限公司杭州供电公司 一种液冷系统的自适应控制方法、装置、设备和存储介质
CN119026352A (zh) * 2024-08-19 2024-11-26 常州刘国钧高等职业技术学校 物联网计算机机房用监控系统及方法
CN119272633A (zh) * 2024-12-06 2025-01-07 广州智业节能科技有限公司 暖通空调建模仿真方法及系统
CN119720847A (zh) * 2024-12-09 2025-03-28 昆明理工大学 一种基于深度学习的侧吹炉窑数智化预测系统
CN120910974A (zh) * 2025-10-10 2025-11-07 中国石油集团东方地球物理勘探有限责任公司 数据中心机房参数调节方法、装置、存储介质和电子设备
CN121346356A (zh) * 2025-10-23 2026-01-16 北京观天执行科技股份有限公司 基于智能优化的多能耦合温度场控制方法及系统

Families Citing this family (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN118940624B (zh) * 2024-07-22 2026-03-17 中国联合网络通信集团有限公司 模型训练方法、装置及可读存储介质
CN118917249A (zh) * 2024-08-16 2024-11-08 国网上海市电力公司 一种数据中心机房制冷设备运行优化方法及相关装置
CN118795792B (zh) * 2024-09-13 2025-01-24 杭州瑞德设计股份有限公司 用于家电的远程控制方法
CN119573212A (zh) * 2024-12-19 2025-03-07 深圳市英维克科技股份有限公司 一种空调的控制方法及电子设备
CN119743945B (zh) * 2025-03-05 2025-06-06 杭州华宏通信设备有限公司 应用于机房水冷空调的温度调控系统
CN119847310A (zh) * 2025-03-19 2025-04-18 国家工业信息安全发展研究中心 一种数据中心节能系统及方法
CN120430034B (zh) * 2025-04-22 2025-12-23 哈尔滨工业大学 一种实现真空环境微区惰性气体保护氛围动态调控的方法及系统
CN120197457B (zh) * 2025-05-27 2026-04-24 北京华科软科技有限公司 一种输电线路二三维联动排位设计方法
CN120733491B (zh) * 2025-07-14 2026-03-20 北京君发科技集团有限公司 一种采用氧化炉进行氧化处理的废气处理系统及方法
CN120507482B (zh) * 2025-07-18 2025-09-19 上海季丰电子股份有限公司 洁净空气污染检测方法、装置、终端设备及存储介质
CN121351362B (zh) * 2025-09-29 2026-04-10 陕西劢承机械设备有限公司 基于3d模型技术的集装袋自动设备虚拟调试系统及方法
CN121604368B (zh) * 2026-01-28 2026-04-17 恒华数字科技集团有限公司 一种基于ai的预制化数据中心制冷调控方法及系统

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109341010A (zh) * 2018-09-19 2019-02-15 新智能源系统控制有限责任公司 一种电制冷机空调系统用供能一体化的控制方法和装置
CN110726218A (zh) * 2019-10-29 2020-01-24 珠海格力电器股份有限公司 空调器及其控制方法、装置、存储介质和处理器
CN110736244A (zh) * 2019-10-29 2020-01-31 珠海格力电器股份有限公司 控制空调的方法、装置和空调
CN110736228A (zh) * 2019-10-29 2020-01-31 珠海格力电器股份有限公司 空调设备的控制方法、装置、存储介质及空调系统

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150134123A1 (en) * 2013-11-14 2015-05-14 Ces Group, Llc Predictive monitoring and control of an environment using cfd
US9644857B1 (en) * 2015-12-01 2017-05-09 Nasser Ashgriz Virtual thermostat for a zonal temperature control

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109341010A (zh) * 2018-09-19 2019-02-15 新智能源系统控制有限责任公司 一种电制冷机空调系统用供能一体化的控制方法和装置
CN110726218A (zh) * 2019-10-29 2020-01-24 珠海格力电器股份有限公司 空调器及其控制方法、装置、存储介质和处理器
CN110736244A (zh) * 2019-10-29 2020-01-31 珠海格力电器股份有限公司 控制空调的方法、装置和空调
CN110736228A (zh) * 2019-10-29 2020-01-31 珠海格力电器股份有限公司 空调设备的控制方法、装置、存储介质及空调系统

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP4375585A4 *

Cited By (21)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116382377B (zh) * 2023-05-12 2023-10-24 安徽中科新源半导体科技有限公司 一种用于基站机柜温控的多联机控制方法及系统
CN116382377A (zh) * 2023-05-12 2023-07-04 安徽中科新源半导体科技有限公司 一种用于基站机柜温控的多联机控制方法及系统
CN117171914B (zh) * 2023-09-05 2024-03-12 三河市皓智精密机械制造有限公司 一种高精密主轴性能测试方法及系统
CN117171914A (zh) * 2023-09-05 2023-12-05 三河市皓智精密机械制造有限公司 一种高精密主轴性能测试方法及系统
CN116989432B (zh) * 2023-09-27 2024-01-16 深圳市美兆环境股份有限公司 一种空调系统节能运行优化调度方法及系统
CN116989432A (zh) * 2023-09-27 2023-11-03 深圳市美兆环境股份有限公司 一种空调系统节能运行优化调度方法及系统
CN117241561A (zh) * 2023-10-13 2023-12-15 北京科技大学 一种数据中心分布式精密空调机组优化运行参数估计方法
CN117412555A (zh) * 2023-10-13 2024-01-16 晋永源科技(深圳)有限公司 一种数据中心机房的空调控温系统
CN117241561B (zh) * 2023-10-13 2024-02-09 北京科技大学 一种数据中心分布式精密空调机组优化运行参数估计方法
CN117154301B (zh) * 2023-10-27 2024-03-12 内蒙古中电储能技术有限公司 一种用于储能电池组的风冷散热控制方法及系统
CN117154301A (zh) * 2023-10-27 2023-12-01 内蒙古中电储能技术有限公司 一种用于储能电池组的风冷散热控制方法及系统
CN117725636A (zh) * 2023-12-22 2024-03-19 西安交通大学 一种数据中心的内部物理场预测方法、系统及设备
CN118274417A (zh) * 2024-05-31 2024-07-02 南通宝雪冷冻设备有限公司 用于工业制冷的全局性调控方法及装置
CN118499977A (zh) * 2024-07-17 2024-08-16 国鲸科技(广东横琴粤澳深度合作区)有限公司 一种温度快速响应的电子制冷系统
CN119026352A (zh) * 2024-08-19 2024-11-26 常州刘国钧高等职业技术学校 物联网计算机机房用监控系统及方法
CN118765100A (zh) * 2024-09-05 2024-10-11 国网浙江省电力有限公司杭州供电公司 一种液冷系统的自适应控制方法、装置、设备和存储介质
CN119272633A (zh) * 2024-12-06 2025-01-07 广州智业节能科技有限公司 暖通空调建模仿真方法及系统
CN119720847A (zh) * 2024-12-09 2025-03-28 昆明理工大学 一种基于深度学习的侧吹炉窑数智化预测系统
CN119720847B (zh) * 2024-12-09 2025-09-23 昆明理工大学 一种基于深度学习的侧吹炉窑数智化预测系统
CN120910974A (zh) * 2025-10-10 2025-11-07 中国石油集团东方地球物理勘探有限责任公司 数据中心机房参数调节方法、装置、存储介质和电子设备
CN121346356A (zh) * 2025-10-23 2026-01-16 北京观天执行科技股份有限公司 基于智能优化的多能耦合温度场控制方法及系统

Also Published As

Publication number Publication date
US20240280286A1 (en) 2024-08-22
CN117413147A (zh) 2024-01-16
EP4375585A1 (en) 2024-05-29
EP4375585A4 (en) 2025-04-30

Similar Documents

Publication Publication Date Title
WO2023010556A1 (zh) 精密空调的动态预测控制方法、装置和系统
CN115167562B (zh) 一种机房温度控制方法和装置
CN109800066B (zh) 一种数据中心节能调度方法及系统
US8155793B2 (en) System and method for controlling air conditioning facilities, and system and method for power management of computer room
JP5511698B2 (ja) 空調機連係制御システム、空調機連係制御方法および空調機連係制御プログラム
US10180261B1 (en) Model based cooling control system
Berezovskaya et al. Modular model of a data centre as a tool for improving its energy efficiency
US20210033097A1 (en) Method for optimizing the efficiency and/or the running performance of a fan or a fan arrangement
Simon et al. Artificial neural network based prediction of control strategies for multiple air-cooling units in a raised-floor data center
CA3005184C (en) System and method for measuring and evaluating building energy performance
CN112784343B (zh) 一种基于数字孪生模型的机房设计方法、装置及设备
CN119642339B (zh) 除湿器的节能控制方法、系统、介质及设备
Khalid et al. Thermal control strategies for reliable and energy-efficient data centers
WO2023033155A1 (ja) 空調制御システム
CN107563043A (zh) 一种室外机安装方案评估方法及装置
JP7824846B2 (ja) 空調制御システム
CN117374478A (zh) 一种风道出风口流量调节方法、装置及储能箱
JP6455937B2 (ja) シミュレーション装置、シミュレーション方法及びプログラム
Zhang et al. Real time thermal management controller for data center
Behl et al. Campus-wide integrated building energy simulation
Volk et al. CoolEmAll—Optimising cooling efficiency in data centres
CN113027799B (zh) 服务器风扇散热测试方法、装置、设备及可读存储介质
Lucchese et al. On server cooling policies for heat recovery: exhaust air properties of an Open Compute Windmill V2 platform
Parekh Comparative rack level cfd analysis of air to hybrid cooling data center
Lucchese et al. Coldspot: a thermal supervisor aimed at server rooms implementing a raised plenum cooling setup

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21952430

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 202180098830.0

Country of ref document: CN

WWE Wipo information: entry into national phase

Ref document number: 2021952430

Country of ref document: EP

WWE Wipo information: entry into national phase

Ref document number: 18681136

Country of ref document: US

ENP Entry into the national phase

Ref document number: 2021952430

Country of ref document: EP

Effective date: 20240131

NENP Non-entry into the national phase

Ref country code: DE