A method, apparatus, and storage medium for energy consumption identification of applications based on graph convolutional networks.

Abstract

This invention discloses an application energy consumption identification method, computer device, and storage medium based on graph convolutional networks. The method includes steps such as inputting an energy consumption matrix, a self-connected adjacency matrix, and a degree matrix into a graph convolutional network for processing, and obtaining the application energy consumption identification result output by the graph convolutional network. By using a graph convolutional network to process data such as the energy consumption matrix, self-connected adjacency matrix, and degree matrix, this invention can obtain application energy consumption identification results. Specifically, the application energy consumption identification results can represent information such as the total energy consumption generated by the application itself and the resources it calls, thereby helping to classify the energy consumption of applications in the 5G core network and locate applications with abnormal energy consumption. When there is a large amount of data interaction between the user terminal and the application, it can accurately classify the energy consumption of the application. Using a graph convolutional neural network can improve the efficiency of application energy consumption classification. This invention has wide applications in the field of communication technology.

Description

Technical Field

[0001] This invention relates to the field of communication technology, and in particular to a method for identifying application power consumption, a computer device, and a storage medium based on graph convolutional networks. Background Technology

[0002] With the rapid development of 5G communication technology, existing research focuses more on how to improve energy efficiency while meeting user experience requirements, and this is primarily achieved within the network. Therefore, requirements, use cases, and solutions are largely confined to the network. Vertical industries and customers cannot obtain energy efficiency-related information from the network. This leads to a lack of detailed energy consumption information for users or network applications, making it difficult to pinpoint the cause when energy consumption anomalies occur. For example, in satellite and terrestrial scenarios, where satellite and terrestrial coverage varies in some areas, energy saving can be considered as a dimension in providing communication services. Users or operators can choose the best way to meet both user experience and energy efficiency. From another perspective, the network can also respond to different energy consumption patterns of applications or adjust network resources accordingly. In conclusion, there is a need for identifying application energy consumption within communication networks.

[0003] However, due to the strong coupling between various hardware and software components such as user terminals, network slices, and applications in communication networks, current technologies struggle to accurately identify the total energy consumption of an application itself and the resources it calls upon.

[0004] Terminology Explanation: AMF: Access and Mobility Management Function; 3GPP: 3rd Generation Partnership Project; UE: User Equipment; RAN: Radio Access Network; AF: Application Function; NRFG: Network Repository Function; PCF: Policy Control Function; ASF: Authentication Server Function. NSSF: Network Slice Selection Function. UDM: Unified Data Management; NEF: Network Exposure Function; NWDAF: Network Data Analytics Function; UDR: Unified Data Repository; NR: New Radio; GCN: Graph Convolution Network. Summary of the Invention

[0005] To address the technical problem that current communication network technologies struggle to accurately identify the total energy consumption of an application and the resources it calls, the present invention aims to provide an application energy consumption identification method, computer device, and storage medium based on graph convolutional networks.

[0006] On one hand, embodiments of the present invention include a method for identifying application power consumption based on graph convolutional networks, the method comprising the following steps: Obtain the energy consumption of user terminals, network slices, and applications in the communication network, and construct an energy consumption matrix; Construct an undirected topology graph using user terminals, network slices, and applications in the communication network as nodes; Based on the undirected topological graph, construct the self-connected adjacency matrix and degree matrix; The energy consumption matrix, the self-connection adjacency matrix, and the degree matrix are input into a graph convolutional network for processing. Obtain the application energy consumption identification results output by the graph convolutional network.

[0007] Furthermore, the step of obtaining the energy consumption of user terminals, network slices, and applications in the communication network, and constructing an energy consumption matrix, includes: Over a period of time, energy consumption is detected for several user terminals, several network slices, and several applications respectively to obtain their respective energy consumption time series; The energy consumption matrix is ​​formed by using the energy consumption time series of each user terminal, the energy consumption time series of each network slice, and the energy consumption time series of each application as rows or columns.

[0008] Furthermore, the step of performing energy consumption detection on the plurality of user terminals, the plurality of network slices, and the plurality of applications includes: Detect the total number of actual nodes; the actual nodes include actual running user terminals, network slices, and applications; When the total number of actual nodes is less than the number corresponding to the size of the graph convolutional network, test nodes are run; the test nodes include user terminals, network slices and / or applications used for testing; Energy consumption is measured for each of the actual nodes and each of the test nodes.

[0009] Furthermore, the running test node includes: The number of test nodes is determined based on the difference between the total number of actual nodes and the number corresponding to the size of the graph convolutional network. Run the corresponding number of test nodes.

[0010] Furthermore, the running test node includes: Obtain the first average energy consumption of all the actual nodes; Determine the application's energy consumption control targets; A second average energy consumption is determined based on the degree of deviation of the first average energy consumption from the application's energy consumption control target; The energy consumption of each test node is determined based on the second average energy consumption. The test node is operated according to the determined energy consumption.

[0011] Furthermore, the construction of an undirected topology graph using user terminals, network slices, and applications in the communication network as nodes includes: Each of the aforementioned user terminals, each of the aforementioned network slices, and each of the aforementioned applications are respectively used as nodes; When any of the user terminals is connected to any of the network slices, an edge is established between the user terminal and the node corresponding to the network slice. When any of the network slices runs any of the applications, an edge is established between the network slice and the node corresponding to the application. Each node and each edge is used as the undirected topology graph.

[0012] Further, the step of constructing a self-connected adjacency matrix and a degree matrix based on the undirected topological graph includes: Construct an adjacency matrix based on the undirected topological graph; Based on the adjacency matrix, obtain the corresponding self-connected adjacency matrix and the degree matrix.

[0013] Furthermore, the application energy consumption identification method based on graph convolutional networks also includes: Obtain the recognition result request sent by the user terminal; In response to the identification result request, the application energy consumption identification result is sent to the user terminal.

[0014] On the other hand, embodiments of the present invention also include a computer device, including a memory and a processor, wherein the memory is used to store at least one program, and the processor is used to load at least one program to execute an application energy consumption identification method based on graph convolutional networks according to the embodiments.

[0015] On the other hand, embodiments of the present invention also include a storage medium storing a processor-executable program, which, when executed by a processor, is used to perform an application power consumption identification method based on a graph convolutional network in the embodiments.

[0016] The beneficial effects of this invention are as follows: The application energy consumption identification method based on graph convolutional networks in the embodiments can obtain application energy consumption identification results by using graph convolutional networks to process data such as energy consumption matrix, self-connected adjacency matrix and degree matrix. Specifically, the application energy consumption identification results can represent information such as the total energy consumption generated by the application itself and the resources it calls, thereby helping to classify the application energy consumption of 5G core network and locate applications with abnormal energy consumption. When there is a large amount of data interaction between user terminal and application, the application can be accurately classified in terms of energy consumption. Using graph convolutional neural networks can improve the efficiency of application energy consumption classification. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of a scenario for the application energy consumption identification method based on graph convolutional networks in the embodiment. Figure 2 This is a schematic diagram illustrating the steps of the application energy consumption identification method based on graph convolutional networks in the embodiment. Figure 3 This is a flowchart illustrating the application energy consumption identification method based on graph convolutional networks in this embodiment. Figure 4 This is a schematic diagram of the network architecture of the convolutional network GCN in the embodiment. Figure 5 This is a schematic diagram of an undirected topology graph in the embodiment. Detailed Implementation

[0018] In this embodiment, the application energy consumption identification method based on graph convolutional networks can be applied to... Figure 1 In the communication network shown. (Refer to...) Figure 1The communication network includes multiple user terminals (UEs). The UEs access the core network through the access network. The core network includes network elements such as NEF, NRF, PCF, UDM, AUSF, AF, and NSSF, and is divided into network slices ①, ②, and ③, which run applications such as information communication services, smart grid, monitoring, Internet of Things, public safety, AR / VR, live streaming, and mobile banking, thereby providing corresponding services / businesses.

[0019] Reference Figure 2 The energy consumption identification method for applications based on graph convolutional networks includes the following steps: S1. Obtain the energy consumption of user terminals, network slices, and applications in the communication network, and construct an energy consumption matrix; S2. Construct an undirected topology graph using user terminals, network slices, and applications in the communication network as nodes; S3. Based on the undirected topological graph, construct the self-connected adjacency matrix and degree matrix; S4. Input the energy consumption matrix, self-connection adjacency matrix, and degree matrix into the graph convolutional network for processing; S5. Obtain the application energy consumption identification results output by the graph convolutional network.

[0020] The steps in the application energy consumption identification method based on graph convolutional networks can be executed by the core network or its NWDAF network elements.

[0021] Specifically, the process of the application energy consumption identification method based on graph convolutional networks is as follows: Figure 3 As shown. (Refer to...) Figure 3 An application energy consumption classification update request can be initiated from the AF network element in the core network to the NWDAF network element, thereby triggering the NWDAF to execute steps S1-S5.

[0022] Reference Figure 3 The process of the application energy consumption identification method based on graph convolutional networks includes: 1. Every 2 hours, AF sends an energy consumption classification update request message to NWDAF, carrying application energy consumption data in time series; 2. When NWDAF receives an energy consumption classification update request, the energy consumption classification model analyzes the data to obtain the energy consumption classification for each application. 3. Return the energy consumption classification results to AF to help locate applications with abnormal energy consumption; 4. At the same time, NWDAF sends the energy consumption classification results to UDR for storage and updates, so that UE can request and obtain the application energy consumption analysis results.

[0023] Reference Figure 3NWDAF runs the energy consumption hierarchical network model. In this embodiment, the energy consumption hierarchical network model is a graph convolutional network (GCN), and the network architecture of the GCN is as follows: Figure 4 As shown. (Refer to...) Figure 4 A Graph Convolutional Network (GCN) can be built using PyG's GCNConv. A GCN consists of two hidden layers, one input layer, one output layer, and two ReLU activation function layers. The input parameters of the GCN are the adjacency matrix A and the degree matrix. With the initial eigenvector H (0) .

[0024] After running the Graph Convolutional Network (GCN) using NWDAF, set the global variables according to Table 1, and then start training the GCN. Specifically, you can use the adjacency matrix A and the degree matrix labeled with hierarchical labels. With the initial eigenvector H (0) As input data to the Graph Convolutional Network (GCN), the adjacency matrix A and the degree matrix are used for training. With the initial eigenvector H (0) The meanings of the corresponding hierarchical labels are: adjacency matrix A, degree matrix With the initial eigenvector H (0) The rating indicates the individual energy consumption of each communication network component, such as user terminals, network slices, and applications. The rating label indicates the total energy consumption generated by the application itself and the resources it calls.

[0025] Table 1

[0026] When training the Graph Convolutional Network (GCN), the error function is calculated by taking the cross-entropy error of all results with hierarchical labels. Specifically, the formula is as follows:

[0027] in It is a collection of nodes with hierarchical labels. Indicates the number of categories. The predicted value is represented by the NWDAF, and the trained model parameters and the optimal optimizer parameters are then stored in the NWDAF. When the NWDAF needs to run the trained graph convolutional network GCN, it calls the trained parameters.

[0028] In this embodiment, when NWDAF performs step S1, which is to obtain the energy consumption of user terminals, network slices, and applications in the communication network and construct the energy consumption matrix, it can specifically perform the following steps: S101. Over a period of time, energy consumption is measured for several user terminals, several network slices, and several applications respectively, and their respective energy consumption time series are obtained; S102. The energy consumption time series of each user terminal, the energy consumption time series of each network slice, and the energy consumption time series of each application are respectively used as rows or columns to form a matrix, which is the energy consumption matrix.

[0029] Steps S101-S102 can be collected by a third party outside the core network, and then the collected data is sent to NWDAF.

[0030] In step S101, a time period can be set. , ...at the sampling time, respectively for and User terminals, and Network slicing and and Applications perform energy consumption monitoring to detect the individual energy consumption of user terminals, network slices, and applications, obtaining their respective energy consumption time series. For example, for user terminals... The detected energy consumption time series are 0.005, 0.005, 0.058, 0.06, 0.062... for user terminals. The detected energy consumption time series are 0.095, 0.098, 0.101, 0.096, 0.005... for network slicing. The detected energy consumption time series is 2, 3.6, 2, 2, 2... for network slices. The detected energy consumption time series is 8, 7.6, 8, 7.8, 8... for the application. The detected energy consumption time series is 1.2, 1.32, 1.44, 2, 1... for the application. The energy consumption time series obtained are 0.99, 0.98, 1.02, 10, 1.21... In each of the above energy consumption time series, the unit of the data is watts.

[0031] In step S102, the energy consumption time series obtained in step S1 are used as rows or columns to form a matrix, which is the energy consumption matrix. For example, using the energy consumption time series as rows, an energy consumption matrix with the following form is formed:

[0032] In this embodiment, when NWDAF performs step S101, which is to perform energy consumption detection on several user terminals, several network slices, and several applications respectively, it can specifically perform the following steps: S10101. Detect the total number of actual nodes; S10102. When the total number of actual nodes is less than the number corresponding to the size of the graph convolutional network, run the test nodes; S10103. Perform energy consumption detection on each actual node and each test node.

[0033] In step S10101, the actual nodes to be detected include the actual running user terminals, network slices, and applications. These user terminals, network slices, and applications are actually used to provide... Figure 1 Among the various communication services / services shown, the total number of actual nodes is the total number of user terminals, network slices, and applications actually running.

[0034] In step S10102, the quantity corresponding to the scale of the graph convolutional network can refer to the scale of the input data (e.g., the size of the matrix) received by the graph convolutional network under the influence of the network parameters and training process.

[0035] In step S10102, the total number of actual nodes will affect the size of the actual energy consumption matrix. If the total number of actual nodes is less than the number corresponding to the scale of the graph convolutional network, then the size of the energy consumption matrix will not match the network parameters of the graph convolutional network, and NWDAF will run test nodes. In this embodiment, the test nodes include at least one of the following: user terminals, network slices, and applications used for testing. For example, specifically, they can be the following: A. The test node includes multiple user terminals; B. The test node includes multiple network slices; C. The test node includes multiple applications; D. The test node includes multiple user terminals and multiple network slices; E. The test node includes multiple user terminals and multiple applications; F. The test node includes multiple user terminals, multiple network slices, and multiple applications.

[0036] During step S10102, NWDAF can call a function that is in an idle state (not actually provided). Figure 1 The user terminals, network slices, and / or applications of the various communication services / services shown can be used as test nodes. NWDAF can also simulate virtual user terminals, network slices, and / or applications by running computer programs as test nodes.

[0037] In step S10103, regardless of whether the test node is real or virtual, NWDAF can treat the test node as a node of the same type as the actual node, thereby performing energy consumption detection.

[0038] In this embodiment, the principle of executing steps S10101-S10103 is as follows: when the total number of actual nodes is less than the number corresponding to the scale of the graph convolutional network, by running test nodes, the size of the energy consumption matrix can be matched with the network parameters of the graph convolutional network, so that the network parameters of the graph convolutional network do not need to be adjusted, and the graph convolutional network can also adapt to different communication network application environments.

[0039] In this embodiment, when NWDAF executes step S10102, which is the step of running test nodes, it can determine the number of test nodes based on the difference between the total number of actual nodes and the number corresponding to the size of the graph convolutional network. Specifically, the number of test nodes run by NWDAF is exactly equal to the number of actual nodes, so that the sum of the number of actual nodes and the number of test nodes equals the number corresponding to the size of the graph convolutional network.

[0040] In this embodiment, when virtual test nodes are created, the energy consumption of each test node is controllable. When NWDAF executes step S10102, which is the step of running the test nodes, it can use the following steps to set the energy consumption of each test node: S1010201. Obtain the first average energy consumption of all actual nodes; S1010202. Determine the application's energy consumption control targets; S1010203. Determine the second average energy consumption based on the degree of deviation of the first average energy consumption from the application's energy consumption control target; S1010204. Determine the energy consumption of each test node based on the second average energy consumption; S1010205. Run the test node according to the determined energy consumption.

[0041] In step S1010201, NWDAF obtains the average energy consumption of all actual nodes, i.e., the first average energy consumption. Specifically, the first average energy consumption can be the energy consumption of all actual nodes at different times. , ...the arithmetic mean of energy consumption detected at each sampling time, or the energy consumption of all actual nodes at each sampling time... or The arithmetic mean of energy consumption detected at certain specific sampling times. In step S1010202, the application energy consumption control target represents the level that the core network hopes to control the overall energy consumption of all applications to reach, which can be represented by data with units such as W.

[0042] In step S1010203, NWDAF determines the degree of deviation of the first average energy consumption from the application's energy consumption control target, that is, whether the first average energy consumption is too large or too small relative to the application's energy consumption control target, thereby determining the second average energy consumption. Specifically, the second average energy consumption can be set to a value outside the range formed by the first average energy consumption and the application's energy consumption control target. For example, when the first average energy consumption is less than the application's energy consumption control target, the second average energy consumption can be set to a value that is less than the first average energy consumption or greater than the application's energy consumption control target.

[0043] In step S1010204, NWDAF sets the energy consumption of each test node with the goal that "the average of the sum of the energy consumption of all actual nodes and the energy consumption of all test nodes is equal to the second average energy consumption". The energy consumption of different test nodes may be equal or unequal.

[0044] In step S1010205, NWDAF sets the energy consumption of each test node according to the energy consumption calculated in step S1010204.

[0045] In this embodiment, the principle of executing steps S1010201-S1010205 is as follows: by configuring the energy consumption of each test node, when NWDAF runs the test node and performs energy consumption detection on all actual nodes and test nodes, the second average energy consumption detected is more deviated from the application energy consumption control target than the first average energy consumption. This is beneficial to adjust the application's energy consumption to approach the application energy consumption control target through the negative feedback control mechanism, thereby achieving energy consumption control of the application.

[0046] In this embodiment, when NWDAF performs step S2, which is to construct an undirected topology graph using user terminals, network slices, and applications in the communication network as nodes, it can specifically perform the following steps: S201. Each user terminal, each network slice, and each application is used as a node; S202. When any user terminal is connected to any network slice, an edge is established between the user terminal and the corresponding node of the network slice; S203. When any application is running on any network slice, establish an edge between the network slice and the node corresponding to the application; S204. Use each node and each edge as an undirected topological graph.

[0047] In this embodiment, the undirected topological graph obtained by performing steps S201-S204 is as follows: Figure 5As shown in the diagram. Assuming it contains a total of 9 user terminals, 3 core network slices, and 8 applications, then the undirected topology graph contains a total of 9 + 3 + 8 = 20 nodes.

[0048] In this embodiment, when NWDAF performs step S3, which is to construct the self-connected adjacency matrix and degree matrix based on the undirected topological graph, it can specifically perform the following steps: S301. Construct an adjacency matrix based on the undirected topological graph; S302. Based on the adjacency matrix, obtain the corresponding self-connected adjacency matrix and degree matrix.

[0049] by Figure 5 Taking the undirected topological graph shown as an example, in step S301, due to the undirected topological graph... Figure 1 It contains 20 nodes, therefore the size of the constructed adjacency matrix is ​​20×20, where U1 identifies user terminal node 1, S1 represents core network slice 1, and A1 and A2 represent application nodes 1 and 2, respectively. The constructed adjacency matrix is ​​shown in the following formula:

[0050] In step S302, the adjacency matrix after adding self-joins is obtained based on the adjacency matrix. Sum-degree matrix Among them, the adjacency matrix As shown in the following formula:

[0051] Degree matrix The degree matrix represents the number of edges connected to a node. As shown in the following formula:

[0052] In this embodiment, adjacency matrix and self-connected adjacency matrix are used. Sum-degree matrix Each column in the table represents U1...S1...A1 A2.

[0053] In step S4, the graph convolutional network (GCN) processes the energy consumption matrix and the self-connection adjacency matrix. Sum-degree matrix During processing, the propagation between layers follows the formula shown below:

[0054] As shown in the above formula, each layer of the Graph Convolutional Network (GCN) performs calculations, adds an activation function σ, and undergoes a nonlinear transformation to obtain a matrix that aggregates the features of adjacent vertices. .in, The first character in a graph convolutional network (GCN) represents the first character. The network parameters of the layer, Indicates the first The feature vector output by the layer, The first part will be input into the graph convolutional network GCN. Process at layer +1. Indicates the first +1 layer output feature vector The initial feature vector is the input data fed into the graph convolutional network (GCN), specifically the energy consumption matrix.

[0055] In step S5, the final application energy consumption identification result can be represented in the following standard format:

[0056] in, The feature vector can specifically be... , The feature vectors processed by each layer of the graph convolutional network GCN. The final output of the graph convolutional network GCN. The energy consumption identification result for an application can represent the total energy consumption generated by the application itself and the resources it calls upon. For example, It can be represented as continuous data or in hierarchical form.

[0057] In this embodiment, in addition to performing steps S1-S5, NWDAF can also perform the following steps: S6. Obtain the recognition result request sent by the user terminal; S7. In response to the identification result request, send the application's energy consumption identification result to the user terminal.

[0058] Before performing steps S6-S7, refer to Figure 3 NWDAF can store the application power consumption identification results in the UDR. In step S6, the UDR in the core network can receive the identification result request sent by the user terminal UE, and the UDR sends the stored application power consumption identification results to the user terminal UE, so that the user terminal UE can know the power consumption of the application it is currently using based on the application power consumption identification results, and thus perform power consumption management such as sleep mode.

[0059] A computer program that executes the application energy consumption identification method based on graph convolutional networks in this embodiment can be written into a computer device or storage medium. When the computer program is read out and run, the application energy consumption identification method based on graph convolutional networks in this embodiment is executed, thereby achieving the same technical effect as the application energy consumption identification method based on graph convolutional networks in the embodiment.

[0060] It should be noted that, unless otherwise specified, when a feature is referred to as "fixed" or "connected" to another feature, it can be directly fixed or connected to the other feature, or indirectly fixed or connected to the other feature. Furthermore, the descriptions of "upper," "lower," "left," and "right" used in this disclosure are only relative to the relative positional relationships of the components of this disclosure in the accompanying drawings. The singular forms "a," "an," and "the" used in this disclosure are also intended to include the plural forms, unless the context clearly indicates otherwise. Moreover, unless otherwise defined, all technical and scientific terms used in this embodiment have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this embodiment specification is only for describing particular embodiments and is not intended to limit the invention. The term "and / or" as used in this embodiment includes any combination of one or more of the associated listed items.

[0061] It should be understood that although the terms first, second, third, etc., may be used to describe various elements in this disclosure, these elements should not be limited to these terms. These terms are only used to distinguish elements of the same type from each other. For example, a first element may also be referred to as a second element without departing from the scope of this disclosure, and similarly, a second element may also be referred to as a first element. The use of any and all instances or exemplary language (“e.g.,” “such as,” etc.) provided in this embodiment is intended only to better illustrate embodiments of the invention and, unless otherwise required, does not impose a limitation on the scope of the invention.

[0062] It should be recognized that embodiments of the present invention can be implemented or carried out by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-transitory computer-readable storage medium. The method can be implemented using standard programming techniques—including a non-transitory computer-readable storage medium configured with a computer program, wherein such a storage medium causes the computer to operate in a specific and predefined manner—according to the methods and drawings described in the specific embodiments. Each program can be implemented in a high-level procedural or object-oriented programming language to communicate with the computer system. However, if desired, the program can be implemented in assembly or machine language. In any case, the language can be a compiled or interpreted language. Furthermore, for this purpose, the program can run on a programmed application-specific integrated circuit (ASIC).

[0063] Furthermore, the procedures described in this embodiment can be performed in any suitable order unless otherwise indicated by this embodiment or clearly contradicted by the context. The procedures (or variations and / or combinations thereof) described in this embodiment can be executed under the control of one or more computer systems configured with executable instructions, and can be implemented by hardware or a combination thereof as code (e.g., executable instructions, one or more computer programs, or one or more applications) that commonly executes on one or more processors. A computer program includes multiple instructions executable by one or more processors.

[0064] Furthermore, the method can be implemented in any suitable type of computing platform, including but not limited to personal computers, minicomputers, mainframes, workstations, networked or distributed computing environments, standalone or integrated computer platforms, or in communication with charged particle tools or other imaging devices, etc. Aspects of the invention can be implemented as machine-readable code stored on a non-transitory storage medium or device, whether removable or integrated into a computing platform, such as a hard disk, optical read and / or write storage medium, RAM, ROM, etc., such that it is readable by a programmable computer, and when the storage medium or device is read by the computer, it can be used to configure and operate the computer to perform the processes described herein. Furthermore, the machine-readable code, or portions thereof, can be transmitted via wired or wireless networks. The invention of this embodiment includes these and other different types of non-transitory computer-readable storage media when such media comprises instructions or programs that implement the steps above in conjunction with a microprocessor or other data processor. When programmed according to the methods and techniques of the invention, the invention also includes the computer itself.

[0065] A computer program can be applied to input data to perform the functions of this embodiment, thereby transforming the input data to generate output data stored in non-volatile memory. The output information can also be applied to one or more output devices, such as a display. In a preferred embodiment of the invention, the transformed data represents physical and tangible objects, including specific visual depictions of physical and tangible objects generated on the display.

[0066] The above are merely preferred embodiments of the present invention. The present invention is not limited to the above-described embodiments. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention, as long as they achieve the technical effects of the present invention by the same means, should be included within the scope of protection of the present invention. Within the scope of protection of the present invention, the technical solutions and / or implementation methods can have various modifications and variations.

Claims

1. A method for identifying application energy consumption based on graph convolutional networks, characterized in that, The application energy consumption identification method based on graph convolutional networks includes: Obtain the energy consumption of user terminals, network slices, and applications in the communication network, and construct an energy consumption matrix; Construct an undirected topology graph using user terminals, network slices, and applications in the communication network as nodes; Based on the undirected topological graph, construct the self-connected adjacency matrix and degree matrix; The energy consumption matrix, the self-connection adjacency matrix, and the degree matrix are input into a graph convolutional network for processing. Obtain the application energy consumption identification results output by the graph convolutional network; The step of obtaining the energy consumption of user terminals, network slices, and applications in the communication network, and constructing an energy consumption matrix, includes: Over a period of time, energy consumption is detected for several user terminals, several network slices, and several applications respectively to obtain their respective energy consumption time series; The energy consumption matrix is ​​formed by using the energy consumption time series of each user terminal, the energy consumption time series of each network slice, and the energy consumption time series of each application as rows or columns, respectively. The construction of an undirected topology graph using user terminals, network slices, and applications in the communication network as nodes includes: Each of the aforementioned user terminals, each of the aforementioned network slices, and each of the aforementioned applications are respectively used as nodes; When any of the user terminals is connected to any of the network slices, an edge is established between the user terminal and the node corresponding to the network slice. When any of the network slices runs any of the applications, an edge is established between the network slice and the node corresponding to the application. Each node and each edge is used as the undirected topology graph; The step of constructing a self-connected adjacency matrix and a degree matrix based on the undirected topological graph includes: Construct an adjacency matrix based on the undirected topological graph; Based on the adjacency matrix, obtain the corresponding self-connected adjacency matrix and the degree matrix.

2. The application energy consumption identification method based on graph convolutional networks according to claim 1, characterized in that, The step of performing energy consumption detection on several user terminals, several network slices, and several applications includes: Detect the total number of actual nodes; the actual nodes include actual running user terminals, network slices, and applications; When the total number of actual nodes is less than the number corresponding to the size of the graph convolutional network, test nodes are run; the test nodes include user terminals, network slices and / or applications used for testing; Energy consumption is measured for each of the actual nodes and each of the test nodes.

3. The application energy consumption identification method based on graph convolutional networks according to claim 2, characterized in that, The running test node includes: The number of test nodes is determined based on the difference between the total number of actual nodes and the number corresponding to the size of the graph convolutional network. Run the corresponding number of test nodes.

4. The application energy consumption identification method based on graph convolutional networks according to claim 2, characterized in that, The running test node includes: Obtain the first average energy consumption of all the actual nodes; Determine the application's energy consumption control targets; A second average energy consumption is determined based on the degree of deviation of the first average energy consumption from the application's energy consumption control target; The energy consumption of each test node is determined based on the second average energy consumption. The test node is operated according to the determined energy consumption.

5. The application energy consumption identification method based on graph convolutional networks according to any one of claims 1-4, characterized in that, The application energy consumption identification method based on graph convolutional networks also includes: Obtain the recognition result request sent by the user terminal; In response to the identification result request, the application energy consumption identification result is sent to the user terminal.

6. A computer device, characterized in that, The method includes a memory and a processor, wherein the memory is used to store at least one program and the processor is used to load at least one program to execute the application energy consumption identification method based on graph convolutional networks according to any one of claims 1-5.

7. A computer-readable storage medium storing a processor-executable program, characterized in that, A processor-executable program, when executed by a processor, is used to perform the application energy consumption identification method based on graph convolutional networks according to any one of claims 1-5.

Images