Data center modeling and operation strategy determination method and device based on spatial-temporal characteristics

Abstract

The invention relates to the technical field of artificial intelligence, and provides a data center modeling and operation strategy determination method and device based on spatial-temporal characteristics. The method comprises the following steps: acquiring historical thermal environment data of a data center; training a preset cascade neural network model by using the historical thermal environment data to obtain a thermal environment simulation model; based on the thermal environment simulation model, constructing an operation optimization target of the data center and a corresponding constraint condition; determining a state variable, an action variable and a reward function in the Markov decision process according to the operation optimization target and the corresponding constraint condition so as to establish a Markov decision process model; and according to the Markov decision process model, in a learning environment corresponding to the thermal environment simulation model, a deep reinforcement learning algorithm is used for training to obtain an operation strategy of the data center. According to the embodiment of the invention, the dynamic thermal environment of the data center can be quickly and accurately described, and an effective refrigeration strategy of the data center can be constructed.

Description

Technical Field

[0001] This specification relates to the field of artificial intelligence technology, and in particular to a method and device for determining data center modeling and operation strategy based on spatiotemporal characteristics. Background Art

[0002] With the advent of the big data era, the whole society has an increasing demand for computing power, and various large models are highly favored. However, the training of these large models often requires a lot of computing power. In order to meet this exponentially growing demand for computing power, the demand for data centers around the world is also increasing. As a typical energy-consuming device, the energy consumption of data centers is increasing, among which the cooling system in the data center accounts for a large amount of energy consumption. Therefore, it is of great engineering significance to study the energy saving of data center cooling equipment, that is, how to control the air conditioning in the computer room so that each device can operate safely while minimizing the possibility of overall energy depletion.

[0003] One of the key points in the process of studying the energy-saving problem of data center refrigeration equipment is the characterization of the dynamic thermal environment of the data center, that is, predicting the distribution of the temperature field of the computer room under different control strategies. Computational fluid dynamics (CFD) is often used to characterize the thermal environment. The CFD method replaces the field of the original continuous physical quantity in the time domain and space domain with a set of variable values ​​at a finite number of discrete points, establishes a set of algebraic equations about the relationship between the field variables at these discrete points, and then solves the algebraic equations to obtain the approximate values ​​of the field variables. However, the CFD method consumes simulation resources extremely, the simulation time is extremely long, and it is not suitable for real-time control optimization. Therefore, there is an urgent need for a data center modeling and operation strategy determination method based on spatiotemporal characteristics to achieve rapid and accurate characterization of the dynamic thermal environment of the data center, and to build an effective data center cooling strategy, thereby reducing the energy consumption of the data center. Summary of the invention

[0004] In view of the high cost and difficulty of implementing the current CFD method, which is not suitable for real-time control optimization, this solution is proposed to overcome the above problems or at least partially solve the above problems.

[0005] On the one hand, some embodiments of this specification aim to provide a data center modeling and operation strategy determination method based on spatiotemporal characteristics, the method comprising:

[0006] Obtain historical thermal environment data of the data center;

[0007] Using the historical thermal environment data to train a preset cascade neural network model, so that the cascade neural network model predicts the thermal environment data at the next moment according to the spatiotemporal characteristics in the historical thermal environment data, and obtains a thermal environment simulation model;

[0008] Based on the thermal environment simulation model, constructing the operation optimization objectives and corresponding constraints of the data center;

[0009] Determine the state variables, action variables and reward function in the Markov decision process according to the operation optimization goal and the corresponding constraint conditions to establish a Markov decision process model;

[0010] According to the Markov decision process model, in the learning environment corresponding to the thermal environment simulation model, a deep reinforcement learning algorithm is applied to train and obtain an operation strategy of the data center.

[0011] Furthermore, obtaining the thermal environment data of the data center further includes:

[0012] Constructing a machine room model; the machine room model includes two rows of cabinet units and one or more row-to-row air conditioning units;

[0013] The computer room models are combined and arranged according to the actual layout of the data center to obtain a data center simulation model;

[0014] The data center simulation model is run under a plurality of preset working conditions to obtain thermal environment data of the data center under the plurality of working conditions.

[0015] Furthermore, the cascaded neural network model includes a cascaded Transformer model and a graph neural network model.

[0016] Furthermore, the historical thermal environment data is used to train a preset cascade neural network model, so that the cascade neural network model predicts the thermal environment data at the next moment according to the spatiotemporal characteristics in the historical thermal environment data, and obtains a thermal environment simulation model, including:

[0017] Inputting the historical thermal environment data into the cascade neural network model, so that the Transformer model extracts the time series characteristics of the thermal environment data before the first moment in the historical thermal environment data, predicts the thermal environment data at the next moment after the first moment, and iterates the Transformer model according to the prediction result;

[0018] Determine the input data of the graph neural network model according to the output result of the Transformer model, so that the graph neural network model extracts spatial features in the input data, predicts the thermal environment data at the next moment of the first moment, and iterates the graph neural network model according to the prediction result, thereby obtaining a thermal environment simulation model;

[0019] or,

[0020] Inputting the historical thermal environment data into the cascade neural network model, so that the graph neural network model extracts the spatial features before the second moment in the historical thermal environment data, predicts the thermal environment data at the next moment after the second moment, and iterates the graph neural network model according to the prediction result;

[0021] The input data of the Transformer model is determined according to the output results of the graph neural network model, so that the Transformer model extracts the time series features in the input data, predicts the thermal environment data at the next moment of the second moment, and iterates the Transformer model according to the prediction results to obtain a thermal environment simulation model.

[0022] Furthermore, based on the thermal environment simulation model, the operation optimization objectives and corresponding constraints of the data center are constructed, including:

[0023] Determining the energy consumption of the thermal environment simulation model according to the thermal environment simulation data and the running time information of the thermal environment simulation model;

[0024] Minimizing the energy consumption is used as the operation optimization goal, and the constraint condition is determined according to the operation strategy operation range allowed by the data center.

[0025] Furthermore, according to the Markov decision process model, in the learning environment corresponding to the thermal environment simulation model, a deep reinforcement learning algorithm is applied to train and obtain an operation strategy of the data center, including:

[0026] Step 0: Initialize network parameters and information buffer; the information buffer is used to store historical thermal environment data;

[0027] Step 1: Get the current state and obtain the current action based on the control strategy sampling;

[0028] Step 2: inputting the historical thermal environment data into the thermal environment simulation model to predict the state at the next moment;

[0029] Step 3: Update the information buffer according to the current action and the next moment state;

[0030] Step 4: Input the current state and current action into the reward function to calculate the reward value;

[0031] Step 5: adding the state transition data consisting of the current state, current action, reward value and next moment state to the replay buffer of this iteration; the replay buffer is used to store the state transition data required for each iteration;

[0032] Step 6: Partial batch sampling is performed on the state transition data in the replay buffer of this iteration to obtain sampled batch data, and reinforcement learning is performed based on the sampled batch data to update network parameters;

[0033] Repeat steps 1 to 6 above until the preset number of iterations is reached.

[0034] Further, reinforcement learning is performed based on the sampled batch data to update network parameters, including:

[0035] Based on the sampled batch data, reinforcement learning is used to update the action value function, state value function and control strategy.

[0036] On the other hand, some embodiments of this specification also provide a data center modeling and operation strategy determination device based on spatiotemporal characteristics, the device comprising:

[0037] A receiving module, used to obtain historical thermal environment data of the data center;

[0038] A simulation module, used to train a preset cascade neural network model using the historical thermal environment data, so that the cascade neural network model predicts the thermal environment data at the next moment according to the spatiotemporal characteristics in the historical thermal environment data, and obtains a thermal environment simulation model;

[0039] A construction module, used to construct an operation optimization target and corresponding constraint conditions of a data center based on the thermal environment simulation model;

[0040] A determination module, used to determine the state variables, action variables and reward function in the Markov decision process according to the operation optimization goal and the corresponding constraint conditions, so as to establish a Markov decision process model;

[0041] The training module is used to apply a deep reinforcement learning algorithm to train the data center's operation strategy according to the Markov decision process model in a learning environment corresponding to the thermal environment simulation model.

[0042] On the other hand, some embodiments of the present specification further provide a computer device, including a memory, a processor, and a computer program stored in the memory, wherein when the computer program is executed by the processor, the instructions of the above method are executed.

[0043] On the other hand, some embodiments of the present specification further provide a computer storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor of a computer device, the instructions of the above method are executed.

[0044] On the other hand, some embodiments of the present specification further provide a computer program product, wherein the computer program product comprises a computer program, and when the computer program is executed by a processor of a computer device, the instructions of the above method are executed.

[0045] One or more technical solutions provided by some embodiments of this specification have at least the following technical effects:

[0046] The embodiments of the present specification automatically obtain historical thermal environment data of a data center, and use a cascade neural network model trained with the historical thermal environment data. During the training process, the cascade neural network model predicts the thermal environment data at the next moment based on the spatiotemporal characteristics in the historical thermal environment data, thereby ensuring that the trained thermal environment simulation model can fully explore the intrinsic connections in the historical thermal environment data from both the temporal and spatial dimensions, and achieve accurate assessment of the future thermal environment. Afterwards, the operation optimization objectives and corresponding constraints of the data center are constructed based on the thermal environment simulation model, the operation optimization process of the data center is converted into a Markov decision process, the state variables, action variables and reward functions in the Markov decision process are determined, and a Markov decision process model is established. In addition, the thermal environment simulation model can provide a high-quality learning environment, and use deep reinforcement learning training to obtain the operation strategy of the data center, thereby ensuring the effectiveness of the operation strategy and facilitating real-time optimization control of the data center.

[0047] The above description is only an overview of the technical solutions of some embodiments of this specification. In order to more clearly understand the technical means of some embodiments of this specification, they can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of some embodiments of this specification more obvious and easy to understand, the specific implementation methods of some embodiments of this specification are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate some embodiments of this specification or technical solutions in the prior art, the following briefly introduces the drawings required for the embodiments or prior art descriptions. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. In the drawings:

[0049] Figure 1 A schematic diagram of an implementation system of a data center modeling and operation strategy determination method based on spatiotemporal characteristics in some embodiments of this specification is shown;

[0050] Figure 2 A flowchart of a method for data center modeling and operation strategy determination based on spatiotemporal characteristics in some embodiments of this specification is shown;

[0051] Figure 3 A schematic diagram of the steps of obtaining thermal environment data of a data center in some embodiments of this specification;

[0052] Figure 4 This is a schematic diagram of a machine room model in some embodiments of this specification;

[0053] Figure 5 A schematic diagram of the first step of obtaining a thermal environment simulation model in some embodiments of this specification;

[0054] Figure 6 A schematic diagram of the second step of obtaining a thermal environment simulation model in some embodiments of this specification;

[0055] Figure 7 A schematic diagram of spatial topological geometric relationships in some embodiments of this specification;

[0056] Figure 8 A schematic diagram of the connection relationship between the Transformer model and the graph neural network model in some embodiments of this specification;

[0057] Fig. 9 A schematic diagram of the steps for determining the operation optimization target and the corresponding constraint conditions in some embodiments of this specification;

[0058] Fig.10 This is a schematic diagram of the transformation framework in some embodiments of this specification;

[0059] Fig.11 A schematic diagram of the steps of applying deep reinforcement learning algorithm training to obtain the operation strategy of a data center in some embodiments of this specification;

[0060] Fig.12 A schematic diagram of the structure of a data center modeling and operation strategy determination device based on spatiotemporal characteristics in some embodiments of this specification;

[0061] Fig.13 This is a schematic diagram of the computer device structure provided in some embodiments of this specification.

[0062] [Description of Reference Numerals]

[0063] 101. Terminal;

[0064] 102. Server;

[0065] 401, cabinet unit;

[0066] 402. Inter-row air conditioning unit;

[0067] 1201, receiving module;

[0068] 1202, simulation module;

[0069] 1203. Construction module;

[0070] 1204. Determine a module;

[0071] 1205. Training module;

[0072] 1302. Computer equipment;

[0073] 1304. Processor;

[0074] 1306. Memory;

[0075] 1308, driving mechanism;

[0076] 1310, input / output interface;

[0077] 1312. Input devices;

[0078] 1314. Output device;

[0079] 1316. Presentation equipment;

[0080] 1318. Graphical user interface;

[0081] 1320, network interface;

[0082] 1322, communication link;

[0083] 1324. Communication bus. DETAILED DESCRIPTION

[0084] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in some embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on some embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this specification.

[0085] It should be noted that the terms "first", "second", etc. in the specification and claims of this document and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of this document described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, device, product or equipment that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or equipment. It should be noted that the acquisition, storage, use, processing, etc. of data in the technical solution of this application comply with the relevant provisions of relevant laws and regulations.

[0086] like Figure 1 The diagram is a schematic diagram of the implementation system of the data center modeling and operation strategy determination method based on spatiotemporal characteristics according to an embodiment of the present invention, which may include: a terminal 101 and a server 102, wherein the terminal 101 and the server 102 communicate with each other through a network, and the network may include a local area network (LAN), a wide area network (WAN), the Internet or a combination thereof, and is connected to a website, a user device (such as a computing device), and a back-end system. A staff member may send a data center operation strategy determination request to the server 102 through the terminal 101, and after receiving the data center operation strategy determination request, the server 102 calls the database for calculation processing to obtain the operation strategy of the data center, and sends the operation strategy of the data center to the terminal 101, so that the staff member processes the business according to the operation strategy of the data center.

[0087] In the embodiments of this specification, the server 102 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDN, Content Delivery Network), and big data and artificial intelligence platforms.

[0088] In an optional embodiment, the terminal 101 may include but is not limited to electronic devices such as self-service terminal devices, desktop computers, tablet computers, laptop computers, smart wearable devices, etc. Optionally, the operating system running on the electronic device may include but is not limited to Android system, IOS system, Linux, Windows, etc. Of course, the terminal 101 is not limited to the above-mentioned electronic devices with a certain entity, and it can also be software running in the above-mentioned electronic devices.

[0089] In addition, it should be noted that Figure 1 What is shown is only an application environment provided by the present disclosure. In actual application, multiple terminals 101 may be included, and this specification does not limit this.

[0090] Figure 2 It is a flowchart of a method for data center modeling and operation strategy determination based on spatiotemporal characteristics provided by an embodiment of the present invention. This specification provides method operation steps as described in the embodiment or flowchart, but may include more or fewer operation steps based on conventional or non-creative labor. The order of steps listed in the embodiment is only one way of executing the steps among many orders, and does not represent the only order of execution. When the system or device product is executed in practice, it can be executed in the order or in parallel according to the method shown in the embodiment or the accompanying drawings. Specifically, Figure 2 As shown, applied to the above-mentioned server side, the method may include:

[0091] S201: Acquire historical thermal environment data of the data center;

[0092] S202: using the historical thermal environment data to train a preset cascade neural network model, so that the cascade neural network model predicts the thermal environment data at the next moment according to the spatiotemporal characteristics in the historical thermal environment data, and obtains a thermal environment simulation model;

[0093] S203: constructing an operation optimization target and corresponding constraint conditions of the data center based on the thermal environment simulation model;

[0094] S204: Determine the state variables, action variables and reward function in the Markov decision process according to the operation optimization goal and the corresponding constraint conditions to establish a Markov decision process model;

[0095] S205: According to the Markov decision process model, in the learning environment corresponding to the thermal environment simulation model, a deep reinforcement learning algorithm is applied to train and obtain an operation strategy of the data center.

[0096] The embodiments of the present specification automatically obtain historical thermal environment data of a data center, and use a cascade neural network model trained with the historical thermal environment data. During the training process, the cascade neural network model predicts the thermal environment data at the next moment based on the spatiotemporal characteristics in the historical thermal environment data, thereby ensuring that the trained thermal environment simulation model can fully explore the intrinsic connections in the historical thermal environment data from both the temporal and spatial dimensions, and achieve accurate assessment of the future thermal environment. Afterwards, the operation optimization objectives and corresponding constraints of the data center are constructed based on the thermal environment simulation model, the operation optimization process of the data center is converted into a Markov decision process, the state variables, action variables and reward functions in the Markov decision process are determined, and a Markov decision process model is established. In addition, the thermal environment simulation model can provide a high-quality learning environment, and use deep reinforcement learning training to obtain the operation strategy of the data center, thereby ensuring the effectiveness of the operation strategy and facilitating real-time optimization control of the data center.

[0097] It can be understood that, in some embodiments, the existing methods for implementing dynamic thermal environment modeling are mainly divided into two categories: mechanism-driven and data-driven. The mechanism-driven method mainly starts from the basic physical laws of temperature field changes and simplifies the complex heat transfer process through physical formulas. The data-driven method mainly realizes the modeling of the temperature field by mining the internal laws collected from actual or simulator data, and does not rely on the actual physical laws. However, the mechanism-driven method is either based on relatively ideal assumptions, which is insufficient to describe complex physical processes, or loses accuracy when applied to large-scale problems due to the increasing complexity of boundary conditions and coupling effects. The data-driven method does not make full use of the inherent characteristics of time and space of the temperature field in the data center, so when applied to large-scale problems, it may lead to a loss of accuracy and robustness.

[0098] Considering that the change of the temperature field of the data center itself contains inherent structural characteristics and potential laws, it is mainly reflected in two levels: time and space: in time, the change of the temperature field of the data center is continuous and inertial, and the temperature distribution at the current moment has an intrinsic relationship with the temperature distribution at multiple historical moments; in space, there is a topological geometric relationship between various devices in the data center. Therefore, by mining the spatiotemporal characteristics in the historical thermal environment data of the data center, it is possible to ensure the prediction of the thermal environment data at the next moment.

[0099] Based on this, the cascaded neural network model is a structure that cascades multiple neural network models according to specific rules, including a cascaded Transformer model and a graph neural network model. The Transformer model is used to extract temporal features in the data, and the graph neural network is used to extract spatial features in the data, thereby ensuring the modeling accuracy of the thermal environment simulation model and the effectiveness of the subsequent data center operation strategy obtained using deep reinforcement learning.

[0100] Furthermore, since the cost of acquiring the real thermal environment data of the data center is high and the working conditions are relatively simple, in some embodiments, the actual thermal environment data of the data center is obtained by referring to the attached Figure 3 , obtaining the thermal environment data of the data center, further comprising:

[0101] S301: construct a machine room model; the machine room model includes a plurality of rows of cabinet units and one or more row-to-row air conditioning units;

[0102] S302: Arrange the computer room models according to the layout of the data center to obtain a data center simulation model;

[0103] S303: Running the data center simulation model under various preset working conditions to obtain thermal environment data of the data center under various working conditions.

[0104] Specifically, the computer room is often provided with a raised floor, and the rows of cabinet units may be two or more rows of cabinet units. Figure 4 The schematic diagram of the computer room model shown in the figure includes two rows of cabinet units 401 and an inter-row air conditioning unit (ACU) 402, wherein the two rows of cabinet units 401 are relatively parallel and distributed, and each row of cabinet units 401 is composed of three cabinets, and the cold channel is closed. The inter-row air conditioning unit 402 inhales hot air from above and sends cold air to the space under the elevated floor from below. The cold air enters the closed cold channel through the opening on the elevated floor, flows through the servers in the cabinet unit 401, takes away the heat generated by the servers, and is discharged from the hot channel behind the cabinet unit 401, and returns to the air conditioning return outlet to form a cycle. And above the cold channel of each cabinet. One or more sensors can be arranged between the hot channels and the air outlets and return outlets of each cabinet. For example, continue as shown in the attached figure. Figure 4 As shown, a sensor is arranged above the cold channel of each cabinet, a sensor is arranged at the upper, middle and lower positions of the hot channel of each cabinet, a temperature sensor is arranged at the return air outlet of the inter-row air conditioning unit 402, and a temperature sensor and an air volume sensor are arranged at the air outlet of the inter-row air conditioning unit 402. It should be noted that the number of cabinets in each row of cabinet units 401 is usually multiple, which is not limited in this article.

[0105] Furthermore, the computer room models can be combined and arranged according to the layout of the data center to obtain a data center simulation model. The layout of the data center is not limited in this article, and the data center simulation model is only a simulation of the physical environment of the space (excluding the thermal environment). After that, the data center simulation model is run under multiple preset working conditions to quickly obtain the thermal environment data of the data center under multiple working conditions.

[0106] Specifically, running the data center simulation model under various preset working conditions requires presetting the equipment specifications and operating rules of the cabinet units and row-to-row air conditioning units according to the actual situation, for example:

[0107] For the inter-row air conditioning unit ACU, a sinking EC fan air conditioner is used, the maximum air conditioning flow is set to 4.8 cubic meters per second, and the maximum power consumption is 2.2kW. The outlet air temperature is controlled by the supply air temperature, and the outlet air temperature control range is 10℃-20℃; the outlet air volume is controlled by the return air temperature, and the return air temperature control range is 20℃-30℃. The return air temperature setting value and the outlet air temperature setting value always maintain a temperature difference of 10℃. In order to make the air conditioning strategy more random, the outlet air temperature fluctuates randomly at intervals of 1℃, 2℃, and 3℃, namely:

[0108]

[0109] in, Set the air outlet temperature of the air conditioner at the current moment. The air outlet temperature of the air conditioner is set at the previous moment, ΔT is the random fluctuation temperature, and its value is 1°C, 2°C or 3°C. The return air temperature of the air conditioner is calculated according to the fixed temperature difference relative to the air outlet temperature.

[0110] For the cabinet unit, the IT load selects three levels: 1000W, 3000W, and 5000W, representing high load, medium load, and low load conditions respectively. Considering the randomness of IT load distribution, each cabinet may take any of the above three levels, so there are 3^6=729 load topologies. In addition, considering the random volatility of the load, the actual value of the IT load adds a Gaussian random noise on the basis of the original three levels:

[0111] IT=IT base +1000×N(μ,σ 2 ) (2)

[0112] Among them, IT represents the load of the cabinet, IT base represents one of high load, medium load or low load, N(μ,σ 2 ) represents the mean μ and the variance σ 2 Gaussian distribution.

[0113] When running the data center simulation model under various preset working conditions, the simulation interval can be set to 60s, the initial temperature field is set to the temperature field in steady state, and the initial IT load and air conditioning outlet and return air temperature are randomly selected according to the above operating rules. It is assumed that the air conditioning outlet temperature and IT load are modified every 10 simulation intervals, i.e., 10 minutes, so as to initialize the temperature field, equipment specifications and operating rules of the data center simulation model, and based on the cold channel temperature, hot channel temperature and actual air conditioning outlet and return air temperature and air volume detected by the sensor,

[0114] See attached Figure 5 In some embodiments, using the historical thermal environment data to train a preset cascade neural network model so that the cascade neural network model predicts the thermal environment data at the next moment according to the spatiotemporal characteristics in the historical thermal environment data, and obtaining the thermal environment simulation model may include:

[0115] S501: Inputting the historical thermal environment data into the cascade neural network model, so that the Transformer model extracts the time series features of the thermal environment data before the first moment in the historical thermal environment data, predicts the thermal environment data at the next moment after the first moment, and iterates the Transformer model according to the prediction result;

[0116] S502: Determine the input data of the graph neural network model according to the output result of the Transformer model, so that the graph neural network model extracts spatial features in the input data, predicts the thermal environment data at the next moment of the first moment, and iterates the graph neural network model according to the prediction result, thereby obtaining a thermal environment simulation model;

[0117] See attached Figure 6 In some embodiments, using the historical thermal environment data to train a preset cascade neural network model so that the cascade neural network model predicts the thermal environment data at the next moment according to the spatiotemporal characteristics in the historical thermal environment data, and obtaining the thermal environment simulation model may include:

[0118] S601: Input the historical thermal environment data into the cascade neural network model, so that the graph neural network model extracts the spatial features before the second moment in the historical thermal environment data, predicts the thermal environment data at the next moment after the second moment, and iterates the graph neural network model according to the prediction result;

[0119] S602: Determine the input data of the Transformer model according to the output result of the graph neural network model, so that the Transformer model extracts the time series features in the input data, predicts the thermal environment data at the next moment of the second moment, and iterates the Transformer model according to the prediction result, thereby obtaining a thermal environment simulation model.

[0120] It can be understood that, in some embodiments, when the cascade neural network extracts the spatiotemporal features of historical thermal environment data, the structure of the cascade neural network can be that the output end of the Transformer model is connected to the input end of the graph neural network model, or that the output end of the graph neural network model is connected to the input end of the Transformer model. This article does not limit the splicing order of the Transformer model and the graph neural network model. Figure 5 And attached Figure 6 Represents two feasible, parallel model training schemes.

[0121] Specifically, the Transformer model is a network model based on the attention mechanism using an encoder-decoder architecture. Its unique multi-head attention mechanism and position-based encoding mechanism can effectively learn the correlation between different time series tracks and between different moments of the same time series track, which is consistent with the temporal structure characteristics of the temperature field of the data center computer room. The Transformer model uses the information of M historical moments to predict the temperature field at the next moment in the future, where the information I at the historical moment t is t The definition is as follows:

[0122]

[0123] in, Represents the temperature of the cold channel temperature measurement point, Represents the temperature of the hot channel temperature measurement point, T supply ,T return They represent the supply air temperature and return air temperature of the air conditioner at the current time t, V represents the air supply volume of the air conditioner, and IT i Represents the IT load value of the current cabinet. Represents the set supply air temperature and return air temperature of the air conditioner at the next moment, Nc represents the number of cold aisle temperature measurement points, and Nh represents the number of hot aisle temperature measurement points. Figure 4 In the embodiment shown, Nc=6, Nh=18.

[0124] Therefore, the input to the encoder in the Transformer model is:

[0125] EncoderInput = {I i} i=1,2,3…M-1,M (4)

[0126] The input of the decoder is the information of the last moment in history, that is:

[0127] DecoderInput = {I M} (5)

[0128] The output of the Transformer model is:

[0129]

[0130] in, T supply,t+1 ,T return,t+1 ,V t+1 They represent the predicted cold channel temperature, hot channel temperature, air conditioning supply air temperature, air conditioning return air temperature and air supply volume at the next moment respectively.

[0131] See attached Figure 7 ,Since the locations of the equipment and different sensors in the data center simulation model have topological geometric relationships, the monitoring points where the sensors are located can be used as nodes of the graph, and the nodes correspond to Figure 7 The dots in Figure 7 The two isolated dots in the lower middle are actually sensors installed at the inter-row air conditioning units, which can detect the air volume and return air volume. The topological connection relationship information between different nodes is used to define the edges of the graph. Combined with the actual physical meaning of the variables detected by each sensor, considering that the nodes of the graph are not necessarily interconnected, the topological connection relationship rules of the computer room are defined as follows:

[0132] (1) The air conditioning return air temperature and outlet air temperature nodes are not connected: the outlet air temperature is affected by the air conditioning control strategy and is not directly affected by the return air temperature;

[0133] (2) The nine cold channel detection nodes are interconnected: they are all located in a closed cold channel and affect each other;

[0134] (3) The 9 hot channel detection nodes on the left are interconnected, and the 9 hot channel detection nodes on the right are interconnected, but the hot channels on the left and right sides are not connected: the hot channel nodes on the same side are close to each other and have a greater impact on each other; the hot channel nodes on the left and right sides are far apart and their impact on each other is negligible;

[0135] (4) The air outlet detection node of the air conditioner is connected to all nodes of the cold channel: the air pumped out by the air conditioner directly enters the closed cold channel, which has a direct impact on the temperature detection point of the cold channel;

[0136] (5) The air conditioning return air detection node is connected to all nodes in the hot channel: the air in the hot channel directly converges at the air conditioning return air outlet, and the temperature of each hot channel detection point directly affects the temperature of the air conditioning return air outlet.

[0137] (6) All cold channel detection points are interconnected with the hot channel detection points: The air in the cold channel is heated to become the air in the hot channel after passing through different servers. Therefore, the temperatures in the cold and hot channels are directly related. This is also the most important dynamic change law of the temperature field that needs to be learned.

[0138] It should be noted that the number of cold channel detection nodes and hot channel detection nodes can be set according to actual needs.

[0139] The weights of the edges in the graph are defined as follows:

[0140]

[0141] Among them, W ij represents the weight of the edge between different sensor detection nodes, d ij Represents the Euclidean distance between different sensor detection nodes.

[0142] According to W ij After obtaining the topological connection relationship information W, the following neural network model is constructed:

[0143] Y=G(X N,F ,W) (8)

[0144] Among them, X N,F is the input of the graph neural network, i.e., the F-dimensional detection data of N sensor nodes, G is the graph neural network, and Y is the output of the graph neural network. Through the graph neural network, the topological geometric relationship between each detection point in the computer room is effectively captured, which is more conducive to the accurate prediction of the temperature field.

[0145] Attach Figure 5 Taking the training scheme shown in the figure as an example, the connection relationship between the Transformer model and the graph neural network model is shown in the figure. Figure 8 Specifically, based on the Figure 4 With attached Figure 7 The distribution of cold channel detection nodes and hot channel detection nodes is shown in the figure. At this time, the output of the Transformer model is 27 dimensions, including 26-dimensional temperature variable information and 1-dimensional air volume information. Considering Figure 3There are only 26 sensor detection nodes in the model, so the first 26-dimensional temperature variable information is intercepted from the output of the Transformer model, and then input into the graph neural network model described in formula (8). Finally, the output of the graph neural network is spliced ​​with the air volume information in the output of the Transformer model as the output of the entire cascade neural network model. Since the thermal environment data includes two types of data, temperature and air volume, when the Transformer model and the graph neural network model are cascaded, the air volume at the next moment can be predicted only by the Transformer model, without the need to connect to the graph neural network model for processing. The finally trained cascade neural network model is the thermal environment simulation model.

[0146] See attached Fig. 9 In some embodiments, based on the thermal environment simulation model, constructing the operation optimization objectives and corresponding constraints of the data center may include:

[0147] S901: determining the energy consumption of the thermal environment simulation model according to the thermal environment simulation data and the running time information of the thermal environment simulation model;

[0148] S902: Taking minimizing the energy consumption as the operation optimization goal, and determining the constraint condition according to the operation strategy operation range allowed by the data center.

[0149] It can be understood that, in some embodiments, the thermal environment simulation data refers to the thermal environment data obtained by simulating the thermal environment simulation model, and the operation optimization target and the corresponding constraint conditions can be expressed as:

[0150]

[0151]

[0152] Among them, S t Represents the observable thermal environment variables at the current moment, A t Represents the control variable at the current moment, which is defined as follows:

[0153]

[0154]

[0155] Where Δt is the simulation interval, J is the objective function, is the energy consumption of the inter-row air conditioning unit, are the upper and lower limit vectors of the air conditioning supply and return air temperature adjustment, are the upper and lower limit vectors of the hot channel safety temperature threshold, and f is the thermal simulation model.

[0156] Solving the above operation optimization objective can be understood as solving a sequential decision problem, the purpose of which is to minimize the objective function by finding the optimal strategy sequence mapping from observable variables to control variables. However, since the thermal environment simulation model is predicted by data from M historical moments, the above operation optimization objective does not have Markov properties. Therefore, it is necessary to solve the above operation optimization objective and transform it into a problem with Markov properties. Fig.10 The transformation framework diagram shown in FIG. M Indicates the current state, S M+1 Indicates the state at the next moment, a M Indicates the current action. The thermal environment simulation model is used to predict the thermal environment data at the next moment as the preliminary prediction process. After that, only the thermal environment data at the latest historical moment and its corresponding predicted thermal environment data at the next moment are selected for Markov decision making to achieve subsequent reinforcement learning.

[0157] Specifically, according to the operation optimization goal and the corresponding constraints, the state variables, action variables and reward function in the Markov decision process are determined using the following formula to establish a Markov decision process model:

[0158] s t =S t (12)

[0159] a t =A t (13)

[0160]

[0161] R=-(R1+λ2R2)+C (16)

[0162] Among them, s t is the state variable, a t is the action variable, R1 and R2 are the air conditioning energy consumption, They are the upper and lower limit vectors of the hot channel safety temperature threshold, is the hot channel safety temperature vector corresponding to the kth node, V is the air volume of the air conditioner, V s , P s is the air volume and power of the air conditioner under standard working conditions, R is the reward function, λ1 and λ2 are hyperparameters for adjusting various penalty weights, and C is a positive constant used to balance the distribution of reward values. The purpose of designing C is to ensure that the final reward can span the positive and negative ranges to facilitate the learning of the agent. Its value is roughly equal to the weighted expected sum of the two penalties under the random control strategy.

[0163] See attached Fig.11In some embodiments, according to the Markov decision process model, in the learning environment corresponding to the thermal environment simulation model, applying a deep reinforcement learning algorithm to train and obtain an operation strategy of the data center may include:

[0164] Step 0: Initialize network parameters and information buffer; the information buffer is used to store historical thermal environment data;

[0165] Step 1: Get the current state and obtain the current action based on the control strategy sampling;

[0166] Step 2: inputting the historical thermal environment data into the thermal environment simulation model to predict the state at the next moment;

[0167] Step 3: Update the information buffer according to the current action and the next moment state;

[0168] Step 4: Input the current state and current action into the reward function to calculate the reward value;

[0169] Step 5: adding the state transition data consisting of the current state, current action, reward value and next moment state to the replay buffer of this iteration; the replay buffer is used to store the state transition data required for each iteration;

[0170] Step 6: Partial batch sampling is performed on the state transition data in the replay buffer of this iteration to obtain sampled batch data, and reinforcement learning is performed based on the sampled batch data to update network parameters;

[0171] Repeat steps 1 to 6 above until the preset number of iterations is reached.

[0172] It can be understood that, in some embodiments, considering that both state variables and action variables are continuous value variables, the control strategy is obtained by training the agent based on the flexible actor-critic algorithm. As shown in the following algorithm flow, the agent collects data by interacting with the environment. The data includes the current state, the current action and the corresponding reward value and the state at the next moment. These data are stored in the replay buffer. Then, a small amount of batch data (i.e., minibatch) is extracted from the replay buffer, and the gradient descent (or gradient ascent) method is used to combine the objective function J of the Q value function. Q , the objective function of the state value function J V And the objective function J of the policy function π Update the neural network parameters ψ, φ,θ, and finally output the calculated control strategy, so as to quickly and accurately determine the operation strategy of the data center. Among them, the algorithm specially adds an information buffer to store historical thermal environment data. The information buffer is independent of the replay buffer. The information buffer is used to ensure that the thermal environment simulation model makes accurate predictions of the thermal environment at the next moment, thereby serving the Markov decision process.

[0173]

[0174]

[0175] It should be noted that, although the operations of the method of the present invention are described in a specific order in the above embodiments and the accompanying drawings, this does not require or imply that the operations must be performed in the specific order, or that all the operations shown must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps.

[0176] Corresponding to the above-mentioned data center modeling and operation strategy determination method based on spatiotemporal features, some embodiments of this specification also provide a data center modeling and operation strategy determination device based on spatiotemporal features, referring to Fig.12 As shown, in some embodiments, the apparatus may include:

[0177] The receiving module 1201 is used to obtain historical thermal environment data of the data center;

[0178] The simulation module 1202 is used to train a preset cascade neural network model using the historical thermal environment data, so that the cascade neural network model predicts the thermal environment data at the next moment according to the spatiotemporal characteristics in the historical thermal environment data, and obtains a thermal environment simulation model;

[0179] A construction module 1203 is used to construct an operation optimization target and corresponding constraint conditions of a data center based on the thermal environment simulation model;

[0180] A determination module 1204 is used to determine the state variables, action variables and reward function in the Markov decision process according to the operation optimization target and the corresponding constraint conditions to establish a Markov decision process model;

[0181] The training module 1205 is used to apply a deep reinforcement learning algorithm to train and obtain an operation strategy of the data center according to the Markov decision process model in a learning environment corresponding to the thermal environment simulation model.

[0182] For the convenience of description, the above device is described by dividing it into various units according to its functions. Of course, when implementing this specification, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0183] It should be noted that in the embodiments of this specification, the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved are all information and data authorized by the user and fully authorized by all parties.

[0184] It should be noted that the computer program product described in this specification is a software product that mainly implements the method described in this specification through a computer program.

[0185] The embodiments of this specification also provide a computer device. Fig.13 As shown, in some embodiments of the present specification, the computer device 1302 may include one or more processors 1304, such as one or more central processing units (CPUs) or graphics processing units (GPUs), and each processing unit may implement one or more hardware threads. The computer device 1302 may also include any memory 1306, which is used to store any kind of information such as code, settings, data, etc. In a specific embodiment, the computer program on the memory 1306 and can be run on the processor 1304, when the computer program is run by the processor 1304, the instructions of the method described in any of the above embodiments may be executed. Non-limitingly, for example, the memory 1306 may include any one or more combinations of the following: any type of RAM, any type of ROM, flash memory device, hard disk, optical disk, etc. More generally, any memory can use any technology to store information. Further, any memory can provide volatile or non-volatile retention of information. Further, any memory can represent a fixed or removable component of the computer device 1302. In one embodiment, when the processor 1304 executes the associated instructions stored in any memory or combination of memories, the computer device 1302 can perform any operation of the associated instructions. The computer device 1302 also includes one or more drive mechanisms 1308 for interacting with any memory, such as a hard disk drive mechanism, an optical disk drive mechanism, etc.

[0186] The computer device 1302 may also include an input / output interface 1310 (I / O) for receiving various inputs (via input devices 1312) and for providing various outputs (via output devices 1314). A specific output mechanism may include a presentation device 1316 and an associated graphical user interface 1318 (GUI). In other embodiments, the input / output interface 1310 (I / O), input device 1312, and output device 1314 may not be included, and the computer device 1302 may be used as a computer device in a network. The computer device 1302 may also include one or more network interfaces 1320 for exchanging data with other devices via one or more communication links 1322. One or more communication buses 1324 couple the components described above together.

[0187] The communication link 1322 may be implemented in any manner, for example, through a local area network, a wide area network (e.g., the Internet), a point-to-point connection, etc., or any combination thereof. The communication link 1322 may include any combination of hardwired links, wireless links, routers, gateway functions, name servers, etc., governed by any protocol or combination of protocols.

[0188] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), computer-readable storage media, and computer program products of some embodiments of the present specification. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processor to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processor generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0189] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processor to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0190] These computer program instructions can also be loaded onto a computer or other programmable data processor so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable device to implement the process. Figure 1A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0191] In a typical configuration, a computer device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0192] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0193] Computer-readable media include permanent and non-permanent, removable and non-removable media that can be used to store information by any method or technology. Information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computer device. As defined in this specification, computer-readable media does not include temporary computer-readable media (transitory media), such as modulated data signals and carrier waves.

[0194] Those skilled in the art will appreciate that the embodiments of this specification may be provided as methods, systems or computer program products. Therefore, the embodiments of this specification may take the form of complete hardware embodiments, complete software embodiments or embodiments combining software and hardware. Moreover, the embodiments of this specification may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0195] The present specification embodiments may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present specification embodiments may also be practiced in distributed computing environments where tasks are performed by remote processors connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0196] It should also be understood that in the embodiments of this specification, the term "and / or" is only a description of the association relationship of the associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0197] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0198] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the embodiments of this specification. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0199] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A data center modeling and operation strategy determination method based on spatiotemporal characteristics, characterized in that: The method comprises: Obtain historical thermal environment data of the data center; Using the historical thermal environment data to train a preset cascade neural network model, so that the cascade neural network model predicts the thermal environment data at the next moment according to the spatiotemporal characteristics in the historical thermal environment data, and obtains a thermal environment simulation model; Based on the thermal environment simulation model, constructing the operation optimization objectives and corresponding constraints of the data center; Determine the state variables, action variables and reward function in the Markov decision process according to the operation optimization goal and the corresponding constraint conditions to establish a Markov decision process model; According to the Markov decision process model, in the learning environment corresponding to the thermal environment simulation model, a deep reinforcement learning algorithm is applied to train and obtain an operation strategy of the data center.

2. The method according to claim 1, characterized in that: Acquiring thermal environment data of a data center further includes: Constructing a machine room model; the machine room model includes a plurality of rows of cabinet units and one or more row-to-row air conditioning units; The computer room models are combined and arranged according to the layout of the data center to obtain a data center simulation model; The data center simulation model is run under a plurality of preset working conditions to obtain thermal environment data of the data center under the plurality of working conditions.

3. The method according to claim 1, characterized in that: The cascade neural network model includes a cascaded Transformer model and a graph neural network model.

4. The method according to claim 3, characterized in that The historical thermal environment data is used to train a preset cascade neural network model, so that the cascade neural network model predicts the thermal environment data at the next moment according to the spatiotemporal characteristics in the historical thermal environment data, and obtains a thermal environment simulation model, including: Inputting the historical thermal environment data into the cascade neural network model, so that the Transformer model extracts the time series characteristics of the thermal environment data before the first moment in the historical thermal environment data, predicts the thermal environment data at the next moment after the first moment, and iterates the Transformer model according to the prediction result; Determine the input data of the graph neural network model according to the output result of the Transformer model, so that the graph neural network model extracts spatial features in the input data, predicts the thermal environment data at the next moment of the first moment, and iterates the graph neural network model according to the prediction result, thereby obtaining a thermal environment simulation model; or, Inputting the historical thermal environment data into the cascade neural network model, so that the graph neural network model extracts the spatial features before the second moment in the historical thermal environment data, predicts the thermal environment data at the next moment after the second moment, and iterates the graph neural network model according to the prediction result; The input data of the Transformer model is determined according to the output results of the graph neural network model, so that the Transformer model extracts the timing features in the input data, predicts the thermal environment data at the next moment of the second moment, and iterates the Transformer model according to the prediction results to obtain a thermal environment simulation model.

5. The method according to claim 1, characterized in that Based on the thermal environment simulation model, the operation optimization objectives and corresponding constraints of the data center are constructed, including: Determining the energy consumption of the thermal environment simulation model according to the thermal environment simulation data and the running time information of the thermal environment simulation model; Minimizing the energy consumption is used as the operation optimization goal, and the constraint condition is determined according to the operation strategy operation range allowed by the data center.

6. The method according to claim 1, characterized in that According to the Markov decision process model, in the learning environment corresponding to the thermal environment simulation model, a deep reinforcement learning algorithm is applied to train an operation strategy of the data center, including: Step 0: Initialize network parameters and information buffer; the information buffer is used to store historical thermal environment data; Step 1: Get the current state and obtain the current action based on the control strategy sampling; Step 2: Input the historical thermal environment data into the thermal environment simulation model to predict the state at the next moment; Step 3: Update the information buffer according to the current action and the next moment state; Step 4: Input the current state and current action into the reward function to calculate the reward value; Step 5: adding the state transition data consisting of the current state, current action, reward value and next moment state to the replay buffer of this iteration; the replay buffer is used to store the state transition data required for each iteration; Step 6: Partial batch sampling is performed on the state transition data in the replay buffer of this iteration to obtain sampled batch data, and reinforcement learning is performed based on the sampled batch data to update network parameters; Repeat steps 1 to 6 above until the preset number of iterations is reached.

7. The method according to claim 6, characterized in that Reinforcement learning is performed based on the sampled batch data to update network parameters, including: Based on the sampled batch data, reinforcement learning is used to update the action value function, state value function and control strategy.

8. A data center modeling and operation strategy determination device based on spatiotemporal characteristics, characterized in that: The device comprises: A receiving module, used to obtain historical thermal environment data of the data center; A simulation module, used to train a preset cascade neural network model using the historical thermal environment data, so that the cascade neural network model predicts the thermal environment data at the next moment according to the spatiotemporal characteristics in the historical thermal environment data, and obtains a thermal environment simulation model; A construction module, used to construct an operation optimization target and corresponding constraint conditions of a data center based on the thermal environment simulation model; A determination module, used to determine the state variables, action variables and reward function in the Markov decision process according to the operation optimization goal and the corresponding constraint conditions, so as to establish a Markov decision process model; The training module is used to apply a deep reinforcement learning algorithm to train the data center's operation strategy according to the Markov decision process model in a learning environment corresponding to the thermal environment simulation model.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory, characterized in that: When the computer program is executed by the processor, the computer program executes the instructions of the method according to any one of claims 1 to 7.

10. A computer storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor of a computer device, the computer program executes the instructions of the method according to any one of claims 1 to 7.

Images