Inter-cell interference modeling method based on weighted graph convolutional network

Abstract

The invention relates to an inter-cell interference modeling method based on a weighted graph convolutional network, and belongs to the technical field of wireless communication. According to the method, related data of a wireless network and a graph neural network are combined, and a core problem of how to accurately model inter-cell interference of all wireless access points in a region according to real-time performance data and engineering parameters of the wireless access points is solved. In order to solve the problem, influence factors of inter-cell interference are analyzed, and a weighted graph convolutional network combining real-time wireless resources and a frequency band coincidence rate is designed, so that the inter-cell interference in a multi-wireless access point scene in a cellular network is accurately obtained.

Description

Technical Field

[0001] The present invention belongs to the technical field of wireless communications, and in particular relates to an inter-cell interference modeling method based on a weighted graph convolutional network. Background Art

[0002] Predicting and modeling interference at wireless access points is of great significance and can significantly improve the performance and management efficiency of wireless network. First, by predicting interference, data transmission rate can be increased and bit error rate can be reduced, thus improving network performance. Optimizing spectrum utilization is another major benefit. Interference prediction allows efficient spectrum allocation and dynamic spectrum management to avoid resource waste and conflicts. Improving user experience is key. By reducing dropped calls and improving service quality, users can enjoy more stable and high-quality communication services. In addition, interference prediction enhances network reliability, prevents potential problems and responds to failures quickly, providing efficient diagnosis and recovery measures. Saving energy consumption is also an important advantage. Optimizing power control can reduce the energy consumption of wireless access points and terminal devices, thereby reducing operating costs. The main reason for modeling interference at wireless access points is to gain a deep understanding of the interference sources and characteristics. Through modeling, the main interference sources can be identified, the time-frequency characteristics and spatial distribution of interference can be analyzed, and a scientific basis can be provided for the formulation of interference suppression measures. Modeling can also optimize network planning and design, reasonably layout wireless access points, avoid areas with severe interference, and formulate scientific spectrum planning strategies. Supporting dynamic resource management is another important reason. Interference modeling allows real-time adjustment of network parameters and load balancing to optimize network resource utilization. Modeling also improves the effectiveness of interference suppression technology by guiding interference suppression strategies and optimizing algorithms to improve the efficiency and effectiveness of interference suppression.

[0003] With the development of big data technology and artificial intelligence, deep neural networks have emerged in the field of wireless communications, such as wireless traffic prediction, network anomaly detection, and wireless network performance optimization. It can autonomously learn complex nonlinear relationships from a large amount of real data, further reducing the complexity of system design while also having a certain generalization ability. These advantages enable deep neural networks to provide more accurate wireless access point inter-cell interference modeling for wireless communication systems. Using graph convolutional networks (GCNs) in deep neural networks to model wireless access point interference has key advantages. GCN can effectively handle complex interference relationships between wireless access points, and achieve accurate interference prediction by learning features such as the spatial layout and interference intensity of wireless access points. GCN achieves effective information aggregation through multi-layer convolution operations, can dynamically update the model to adapt to changes in network topology, and supports real-time interference management and spectrum allocation optimization. Ultimately, GCN modeling not only optimizes interference suppression strategies, improves network performance and user experience, but also effectively utilizes spectrum resources to promote the intelligence and sustainable development of wireless networks. Summary of the invention

[0004] The purpose of the present invention is to provide a method for modeling inter-cell interference based on a weighted graph convolutional network, which uses a weighted graph convolutional network in deep learning to solve the problem of interference prediction of different wireless access points in a multi-wireless access point scenario.

[0005] To achieve the above object, the technical solution of the present invention is: a method for modeling inter-cell interference based on a weighted graph convolutional network, comprising the following steps:

[0006] Step S1, constructing a cellular network scenario with multiple wireless access points; combining wireless network related data, screening and defining key parameters of each wireless access point;

[0007] Step S2: constructing graph structure data required for the interference model to accurately model the interference model for the interference;

[0008] Step S3, construct a cell-to-cell interference model based on a weighted graph convolutional network. The cell-to-cell interference model based on a weighted graph convolutional network takes graph structure data as input and outputs cell-to-cell interference parameters through a multi-layer graph convolutional network with weights. The cell-to-cell interference model based on a weighted graph convolutional network is trained with the help of real data until convergence.

[0009] In an embodiment of the present invention, in step S1, the key parameters of each wireless access point include engineering parameters and real-time parameters of each wireless access point.

[0010] In one embodiment of the present invention, in step S2, the graph structure data includes node features and structure features, which are calculated through engineering parameters and real-time parameters.

[0011] In one embodiment of the present invention, step S1 specifically includes the following steps:

[0012] Step S11, defining the parameters of the wireless access points in the cellular network; assuming that there are M wireless access points in the cellular network scenario, the engineering parameter D of the mth wireless access point m Includes rated resources R m , center carrier frequency signal f m and position coordinates (x m ,y m ), rated resource R m Including rated bandwidth B m and the maximum transmit power P m ; The real-time parameters of the mth wireless access point in the time interval t include the average occupancy rate of the physical resource block PRB Maximum interference level of cell resource block RB and average interference level

[0013] Step S12: Determine the key parameters required for the interference model:

[0014] The cumulative interference received by any user u within the range of the wireless access point is

[0015]

[0016] in, is the distance between the user and the interfering wireless access point, h mu is the channel between the wireless access point and the user, then the interference to the wireless access point m is expressed as

[0017]

[0018] Without considering the user, it can be seen from the above formula that the interference to the wireless access point depends on the parameters of other wireless access points sharing the same channel with it, including the transmission power, cell bandwidth and spatial location information of the wireless access point; therefore, through f m Ability to determine the set of co-frequency wireless access points that have interference, through and R m Ability to determine real-time bandwidth of wireless access points The interference model is built in combination with other parameters in step S11.

[0019] In one embodiment of the present invention, in step S2, since the relationship between wireless access points in the cellular network can be naturally represented as a graph structure, in which each wireless access point is a node, and the edges connecting them can represent the spatial position relationship and the interference relationship; therefore, the graph structure data of the cellular network scene is constructed according to the key parameters in step S1, and the node characteristics of the graph structure data are the real-time bandwidth resources of the wireless access point. and the maximum transmit power P m , that is, each wireless access point node has a two-dimensional feature vector; the structural characteristics of the graph structure data are calculated by the distance between wireless access points and the frequency band overlap rate.

[0020] In one embodiment of the present invention, the specific method of calculating the structural features of the graph structure data by the distance between wireless access points and the frequency band overlap rate is as follows:

[0021] Considering the real wireless communication scenario of acquiring data, the wireless access point uses the idle spectrum within the bandwidth range for random allocation. Assuming that the physical resource blocks PRB used by the wireless access point at each moment obey uniform distribution, the probability of frequency band overlap between the wireless access point m and other wireless access points with the same frequency is expressed as

[0022]

[0023] By analyzing all wireless access points in the cellular network scenario through the above formula, a frequency band overlap probability matrix at time t can be derived, which is expressed as

[0024]

[0025] Among them, the diagonal elements are all set to 0, that is, the inter-cell interference does not take into account the internal interference of the wireless access point itself; Finally, the above formula is combined with the spatial position relationship to obtain the weighted adjacency matrix A t , where the elements Expressed as

[0026]

[0027] in, is the distance between wireless access points, μ is the path loss index; the matrix A is derived from the above formula t , and use it as the structural feature of graph structured data.

[0028] In one embodiment of the present invention, step S3 specifically includes the following steps:

[0029] Step S31, determine the inter-cell interference model; assume that the inter-cell interference follows a normal distribution within each hour, therefore, combine the average interference matrix composed of all wireless access points The interference standard deviation is defined as The calculation formula is Further determine the interference parameters of the model's output, namely From step S1, it can be seen that the interference of the wireless access point is closely related to the engineering parameters and real-time parameters of other wireless access points with the same frequency. Therefore, the predicted interference parameters are expressed as

[0030]

[0031] in, represents the bandwidth matrix allocated to the wireless access point at time t, It is a matrix composed of the maximum transmit power of all wireless access points; It is an inter-cell interference model, which can calculate the average interference and interference standard deviation of wireless access point m according to the real-time frequency band range and engineering parameters of all wireless access points in the area;

[0032] Step S32: construct an inter-cell interference model based on a weighted graph convolutional network; taking the inter-cell interference model as the target, the graph structure data derived in step S2 is used as the input X of the weighted graph convolutional network. t , a multi-layer weighted graph convolution is used to construct the interference model, and its hierarchical propagation law is as follows:

[0033] H(l+1) =λ(A t H (l) W (l) )

[0034] Among them, W (l) is the weight matrix of layer l, H (l) is the node feature matrix of the previous layer. When l = 0, H (0) =X t , λ(·) is the activation function. Since the inter-cell interference is not only affected by the wireless access points in the area, but there are still adjacent wireless access points with the same frequency outside the node area, the normalized adjacency matrix is ​​not used;

[0035] Step S33, design the loss function of the weighted graph convolutional network; take the graph structure data as input, pass through the multi-layer weighted graph convolutional network, and finally output the high-dimensional features through a pooling layer to obtain interference parameters and σ t ; The network training process is supervised learning, using existing data sets for training, and the loss function is defined as:

[0036]

[0037] in, is the inter-cell interference model; after network training through the above formula, a reliable inter-cell interference model based on weighted graph convolutional network can be obtained.

[0038] The present invention also provides an inter-cell interference modeling system based on a weighted graph convolutional network, comprising a memory, a processor, and computer program instructions stored in the memory and capable of being executed by the processor. When the processor executes the computer program instructions, the method steps described above can be implemented.

[0039] The present invention also provides a computer-readable storage medium, on which computer program instructions that can be executed by a processor are stored. When the processor executes the computer program instructions, the method steps described above can be implemented.

[0040] The present invention also provides an inter-cell interference modeling device based on a weighted graph convolutional network, comprising:

[0041] The first module builds a cellular network scenario with multiple wireless access points; combines wireless network related data to screen and define key parameters of each wireless access point;

[0042] The second module builds the graph structure data required for the interference model to accurately model the interference model for interference;

[0043] The third module builds an inter-cell interference model based on a weighted graph convolutional network. The inter-cell interference model based on a weighted graph convolutional network takes graph structure data as input and outputs inter-cell interference parameters through a multi-layer graph convolutional network with weights. The inter-cell interference model based on a weighted graph convolutional network is trained with the help of real data until convergence.

[0044] Compared with the prior art, the present invention has the following beneficial effects: The method of the present invention combines "wireless network" related data and graph neural network to solve a core problem: how to accurately model the inter-cell interference of all wireless access points in the area based on the real-time performance data and engineering parameters of the wireless access points. In response to this problem, the present invention analyzes the influencing factors of inter-cell interference, designs a weighted graph convolutional network that combines real-time wireless resources and frequency band overlap rate, and uses the weighted graph convolutional network in deep learning to solve the interference prediction problem of different wireless access points in multi-wireless access point scenarios, so as to accurately obtain the inter-cell interference in multi-wireless access point scenarios in cellular networks. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 The diagram is a schematic diagram of a multi-wireless access point scenario using inter-cell interference modeling in an example of the present invention.

[0046] Figure 2 FIG. 4 is a diagram of an interference model architecture based on WGCN in an example of the present invention.

[0047] Figure 3 A schematic diagram of real-time losses in the process of training interference models using different graph neural networks in an example of the present invention.

[0048] Figure 4 It is a schematic diagram of interference parameters and real data output by the interference model of different benchmark algorithms compared with the algorithm proposed in the example of the present invention. DETAILED DESCRIPTION

[0049] The technical solution of the present invention is described in detail below in conjunction with the accompanying drawings.

[0050] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present application belongs.

[0051] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0052] See also Figure 1-4 As shown, the present invention provides a method for modeling inter-cell interference based on a weighted graph convolutional network, and the implementation process includes the following steps:

[0053] Step S1, construct a cellular network scenario with multiple wireless access points; combine "wireless network" related data, screen and define key parameters of each wireless access point, including engineering parameters and real-time parameters of each wireless access point.

[0054] Step S2: construct graph structure data required for the interference model; the graph structure data includes node features and structure features, which are calculated through engineering parameters and real-time parameters to accurately model the interference.

[0055] Step S3: construct an inter-cell interference model based on a weighted graph convolutional network; the model takes graph structure data as input and outputs inter-cell interference parameters through a multi-layer graph convolutional network with weights. The network model is trained with real data until convergence.

[0056] Step S1 specifically includes the following steps:

[0057] Step S11, defining the parameters of the wireless access points in the cellular network; assuming that there are M wireless access points in the scenario, the engineering parameter D of the mth wireless access point m Includes rated resources R m (Including rated bandwidth B m and the maximum transmit power P m ), center carrier frequency signal f m and position coordinates (x m ,y m ); The real-time parameters of the mth wireless access point in the time interval t include the average occupancy rate of the physical resource block (PRB) Maximum interference level of cell resource block (RB) and average interference level

[0058] Step S12, determine the key parameters required for the interference model; combined with theoretical analysis, it can be known that the cumulative interference received by any user u within the range of the wireless access point is

[0059]

[0060] Where d is the distance between the user and the interfering wireless access point, h mu is the channel between the interfering wireless access point and the user. Then the interference to the wireless access point m can be expressed as

[0061]

[0062] Without considering the user, it can be seen from the above formula that the interference to the wireless access point depends on the parameters of other wireless access points that share the same channel with it, including the transmission power, cell bandwidth and spatial location information of the wireless access point. Therefore, through f m Ability to determine the set of co-frequency wireless access points that have interference, through and R m Ability to determine real-time bandwidth of wireless access points Combined with other parameters in S11 for interference modeling.

[0063] Step S2 is specifically implemented as follows:

[0064] Graph structure data required for model construction; since the relationship between wireless access points in a mobile network can be naturally represented as a graph structure, where each wireless access point is a node and the edges connecting them can represent spatial position relationships and interference relationships. Therefore, the graph structure of the cellular network scenario is constructed based on the key parameters in S1, and its node features are the real-time bandwidth resources of the wireless access point. and the maximum transmit power P m , that is, each wireless access point node has a two-dimensional feature vector; the structural feature is calculated by the distance between wireless access points and the frequency band overlap rate. Specifically, considering the real wireless communication scenario in which data is obtained, the wireless access points use the idle spectrum within the bandwidth range for random allocation. Assuming that the PRB used by the wireless access point at each moment follows a uniform distribution, the frequency band overlap probability between the wireless access point m and other wireless access points with the same frequency is expressed as

[0065]

[0066] By analyzing all wireless access points in the scene through the above formula, we can derive a frequency band overlap probability matrix at time t, which is expressed as

[0067]

[0068] The diagonal elements are all set to 0, that is, the inter-cell interference does not take into account the internal interference of the wireless access point itself. Finally, the above formula is combined with the spatial position relationship to obtain the weighted adjacency matrix A t , where the elements It can be expressed as

[0069]

[0070] where d m,n is the distance between wireless access points, μ is the path loss index. The matrix A is derived from the above formula t , and use it as the structural feature of graph data.

[0071] Step S3 specifically includes the following steps:

[0072] Step S31, determine the inter-cell interference model; assume that the inter-cell interference follows a normal distribution within each hour, therefore, combine the average interference matrix composed of all wireless access points The interference standard deviation is defined as The calculation formula is Further determine the interference parameters of the model's output, namely From step S1, it can be seen that the interference of the wireless access point is closely related to the engineering parameters and real-time parameters of other wireless access points with the same frequency. Therefore, the predicted interference parameters are expressed as

[0073]

[0074] in, represents the bandwidth matrix allocated to the wireless access point at time t, It is a matrix composed of the maximum transmit power of all wireless access points; It is an inter-cell interference model, which can calculate the average interference and interference standard deviation of wireless access point m according to the real-time frequency band range and engineering parameters of all wireless access points in the area;

[0075] Step S32: construct an interference model based on a weighted graph convolutional neural network; taking the system model as the target, the graph structure data derived in S2 is used as the input X of the weighted graph convolutional network. t , a multi-layer weighted graph convolution is used to construct the interference model, and its hierarchical propagation law is as follows

[0076] H (l+1) =λ(A t H (l) W (l) )

[0077] Where W is the weight matrix of layer l, H (l) is the node feature matrix of the previous layer. When l = 0, H(0) =X t , λ(·) is the activation function. Since the inter-cell interference is not only affected by the wireless access points in the area, but also by the adjacent wireless access points with the same frequency outside the node area, the normalized adjacency matrix is ​​not used.

[0078] Step S33, design the loss function of the weighted graph convolutional network; take the graph structure data as input, pass through the multi-layer weighted graph convolutional network, and finally output the high-dimensional features through a pooling layer to obtain interference parameters and σ t The network training process is supervised learning, using existing data sets for training, and the loss function is defined as

[0079]

[0080] in is the inter-cell interference model. After network training through the above formula, a reliable inter-cell interference model can be obtained.

[0081] In order to allow those skilled in the art to further understand the inter-cell interference modeling based on weighted graph convolutional network proposed by the present invention, a detailed description is given below in conjunction with a specific embodiment. This embodiment is implemented based on the technical solution of the present invention.

[0082] like Figure 1 As shown, it is a schematic diagram of a multi-wireless access point scenario using inter-cell interference modeling.

[0083] The model mainly includes multiple independent wireless access points serving different users and devices. Each wireless access point has different engineering parameters. Combined with the different real-time data of each wireless access point collected by the background, interference modeling is performed through the WGCN network to obtain the interference value of each wireless access point.

[0084] like Figure 2 As shown, it is an interference model architecture diagram based on WGCN.

[0085] The algorithm uses a weighted graph convolutional neural network, including node features, structural features, weighted graph convolution layers and pooling layers. It takes graph post-section data as input, its node features are the real-time resources and transmission power of the same-frequency wireless access points, and its structural features are the adjacency matrix calculated based on engineering parameters and real-time PRB occupancy, and then outputs the average interference and interference standard deviation of all wireless access points.

[0086] like Figure 3 As shown, it is a schematic diagram of real-time loss during the training process of the interference calculation model.

[0087] In order to compare the performance of interference models constructed by different graph neural network algorithms, weighted graph convolutional networks, graph convolutional networks, and graph isomorphic networks were trained in the same test data set and the corresponding losses were plotted. All networks here use mean square error as the loss function. The horizontal axis of the graph is the training round and the vertical axis is the loss. In the figure, compared with other algorithms, the proposed WGCN shows a gentle and stable downward trend during the training process, and obtains the lowest loss value after training convergence, achieving the best performance. Therefore, this algorithm is used for inter-cell interference modeling.

[0088] like Figure 4 As shown, it is a comparison chart between the interference parameters output by the wireless access point through different interference models and the actual data.

[0089] The experimental figure shows the average interference and interference standard deviation predicted by different algorithms for interference modeling of a wireless access point in the scenario within one week. The dark blue solid line in the figure represents the real interference, and the red solid line and the yellow solid line represent the output interference of WGCN and GCN respectively. The output interference parameters of the proposed WGCN and GCN are close to the real data, but the performance of WGCN is still better than that of GCN, and the proposed model can still output the average interference and interference standard deviation more accurately. In addition, the average interference of the two wireless access points has a certain periodicity, that is, the interference value is usually higher in the afternoon and evening of the day. These moments often correspond to the peak period of mobile network use. When the load is high, the wireless access point will schedule more resources, resulting in greater interference between adjacent wireless access points with the same frequency. Comparing the two parameters, the average interference changes very little over time. The average interference change of the wireless access point in the figure remains within 4dBm, while the change of the interference standard deviation is more intense, indicating that even in the case of good interference level, some unavoidable high interference may still occur. In summary, based on spatial location information, WGCN improves the modeling accuracy of inter-cell interference parameters by using the frequency band overlap rate between wireless access points as a structural feature, making the algorithm achieve better performance than other algorithms.

[0090] The present invention also provides a computer-readable storage medium, on which computer program instructions that can be executed by a processor are stored. When the processor executes the computer program instructions, the method steps described above can be implemented.

[0091] The present invention also provides an inter-cell interference modeling device based on a weighted graph convolutional network, comprising:

[0092] The first module builds a cellular network scenario with multiple wireless access points; combines wireless network related data to screen and define key parameters of each wireless access point;

[0093] The second module builds the graph structure data required for the interference model to accurately model the interference model for interference;

[0094] The third module builds an inter-cell interference model based on a weighted graph convolutional network. The inter-cell interference model based on a weighted graph convolutional network takes graph structure data as input and outputs inter-cell interference parameters through a multi-layer graph convolutional network with weights. The inter-cell interference model based on a weighted graph convolutional network is trained with the help of real data until convergence.

[0095] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0096] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. 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 processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device 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.

[0097] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising 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.

[0098] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0099] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any technician familiar with the profession may use the above disclosed technical content to change or modify it into an equivalent embodiment with equivalent changes. However, any simple modification, equivalent change and modification made to the above embodiment according to the technical essence of the present invention without departing from the technical solution of the present invention still belongs to the protection scope of the technical solution of the present invention.

Claims

1. A method for modeling inter-cell interference based on a weighted graph convolutional network, characterized in that: The steps include: Step S1, constructing a cellular network scenario with multiple wireless access points; combining wireless network related data, screening and defining key parameters of each wireless access point; Step S2: constructing graph structure data required for the interference model to accurately model the interference model for the interference; Step S3, construct a cell-to-cell interference model based on a weighted graph convolutional network. The cell-to-cell interference model based on a weighted graph convolutional network takes graph structure data as input and outputs cell-to-cell interference parameters through a multi-layer graph convolutional network with weights. The cell-to-cell interference model based on a weighted graph convolutional network is trained with the help of real data until convergence.

2. According to claim 1, a method for modeling inter-cell interference based on a weighted graph convolutional network is characterized in that: In step S1, the key parameters of each wireless access point include engineering parameters and real-time parameters of each wireless access point.

3. According to claim 2, a method for modeling inter-cell interference based on a weighted graph convolutional network is characterized in that: In step S2, the graph structure data includes node features and structure features, which are calculated through engineering parameters and real-time parameters.

4. The method for modeling inter-cell interference based on weighted graph convolutional network according to claim 1, characterized in that: Step S1 specifically includes the following steps: Step S11, defining the parameters of the wireless access points in the cellular network; assuming that there are M wireless access points in the cellular network scenario, the engineering parameter D of the mth wireless access point m Includes rated resources R m , center carrier frequency signal f m and position coordinates (x m ,y m ), rated resource R m Including rated bandwidth B m and the maximum transmit power P m ; The real-time parameters of the mth wireless access point in the time interval t include the average occupancy rate of the physical resource block PRB Maximum interference level of cell resource block RB and average interference level Step S12: Determine the key parameters required for the interference model: The cumulative interference received by any user u within the range of the wireless access point is in, is the distance between the user and the interfering wireless access point, h mu is the channel between the wireless access point and the user, then the interference to the wireless access point m is expressed as Without considering the user, it can be seen from the above formula that the interference to the wireless access point depends on the parameters of other wireless access points sharing the same channel with it, including the transmission power, cell bandwidth and spatial location information of the wireless access point; therefore, through f m Ability to determine the set of co-frequency wireless access points that have interference, through and R m Ability to determine real-time bandwidth of wireless access points The interference model is built in combination with other parameters in step S11.

5. The method for modeling inter-cell interference based on weighted graph convolutional network according to claim 4, characterized in that: In step S2, since the relationship between wireless access points in the cellular network can be naturally represented as a graph structure, in which each wireless access point is a node, and the edges connecting them can represent the spatial position relationship and interference relationship; therefore, the graph structure data of the cellular network scene is constructed according to the key parameters in step S1, and the node characteristics of the graph structure data are the real-time bandwidth resources of the wireless access point. and the maximum transmit power P m , that is, each wireless access point node has a two-dimensional feature vector; the structural characteristics of the graph structure data are calculated by the distance between wireless access points and the frequency band overlap rate.

6. The method for modeling inter-cell interference based on weighted graph convolutional network according to claim 5, characterized in that: The specific way to calculate the structural features of graph structure data through the distance between wireless access points and the frequency band overlap rate is as follows: Considering the real wireless communication scenario of acquiring data, the wireless access point uses the idle spectrum within the bandwidth range for random allocation. Assuming that the physical resource blocks PRB used by the wireless access point at each moment obey uniform distribution, the probability of frequency band overlap between the wireless access point m and other wireless access points with the same frequency is expressed as By analyzing all wireless access points in the cellular network scenario through the above formula, a frequency band overlap probability matrix at time t can be derived, which is expressed as Among them, the diagonal elements are all set to 0, that is, the inter-cell interference does not take into account the internal interference of the wireless access point itself; Finally, the above formula is combined with the spatial position relationship to obtain the weighted adjacency matrix A t , where the elements Expressed as in, is the distance between wireless access points, μ is the path loss index; the matrix A is derived from the above formula t , and use it as the structural feature of graph structured data.

7. The method for modeling inter-cell interference based on weighted graph convolutional network according to claim 6, characterized in that: Step S3 specifically includes the following steps: Step S31, determine the inter-cell interference model; assume that the inter-cell interference follows a normal distribution within each hour, therefore, combine the average interference matrix composed of all wireless access points The interference standard deviation is defined as The calculation formula is Further determine the interference parameters of the model's output, namely From step S1, it can be seen that the interference of the wireless access point is closely related to the engineering parameters and real-time parameters of other wireless access points with the same frequency. Therefore, the predicted interference parameters are expressed as in, represents the bandwidth matrix allocated to the wireless access point at time t, It is a matrix composed of the maximum transmit power of all wireless access points; It is an inter-cell interference model, which can calculate the average interference and interference standard deviation of wireless access point m according to the real-time frequency band range and engineering parameters of all wireless access points in the area; Step S32: construct an inter-cell interference model based on a weighted graph convolutional network; taking the inter-cell interference model as the target, the graph structure data derived in step S2 is used as the input X of the weighted graph convolutional network. t , a multi-layer weighted graph convolution is used to construct the interference model, and its hierarchical propagation law is as follows: H (l+1) =λ(A t H (l) W (l) ) Among them, W (l) is the weight matrix of layer l, H (l) is the node feature matrix of the previous layer. When l = 0, H (0) =X t , λ(·) is the activation function. Since the inter-cell interference is not only affected by the wireless access points in the area, but there are still adjacent wireless access points with the same frequency outside the node area, the normalized adjacency matrix is ​​not used; Step S33, design the loss function of the weighted graph convolutional network; take the graph structure data as input, pass through the multi-layer weighted graph convolutional network, and finally output the high-dimensional features through a pooling layer to obtain interference parameters and σ t ; The network training process is supervised learning, using existing data sets for training, and the loss function is defined as: in, is the inter-cell interference model; after network training through the above formula, a reliable inter-cell interference model based on weighted graph convolutional network can be obtained.

8. A cell-to-cell interference modeling system based on a weighted graph convolutional network, characterized in that: The method comprises a memory, a processor, and computer program instructions stored in the memory and executable by the processor. When the processor executes the computer program instructions, the method steps as claimed in any one of claims 1 to 7 can be implemented.

9. A computer-readable storage medium storing computer program instructions that can be executed by a processor, wherein when the processor executes the computer program instructions, the method steps according to any one of claims 1 to 7 can be implemented.

10. A device for modeling inter-cell interference based on a weighted graph convolutional network, characterized in that: include: The first module builds a cellular network scenario with multiple wireless access points; Combined with wireless network related data, filter and define key parameters of each wireless access point; The second module builds the graph structure data required for the interference model to accurately model the interference model for interference; The third module builds an inter-cell interference model based on a weighted graph convolutional network. The inter-cell interference model based on a weighted graph convolutional network takes graph structure data as input and outputs inter-cell interference parameters through a multi-layer graph convolutional network with weights. The inter-cell interference model based on a weighted graph convolutional network is trained with the help of real data until convergence.

Images