CN112616292B - Data center energy efficiency optimization control method based on neural network model - Google Patents

Data center energy efficiency optimization control method based on neural network model Download PDF

Info

Publication number
CN112616292B
CN112616292B CN202011360740.8A CN202011360740A CN112616292B CN 112616292 B CN112616292 B CN 112616292B CN 202011360740 A CN202011360740 A CN 202011360740A CN 112616292 B CN112616292 B CN 112616292B
Authority
CN
China
Prior art keywords
temperature
data
cabinet
data center
power
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.)
Active
Application number
CN202011360740.8A
Other languages
Chinese (zh)
Other versions
CN112616292A (en
Inventor
方遒
周佳康
王智宇
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.)
Hunan University
Original Assignee
Hunan University
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 Hunan University filed Critical Hunan University
Priority to CN202011360740.8A priority Critical patent/CN112616292B/en
Publication of CN112616292A publication Critical patent/CN112616292A/en
Application granted granted Critical
Publication of CN112616292B publication Critical patent/CN112616292B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H05—ELECTRIC TECHNIQUES NOT OTHERWISE PROVIDED FOR
    • H05K—PRINTED CIRCUITS; CASINGS OR CONSTRUCTIONAL DETAILS OF ELECTRIC APPARATUS; MANUFACTURE OF ASSEMBLAGES OF ELECTRICAL COMPONENTS
    • H05K7/00—Constructional details common to different types of electric apparatus
    • H05K7/20—Modifications to facilitate cooling, ventilating, or heating
    • H05K7/20709—Modifications to facilitate cooling, ventilating, or heating for server racks or cabinets; for data centers, e.g. 19-inch computer racks
    • H05K7/20836—Thermal management, e.g. server temperature control
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
    • G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
    • G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
    • G05B13/042—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators in which a parameter or coefficient is automatically adjusted to optimise the performance
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Software Systems (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • Data Mining & Analysis (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Mathematical Physics (AREA)
  • General Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computer Hardware Design (AREA)
  • Thermal Sciences (AREA)
  • Microelectronics & Electronic Packaging (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • Automation & Control Theory (AREA)
  • Air Conditioning Control Device (AREA)

Abstract

The invention discloses a data center energy efficiency optimization control method based on a neural network model, which comprises the following steps of: 1) carrying out data center key point temperature sensing and acquisition work; 2) storing the collected data center temperature data, and displaying the data in the database to a Web browser; 3) based on a large amount of data collected under different working conditions, learning the relation between airflow modes and power states of different machine rooms and the temperature distribution of key points of the machine rooms by using a neural network model, and establishing a nonlinear prediction model of the temperature of the key points of the data machine rooms; 4) on the basis of the nonlinear temperature prediction model, the influence of various factors such as power consumption, refrigeration performance and safety of equipment in a machine room is considered, and the energy consumption of a refrigeration system in the machine room is optimally regulated and controlled. Compared with the prior art, the method has the advantages of safe and reliable data transmission, suitability for the operation conditions of the data center in different airflow modes, capability of optimizing the energy efficiency in real time, high speed, high precision, simplicity in operation and the like.

Description

Data center energy efficiency optimization control method based on neural network model
Technical Field
The invention relates to an energy efficiency optimization method, in particular to a data center energy efficiency optimization control method based on a neural network model.
Background
With the explosive growth of digital information and the recent rapid development of cloud computing, a data center has become a core infrastructure of future IT technologies, and the use of big data services and cloud terminals is more and more common in daily life. The server belongs to high-precision equipment, and any equipment in the server has a problem, which causes great loss, such as interruption of bank service, loss of cloud storage data and the like. The data center machine room is out of order to cause the loss of user stored data, the communication machine room is out of order to cause communication interruption, and huge loss is caused to the security and stability of the whole enterprise and the country. In order to maintain a secure environment for the equipment, each data center is using its thermal control system, which accounts for more than 40% of the energy consumption of the entire data center. Therefore, the research on the thermal control system of the data center is all important for economic energy consumption and safe operation.
In order to solve the problems, more and more enterprises introduce a data center environment monitoring system to monitor and early warn various safety indexes in a machine room in real time, so that the pressure of workers in the machine room is relieved while the coordinated monitoring of data center environment equipment is ensured, and the safety management of the machine room is realized. Some enterprises and research departments use traditional temperature monitoring system who walks the line formula to the characteristics design of data center computer lab, the problem of partial artifical inspection timeliness and accuracy difference has been solved, but also brought new difficulty, if need walk the line in a large number, cause whole to walk the disorderly and maintain the difficulty of line planning, can not be suitable for in some place that has already wired the maturity, when wanting to change the collection point, need walk the line again, the cost is improved, and use a period of time line to maintain again after ageing must use a large amount of manpower and materials, the effect of overall solution problem is not good. As a developing technology, the wireless sensing network becomes a field with wide development prospect along with the technical progress in the aspects of wireless communication, sensors and microcomputers, and provides a new thought for solving the problems of the traditional wire-running temperature monitoring system. For a data center temperature prediction model, the previous research takes a steady-state flow field as an assumption premise, and provides a data center rapid temperature estimation model, which has the main defect that the assumption condition is too ideal, and when the flow field deviates from the designed state, the model accuracy is reduced or the model parameters need to be readjusted. In practice, airflow in a data center is affected by equipment such as cabinet fans and air conditioners, and airflow distribution is often not stable and constant, so that it is very necessary to establish a machine room temperature prediction model suitable for different airflow modes.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide a data center energy efficiency optimization control method based on a neural network model.
The purpose of the invention can be realized by the following technical scheme:
a data center energy efficiency optimization control method based on a neural network model comprises the following steps:
taking a terminal node of a data center as a key point of the data center, and carrying out temperature sensing and acquisition work of the key point of the data center through a key point temperature sensing and acquisition work system; the terminal nodes are the air inlet and outlet of the data center cabinet and the air conditioner;
acquiring temperature data of key points of the data center under different working conditions, storing the temperature data into a database to form a temperature database, and preprocessing and analyzing the temperature data in the database;
thirdly, learning the relation between different airflow modes, power states and the temperature distribution of key points of the data center by using a neural network model based on the acquired temperature database, and establishing a nonlinear temperature prediction model of the temperature of the key points of the data center machine room;
and step four, on the basis of establishing the nonlinear temperature prediction model, establishing an energy consumption model of the air conditioner and the fan refrigeration equipment, designing an energy consumption optimization problem, and solving the energy consumption optimization problem, thereby optimally regulating and controlling the energy consumption of the refrigeration system.
In a further improvement, in the step one, the temperature sensing and collecting work of the key points of the data center by the key point temperature sensing and collecting work system comprises the following steps:
11) the method comprises the steps that temperature sensing and collection of a data center are achieved through a single bus containing a temperature sensor with calibrated digital signal output, the temperature sensor is installed on a terminal node and guarantees normal operation of temperature collection, the temperature sensor is directly connected with a router, and the temperature sensor stores data and sends the data to the router after the router sends a command;
12) the method comprises the following steps that a single chip microcomputer chip carrying ZigBee communication is used as a core processor and a CPU of a router, the single chip microcomputer chip is used as a relay station, each cabinet is responsible for expansion of terminal nodes, data relay and management and command of the uppermost layer of the cabinet, meanwhile, capacity expansion is carried out in the signal coverage area of the router by self, and the router is communicated with a coordinator in a ZigBee wireless sensing mode; the temperature sensor, the router and the coordinator form a ZigBee network;
13) the coordinator initializes and maintains the network, selects the channel used by the network, manages the node, distributes the address, distributes and updates the security key, and sends data with the management center through the serial port;
14) the upper computer management center is a website carried on the server and receives data from the coordinator through a serial port;
15) the temperature sensor of the terminal node, the router, the coordinator and the upper computer management center are combined with a ZigBee wireless sensing network which is in mutual communication to form a complete key point temperature sensing and collecting work system, the terminal node is used as an executor of temperature sensing and collecting, the router is used as a relay, after the ZigBee network is built, the router transmits collected data to the coordinator, the coordinator completes regulation and control in a networking mode of the ZigBee network, and finally the coordinator sends the data to the management center through a serial port to conduct key point temperature sensing and collecting work of the data center.
In the second step, the temperature data preprocessing and the visualization of numerical analysis are completed in a Web browser, and the predicted value and the true value of the data are visually compared; the visualization comprises the following steps:
21) reading and storing serial port data into variables: the serial port communication uses asynchronous communication, calls a function to read temperature data, performs 10-system conversion on the received data because the serial port transmission is 16-system transmission, and stores the converted value into a variable;
22) kalman filtering is carried out on the variable, Gaussian noise is filtered out to reduce errors, and generation of error data is prevented;
23) storing the filtered variables into a document: obtaining the filtered variables, calling corresponding text documents in a target folder, and storing the values of the variables and the corresponding current time into the documents;
24) data visualization: establishing a webpage on a Web browser platform, calling data in a document to display, constructing a user interface of machine learning application, displaying a plurality of data center key point temperature data in the webpage, and visually observing the change trend of the data center key point temperature data through an image module, so as to facilitate the subsequent modeling analysis of a predicted value and a true value; the image module includes a line graph and a bar graph.
In a further improvement, in the third step, the nonlinear temperature prediction model includes the following steps:
31) input and output parameters of the nonlinear temperature prediction model are set as follows: the data center machine room is provided with N cabinets and M air conditioners, and the key point temperature of the data center machine room is the cabinet inlet temperature Tri，iCabinet outlet temperature Tro，iAnd inlet temperature T of the air conditionerci，jOutlet temperature T of air conditionerco，jAnd the temperature of each key point is determined by three factors, namely the air flow rate f around the cabinetiCabinet power Pr，iAnd air conditioner set temperature Tref，jWherein the cabinet peripheral air flow rate fiProportional to the rotation speed of the fan of the cabinet, and the power P of the cabinetr，iThe distributed workload is related, wherein i is more than or equal to 1 and less than or equal to N, j is more than or equal to 1 and less than or equal to M, so that the input of the temperature prediction model is the air flow rate of N cabinets, the power of N cabinets and the set temperature of M air conditioners, the number of the input is 2N + M, the output is the temperature of each key point of the machine room, the number of the output is 2N +2M, namely the nonlinear temperature prediction model is the input of 2N + M, and the output of 2N +2M
Figure GDA0003463635040000041
Model, wherein T is the key point temperature T of the data center machine roomri，i、Tro，i、Tci，j、Tco，jThe set of (a) and (b),
Figure GDA0003463635040000042
is a corresponding rule;
32) establishing a database: cabinet ambient air flow rate fiAnd air conditioner set temperature TrefRespectively is fmax、TrefmaxLower limit of 0, Trefmin(ii) a Power P of cabinetr，iAre each Prmin、PrmaxThe power of one cabinet, the ambient air flow rate and the set temperature of the air conditioner are changed within the range at each time, the settings of other cabinets are kept unchanged, and the air conditioner is collected at different fiAnd Pr，iAnd different Tref，jThe temperature data of the key point, the temperature T of the key point to be collected and the corresponding air flow rate fiCabinet power Pr，iAnd air conditioner set temperature Tref，jTogether form a database;
33) training data volume selection: the amount of data trained was (N + M) × 2000;
34) selecting a model training method: training a neural network to obtain a final nonlinear temperature prediction model; the neural network comprises a multilayer BP neural network and an ELM neural network; the neural network parameter selection method comprises an empirical formula method and a trial and error method.
In a further improvement, in the fourth step, the energy consumption optimization regulation and control of the refrigeration system comprises the following steps:
41) obtaining the power of a refrigerating system, wherein the power of the refrigerating system of the data center comprises the power of M air conditioners, the power of fans of N cabinets and the power P of the air conditionersc，jIs determined by the following formula:
Figure GDA0003463635040000043
where ρ is the air density, CpIs the specific heat capacity of air, fjThe air flow rate around the air conditioner is determined by the air conditioner, Tco，jFor air-conditioner outlet temperature, i.e. air-conditioner set temperature, Tci，jIs the air conditioner inlet temperature, COP (T)co，j) For the cooling efficiency of the air-conditioning node, from Tco，jIt is decided that the formula is as follows,
COP(Tco，j)＝0.0068·Tco，j 2+0.0008·Tco，j+0.458
cabinet fan power Pf，iIs determined by the following formula:
Pf，i＝a0fi 3+a2fi 2+a3fi
Pf，irelated only to the flow rate of air around the cabinet, a0、a2、a3All are constants, are obtained by data fitting, the temperature units are all ℃, the power units are all W, and the power of the refrigeration system of the data center is finally defined as the power P of the air conditionerc，jAnd the power P of the cabinet fanf，iSum of total power P of the refrigerating systemcoolingThe calculation formula of (a) is as follows:
Figure GDA0003463635040000051
42) establishing energy consumption optimization problem, and obtaining total power P of the refrigeration system according to the decision formula of each powercoolingAnd the peripheral air flow rate f of the cabinetiAnd the temperature T of the inlet and the outlet of the air conditionerci，j、Tco，jRelated, non-linear temperature prediction model is combined, and the temperature T of the inlet and the outlet of the air conditionerci，j、Tco，jBy the cabinet ambient air velocity fiCabinet power Pr，iAnd air conditioner set temperature Tref，jPredicted total power P of the refrigerating systemcoolingBy optimizing the cabinet ambient air flow rate fiAnd air conditioner set temperature Tref，jTo reduce, the energy consumption optimization problem is defined as:
the known data center machine room has N cabinets and M air conditioners, and the power P of each cabinet is P ═ Pr，1，Pr，2，…，Pr，N]THas determined that the cabinet ambient air flow rate F ═ F is found in the given constraints1，f2，…，fN]TAnd air conditioner set temperature Tref＝[Tref，1，Tref，2，…，Tref，M]TTo minimize the total power P of the refrigeration systemcooling；
43) Solving an energy consumption optimization problem: solving by using a particle swarm optimization algorithm, wherein the motion of the particles is described by positions and speeds, and the positions represent the air flow rate F around the cabinet and the set temperature T of the air conditionerrefThe position and the speed are initialized randomly, the constraint in the energy consumption optimization problem defines the boundary of the position, the position of the particle changes in each step, the particle moves to the best position along with the continuous update of the speed and the position, and when the iteration times reach the specified limit or the current particle swarm meets the predefined convergence condition, the optimization stops to obtain the best position result;
44) the optimal solution F and T obtained by solvingrefAnd recording, correspondingly adjusting the rotating speed of the cabinet fan and the set temperature of the air conditioner, so that the total power of the refrigerating system can be reduced to the minimum, and finally, the optimal regulation and control of the energy consumption of the refrigerating equipment are realized.
In a further refinement, the constraints in the energy consumption optimization problem in step 42) are:
51) the safe operation conditions of the cabinet are that the inlet temperature is not more than 27 ℃, the outlet temperature is not more than 35 ℃, namely Tri,i≤27℃、Tro,i≤35℃；
52) Cabinet ambient air flow rate fiThe constraint range of (1) is not less than 0 and not more than fi≤fmax；
53) Air conditioner set temperature Tref,jHas a constraint range of Trefmin≤Tref,j≤Trefmax。
The method for the energy efficiency optimization control of the data center mainly comprises the steps of key point temperature sensing and acquisition, data storage processing and visualization, temperature prediction model establishment, optimization control problem solving and the like. The method comprises a whole set of complete flow from key point temperature data acquisition to realization of optimal regulation, can be directly applied to real-time energy consumption optimal regulation of a data center, and the finally obtained optimal regulation scheme can be compared with the traditional control method to obtain the superiority of the optimal regulation scheme.
Compared with the prior art, the invention has the following advantages:
1) the invention not only has the data center energy efficiency optimization control method of the neural network temperature prediction model, but also comprises complete system design, and has simple operation and clear flow.
2) The transmission mode of the key point temperature data is a ZigBee wireless sensor network, the network contains an encryption algorithm, the safety of data transmission is ensured, and the authenticity and the reliability of the data can be ensured by the transmission mode adopted by the network.
3) The temperature prediction model can be suitable for the operation condition of the data center under different airflow modes, and has wider practicability compared with other technologies.
4) Once the neural network temperature prediction model is trained, the inlet and outlet air temperatures of each node of the data center can be predicted at a high speed, and compared with other estimation methods, the result accuracy of the prediction model is high.
5) The data in the database are displayed in the Web browser, the Web browser can be logged in at any time to check the temperature data, and real-time energy efficiency optimization is easy to perform.
Drawings
FIG. 1 is a process flow diagram of the present invention;
FIG. 2 is a schematic diagram of a network architecture of a temperature sensing and acquisition system;
FIG. 3 is a key point temperature sensing and acquisition implementation diagram;
FIG. 4 is a block diagram of a data center temperature monitoring visualization system;
FIG. 5 is a diagram of a data center equipment layout according to the present invention;
FIG. 6 is a schematic diagram of an energy consumption optimization problem;
FIG. 7 is a particle swarm algorithm for solving the optimization problem.
Detailed Description
The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.
A data center energy efficiency optimization control method based on a neural network model comprises the following steps:
1) the method comprises the following steps that an integrated temperature sensor and a single chip microcomputer carrying a ZigBee wireless sensing network are used for realizing key point temperature sensing and acquisition work of a data center;
11) the single bus is used for realizing temperature sensing and acquisition of the data center by using a temperature sensor with calibrated digital signal output, the temperature sensor is arranged on a terminal node to ensure normal operation of temperature acquisition, and the temperature sensor is directly connected with the router, stores data and sends the data to the router after the router sends a command to the temperature sensor;
12) the router is characterized in that a singlechip chip carrying a ZigBee communication part is used as a core processor, the router is a singlechip which is used as a CPU and mainly used as a relay station, each cabinet is responsible for expansion of terminal nodes, data relay and management and command of the uppermost layer of the cabinet, the router can be ensured to be automatically expanded in the coverage area of routing signals, and the router is communicated with the coordinator in a ZigBee wireless sensing mode;
13) the highest manager of the whole network is a coordinator, which basically has the same composition with the router, can be directly used as the router and can directly communicate with the terminal node, the coordinator mainly initializes and maintains the network, selects a channel used by the network, manages the node, allocates an address, distributes and updates a security key, and the coordinator sends data with a management center through a serial port;
14) the management center is a website loaded on the server and can receive data from the coordinator through a serial port;
15) the terminal nodes, the router, the coordinator and the upper computer management center are integrated into a complete working system, the ZigBee wireless sensing network is used for mutual communication of the terminal nodes, the router serves as an executor of temperature sensing and collection, after the ZigBee network is constructed, the router transmits collected data to the coordinator, the coordinator completes regulation and control in a networking mode of the wireless sensing network, the coordinator finally sends the data to the management center through a serial port, and the key point temperature sensing and collection work is completed.
2) The method comprises the steps of collecting and storing a large amount of temperature data of key points of a data center under different working conditions, preprocessing and numerically analyzing the data in a database, achieving visualization in a Web browser, and visually comparing predicted values and actual values of the data;
21) reading and storing serial port data into variables: because the data center has more equipment, temperature information at multiple positions of the data center needs to be sent to the main server, serial port communication uses asynchronous communication, the requirement on clock beat of the asynchronous communication is not high, the data center is suitable for more occasions, a function is called to read temperature data, because the serial port transmission is 16-system transmission, 10-system conversion needs to be carried out on the received data, and the converted value is stored in a variable;
22) kalman filtering the variable: because the transmission speed is high, interference signals are likely to be generated, a Kalman filtering algorithm is adopted to filter data one by one, Gaussian noise is filtered to reduce errors, and error data is prevented from being generated;
23) storing the filtered variables into a document: when the filtered variables are obtained, calling corresponding text documents in a target folder, and storing the values of the variables and the corresponding current time into the documents;
24) data visualization: the data visualization means comprises a mobile phone APP, a Web browser display and the like, webpage establishment can be carried out on a Web browser platform, data in a document are called to be displayed, a user interface of machine learning application is quickly established, temperature data of a plurality of machine cabinet key points are displayed in the webpage, the change trend of the temperature data is more visually observed through modules such as a line graph and the like, and subsequent modeling analysis of a predicted value and a real value is facilitated.
3) Based on an acquired temperature database, learning the relation between different airflow modes and power states and the temperature distribution of the machine room key points by using a neural network model (BP, ELM and the like), and establishing a nonlinear temperature prediction model of the temperature of the machine room key points of the data center;
31) setting input and output parameters of a temperature prediction model: the data center machine room is provided with N cabinets and M air conditioners, and the key point temperature of the data center machine room is the cabinet inlet and outlet temperature Tri,i、Tro,iAnd the inlet and outlet temperature T of the air conditionerci,j、Tco,jAnd the temperature of each key point is determined by three factors, namely the air flow rate f around the cabinetiCabinet power Pr，iAnd air conditioner set temperature Tref，jWherein the cabinet peripheral air flow rate fiProportional to the rotation speed of the fan of the cabinet, and the power P of the cabinetr，iThe distributed workload is related, wherein i is more than or equal to 1 and less than or equal to N, j is more than or equal to 1 and less than or equal to M, the input of the temperature prediction model is the air flow rate of N cabinets, the power of N cabinets and the set temperature of M air conditioners, the number of the input is 2N + M, the output is the temperature of each key point of the machine room, the number of the output is 2N +2M, namely, the temperature prediction model is the input of 2N + M, and the output of 2N +2M
Figure GDA0003463635040000091
Model, wherein T is the key point temperature T of the data center machine roomri，i、Tro，i、Tci，j、Tco，jThe set of (a) and (b),
Figure GDA0003463635040000092
is a corresponding rule;
32) establishing a database: cabinet ambient air flow rate fiAnd air conditioner set temperature TrefCan be manually regulated and controlled and has a certain regulation and control range, and the upper limits are respectively fmax、TrefmaxLower limit of 0, TrefminPower of cabinet Pr，iAre each Prmin、PrmaxThe power of one cabinet, the ambient air flow rate and the set temperature of the air conditioner are changed within the range at each time, the settings of other cabinets are kept unchanged, and the air conditioner is collected at different fiAnd Pr，iAnd different Tref，jThe temperature data of the key point, the temperature T of the key point to be collected and the corresponding air flow rate fiCabinet power Pr，iAnd air conditioner set temperature Tref，jTogether form a database;
33) training data volume selection: for the conventional data center machine room at present, the number of cabinets and air conditioners is large, that is, the input and output parameters of a temperature prediction model are large, the data volume for training must also meet the requirement of the temperature prediction model training, the trained result can be satisfied, the training result is not that the more the training data is, the better the training data is, the training method is also related to the selection of the training method, the data volume can be obtained by experience, taking the data center machine room with 4 cabinets and 1 air conditioner as an example, the data volume for training is 10000 groups, and the increased data volume for training of N, M should be increased along with corresponding multiples;
34) selecting a model training method: the neural network training can be adopted, for example, the multilayer BP neural network and the ELM and the like can be adopted, different types of neural networks have different parameters which need to be set, different parameter settings can influence the training result, an empirical formula or a trial and error method can be used for setting the optimal parameters, the applicability of the empirical formula is wider, the precision of the trial and error method is high, the multilayer BP neural network can obtain the optimal number of neurons by the trial and error method only aiming at one input and output condition, and finally the training method is used for obtaining the temperature prediction model.
4) On the basis of establishing a nonlinear temperature prediction model, considering the influence of various factors such as power consumption, performance and safety of equipment in a machine room, establishing an energy consumption model of air-conditioning and fan refrigeration equipment and designing an energy consumption optimization problem, realizing a rapid solution algorithm of the optimization problem, and optimally regulating and controlling the energy consumption of the refrigeration equipment.
41) Obtaining the power of key refrigerating equipment, wherein the power of a refrigerating system of the data center comprises the power of M air conditioners, the power of fans of N cabinets and the power P of the air conditionersc，jIs determined by the following formula:
Figure GDA0003463635040000101
where ρ is the air density, CpIs the specific heat capacity of air, fjFor the ambient air flow rate of the air conditioner, the parameter is determined by the air conditioner itself, Tco，jFor air-conditioner outlet temperature, i.e. air-conditioner set temperature, Tci，jIs the air conditioner inlet temperature, COP (T)co，j) For the cooling efficiency of the air-conditioning node, from Tco，jIt is decided that the formula is as follows,
COP(Tco，j)＝0.0068·Tco，j 2+0.0008·Tco，j+0.458
cabinet fan power Pf，iIs determined by the following formula:
Pf，i＝a0fi 3+a2fi 2+a3fi
the power is only related to the air flow rate around the cabinet, a0、a2、a3All are constants, are obtained by data fitting, the temperature units are all ℃, the power units are all W, and the power of the refrigeration system of the data center is finally defined as the power P of the air conditionerc，jAnd the power P of the cabinet fanf，iSum of total power P of the refrigerating systemcoolingThe calculation formula of (a) is as follows:
Figure GDA0003463635040000102
42) establishing energy consumption optimization problem, and obtaining total power P of the refrigeration system according to the decision formula of each powercoolingAnd the peripheral air flow rate f of the cabinetiAnd the temperature T of the inlet and the outlet of the air conditionerci，j、Tco，jRelated, non-linear temperature prediction model is combined, and air conditioner inlet temperature Tci，j、Tco，jThe flow velocity f can be controlled by the surrounding air flow velocity of the cabinetiCabinet power Pr，iAnd air conditioner set temperature Tref，jIt is predicted that the allocation of work tasks will change the cabinet power allocation, in most practical cases the cabinet power Pr，iThe distribution being regarded as uncontrolled, the total power P of the refrigeration systemcoolingBy optimizing the cabinet ambient air flow rate f onlyiAnd air conditioner set temperature Tref，jTo reduce, the energy consumption optimization problem can be defined as:
the known data center machine room has N cabinets and M air conditioners, and the power P of each cabinet is P ═ Pr，1，Pr，2，…，Pr，N]THas determined that the cabinet ambient air flow rate F ═ F is found in the given constraints1，f2，…，fN]TAnd air conditioner set temperature Tref＝[Tref，1，Tref，2，…，Tref，M]TTo minimize the total power P of the refrigeration systemcooling；
43) Solving an energy consumption optimization problem: considering the complexity and non-linearity of the energy consumption optimization problem, the solution can be made using a Particle Swarm Optimization (PSO) algorithm, in which the motion of the particles is described by position and velocity, for which the position represents the cabinet ambient air flow rate F and the air conditioner set temperature TrefThe position and the speed are initialized randomly, the constraint in the energy consumption optimization problem defines the boundary of the position, the position of the particle is changed by the speed of the particle in each step, the particle moves to the best position along with the continuous update of the speed and the position, and when the iteration times reach the specified limit or the current particle swarm meets the predefined convergence condition, the optimization stops to obtain the best position result; the constraint conditions are as follows:
51) according to the American Society of Mechanical Engineers (ASME) temperature standard, the conditions for safe operation of the cabinet are that the inlet temperature is not more than 27 ℃ and the outlet temperature is not more than 35 ℃, i.e., Tri,i≤27℃、Tro,i≤35℃；
52) Cabinet ambient air flow rate fiThe constraint range of (1) is not less than 0 and not more than fi≤fmax；
53) Air conditioner set temperature Tref，jHas a constraint range of Trefmin≤Tref,j≤Trefmax。
44) The optimal solution F and T obtained by solvingrefAnd recording, correspondingly adjusting the rotating speed of the cabinet fan and the set temperature of the air conditioner, so that the total power of the refrigerating system can be reduced to the minimum, and finally, the optimal regulation and control of the energy consumption of the refrigerating equipment are realized.
As shown in fig. 1, the method comprises the following steps:
step 1: the temperature sensing and collecting work, the structural schematic diagram of the temperature sensing and collecting system is shown in fig. 2:
11) determining four parts of a terminal node, a router, a coordinator and an upper computer management center, installing a temperature sensor with a single bus containing calibrated digital signal output on a key node of a cabinet as the terminal node to sense and collect temperature, assembling a singlechip chip carrying a ZigBee communication part and other components into the router and the coordinator, and using a Web browser at a PC end as the upper computer management center;
12) determining a communication mode among the components, directly connecting the terminal nodes with a router, communicating the router with a coordinator through wireless communication formed after networking is successful, connecting the coordinator with a management center through a USB (universal serial bus), and obtaining temperature data by the management center through a USB transmission serial port;
13) according to the specific scheme of temperature sensing and acquisition implemented in fig. 3, the temperature sensing and acquisition work is completed.
Step 2: storing, processing and visualizing the data center key point temperature data acquired in the step 1, wherein the main steps are as shown in fig. 4:
21) reading and storing serial port data into a variable, wherein the serial port data transmission is 16-system transmission, 10-system conversion needs to be carried out on received data, and a converted numerical value is stored into the variable;
22) because the transmission speed is high, interference signals are likely to be generated, a Kalman filtering algorithm is adopted to filter variables, Gaussian noise is filtered to reduce errors, and error data is prevented from being generated;
23) when the filtered variables are obtained, calling corresponding text documents in a target folder, and storing the values of the variables and the corresponding current time into the documents;
24) and displaying the temperature data of the key points of the cabinet on a Web browser, and more intuitively observing the change trend of the cabinet through modules such as a line graph and the like, so that the subsequent modeling analysis of a predicted value and a real value is facilitated.
And step 3: establishing a key point temperature prediction model of the data center, taking fig. 5 as an example, wherein the data center is provided with 4 cabinets and 1 air conditioner, the temperature estimation model is a 9-input 10-output model, and training data is selected and trained by adopting a neural network to obtain the temperature prediction model.
And 4, step 4: the method comprises the following main steps of optimizing and regulating energy consumption of a data center:
31) obtaining the power P of the air conditionerc,jPower P of cabinet fanf,iAnd total power P of the refrigeration systemcooling；
32) Referring to FIG. 6, a schematic diagram of an energy consumption optimization problem is shown, in which cabinet fans are adjusted to change the flow rate f of air around the cabinetiAdjusting the set temperature T of the air conditionerref,jSo that the total refrigerating power P of the data center can be changedcooling，；
33) Finding cabinet ambient air velocity f using Particle Swarm Optimization (PSO) algorithm of FIG. 7iAnd air conditioner set temperature Tref,jTo minimize the total power P of the refrigeration systemcoolingThe optimal solution f obtained by solvingiAnd Tref,jAnd recording, correspondingly adjusting the rotating speed of the cabinet fan and the set temperature of the air conditioner, so that the total power of the refrigerating system can be reduced to the minimum, and finally, the optimal regulation and control of the energy consumption of the refrigerating equipment are realized.

Claims (3)

1. A data center energy efficiency optimization control method based on a neural network model is characterized by comprising the following steps:
taking a terminal node of a data center as a key point of the data center, and carrying out temperature sensing and acquisition work of the key point of the data center through a key point temperature sensing and acquisition work system; the terminal nodes are the air inlet and outlet of the data center cabinet and the air conditioner;
acquiring temperature data of key points of the data center under different working conditions, storing the temperature data into a database to form a temperature database, and preprocessing and analyzing the temperature data in the database;
thirdly, learning the relation between different airflow modes, power states and the temperature distribution of key points of the data center by using a neural network model based on the acquired temperature database, and establishing a nonlinear temperature prediction model of the temperature of the key points of the data center machine room; the method for establishing the nonlinear temperature prediction model comprises the following steps:
31) input and output parameters of the nonlinear temperature prediction model are set as follows: the data center machine room is provided with N cabinets and M air conditioners, and the key point temperature of the data center machine roomTemperature is the cabinet inlet temperature Tri，iCabinet outlet temperature Tro，iAnd inlet temperature T of the air conditionerci，jOutlet temperature T of air conditionerco，jAnd the temperature of each key point is determined by three factors, namely the air flow rate f around the cabinetiCabinet power Pr，iAnd air conditioner set temperature Tref，jWherein the cabinet peripheral air flow rate fiProportional to the rotation speed of the fan of the cabinet, and the power P of the cabinetr，iThe distributed workload is related, wherein i is more than or equal to 1 and less than or equal to N, j is more than or equal to 1 and less than or equal to M, the input of the temperature prediction model is the air flow rate of N cabinets, the power of N cabinets and the set temperature of M air conditioners, the number of the input is 2N + M, the output is the temperature of each key point of the machine room, the number of the output is 2N +2M, namely the nonlinear temperature prediction model is the input of 2N + M and the output of 2N +2M
Figure FDA0003463635030000011
Model, wherein T is the key point temperature T of the data center machine roomri，i、Tro，i、Tci，j、Tco，jThe set of (a) and (b),
Figure FDA0003463635030000012
is a corresponding rule;
32) establishing a database: cabinet ambient air flow rate fiAnd air conditioner set temperature TrefRespectively is fmax、TrefmaxLower limit of 0, Trefmin(ii) a Power P of cabinetr，iAre each Prmin、PrmaxThe power of one cabinet, the ambient air flow rate and the set temperature of the air conditioner are changed within the range at each time, the settings of other cabinets are kept unchanged, and the air conditioner is collected at different fiAnd Pr，iAnd different Tref，jThe temperature data of the key point, the temperature T of the key point to be collected and the corresponding air flow rate fiCabinet power Pr，iAnd air conditioner set temperature Tref，jTogether form a database;
33) training data volume selection: the amount of data trained was (N + M) × 2000;
34) selecting a model training method: training a neural network to obtain a final nonlinear temperature prediction model; the neural network comprises a multilayer BP neural network and an ELM neural network; the neural network parameter selection method comprises an empirical formula method and a trial and error method;
on the basis of establishing a nonlinear temperature prediction model, establishing an energy consumption model of air conditioner and fan refrigeration equipment, designing an energy consumption optimization problem, and solving the energy consumption optimization problem so as to optimally regulate and control the energy consumption of a refrigeration system;
the method for optimizing and regulating the energy consumption of the refrigerating system comprises the following steps:
41) obtaining the power of a refrigerating system, wherein the power of the refrigerating system of the data center comprises the power of M air conditioners, the power of fans of N cabinets and the power P of the air conditionersc，jIs determined by the following formula:
Figure FDA0003463635030000021
where ρ is the air density, CpIs the specific heat capacity of air, fjThe air flow rate around the air conditioner is determined by the air conditioner, Tco，jFor the outlet temperature, T, of the air conditionerci，jIs the air conditioner inlet temperature, COP (T)co，j) For the cooling efficiency of the air-conditioning node, from Tco，jIt is decided that the formula is as follows,
COP(Tco，j)＝0.0068·Tco，j 2+0.0008·Tco，j+0.458
cabinet fan power Pf，iIs determined by the following formula:
Pf，i＝a0fi 3+a2fi 2+a3fi
Pf，ionly the air flow rate f around the cabinetiIn connection with, a0、a2、a3All are constants, are obtained by data fitting, the temperature units are all ℃, the power units are all W, and the power of the refrigeration system of the data center is finally defined as the power P of the air conditionerc，jAnd the power P of the cabinet fanf，iSum of total power P of the refrigerating systemcoolingThe calculation formula of (a) is as follows:
Figure FDA0003463635030000022
42) establishing energy consumption optimization problem, and obtaining total power P of the refrigeration system according to the decision formula of each powercoolingAnd the peripheral air flow rate f of the cabinetiAnd the temperature T of the inlet and the outlet of the air conditionerci，j、Tco，jRelated, non-linear temperature prediction model is combined, and the temperature T of the inlet and the outlet of the air conditionerci，j、Tco，jBy the cabinet ambient air velocity fiCabinet power Pr，iAnd air conditioner set temperature Tref，jPredicted total power P of the refrigerating systemcoolingBy optimizing the cabinet ambient air flow rate fiAnd air conditioner set temperature Tref，jTo reduce, the energy consumption optimization problem is defined as:
the known data center machine room has N cabinets and M air conditioners, and the power P of each cabinet is P ═ Pr，1，Pr，2，…，Pr，N]THas determined that the cabinet ambient air flow rate F ═ F is found in the given constraints1，f2，…，fN]TAnd air conditioner set temperature Tref＝[Tref，1，Tref，2，…，Tref，M]TTo minimize the total power P of the refrigeration systemcooling；
The given constraints are:
51) the safe operation conditions of the cabinet are that the inlet temperature is not more than 27 ℃, the outlet temperature is not more than 35 ℃, namely Tri，i≤27℃、Tro，i≤35℃；
52) Cabinet ambient air flow rate fiThe constraint range of (1) is not less than 0 and not more than fi≤fmax；
53) Air conditioner set temperature Tref，jHas a constraint range of Trefmin≤Tref，j≤Trefmax
43) High energy consumptionSolving a problem: solving by using a particle swarm optimization algorithm, wherein the motion of the particles is described by positions and speeds, and the positions represent the air flow rate F around the cabinet and the set temperature T of the air conditionerrefThe position and the speed are initialized randomly, the constraint in the energy consumption optimization problem defines the boundary of the position, the position of the particle is changed by the speed of the particle in each step, the particle moves to the best position along with the continuous update of the speed and the position, and when the iteration times reach the specified limit or the current particle swarm meets the predefined convergence condition, the optimization stops to obtain the best position result;
44) the optimal solution F and T obtained by solvingrefAnd recording, correspondingly adjusting the rotating speed of the cabinet fan and the set temperature of the air conditioner, so that the total power of the refrigerating system can be reduced to the minimum, and finally, the optimal regulation and control of the energy consumption of the refrigerating equipment are realized.
2. The energy efficiency optimization control method of the data center based on the neural network model, according to claim 1, wherein in the first step, the temperature sensing and collecting work of the key points of the data center through the key point temperature sensing and collecting work system comprises the following steps:
11) the method comprises the steps that temperature sensing and collection of a data center are achieved through a single bus containing a temperature sensor with calibrated digital signal output, the temperature sensor is installed on a terminal node and guarantees normal operation of temperature collection, the temperature sensor is directly connected with a router, and the temperature sensor stores data and sends the data to the router after the router sends a command;
12) the method comprises the following steps that a single chip microcomputer chip carrying ZigBee communication is used as a core processor and a CPU of a router, the single chip microcomputer chip is used as a relay station, each cabinet is responsible for expansion of terminal nodes, data relay and management and command of the uppermost layer of the cabinet, meanwhile, capacity expansion is carried out in the signal coverage area of the router by self, and the router is communicated with a coordinator in a ZigBee wireless sensing mode; the temperature sensor, the router and the coordinator form a ZigBee network;
13) the coordinator initializes and maintains the network, selects the channel used by the network, manages the nodes, allocates the addresses, distributes and updates the security keys, and sends data through the serial port and the upper computer management center;
14) the upper computer management center is a website carried on the server and receives data from the coordinator through a serial port;
15) the temperature sensor of the terminal node, the router, the coordinator and the upper computer management center are combined with a ZigBee wireless sensing network which is in mutual communication to form a complete key point temperature sensing and collecting work system, the terminal node is used as an executor of temperature sensing and collecting, the router is used as a relay, after the ZigBee network is built, the router transmits collected data to the coordinator, the coordinator completes regulation and control in a networking mode of the ZigBee network, and finally the coordinator sends the data to the management center through a serial port to conduct key point temperature sensing and collecting work of the data center.
3. The energy efficiency optimization control method for the data center based on the neural network model is characterized in that in the second step, the temperature data preprocessing and the numerical analysis are visualized in a Web browser, and the predicted value and the actual value of the data are visually compared; the visualization comprises the following steps:
21) reading and storing serial port data into variables: the serial port communication uses asynchronous communication, calls a function to read temperature data, performs 10-system conversion on the received data because the serial port transmission is 16-system transmission, and stores the converted value into a variable;
22) kalman filtering is carried out on the variable, Gaussian noise is filtered out to reduce errors, and generation of error data is prevented;
23) storing the filtered variables into a document: obtaining the filtered variables, calling corresponding text documents in a target folder, and storing the values of the variables and the corresponding current time into the documents;
24) data visualization: establishing a webpage on a Web browser platform, calling data in a document to display, constructing a user interface of machine learning application, displaying a plurality of data center key point temperature data in the webpage, and visually observing the change trend of the data center key point temperature data through an image module, so as to facilitate the subsequent modeling analysis of a predicted value and a true value; the image module includes a line graph and a bar graph.
CN202011360740.8A 2020-11-27 2020-11-27 Data center energy efficiency optimization control method based on neural network model Active CN112616292B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011360740.8A CN112616292B (en) 2020-11-27 2020-11-27 Data center energy efficiency optimization control method based on neural network model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011360740.8A CN112616292B (en) 2020-11-27 2020-11-27 Data center energy efficiency optimization control method based on neural network model

Publications (2)

Publication Number Publication Date
CN112616292A CN112616292A (en) 2021-04-06
CN112616292B true CN112616292B (en) 2022-04-22

Family

ID=75228028

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011360740.8A Active CN112616292B (en) 2020-11-27 2020-11-27 Data center energy efficiency optimization control method based on neural network model

Country Status (1)

Country Link
CN (1) CN112616292B (en)

Families Citing this family (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113254876B (en) * 2021-04-28 2024-05-14 浙江智马达智能科技有限公司 Air conditioner energy consumption data determining method, device, equipment and storage medium
CN113225994B (en) * 2021-05-08 2022-06-24 广州远正智能科技股份有限公司 Intelligent air conditioner control method facing data center
CN114002952A (en) * 2021-09-29 2022-02-01 苏州浪潮智能科技有限公司 Cluster centralized heat dissipation regulation and control system and method
CN114234384B (en) * 2021-12-27 2023-08-29 中铁第四勘察设计院集团有限公司 Air conditioning optimization control method and system for railway passenger station
CN114935892B (en) * 2022-06-10 2023-10-03 杭州电子科技大学 Air flow rate self-adaptive control modeling method for air water generator
CN116193819B (en) * 2023-01-19 2024-02-02 中国长江三峡集团有限公司 Energy-saving control method, system and device for data center machine room and electronic equipment
CN115776813B (en) * 2023-02-14 2023-04-28 南京金宁能源科技有限公司 Efficient machine room control method and device
CN116840605B (en) * 2023-08-28 2023-11-17 北京英沣特能源技术有限公司 Efficient refrigeration machine room energy efficiency prediction method
CN117412555B (en) * 2023-10-13 2024-06-18 三通(常州)电子科技有限公司 Air conditioner temperature control system of data center computer lab

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102620378A (en) * 2011-01-27 2012-08-01 国际商业机器公司 Method and system for data center energy saving controlling
CN109890176A (en) * 2019-03-01 2019-06-14 北京慧辰资道资讯股份有限公司 A kind of method and device based on artificial intelligence optimization's energy consumption of machine room efficiency
CN110805997A (en) * 2019-11-14 2020-02-18 中金新源（天津）科技有限公司 Energy-saving control method for central air-conditioning system
CN111144543A (en) * 2019-12-30 2020-05-12 中国移动通信集团内蒙古有限公司 Data center air conditioner tail end temperature control method, device and medium
CN111814388A (en) * 2020-06-12 2020-10-23 菲尼克斯(上海)环境控制技术有限公司 CFD simulation verification method for lower air supply data center based on neural network

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102620378A (en) * 2011-01-27 2012-08-01 国际商业机器公司 Method and system for data center energy saving controlling
CN109890176A (en) * 2019-03-01 2019-06-14 北京慧辰资道资讯股份有限公司 A kind of method and device based on artificial intelligence optimization's energy consumption of machine room efficiency
CN110805997A (en) * 2019-11-14 2020-02-18 中金新源（天津）科技有限公司 Energy-saving control method for central air-conditioning system
CN111144543A (en) * 2019-12-30 2020-05-12 中国移动通信集团内蒙古有限公司 Data center air conditioner tail end temperature control method, device and medium
CN111814388A (en) * 2020-06-12 2020-10-23 菲尼克斯(上海)环境控制技术有限公司 CFD simulation verification method for lower air supply data center based on neural network

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
《基于ZigBee技术的中央空调无线监控系统设计》;陈勇强;《中国优秀博硕士学位论文全文数据库(硕士)工程科技Ⅱ辑 2018年》;20180415(第4期);第二、三章 *
《基于热感知的数据中心能耗优化策略研究》;谷丽君;《中国优秀博硕士学位论文全文数据库(博士)信息科技辑 2020年》;20200115(第1期);全文 *

Also Published As

Publication number Publication date
CN112616292A (en) 2021-04-06

Similar Documents

Publication Publication Date Title
CN112616292B (en) Data center energy efficiency optimization control method based on neural network model
US10619882B2 (en) Building management system with scorecard for building energy and equipment performance
US10082779B2 (en) Energy management computer system
US20210200713A1 (en) Systems and methods for generating a data structure from multiple bim files
US20190042629A1 (en) Building management system with timeseries processing
US20200200420A1 (en) Building management system with central plantroom dashboards
CN107223195A (en) Variable air quantity for HVAC system is modeled
CN103996095A (en) Building energy control system and method
CN107860102A (en) A kind of method and device for controlling central air-conditioning
Ghofrani et al. Prediction of building indoor temperature response in variable air volume systems
US11236917B2 (en) Building control system with zone grouping based on predictive models
Kriksciuniene et al. Data Analysis in the Intelligent Building Environment.
CN117234301A (en) Server thermal management method based on artificial intelligence
Li et al. Thermal error modeling of electric spindle based on particle swarm optimization-SVM neural network
Hou et al. Development of event-driven optimal control for central air-conditioning systems
Zhao et al. The application of advanced control technologies in air conditioning system–a review
Tashiro et al. A network model for prediction of temperature distribution in data centers
CN106403080A (en) Central air-conditioning cold station system and control method
Chassiakos et al. BIM and lean-business process reengineering for energy management optimization of existing building stock
Lim et al. A Preliminary Review of Building Informatics for Sustainable Energy Management
Ahmed et al. A review on enhancing energy efficiency and adaptability through system integration for smart buildings
Mörzinger et al. A modular, holistic optimization approach for industrial appliances
CN111210060A (en) Method for predicting temperature of machine room during working day
Qu Research on Environment Monitoring of Network Computer Rooms in Colleges and Universities Based on The Internet of Things Technology
Wang et al. Trends In Digital Twinning for Design, Operation And Maintenance In Non-Residential Buildings: A Short Review from Literature

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant