A star flash-wapi heterogeneous power network fusion networking method and system based on a graph neural network

Abstract

The application discloses a star flash-WAPI heterogeneous power network fusion networking method based on a graph neural network, which realizes efficient fusion of the star flash and the WAPI by optimizing node selection and spectrum resource allocation of terminals. The method considers star flash-WAPI heterogeneous power network fusion networking in a power scene, wherein the power terminal is multi-mode, simultaneously integrates star flash and WAPI communication capabilities, and can access network nodes according to communication requirements; the network node is also multi-mode, and provides the terminal with two communication modes of the star flash and the WAPI; a self-organizing network is constructed based on the multi-mode network node, ensuring the safety, reliability and low latency of data transmission; under the constraint conditions of the upper limit of the number of nodes served in each node and the limited spectrum resource, the target is to maximize the total rate of the system; a network controller collects star flash-WAPI heterogeneous power network condition information and node information and intelligently allocates network nodes and spectrum resources for the terminal by using a graph neural network (GNN) method, so as to maximize the total rate of the system.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of electric power, more particularly, to a star flash-WAPI heterogeneous electric power network fusion networking method and system based on a graph neural network. BACKGROUND

[0002] With the rapid development of wireless communication technology, star flash and WAPI network as two important wireless communication technologies, each plays an important role in different application scenarios.

[0003] Star flash is a new type of wireless short-range communication technology, which has the characteristics of low power consumption, low latency, high speed, high reliability, etc., and is suitable for power network scenarios. While WAPI network is a wireless local area network security technology, which has high security, strong authentication capability and other characteristics, and is widely used in scenarios that require high security requirements such as power network.

[0004] However, with the complexity and diversification of power network scenarios, a single wireless network technology often cannot meet all its needs. Therefore, the fusion of star flash network and WAPI network has become an effective solution. This fusion networking scheme not only makes full use of the low power consumption and low latency advantages of star flash, but also takes advantage of the high security of WAPI network to provide more comprehensive and reliable wireless communication services for power terminals.

[0005] In star flash-WAPI heterogeneous power network fusion networking, spectrum resource allocation and terminal node selection are two key issues. Reasonable spectrum resource allocation can improve spectrum utilization efficiency, reduce network congestion, and improve terminal experience. Optimizing terminal node selection can ensure that terminals can connect to the most suitable network nodes, thereby improving the overall performance of the network. Traditional spectrum resource allocation and terminal node selection methods are often based on fixed rules or simple algorithms, and do not consider the "maximum system rate" principle, making it difficult to adapt to complex network environments and dynamic terminal needs.

[0006] In view of the above problems, there is an urgent need for a star flash-WAPI heterogeneous power network fusion networking method and system based on a graph neural network. SUMMARY

[0007] To solve the problems existing in the prior art, the present application provides a star flash-WAPI heterogeneous power network fusion networking method and system based on a graph neural network.

[0008] The present application adopts the following technical solutions.

[0009] The first aspect of the application relates to a star flash-WAPI heterogeneous power network fusion networking method based on a graph neural network, the method comprising the following steps: a multi-mode network is composed of a network controller, a WAPI authentication server, multi-mode nodes and multi-mode terminals; a multi-mode network model is constructed with the maximum system total rate as the target and the upper limit of node service quantity and the upper limit of spectrum resources as constraints; the multi-mode network model of the star flash-WAPI heterogeneous power network and the network mode division result of the multi-mode terminal and the multi-mode node are solved by using an attention graph neural network, and the network mode of the multi-mode terminal and the multi-mode node is intelligently divided, and the network node of the multi-mode terminal is intelligently allocated.

[0010] The multi-mode network model is constructed with the maximum system total rate as the target and the upper limit of node service quantity and the upper limit of spectrum resources as constraints, including:

[0011] The objective function of the multi-mode network model is:

[0012]

[0013] In the formula, is the terminal whether to select the node The mode provides service on the channel

[0014] is the highest communication rate when .

[0015] The multi-mode network model is constructed with the maximum system total rate as the target and the upper limit of node service quantity and the upper limit of spectrum resources as constraints, including:

[0016] The constraint condition of the multi-mode network model is:

[0017]

[0018]

[0019] In the formula, is the upper limit of node service quantity,

[0020] is the upper limit of spectrum resources,

[0021] is the total number of multi-mode nodes,

[0022] is the total number of multi-mode terminals.

[0023] ​​The attention graph neural network is used for solving the multi-mode network model of the star-flash-WAPI heterogeneous power network and the network mode division result of the multi-mode terminal and multi-mode node, and intelligently dividing the network mode of the multi-mode terminal and multi-mode node, and intelligently allocating the network node of the multi-mode terminal, including:

[0024] The given problem Hamiltonian is used to convert the multi-mode network model into:

[0025]

[0026] In the formula, , is an arbitrary binary decision variable , is a coefficient matrix between arbitrary binary decision variables.

[0027] The attention graph neural network is used for solving the multi-mode network model of the star-flash-WAPI heterogeneous power network and the network mode division result of the multi-mode terminal and multi-mode node, and intelligently dividing the network mode of the multi-mode terminal and multi-mode node, and intelligently allocating the network node of the multi-mode terminal, including: using a GNN solver to solve the multi-mode network model, capturing complex coupling relationships through a multi-layer GATv2Conv, and introducing a JK module to implement multi-scale feature fusion.

[0028] The attention graph neural network is used for solving the multi-mode network model of the star-flash-WAPI heterogeneous power network and the network mode division result of the multi-mode terminal and multi-mode node, and intelligently dividing the network mode of the multi-mode terminal and multi-mode node, and intelligently allocating the network node of the multi-mode terminal, including: the GNN solver includes L-1 layers of GATv2Conv and 1 layer of output layer; a JK module is obtained by using a pooling operation to aggregate, and feature fusion between layers is performed.

[0029] The attention graph neural network is used for solving the multi-mode network model of the star-flash-WAPI heterogeneous power network and the network mode division result of the multi-mode terminal and multi-mode node, and intelligently dividing the network mode of the multi-mode terminal and multi-mode node, and intelligently allocating the network node of the multi-mode terminal, including: the quadratic decision variable is relaxed into a continuous probability parameter , and a differentiable loss function is generated as:

[0030]

[0031] The nodes of the GNN are unsupervisedly trained using the differentiable loss function, and ;

[0032] The greedy algorithm is applied to map the continuous probability parameter back to the optimal quadratic decision variable .

[0033] In a second aspect, the application relates to a star flash-WAPI heterogeneous power network fusion networking system based on a graph neural network, which is implemented by the star flash-WAPI heterogeneous power network fusion networking method based on a graph neural network.

[0034] In a third aspect, the application relates to a terminal, which comprises a processor and a storage medium.

[0035] In a fourth aspect, the application relates to a computer readable storage medium, which stores a computer program.

[0036] Compared with the prior art, the star flash-WAPI heterogeneous power network fusion networking method and system based on a graph neural network can realize efficient fusion of star flash and WAPI networks by dynamically allocating spectrum resources and optimizing terminal node selection, and improve the total system rate.

[0037] The application has the following beneficial effects:

[0038] 1. The method can maximize the total system rate by allocating spectrum resources in wireless backhaul links and wireless access networks.

[0039] 2. The method can intelligently allocate spectrum resources and optimize terminal node selection based on a graph neural network to realize dynamic adaptation of resources in star flash-WAPI heterogeneous power network fusion networking. BRIEF DESCRIPTION OF DRAWINGS

[0040] Figure 1 FIG. 1 is a star flash-WAPI heterogeneous power network fusion networking architecture diagram.

[0041] Figure 2A flowchart of a star flash-WAPI heterogeneous power network fusion networking method based on a graph neural network. DETAILED DESCRIPTION

[0042] To make the purpose, technical solutions and advantages of the present application clearer and more accurate, the technical solutions of the present application are described in detail below through multiple specific embodiments. The embodiments adopted by the present application are only used to explain the present application and are not used to limit the content of the present application.

[0043] Figure 1 A star flash-WAPI heterogeneous power network fusion networking architecture diagram is shown in FIG. 1. Figure 1 In a first aspect of the present application, a star flash-WAPI heterogeneous power network fusion networking method based on a graph neural network is provided, which comprises the following steps:

[0044] Step 201, a multi-mode network is composed of a network controller, a WAPI authentication server, multi-mode nodes and multi-mode terminals.

[0045] A network controller (AC), a WAPI authentication server (AS), multi-mode nodes and multi-mode terminals are deployed in the star flash-WAPI heterogeneous power network. In the power scenario, star flash-WAPI heterogeneous power network fusion networking is carried out, in which the terminal is multi-mode, integrates star flash and WAPI communication capabilities, and can access the network node according to the communication demand. The network node is also multi-mode, and provides the terminal with two communication modes of star flash and WAPI.

[0046] The star flash mode of the node provides the terminal with low-latency and high-reliability wireless short-range communication services through the communication interface of its access layer, and the WAPI mode of the node ensures the security of data transmission through AS identity authentication and encrypted communication.

[0047] The star flash-WAPI heterogeneous power network fusion networking method based on a graph neural network provided by the embodiment of the present application can optimize resource allocation and terminal node selection, improve the system total rate, and meet the requirements of high-performance, high-efficiency and high-reliability networking communication technology in the case of star flash-WAPI complex heterogeneous power network fusion network. The star flash-WAPI heterogeneous power network fusion networking architecture is shown in FIG. 1, which assumes that the star flash-WAPI heterogeneous power network is composed of two network controllers (one main and one backup), a WAPI authentication server (one main and one backup), multi-mode nodes, and multi-mode terminals. The multi-mode terminal can be connected with the node according to the demand. Figure 1

[0048] Step 202, a multi-mode network model is constructed with the maximum system total rate as the target and the upper limit of node service quantity and the upper limit of spectrum resource as the constraint.

[0049] Figure 2 ​The flow chart of the star flash-WAPI heterogeneous power network fusion networking method based on the graph neural network. As shown in Figure 2 , the multi-mode node provides star flash and WAPI communication services for the terminal, ensures the security, reliability and low latency of data transmission, and under the constraints of the upper limit of the number of nodes served and the limited spectrum resources, the goal is to maximize the total rate of the system.

[0050] The multi-mode node is represented as , , the multi-mode terminal is represented as , According to the communication demand, when the terminal has demand for low latency and high reliability, it preferentially accesses the star flash mode of the node; when the terminal has demand for security and identity authentication, it preferentially accesses the WAPI mode of the node.

[0051] The star flash mode of the node provides low latency and high reliability wireless short-range communication services for the terminal through its access layer communication interface, and the WAPI mode of the node ensures the security of data transmission through identity authentication and encrypted communication by AS.

[0052] Each multi-mode node has a channel to provide network services for the terminal. The channel allocation of each terminal and the association of the node are defined as binary variables . indicates that the terminal selects the mode of the node to provide services on the channel , otherwise , wherein the star flash mode of the node is defined as , and the WAPI mode of the node is defined as . The highest communication rate that can be achieved is:

[0053]

[0054] In the formula, is the channel bandwidth, is the signal-to-noise ratio of the terminal, which can be represented as:

[0055]

[0056] wherein, is the transmission power of the node on the channel in the mode, is the channel gain from the node to the terminal , a power of additive white Gaussian noise at the terminal receiver, represents the network mode of the terminal other than the associated terminal transmission causes co-channel interference.

[0057] The AC selects the optimal network node for the terminal and allocates the channel based on the graph neural network, according to the principle of "maximum total rate of the system", that is:

[0058] The objective function of the multi-mode network model is:

[0059]

[0060] wherein, the terminal whether to select the node the mode provides services on the channel ,

[0061] the highest communication rate when .

[0062] The constraint condition of the multi-mode network model is:

[0063]

[0064]

[0065] wherein, is the upper limit of the number of node services,

[0066] is the upper limit of the spectrum resource,

[0067] is the total number of multi-mode nodes,

[0068] is the total number of multi-mode terminals.

[0069] C1 represents the maximum number of terminals that can be associated with a node constraint, C2 represents that a terminal can only be associated with one node, and C3 represents the binary decision variable of the association of the terminal and the node.

[0070] Step 203, using the attention graph neural network to solve the multi-mode network model of the star flash-WAPI heterogeneous power network and the network mode division result of the multi-mode terminal and the multi-mode node, and intelligently dividing the network mode for the multi-mode terminal and the multi-mode node, and intelligently allocating the network node for the multi-mode terminal.

[0071] The AC collection network state information and node information and intelligently allocates network nodes for the terminal by using an attention graph neural network (GNN) method, so that the total rate of the system is maximized and the optimal network node is selected for the terminal.

[0072] The attention graph neural network is used to solve the multi-mode network model of the star flash-WAPI heterogeneous power network and the network mode division result of the multi-mode terminal and the multi-mode node, and intelligently divide the network mode for the multi-mode terminal and the multi-mode node, and intelligently allocate network nodes for the multi-mode terminal, including:

[0073] The given problem Hamiltonian is used to convert the multi-mode network model into:

[0074]

[0075] In the formula, , is an arbitrary binary decision variable , is a coefficient matrix between arbitrary binary decision variables.

[0076] The binary variable defined by the channel allocation of each terminal and the association of the node is represented as a node in the graph, and the association relationship between them is represented by an edge.

[0077] An undirected graph is created, and each binary decision variable is associated with each vertex. If there is an interaction (element in the QUBO matrix) between the decision variable and , an edge is created between the corresponding two points.

[0078] Then, for the given problem Hamiltonian and the graph, a GNN solver is used for solving. This method considers combining the multi-head dynamic attention mechanism with GNN. Since multiple layers of GATv2Conv are needed in the complex graph to capture complex coupling relationships, this method introduces a JK module to solve the over-smoothing problem in deep graph neural networks through multi-scale feature fusion. The neural network model has layers, which are composed of layer GATv2Conv and one output layer, and the feature fusion between layers is performed by the JK module.

[0079] The GNN training strategy is as follows:

[0080] First, a differentiable loss function is generated according to the standard backpropagation requirement, and the following method is used to relax the quadratic decision variable into a continuous probability parameter The relaxation method is applied to the problem Hamiltonian, and a differentiable loss function is generated as follows:

[0081]

[0082]

[0083] The nodes of the GNN are unsupervisedly trained using the differentiable loss function, and ; forward propagation, loss calculation, back propagation and parameter updating are performed until the pre-set iteration stopping condition is met, and the final result is obtained. Finally, after the unsupervised training process is completed, the greedy algorithm is applied to map the continuous probability parameters back to the optimal binary decision variables .

[0084] The second aspect of the application relates to a star flash-WAPI heterogeneous power network fusion networking system based on a graph neural network. The system uses the star flash-WAPI heterogeneous power network fusion networking method based on a graph neural network described in the first aspect of the application. The system is a multi-mode network composed of a network controller, a WAPI authentication server, multi-mode nodes and multi-mode terminals. The system is used to construct a multi-mode network model with the maximum system total rate as the target and the upper limit of node service quantity and the upper limit of spectrum resources as the constraints. The attention graph neural network is used to solve the multi-mode network model of the star flash-WAPI heterogeneous power network and the network mode division result of the multi-mode terminal and the multi-mode node, and the multi-mode terminal and the multi-mode node are intelligently divided into network modes, and the multi-mode terminal is intelligently allocated to network nodes.

[0085] The third aspect of the application relates to a terminal comprising a processor and a storage medium. The storage medium is used to store instructions, and the processor is used to operate according to the instructions to perform the steps of the method described in the first aspect of the application.

[0086] The fourth aspect of the application relates to a computer-readable storage medium having a computer program stored thereon, which is executed by a processor to implement the steps of the method described in the first aspect of the application.

[0087] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and not to limit them. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions of the present application still include modifications or equivalent replacements of the specific embodiments of the present application. Any modification or equivalent replacement without departing from the spirit and scope of the present application should be covered in the protection scope of the claims of the present application.

Claims

1. A method for fusion networking of star-flash-WAPI heterogeneous power networks based on graph neural networks, characterized in that, The method includes the following steps: A multi-mode network consists of a network controller, a WAPI authentication server, multi-mode nodes, and multi-mode terminals. With the goal of maximizing the total system rate and constraints such as the upper limit of the number of node services and the upper limit of spectrum resources, a multi-mode network model is constructed. The network controller collects network status information and node information, and uses an attention graph neural network and a given problem Hamiltonian to solve for the multi-mode network model of the star-flash-WAPI heterogeneous power network and the network mode partitioning results of multi-mode terminals and multi-mode nodes. It intelligently partitions network modes for multi-mode terminals and multi-mode nodes, and intelligently allocates network nodes for multi-mode terminals. The network nodes are multi-mode, providing both star-flash and WAPI communication modes for multi-mode terminals.

2. The method for fusion networking of star-flash-WAPI heterogeneous power networks based on graph neural networks according to claim 1, characterized in that: The construction of a multi-mode network model, with the goal of maximizing the total system rate and constraints such as the upper limit of the number of node services and the upper limit of spectrum resources, includes: The objective function of the multimodal network model is: In the formula, For the terminal Select node? of Mode in channel Provide services online. for The highest communication rate at that time, This indicates that terminal k selects node q in mode s to provide service on channel m, and defines the channel allocation for each terminal and the association between nodes as binary variables. .

3. The method for fusion networking of star-flash-WAPI heterogeneous power networks based on graph neural networks according to claim 2, characterized in that: The construction of a multi-mode network model, with the goal of maximizing the total system rate and constraints such as the upper limit of the number of node services and the upper limit of spectrum resources, includes: The constraints of the multimodal network model are: In the formula, The maximum number of nodes that can be served. This is the upper limit of spectrum resources. The total number of multi-mode nodes. This represents the total number of multi-mode terminals.

4. The method for fusion networking of star-flash-WAPI heterogeneous power networks based on graph neural networks according to claim 3, characterized in that: The method utilizes attention graph neural networks to solve for the multi-mode network model and network pattern partitioning results of the star-flash-WAPI heterogeneous power network, and intelligently partitions network patterns for multi-mode terminals and nodes, and intelligently allocates network nodes for multi-mode terminals, including: Using the given problem Hamiltonian, solving the multimodal network model is transformed into: In the formula, , For any binary decision variable , Let be the coefficient matrix between any binary decision variables.

5. The method for fusion networking of star-flash-WAPI heterogeneous power networks based on graph neural networks according to claim 4, characterized in that: The method of using an attention graph neural network to solve the multi-mode network model and network pattern partitioning results of the star-flash-WAPI heterogeneous power network, and intelligently allocating network nodes for multi-mode terminals, includes: The multi-mode network model is solved using a GNN solver, and complex coupling relationships are captured through a multi-layer GATv2Conv. A JK module is introduced to implement multi-scale feature fusion.

6. The method for fusion networking of star-flash-WAPI heterogeneous power networks based on graph neural networks according to claim 5, characterized in that: The method utilizes attention graph neural networks to solve for the multi-mode network model and network pattern partitioning results of the star-flash-WAPI heterogeneous power network, and intelligently partitions network patterns for multi-mode terminals and nodes, and intelligently allocates network nodes for multi-mode terminals, including: The GNN solver includes an L-1 layer of GATv2Conv and an output layer; it uses pooling operations to aggregate and obtain JK modules for feature fusion between layers.

7. The method for fusion networking of star-flash-WAPI heterogeneous power networks based on graph neural networks according to claim 6, characterized in that: The method utilizes attention graph neural networks to solve for the multi-mode network model and network pattern partitioning results of the star-flash-WAPI heterogeneous power network, and intelligently partitions network patterns for multi-mode terminals and nodes, and intelligently allocates network nodes for multi-mode terminals, including: secondary decision variables Relaxation is a continuous probability parameter The differentiable loss function is generated as follows: Unsupervised training of GNN nodes is performed using a differentiable loss function to obtain... ; Apply a greedy algorithm to the continuous probability parameters Mapping back to optimal quadratic decision variables .

8. A star-flash-WAPI heterogeneous power network convergence networking system based on graph neural networks, characterized in that: The system is implemented using the star-flash-WAPI heterogeneous power network fusion networking method based on graph neural network as described in any one of claims 1-7; The system is a multi-mode network consisting of a network controller, a WAPI authentication server, multi-mode nodes, and multi-mode terminals. The system is used to construct a multi-mode network model with the goal of maximizing the total system rate and constraints such as the upper limit of the number of node services and the upper limit of spectrum resources. Attention graph neural networks are used to solve the multi-mode network model and network pattern partitioning results of multi-mode terminals and multi-mode nodes in the star-flash-WAPI heterogeneous power network. Network patterns are intelligently partitioned for multi-mode terminals and multi-mode nodes, and network nodes are intelligently allocated for multi-mode terminals.

9. A terminal, comprising a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method according to any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the method according to any one of claims 1-7.

Images