Method, device and equipment for handling failure of cross-operator shared network and medium

Abstract

The application provides a fault processing method and device of a cross-operator shared network, equipment and a medium, relates to the technical field of communication, and comprises the following steps: taking each operator physical cell as a graph node, constructing a heterogeneous multi-relationship graph containing same-operator collaborative edges and cross-operator coupling edges according to the operator type of adjacent cells; initializing the graph node, determining the relationship edge weight, constructing a heterogeneous multi-relationship graph neural network and generating a digital twin of the cross-operator shared network; when detecting that the performance of the physical cell is deteriorated, screening suspicious operations that exist potential interference to the deteriorated cell, respectively performing fact scenario and counter-fact scenario simulation deduction in the digital twin, obtaining fact performance prediction values and counter-fact performance prediction values and performing cause-effect analysis; generating a fault processing instruction based on the cause-effect analysis result and delivering the fault processing instruction to the corresponding network to repair the performance of the deteriorated cell. The application can realize cause-effect positioning and closed-loop processing of the fault of the cross-operator shared network.

Description

Technical Field

[0001] This invention relates to the field of communication technology, and specifically to a fault handling method, apparatus, equipment, and medium for cross-carrier shared networks. Background Technology

[0002] With the advancement of large-scale mobile communication network construction, sharing resources such as Active Antenna Unit (AAU) / Remote Radio Unit (RRU) hardware, antennas, site locations, and spectrum among multiple operators has become the mainstream approach to reduce deployment costs and improve coverage efficiency. However, current shared networks mostly adopt a network topology of physical layer sharing and logical independence, which is prone to problems such as cross-domain interference coupling, interconnected faults, and different inductance at the same site.

[0003] The relevant technologies mainly use static parameter configuration, threshold alarm monitoring and manual experience to solve the above problems. However, they are difficult to accurately distinguish the source of the fault, cannot quantify the interference between operations, and have not formed a self-governing fault closed loop, resulting in slow fault location, high misjudgment rate and low operation and maintenance efficiency. Summary of the Invention

[0004] The present invention aims to provide a fault handling method, apparatus, equipment and medium for cross-carrier shared networks, so as to at least solve the problems of slow fault location, high misjudgment rate and low operation and maintenance efficiency in the prior art.

[0005] The technical solution of the present invention to solve the above-mentioned technical problems is as follows: This invention provides a fault handling method for cross-carrier shared networks, comprising: Using the physical cells of each operator as graph nodes, and constructing the relationship edges between adjacent graph nodes based on the operator type of adjacent physical cells, a heterogeneous multi-relationship graph is obtained. The relationship edges include cooperative edges within the same operator and coupling edges across operators. The graph nodes are initialized with information, the edge weights of each relation edge are determined, and a heterogeneous multi-relation graph neural network containing relation graph convolutional layers is constructed. A digital twin of a cross-carrier shared network is generated based on the heterogeneous multi-relation graph neural network. When any physical cell performance degradation is detected, suspicious operations that may interfere with the degraded cell are screened out. In the digital twin, factual scenario simulation and counterfactual scenario simulation are performed on the suspicious operations to obtain the corresponding factual performance prediction value and counterfactual performance prediction value. A causal effect analysis is performed on the suspicious operation based on the factual performance prediction value and the counterfactual performance prediction value. Based on the causal effect analysis results, fault handling instructions are generated and sent to the corresponding network to repair the performance of the degraded cell.

[0006] The technical solution provided by this invention brings at least the following beneficial effects: In the fault handling method for cross-operator shared networks provided by this invention, physical cells of operators are used as graph nodes, and the relationship edges between adjacent graph nodes are constructed according to the type of operator they belong to. This forms a heterogeneous multi-relationship graph adapted to cross-operator shared networks, clearly representing the cooperative relationship and interference coupling relationship between cells of different operators. By initializing graph nodes, determining edge weights, and constructing a heterogeneous multi-relationship graph neural network based on the convolutional layer of the relationship graph, the graph structure can be used to uniformly model the state of multi-operator networks, thereby generating a digital twin consistent with the real network and achieving accurate simulation of the operating state of the shared network. By simulating and deducing factual and counterfactual scenarios of suspicious operations within the digital twin and performing causal effect analysis, key factors leading to cell performance degradation can be identified from suspicious operations, improving the accuracy of fault location. Furthermore, fault handling instructions can be generated and issued for execution based on the causal effect analysis results, realizing autonomous closed-loop processing of faults in cross-operator co-constructed and shared networks.

[0007] Based on the above technical solution, the present invention can be further improved as follows.

[0008] Furthermore, the step of constructing the relationship edges between adjacent graph nodes based on the operator type of adjacent physical cells includes: if the adjacent physical cells belong to the same operator, then constructing the same operator collaborative edge between the corresponding adjacent graph nodes; if the adjacent physical cells belong to different operators, then constructing the cross-operator coupling edge between the corresponding adjacent graph nodes.

[0009] The beneficial effects of this scheme are as follows: By constructing relation edges differently according to the operator type of adjacent physical cells, it is possible to distinguish and model the collaborative relationship within the same operator and the coupling interference relationship between different operators. This makes the heterogeneous multi-relationship graph structure fit the real operating characteristics of cross-operator shared networks, thereby improving the modeling accuracy of subsequent heterogeneous multi-relationship graph neural network models.

[0010] Furthermore, when constructing the heterogeneous multi-relationship graph, the method further includes: if multiple physical cells are affected by the same environmental factor, then a common virtual node corresponding to the environmental factor is introduced into the heterogeneous relationship graph; and an environmental influence edge is constructed between the common virtual node and the graph node corresponding to the cell affected by the environmental factor.

[0011] The beneficial effects of this scheme are as follows: By introducing common virtual nodes and constructing environmental impact edges in a heterogeneous multi-relationship graph, the common interference effect of environmental factors on multiple physical cells can be incorporated into the graph structure modeling, fully reflecting the multiple sources of cell performance degradation in the shared network, and further improving the accuracy of network modeling and fault causal analysis.

[0012] Furthermore, the information initialization of the graph node includes: configuring node attribute information for the graph node, the node attribute information including the cell identifier of the corresponding physical cell and the operator identity identifier to which it belongs; collecting network operation parameters of each physical cell and preprocessing them to obtain structured feature data, and generating a node characteristic vector of the corresponding graph node based on the structured feature data; binding the node attribute information and the node feature vector to the corresponding graph node to complete the information initialization of the graph node.

[0013] The beneficial effects of adopting this scheme are as follows: by configuring cell identifiers and operator identity identifiers, the identity of each graph node can be clearly defined; by converting the preprocessed network operation parameters into node feature vectors, the cell operation status can be quantitatively represented; by binding the two types of information to the corresponding graph nodes, the heterogeneous multi-relationship graph can simultaneously possess identity attributes and operation characteristics, thereby providing complete and standardized data support for subsequent graph convolution operations, network modeling, and fault inference.

[0014] Furthermore, for the cross-carrier coupling edge, determining the edge weight of each of the relationship edges includes: obtaining the number of used physical resource blocks and the transmit power of the coupled adjacent physical cells corresponding to the cross-carrier coupling edge; obtaining the total number of physical resource blocks and the maximum allowed transmit power of the coupled adjacent physical cells; determining the resource load ratio based on the number of used physical resource blocks and the total number of physical resource blocks of the coupled adjacent physical cells; determining the transmit power ratio based on the transmit power and the maximum allowed transmit power of the coupled adjacent physical cells; and dynamically calculating the edge weight of the cross-carrier coupling edge based on the resource load ratio and the transmit power ratio.

[0015] The beneficial effects of this scheme are as follows: By dynamically calculating the cross-operator coupling edge weights by combining the cell resource occupancy and transmission power status, the calculated edge weights can closely match the actual network operation status, truly reflect the degree of interference coupling between cells of different operators, and provide a reliable basis for subsequent graph convolution operations, simulations, and fault analysis.

[0016] Furthermore, the construction of the heterogeneous multi-relation graph neural network including the relation graph convolutional layer includes: using the relation graph convolutional layer as the basic operation layer of the network, building the heterogeneous multi-relation graph neural network based on the heterogeneous multi-relation graph; using the node feature vector as the initial input feature of each graph node, performing weighted graph convolution aggregation operation on the graph nodes based on the type of the relation edge and the edge weight, and completing the construction of the heterogeneous multi-relation graph neural network.

[0017] The beneficial effects of this scheme are as follows: By building a network based on the convolutional layer of the relation graph and combining the relation edge type and edge weight to carry out weighted graph convolutional aggregation, feature transfer and fusion can be completed according to the network topology and correlation strength, fully exploring the correlation features between cells under different types of relations, improving the network's ability to learn the shared network state across operators, and optimizing the representation effect of heterogeneous multi-relation graph neural networks.

[0018] Furthermore, the step of generating a digital twin of a cross-carrier shared network based on the heterogeneous multi-relationship graph neural network includes: real-time acquisition of network operation parameters of the real shared network across carriers, data synchronization of the heterogeneous multi-relationship graph neural network, and obtaining the digital twin consistent with the state of the real shared network through network simulation.

[0019] The beneficial effects of this scheme are as follows: By collecting real network operating parameters in real time and synchronizing the data, the generated digital twin can maintain a high degree of consistency with the real shared network state, realize the dynamic reproduction of the network operating status, and provide a reliable virtual operating environment for subsequent fault simulation, performance prediction and causal analysis.

[0020] Furthermore, the step of performing factual scenario simulation and counterfactual scenario simulation in the digital twin for the suspicious operation to obtain the corresponding factual performance prediction value and counterfactual performance prediction value includes: performing network forward propagation in the digital twin based on all current graph node features and current edge weights to obtain the factual performance prediction value of the degraded cell; rolling back the network parameters corresponding to the suspicious operation in the digital twin to before the operation was executed, recalculating the edge weights of each cross-operator coupling edge based on the rolled-back network parameters, and performing network forward propagation to obtain the counterfactual performance prediction value of the degraded cell.

[0021] The beneficial effects of adopting this scheme are as follows: By conducting simulations of the factual scenario and the counterfactual scenario after parameter rollback, the network state changes before and after the execution of suspicious operations can be accurately simulated, the impact of suspicious operations on cell performance can be intuitively quantified, and objective evidence can be provided for subsequent causal effect analysis, effectively improving the accuracy of fault root cause identification.

[0022] Furthermore, the step of performing causal effect analysis on the suspicious operation based on the factual performance prediction value and the counterfactual performance prediction value includes: calculating the difference between the factual performance prediction value and the counterfactual performance prediction value, and recording it as the causal effect value; if the causal effect value is greater than a preset degradation threshold, then the corresponding suspicious operation is determined to be the target causal operation.

[0023] The beneficial effects of this scheme are as follows: by calculating the difference between two sets of performance prediction values ​​to obtain the causal effect value, and combining it with a preset threshold to determine the target causal operation, the correlation between suspicious operations and cell performance degradation can be quantified, reducing the probability of misjudgment and improving the accuracy of fault location.

[0024] Furthermore, the step of generating a fault handling instruction based on the target causal operation and sending it to the corresponding network includes: constructing a target optimization function based on the performance loss terms of the degraded cell and the interfering cell; the target optimization function includes a first performance loss term of the degraded cell and a second performance loss term of the interfering cell performing the target causal operation; backpropagating the gradient of the target optimization function in the digital twin to obtain a parameter adjustment amount, wherein the parameter adjustment amount is an equilibrium optimization value that minimizes the performance loss of the degraded cell and minimizes the impact on the interfering cell; and generating the fault handling instruction based on the parameter adjustment amount and sending it to the corresponding network.

[0025] The beneficial effects of this approach are as follows: By constructing a target optimization function that includes performance loss terms for both cells, and determining parameter adjustment amounts based on this target optimization function, the operating status of both degraded and interfering cells can be taken into account. This approach can maximize the repair of degradation problems while reducing the negative impact on interfering cells, thus ensuring the overall operational quality of the cross-operator shared network.

[0026] Furthermore, the step of generating fault handling instructions based on the causal effect analysis results and sending them to the corresponding network includes: generating the fault handling instructions based on the causal effect analysis results; executing the fault handling instructions in the digital twin and verifying the execution results; if the verification is successful, sending the fault handling instructions to the corresponding real network to perform performance repair on the degraded cell.

[0027] The beneficial effects of adopting this scheme are as follows: By simulating the execution of fault handling instructions within a digital twin and verifying their success before sending them to the real network, problems in the fault adjustment scheme can be identified in advance, avoiding secondary faults and ensuring the stable operation of the cross-carrier shared network.

[0028] Correspondingly, the present invention provides a fault handling device for a cross-carrier shared network, comprising: The heterogeneous graph construction module is used to use the physical cells of each operator as graph nodes, and construct the relationship edges between adjacent graph nodes based on the operator type of adjacent physical cells to obtain a heterogeneous multi-relationship graph. The relationship edges include cooperative edges within the same operator and coupling edges across operators. The twin construction module is used to initialize the information of the graph nodes, determine the edge weights of each relation edge, and construct a heterogeneous multi-relation graph neural network containing relation graph convolutional layers, and generate a digital twin of a cross-carrier shared network based on the heterogeneous multi-relation graph neural network; The simulation and deduction module is used to filter out suspicious operations that may interfere with the degraded cell when the performance degradation of any of the physical cells is detected. In the digital twin, the module performs factual scenario simulation and deduction and counterfactual scenario simulation and deduction for the suspicious operations respectively to obtain the corresponding factual performance prediction value and counterfactual performance prediction value. The causal analysis module is used to perform causal effect analysis on the suspicious operation based on the factual performance prediction value and the counterfactual performance prediction value. The fault handling module is used to generate fault handling instructions based on the causal effect analysis results and send them to the corresponding network to repair the performance of the degraded cell.

[0029] The present invention also provides an electronic device, comprising: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to execute the instructions to implement the above-described fault handling method for cross-carrier shared networks.

[0030] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described fault handling method for cross-carrier shared networks. Attached Figure Description

[0031] Figure 1 A flowchart illustrating a fault handling method for a cross-carrier shared network provided by the present invention; Figure 2 A topological diagram of heterogeneous multi-relationship graphs in a fault handling method for a cross-carrier shared network provided by the present invention; Figure 3 A schematic diagram of the nodes in a heterogeneous multi-relationship graph in a fault handling method for a cross-carrier shared network provided by the present invention; Figure 4 A schematic diagram of the modeling of the digital twin in a fault handling method for a cross-carrier shared network provided by the present invention; Figure 5 A schematic block diagram of a fault handling device for a cross-carrier shared network provided by the present invention; Figure 6This is a schematic diagram of an electronic device provided by the present invention. Detailed Implementation

[0032] The principles and features of the present invention are described below. The examples given are only for explaining the present invention and are not intended to limit the scope of the present invention.

[0033] The relevant technologies mainly use static parameter configuration, threshold alarm monitoring and manual experience to solve the above problems. However, they are difficult to accurately distinguish the source of the fault, cannot quantify the interference between operations, and have not formed a self-governing fault closed loop, resulting in slow fault location, high misjudgment rate and low operation and maintenance efficiency.

[0034] To address the aforementioned problems, this invention proposes a fault handling method, apparatus, device, and medium for cross-carrier shared networks. The technical solutions of the embodiments of this disclosure are described in detail below: In one embodiment of the present invention, a fault handling method for a cross-carrier shared network is provided. (See reference...) Figure 1 As shown, the fault handling method for this cross-carrier shared network specifically includes the following steps: S110: Using the physical cells of each operator as graph nodes, and constructing the relationship edges between adjacent graph nodes based on the operator type of adjacent physical cells, a heterogeneous multi-relationship graph is obtained. The relationship edges include collaborative edges within the same operator and coupling edges across operators. S120: Initialize the information of the graph nodes, determine the edge weights of each relation edge, and construct a heterogeneous multi-relation graph neural network containing relation graph convolutional layers. Generate a digital twin of the cross-carrier shared network based on the heterogeneous multi-relation graph neural network. S130: When performance degradation of any physical cell is detected, suspicious operations that may interfere with the degraded cell are screened out. In the digital twin, factual scenario simulation and counterfactual scenario simulation are performed for the suspicious operations to obtain the corresponding factual performance prediction value and counterfactual performance prediction value. S140: Conduct causal effect analysis on suspicious operations based on factual performance predictions and counterfactual performance predictions; S150: Generates fault handling instructions based on causal effect analysis results and sends them to the corresponding network to repair the performance of degraded cells.

[0035] In the fault handling method for cross-operator shared networks provided in the above embodiments, the physical cells of operators are used as graph nodes, and the relationship edges between adjacent graph nodes are constructed according to the type of operator they belong to. This can form a heterogeneous multi-relationship graph adapted to cross-operator shared networks, clearly representing the cooperative relationship and interference coupling relationship between cells of different operators. By initializing the graph nodes, determining the edge weights, and constructing a heterogeneous multi-relationship graph neural network based on the convolutional layer of the relationship graph, the graph structure can be used to uniformly model the state of multi-operator networks, thereby generating a digital twin consistent with the real network and realizing accurate simulation of the operating state of the shared network. By simulating and deducing factual and counterfactual scenarios of suspicious operations in the digital twin and performing causal effect analysis, the key factors leading to cell performance degradation can be identified from suspicious operations, improving the accuracy of fault location. Furthermore, fault handling instructions can be generated and issued for execution based on the causal effect analysis results, realizing autonomous closed-loop processing of faults in cross-operator co-constructed and shared networks.

[0036] The above steps will now be described in more detail in another embodiment.

[0037] In S110, physical cells of each operator are used as graph nodes, and the relationship edges between adjacent graph nodes are constructed based on the operator type of adjacent physical cells, resulting in a heterogeneous multi-relationship graph. The relationship edges include collaborative edges within the same operator and coupling edges across operators.

[0038] The aforementioned operators, namely Public Land Mobile Network (PLMN) operators, are operating entities that possess independent spectrum resources, independent core networks, independent radio access networks, and hold communication operation licenses; this embodiment is used to realize fault location and autonomous closed-loop processing of cross-operator shared networks.

[0039] The aforementioned cross-operator shared network refers to a mobile communication network in which two or more operators with independent PLMNs share facilities at the physical layer and operate independently at the logical layer by sharing site locations, antennas, wireless hardware equipment such as AAU / RRU, spectrum resources or transmission resources.

[0040] The aforementioned physical cell is the smallest wireless access and coverage unit formed by the corresponding base station in the mobile communication network through independent radio frequency and carrier.

[0041] The aforementioned heterogeneous multi-relationship graph is a relational topology graph composed of multiple graph nodes and the relationship edges between adjacent graph nodes, used to characterize the mutual influence relationships between physical cells in a cross-carrier shared network.

[0042] The graph nodes mentioned above are topology nodes obtained by abstracting and modeling physical cells in a heterogeneous multi-relationship graph. Each graph node uniquely corresponds to a physical cell.

[0043] The aforementioned relation edges are topological edges that connect adjacent graph nodes in a heterogeneous multi-relation graph, used to characterize the association properties between corresponding adjacent graph nodes.

[0044] For example, the above-mentioned construction of the relationship edges between adjacent graph nodes based on the operator type of adjacent physical cells can be implemented as follows: if adjacent physical cells belong to the same operator, then a same-operator collaborative edge is constructed between the corresponding adjacent graph nodes; if adjacent physical cells belong to different operators, then a cross-operator coupling edge is constructed between the corresponding adjacent graph nodes.

[0045] Among them, the aforementioned cooperating edge with the same operator refers to the relationship edge between the corresponding graph nodes of adjacent physical cells belonging to the same operator, which is used to characterize the cooperative relationship between adjacent physical cells such as coordinated scheduling, neighbor cell handover, and load balancing; for example, if adjacent physical cells all belong to operator A, then the relationship edge between the corresponding graph nodes of the adjacent physical cells is the cooperating edge with the same operator.

[0046] The aforementioned cross-operator coupling edge refers to the relationship edge between graph nodes corresponding to adjacent physical cells belonging to different operators. It is used to characterize the mutual influence of performance such as interference coupling and resource competition between adjacent physical cells due to the sharing of hardware, spectrum, site and other resources. For example, if one adjacent physical cell belongs to operator A and the other belongs to operator B, then the relationship edge between the graph nodes corresponding to the adjacent physical cells is a cross-operator coupling edge.

[0047] In real-world scenarios, network service quality can also be affected by external events (such as heavy rain, concerts, etc.). Therefore, in a preferred embodiment, when constructing a heterogeneous multi-relationship graph, this embodiment further introduces environmental influencing factors to make the subsequently established heterogeneous multi-relationship graph neural network more closely resemble the real network operating state.

[0048] For example, the above process can be implemented as follows: if multiple physical cells are affected by the same environmental factor, a common virtual node corresponding to the environmental factor is introduced into the heterogeneous relationship graph; an environmental influence edge is constructed between the common virtual node and the graph node corresponding to the cell affected by the environmental factor.

[0049] Among them, the aforementioned environmental factors refer to external events that can have a common impact on the signal coverage, network performance, or service quality of multiple physical cells, including influencing factors that are not caused by the network equipment itself, such as weather conditions, large-scale events, and sudden interference.

[0050] The aforementioned public virtual nodes refer to virtual nodes added to the heterogeneous multi-relationship graph to characterize the impact of the aforementioned environmental factors on the relevant physical cells. They do not correspond to real physical cells and are only used to carry environmental interference information and associate all physical cells affected by the corresponding environmental factors.

[0051] The aforementioned environmental impact edges are the relationship edges connecting the public virtual nodes and the corresponding graph nodes of the affected cells, used to characterize the common impact relationship of external environmental factors on physical cells.

[0052] The process of constructing heterogeneous multi-relationship graphs will be described in detail below in a specific implementation of this embodiment: S1101: Obtain topology information of cross-carrier shared networks. Specifically, this involves obtaining information such as the cell identifier, carrier identity, neighbor cell relationship list, and whether AAU / RRU / site address are shared for each physical cell from the network management and configuration platforms of each carrier. S1102: Model each physical cell as a corresponding independent graph node in a heterogeneous multi-relationship graph; S1103: Iterate through each pair of adjacent physical cells and construct the relationship edges between the adjacent graph nodes according to their operator type: If adjacent physical cells belong to the same PLMN and have a neighboring relationship (such as interconnection through the base station interface (Xn interface), then the corresponding adjacent graph nodes are connected by solid lines to obtain the above-mentioned cooperating edge with the same operator. This edge is used to transmit normal cooperation information (such as same frequency handover and load balancing). If two nodes belong to different PLMNs but share the same AAU / RRU hardware (i.e., physical co-construction), then the corresponding adjacent graph nodes are connected by dashed lines to obtain the above cross-operator coupling edge, which is used to characterize the interference coupling relationship between adjacent physical cells. S1104: If multiple physical cells are simultaneously affected by the same environmental factor (such as rainstorms, large-scale events, and other external events), a common virtual node is introduced in the heterogeneous multi-relationship graph to carry the environmental interference information of the external event. The common virtual node is connected to the graph nodes corresponding to each physical cell affected by the external event by a dashed line edge, which is used to model the common environmental interference in the operator's shared network. S1105: Combine all graph nodes with the constructed relation edges to obtain, as shown below. Figure 2 The diagram shown is a heterogeneous multi-relationship diagram corresponding to the cross-carrier shared network topology.

[0053] like Figure 2As shown, the specific implementation described above models the physical cells of different operators' PLMN-A and PLMN-B as graph nodes, and models network interactions through three types of relationship edges: solid lines connecting the same operator's cell within the same PLMN for collaborative edge; dashed lines connecting cells of different operators across PLMNs for coupling edge; and adding a common virtual node Env, and constructing environmental impact edges in the form of dotted lines. This common virtual node is then connected to the graph nodes corresponding to each physical cell affected by external events, thereby constructing a heterogeneous multi-relationship graph covering internal network collaboration, cross-network interference, and external environmental impact.

[0054] In S120, information is initialized for graph nodes, edge weights of each relation edge are determined, and a heterogeneous multi-relation graph neural network containing relation graph convolutional layers is constructed. Based on the heterogeneous multi-relation graph neural network, a digital twin of a cross-operator shared network is generated.

[0055] The edge weights mentioned above refer to the quantized values ​​configured for corresponding relationship edges in a heterogeneous multi-relationship graph, used to quantify the degree of mutual influence between corresponding adjacent graph nodes.

[0056] The aforementioned relational graph convolutional layer is a network operation layer designed for the above-mentioned heterogeneous multi-relational graph. It can aggregate the features of neighboring nodes based on the connection relationship and edge weight of the graph nodes, and mine the correlation patterns and mutual influences between corresponding physical cells.

[0057] The aforementioned heterogeneous multi-relation graph neural network is a deep learning network built by fusing the topology, node features, and edge weights of heterogeneous multi-relation graphs as inputs and integrating convolutional layers of the relation graph. It is used to learn the network features and association rules carried by different graph nodes and relation edges.

[0058] The aforementioned digital twin is a digital mirror image that corresponds one-to-one with the real cross-carrier shared network. Based on the above heterogeneous multi-relationship graph neural network, it can completely reproduce the network topology, physical cell attributes, node association relationships, edge weights and real-time operating status of the real cross-carrier shared network.

[0059] For example, the above-mentioned information initialization of graph nodes can be achieved as follows: configure node attribute information for graph nodes, including the cell identifier of the corresponding physical cell and the operator identity identifier to which it belongs; collect network operation parameters of each physical cell and preprocess them to obtain structured feature data, and generate node characteristic vectors of the corresponding graph nodes based on the structured feature data; bind the node attribute information and node feature vectors to the corresponding graph nodes to complete the information initialization of graph nodes.

[0060] In one specific implementation of this embodiment, the definition and information initialization process of graph nodes can be achieved as follows: each physical cell is modeled as a corresponding graph node v in a heterogeneous multi-relationship graph. i vi This process involves representing graph node i and binding it with its corresponding cell identifier and operator identity identifier to initialize the graph node's identity information. Real-time network operation data is collected from the network management and configuration platforms, network element devices, and user terminals of various operators. For example, the network management and configuration platforms may include an Operation Support System (OSS) and a Business Support System (BSS), the network element devices may be gNodeB base stations, and the network operation data may include key performance indicators (such as throughput, latency, and call drop rate), network configuration parameters (such as frequency, bandwidth, and transmit power), resource usage status (such as PRB utilization), and device alarm logs. The collected multi-source heterogeneous data undergoes time alignment, unit unification, and feature standardization processing to form structured feature data. Based on this structured feature data, node characteristic vectors for the corresponding graph nodes are generated. Complete the initialization of feature information for graph nodes, where x i d represents the node feature vector of node i in the graph, where d is the dimension of the node feature vector, which can be set according to actual needs.

[0061] Specifically, the graph nodes obtained from the above information initialization can be as follows: Figure 3 As shown: Circular nodes Cell ID101 / PLMN-A represent physical cell 101 in operator A, and Cell ID205 / PLMN-B represent physical cell 205 in operator B. Each node is labeled with its cell ID and operator identifier. The rectangle (feature vector box) is used to label the feature vector x corresponding to node i in the graph. i =[KPI,PRB,Power,...]∈R d， Among them, KPI stands for Key Performance Indicators, which reflects the quality of service in the cell; PRB stands for Physical Resource Block Utilization, which reflects the wireless resource load in the cell; and Power stands for Transmit Power, which reflects the radio frequency signal strength and interference level in the cell.

[0062] In real-world scenarios, since the interference coupling relationship between adjacent physical cells represented by the aforementioned cross-carrier coupling edge changes in real time with network operation, in a preferred embodiment, the edge weight of the cross-carrier coupling edge can be dynamically calculated by combining the transmission power and physical resource block occupancy of adjacent physical cells, so that the heterogeneous multi-relationship graph neural network subsequently constructed is more in line with the real network operation.

[0063] For example, the process of dynamically calculating the weight of a cross-operator coupling edge can be implemented as follows: Obtain the number of used physical resource blocks and transmit power of the coupled adjacent physical cells corresponding to the cross-operator coupling edge; obtain the total number of physical resource blocks and the maximum allowed transmit power of the coupled adjacent physical cells; determine the resource load ratio based on the number of used physical resource blocks and the total number of physical resource blocks of the coupled adjacent physical cells; determine the transmit power ratio based on the transmit power and the maximum allowed transmit power of the coupled adjacent physical cells; and dynamically calculate the edge weight of the cross-operator coupling edge based on the resource load ratio and the transmit power ratio. The specific dynamic calculation formula is as follows: in, Let be the edge weight of the cross-operator coupling edge between graph node i and graph node j, where graph node i and graph node j are adjacent graph nodes and belong to different operators; It is an adjustable hyperparameter, with a value range of [value range missing]. It can be set based on actual needs to flexibly adjust the proportion of the impact of resource contention and power interference on coupling strength to adapt to different network scenarios. For example, its value can be increased in high-load scenarios and decreased in interference-dominated scenarios. Let i be the number of physical resource blocks used in the physical cell corresponding to graph node i; Let j be the number of physical resource blocks used in the physical cell corresponding to graph node j; The total number of physical resource blocks in the cell (the maximum number of PRBs that can be allocated to a single cell) is determined by the system bandwidth. It reflects the total resource usage of two cells and is used to quantify the coupling effect caused by load / resource competition. For example, when the PRB utilization of two cells increases at the same time, the competition for resources on the same frequency intensifies, the interference between adjacent channels increases, and the weight of the coupling side increases accordingly. Let i be the current transmit power of the physical cell corresponding to node i in the graph; Let j be the current transmit power of the physical cell corresponding to node j in the graph; The maximum permissible transmit power of the cell (the maximum transmit power of a single cell as defined by the equipment hardware or specifications). It reflects the degree of superposition of signal transmission strength between two cells and is used to quantify the coupling effect caused by radio frequency signal power. For example, when the power of two cells increases at the same time, the signal coverage expands, the adjacent channel / co-channel interference to adjacent cells is enhanced, and the coupling side weight increases accordingly.

[0064] Furthermore, since the correlation between the aforementioned operator-coordinated edge representations is relatively stable, the edge weights can be determined by static setting or periodic calculation, specifically by combining the neighbor cell handover frequency and load sharing ratio; the aforementioned environmental impact edge weights can be set by combining the environmental event level and signal attenuation amplitude.

[0065] After initializing the graph nodes and determining the edge weights of each relation edge through the above process, this embodiment can further implement the above-mentioned construction of a heterogeneous multi-relation graph neural network containing relation graph convolutional layers using the relation graph convolutional layers as the basic operation layers of the network; constructing a heterogeneous multi-relation graph neural network based on the heterogeneous multi-relation graph; using the node feature vectors as the initial input features of each graph node; performing weighted graph convolution aggregation operations on the graph nodes based on the type and weight of the relation edges to complete the construction of the heterogeneous multi-relation graph neural network.

[0066] The process of constructing a heterogeneous multi-relationship graph neural network will be explained in detail below in a specific implementation: This specific implementation uses an improved Relational Graph Convolutional Network (R-GCN) as the base graph neural network layer, relying on a message passing mechanism to complete the iterative update of node features: in, The feature state of graph node i in the (l+1)th layer; Let j be the feature state of graph node j in the l-th layer; It is a set of all relation edge types in a heterogeneous multi-relation graph (this specific implementation includes three types: edge cooperating with the same operator, edge coupling across operators, and edge affecting the environment). This indicates that the summation operation is performed by traversing all relation edge types. This represents summing up all neighboring nodes of node i in the graph for the current relation edge type r. This is the normalization coefficient, used to perform scale normalization on the aggregated features; This is the learnable weight matrix corresponding to the current relation edge type; The message that a neighboring node passes to graph node i; This is an activation function used to perform a non-linear transformation on the aggregated overall information. For example, this activation function can be a Rectified Linear Unit (ReLU).

[0067] like Figure 4 As shown, in the above operation process, each graph node receives feature information transmitted by its neighboring nodes, combines relationship weights and weight matrices to complete multi-dimensional information aggregation, and obtains a new node state after normalization and nonlinear transformation. Through multi-layer network stacking, each graph node gradually integrates global information from the entire network to generate the final node embedding representation, i.e., a node intelligent profile. This profile can comprehensively reflect the node's own state and the complex relationship between the network as a whole, providing core support for subsequent fault attribution and operation and maintenance strategy generation.

[0068] Furthermore, after constructing the heterogeneous multi-relationship graph neural network through the above process, this embodiment can also combine real-time operating data of the actual network to generate a digital twin of the cross-carrier shared network. For example, this digital twin can be generated by the following method: real-time acquisition of network operating parameters of the real-world shared network across carriers, data synchronization of the heterogeneous multi-relationship graph neural network, and obtaining a digital twin consistent with the state of the real-world shared network through network simulation.

[0069] The aforementioned digital twin can synchronously map the topology, node attributes, various relationships, and real-time operating status of a real cross-carrier shared network, achieving a full digital replication of all elements of the cross-carrier shared network, thereby providing reliable support for subsequent fault location and closed-loop autonomy.

[0070] In S130, when performance degradation of any physical cell is detected, suspicious operations that may interfere with the degraded cell are screened out. In the digital twin, factual scenario simulation and counterfactual scenario simulation are performed on the suspicious operations to obtain the corresponding factual performance prediction value and counterfactual performance prediction value.

[0071] The aforementioned degraded cells are physical cells in cross-operator shared networks where service quality has declined; for example, such degraded cells may be physical cells where key performance indicators (throughput, latency, drop rate, call completion rate, etc.) deviate from the normal threshold range.

[0072] The aforementioned suspicious operations are network element behaviors, parameter adjustments, or state changes in cross-operator shared networks that may directly or indirectly interfere with degraded cells, thereby causing their performance degradation. For example, if the system detects that physical cell 101 in PLMN-B has an abnormal performance, and physical cell 205 in PLMN-B performed a power increase operation before the abnormal performance occurred, then the power increase operation is a suspicious operation that may have caused the abnormal performance of physical cell 101.

[0073] For example, when performance degradation of any physical cell is detected, the process of filtering out suspicious operations that may potentially interfere with the degraded cell can be implemented as follows: real-time collection of KPI data, equipment alarms, and resource status of each physical cell in the cross-operator shared network, and comparison with preset normal thresholds or historical baseline data; if multiple or single core indicators of a physical cell are detected to exceed the normal range, the physical cell is determined to be a degraded cell; taking the degraded cell as the center, traversing the target neighboring cells in the heterogeneous multi-relationship graph that are associated with the degraded cell through cross-operator coupling edges, and filtering out all suspicious operations (such as neighboring cell power increase, PRB utilization increase, parameter modification, etc.) based on the interference mechanism.

[0074] The above-mentioned scenario simulation is a simulation operation that replicates the current real network operating state within a digital twin. The simulation process keeps all configurations such as suspicious operations, network parameters, topology relationships, and service load unchanged, simulates the complete network operation process, and finally outputs the performance prediction results corresponding to the degraded cells (i.e., the above-mentioned factual performance prediction values).

[0075] The above counterfactual scenario simulation is a simulation operation performed in a digital twin after intervention is applied to suspicious operations. The simulation process cancels the corresponding suspicious operations, keeps the other configurations (such as network parameters, topology, service load, transmission power, etc.) unchanged, simulates the network operation process, and finally outputs the performance prediction results corresponding to the degraded cell (i.e. the above counterfactual performance prediction values).

[0076] For example, the above-mentioned execution of factual scenario simulation and counterfactual scenario simulation for suspicious operations in the digital twin to obtain the corresponding factual performance prediction value and counterfactual performance prediction value can be achieved as follows: Based on all current graph node features and current edge weights, network forward propagation is performed in the digital twin to obtain the factual performance prediction value of the degraded cell; in the digital twin, the network parameters corresponding to the suspicious operation are rolled back to before the operation was executed, the edge weights of each cross-operator coupling edge are recalculated based on the rolled-back network parameters, and network forward propagation is performed to obtain the counterfactual performance prediction value of the degraded cell.

[0077] Specifically, the above process is implemented as follows: A trained cross-carrier shared network digital twin is invoked, and all parameters, including the current real topology, graph node characteristics, relational edge weights, device configuration, user load, and environmental status of the cross-carrier shared network, are simultaneously loaded to ensure that the virtual environment is completely consistent with the real network. All selected suspicious operations are retained in the digital twin without modifying any parameters or operating status. The digital twin is then driven to perform simulation calculations according to the network operating logic, simulating the cell operating status under the current real operating conditions. Various KPI indicators of degraded cells are collected and recorded to obtain the factual performance prediction value. Only the corresponding suspicious operation is undone in the digital twin (all other network parameters, topology, load, and environment remain unchanged). The digital twin is then driven to perform simulation calculations again, simulating the operating conditions without the suspicious operation, collecting and recording various KPI indicators of degraded cells to obtain the counterfactual performance prediction value.

[0078] The execution process of the above-mentioned factual scenario simulation and counterfactual scenario simulation will be described in detail below with reference to specific embodiments: When the system detects an abnormal performance of a physical cell within the PLMN-B (such as a 30% decrease in throughput), it identifies that physical cell as a degraded cell and performs the following operations: Perform real-world transfer (i.e., the above-mentioned factual scenario simulation): use all current node feature vectors. and edge weight A heterogeneous multi-relational graph neural network (HMR-GNN) is used for forward propagation to obtain the factual performance prediction values ​​for the degraded cell. .

[0079] The most recent power boost operation in PLMN-A is considered a suspicious operation, and a counterfactual world propagation is performed (i.e., the aforementioned counterfactual scenario simulation): the power characteristics of relevant nodes within PLMN-A are analyzed. Rollback to the previous cycle value; the previous cycle refers to the most recent data collection cycle before this power boost operation was triggered; recalculate the weights of all cross-carrier coupling edges. Perform HMR-GNN forward propagation again to obtain counterfactual performance predictions. .

[0080] In S140, a causal effect analysis is performed on the suspicious operation based on the factual performance prediction value and the counterfactual performance prediction value.

[0081] The above causal effect analysis is an analytical process that compares two sets of predicted values ​​to quantify the impact of suspicious operations on cell performance degradation, determine whether there is a direct causal relationship between the two, and is used to locate the root cause of the fault.

[0082] For example, the above-mentioned causal effect analysis of suspicious operations based on factual performance prediction values ​​and counterfactual performance prediction values ​​can be achieved as follows: calculate the difference between factual performance prediction values ​​and counterfactual performance prediction values, and record it as the causal effect value; if the causal effect value is greater than a preset degradation threshold, then the corresponding suspicious operation is determined to be the target causal operation.

[0083] The above-mentioned target causality determination operation will be described in detail below in a specific embodiment: The actual performance prediction value obtained from the above calculation and counterfactual performance prediction value For example, perform the following causal effect calculation: in, The above causal effect value; if (The default degradation threshold here is -0.25, which means that the performance degradation exceeds 25%). Then, the most recent power boost operation of the PLMN-A is determined to be a strong causal factor of the performance degradation, and is recorded as the target causal operation.

[0084] In S150, fault handling instructions are generated based on the causal effect analysis results and sent to the corresponding network to repair the performance of the degraded cell.

[0085] The aforementioned fault handling instructions are control instructions generated based on the root causes of interference derived from causal effect analysis. They are used to restore the performance of degraded cells by adjusting cell parameters, resolving abnormal operations, and optimizing resource allocation. For example, the fault handling instructions may include the operation object, the action to be performed, and the parameter adjustment amount.

[0086] For example, the above-mentioned generation of fault handling instructions based on causal effect analysis results and issuance to the corresponding network can be achieved as follows: constructing a target optimization function based on the performance loss terms of the degraded cell and the interfering cell; the target optimization function includes a first performance loss term of the degraded cell and a second performance loss term of the interfering cell performing the target causal operation; backpropagating the gradient of the target optimization function in the digital twin to obtain the parameter adjustment amount, which is the equilibrium optimization value that minimizes the performance loss of the degraded cell and minimizes the impact on the interfering cell; generating fault handling instructions based on the parameter adjustment amount and issuing them to the corresponding network.

[0087] Specifically, the objective function described above can be as follows: in, This is the first performance loss term for degraded cells in PLMN-B; This is the second performance loss term for the interfering cell performing the target causal operation in PLMN-A; , This is a weighting coefficient used to balance the priorities of PLMN-B and PLMN-A, and can be flexibly set based on the actual scenario. The current value of the parameter to be optimized; The parameter adjustment amount is determined for the parameters to be optimized; backpropagation calculation is performed using a heterogeneous multi-relation graph neural network to solve the gradient value of the objective optimization function with respect to the adjustable parameters of each network. Based on the gradient magnitude, the key adjustable parameters that have the greatest impact on the performance of the degraded cell are identified, and their corresponding parameter adjustment amounts are calculated.

[0088] By using the aforementioned objective optimization function, an optimal set of parameter adjustment schemes can be found in cross-operator shared networks, prioritizing the recovery of degraded cell performance while minimizing the impact on the network service quality of another operator.

[0089] Preferably, the above-mentioned generation of fault handling instructions based on the causal effect analysis results and the issuance of such instructions to the corresponding network can achieve the following: generating fault handling instructions based on the causal effect analysis results; executing the fault handling instructions in the digital twin and verifying the execution results; if the verification is successful, issuing the fault handling instructions to the corresponding real network to repair the performance of the degraded cell.

[0090] Specifically, assuming that the key adjustable parameter that has the greatest impact on the performance of degraded cells in PLMN-B is identified by the above objective optimization function as the downlink power of physical cell 101 in PLMN-A, and the calculated parameter adjustment amount is 3dBm, then a fault handling instruction can be generated: reduce the downlink transmit power of physical cell 101 in PLMN-A by 3dBm; submit the fault handling instruction to the digital twin for operation and verify the effect; after successful verification, send the fault handling instruction to the OSS system of PLMN-A for execution through the northbound interface; for example, the above verification process can be as follows: if the performance indicators of the degraded cell return to the normal level and do not have a negative impact on other network elements, then the fault handling instruction is deemed to have passed the verification.

[0091] Preferably, after executing the fault handling command in a real cross-carrier shared network, this embodiment can also collect the actual operating indicators and parameter status of each physical cell, update the node feature vector of HMR-GNN based on the measured data, and recalculate and update the weights of various relation edges. In this way, the model can be learned online based on the newly added real samples, and the model's prediction, analysis and decision-making capabilities can be continuously improved.

[0092] Accordingly, the present invention provides a fault handling device for a cross-carrier shared network, with reference to... Figure 5 As shown, the fault handling device 500 for the cross-carrier shared network can include a heterogeneous graph construction module 510, a twin construction module 520, a simulation and deduction module 530, a causal analysis module 540, and a fault handling module 550. Among them: The heterogeneous graph construction module 510 is used to construct the relationship edges between adjacent graph nodes based on the physical cells of each operator and the operator type of adjacent physical cells, so as to obtain a heterogeneous multi-relationship graph. The relationship edges include the same operator collaborative edge and the cross-operator coupling edge. The twin construction module 520 is used to initialize the information of graph nodes, determine the edge weights of each relation edge, and construct a heterogeneous multi-relation graph neural network containing relation graph convolutional layers. Based on the heterogeneous multi-relation graph neural network, a digital twin of a cross-operator shared network is generated. The simulation and deduction module 530 is used to filter out suspicious operations that may interfere with the degraded cell when the performance of any physical cell is detected. In the digital twin, it performs factual scenario simulation and deduction and counterfactual scenario simulation and deduction for the suspicious operations respectively, and obtains the corresponding factual performance prediction value and counterfactual performance prediction value. Causal analysis module 540 is used to perform causal effect analysis on suspicious operations based on factual performance prediction values ​​and counterfactual performance prediction values; The fault handling module 550 is used to generate fault handling instructions based on the causal effect analysis results and send them to the corresponding network to repair the performance of the degraded cell.

[0093] In one implementation of this embodiment, the heterogeneous graph construction module includes a cross-PLMN data acquisition and fusion unit. Specifically, this cross-PLMN data acquisition and fusion unit is used to: collect network operation data in real time from the network management and configuration platforms, network element devices, and user terminals of various operators. For example, the network management and configuration platform may include an Operation Support System (OSS) and a Business Support System (BSS), the network element device may be a gNodeB base station, and the network operation data may include key performance indicators (such as throughput, latency, and call drop rate), network configuration parameters (such as frequency, bandwidth, and transmit power), resource usage status (such as PRB utilization), and device alarm logs, etc.; and perform time alignment, unit unification, and feature standardization processing on the collected multi-source heterogeneous data to form structured data.

[0094] In one implementation of this embodiment, the heterogeneous graph construction module is specifically used to: if adjacent physical cells belong to the same operator, construct a same-operator collaborative edge between the corresponding adjacent graph nodes; if adjacent physical cells belong to different operators, construct a cross-operator coupling edge between the corresponding adjacent graph nodes.

[0095] In one implementation of this embodiment, the heterogeneous graph construction module is specifically used to: if multiple physical cells are affected by the same environmental factor, introduce a common virtual node corresponding to the environmental factor in the heterogeneous relationship graph; and construct an environmental influence edge between the common virtual node and the graph node corresponding to the cell affected by the environmental factor.

[0096] In one implementation of this embodiment, the twin construction module is specifically used to: configure node attribute information for graph nodes, the node attribute information including the cell identifier of the corresponding physical cell and the operator identity identifier to which it belongs; collect network operation parameters of each physical cell and preprocess them to obtain structured feature data, and generate node characteristic vectors of the corresponding graph nodes based on the structured feature data; bind the node attribute information and node feature vectors to the corresponding graph nodes to complete the information initialization of the graph nodes.

[0097] In one implementation of this embodiment, the twin construction module is specifically used to: obtain the number of used physical resource blocks and transmit power of the coupled adjacent physical cells corresponding to the cross-operator coupling edge; obtain the total number of physical resource blocks and the maximum allowed transmit power of the coupled adjacent physical cells; determine the resource load ratio based on the number of used physical resource blocks and the total number of physical resource blocks of the coupled adjacent physical cells; determine the transmit power ratio based on the transmit power and the maximum allowed transmit power of the coupled adjacent physical cells; and dynamically calculate the edge weight of the cross-operator coupling edge based on the resource load ratio and the transmit power ratio.

[0098] In one implementation of this embodiment, the aforementioned twin construction module is specifically used to: use the relation graph convolutional layer as the basic operation layer of the network, build a heterogeneous multi-relation graph neural network based on the heterogeneous multi-relation graph; use the node feature vector as the initial input feature of each graph node, and perform weighted graph convolution aggregation operation on the graph nodes based on the type and weight of the relation edges to complete the construction of the heterogeneous multi-relation graph neural network.

[0099] In one implementation of this embodiment, the aforementioned twin construction module is specifically used to: collect network operation parameters of real-world shared networks across operators in real time, synchronize data with heterogeneous multi-relationship graph neural networks, and obtain a digital twin consistent with the state of the real-world shared network through network simulation.

[0100] In one implementation of this embodiment, the simulation and deduction module includes a counterfactual causal inference engine. Specifically, the counterfactual causal inference engine is used to: perform network forward propagation in the digital twin based on all current graph node features and current edge weights to obtain the factual performance prediction value of the degraded cell; roll back the network parameters corresponding to the suspicious operation in the digital twin to before the operation was executed; recalculate the edge weights of each cross-operator coupling edge based on the rolled-back network parameters; and perform network forward propagation to obtain the counterfactual performance prediction value of the degraded cell.

[0101] In one implementation of this embodiment, the causal analysis module is specifically used to: calculate the difference between the factual performance prediction value and the counterfactual performance prediction value, and record it as the causal effect value; if the causal effect value is greater than a preset degradation threshold, then the corresponding suspicious operation is determined to be the target causal operation.

[0102] In one implementation of this embodiment, the fault handling module is specifically used for: constructing a target optimization function based on the performance loss terms of the degraded cell and the interfering cell; the target optimization function includes a first performance loss term of the degraded cell and a second performance loss term of the interfering cell performing the target causal operation; backpropagating the gradient of the target optimization function in the digital twin to obtain the parameter adjustment amount, the parameter adjustment amount being the equilibrium optimization value that minimizes the performance loss of the degraded cell and minimizes the impact on the interfering cell; generating a fault handling instruction based on the parameter adjustment amount and sending it to the corresponding network.

[0103] In one implementation of this embodiment, the fault handling module is specifically used to: generate fault handling instructions based on the causal effect analysis results; execute the fault handling instructions in the digital twin and verify the execution results; if the verification is successful, send the fault handling instructions to the corresponding real network to repair the performance of the degraded cell.

[0104] It should be noted that the specific implementation details of the fault handling device for the cross-carrier shared network have been described in detail in the corresponding section of the fault handling method for the cross-carrier shared network, so they will not be repeated here.

[0105] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to embodiments of this disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0106] An electronic device according to the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements any of the above-mentioned fault handling methods for cross-carrier shared networks. That is, an electronic device according to the present invention may include, but is not limited to: a processor and a memory; the memory is used to store the computer program; the processor is used to execute the fault handling method for cross-carrier shared networks shown in any embodiment of the present invention by calling the computer program.

[0107] In one alternative embodiment, an electronic device is provided, such as Figure 6 As shown, Figure 6The illustrated electronic device 6000 includes a processor 6001 and a memory 6003. The processor 6001 and the memory 6003 are connected, for example, via a bus 6002. Optionally, the electronic device 6000 may further include a transceiver 6004, which can be used for data interaction between the electronic device and other electronic devices, such as sending and / or receiving data. It should be noted that in practical applications, the transceiver 6004 is not limited to one type, and the structure of the electronic device 6000 does not constitute a limitation on the present invention.

[0108] Processor 6001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this invention. Processor 6001 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0109] Bus 6002 may include a path for transmitting information between the aforementioned components. Bus 6002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. Bus 6002 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 6 The bus 6002 is represented by only one thick line, but this does not mean that there is only one bus or one type of bus.

[0110] The memory 6003 may be ROM (Read Only Memory) or other types of static storage devices capable of storing static information and instructions, RAM (Random Access Memory) or other types of dynamic storage devices capable of storing information and instructions, or EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto.

[0111] The memory 6003 stores application code (computer program) for executing the present invention, and its execution is controlled by the processor 6001. The processor 6001 executes the application code stored in the memory 6003 to implement the content shown in the foregoing method embodiments.

[0112] Among them, electronic devices can also be terminal devices, which can be any device that can install applications, including at least one of smartphones, tablets, laptops, desktop computers, smart speakers, smartwatches, smart TVs, and smart in-vehicle devices.

[0113] It should be noted that, Figure 6 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the invention.

[0114] The present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements any of the aforementioned fault handling methods for cross-carrier shared networks.

[0115] Alternatively, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a compact disc read-only memory (CD-ROM), magnetic tape, a floppy disk, and an optical data storage device, etc.

[0116] In an exemplary embodiment, a computer program product or computer program is also provided, which includes computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the electronic device to perform the aforementioned fault handling method for a cross-carrier shared network.

[0117] Computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as "C" or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0118] It should be understood that the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of methods and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0119] The computer-readable storage medium provided by this invention can be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EEPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0120] The aforementioned computer-readable storage medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to perform the method shown in the above embodiments.

[0121] The above description is merely a preferred embodiment of the present invention and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of disclosure in this invention is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-disclosed concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this invention.

[0122] It should be noted that the terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and represent a limitation on a specific order or sequence. Where appropriate, the order of use for similar objects can be interchanged so that the embodiments of this application described herein can be implemented in an order other than that shown or described.

[0123] Those skilled in the art will recognize that this invention can be implemented as a system, method, or computer program product. Therefore, this invention can be specifically implemented in the following forms: it can be entirely hardware, entirely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software, generally referred to herein as a "circuit," "module," or "system." Furthermore, in some embodiments, this invention can also be implemented as a computer program product contained in one or more computer-readable media, which includes computer-readable program code.

[0124] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A fault handling method for a cross-carrier shared network, characterized in that, include: Using the physical cells of each operator as graph nodes, and constructing the relationship edges between adjacent graph nodes based on the operator type of adjacent physical cells, a heterogeneous multi-relationship graph is obtained. The relationship edges include cooperative edges within the same operator and coupling edges across operators. The graph nodes are initialized with information, the edge weights of each relation edge are determined, and a heterogeneous multi-relation graph neural network containing relation graph convolutional layers is constructed. A digital twin of a cross-carrier shared network is generated based on the heterogeneous multi-relation graph neural network. When any physical cell performance degradation is detected, suspicious operations that may interfere with the degraded cell are screened out. In the digital twin, factual scenario simulation and counterfactual scenario simulation are performed on the suspicious operations to obtain the corresponding factual performance prediction value and counterfactual performance prediction value. A causal effect analysis is performed on the suspicious operation based on the factual performance prediction value and the counterfactual performance prediction value. Based on the causal effect analysis results, fault handling instructions are generated and sent to the corresponding network to repair the performance of the degraded cell.

2. The fault handling method for a cross-carrier shared network according to claim 1, characterized in that, The construction of relationship edges between adjacent graph nodes based on the operator type of adjacent physical cells includes: If the adjacent physical cells belong to the same operator, then the same operator collaborative edge is constructed between the corresponding adjacent graph nodes; If the adjacent physical communities belong to different operators, then the cross-operator coupling edge is constructed between the corresponding adjacent graph nodes.

3. The fault handling method for a cross-carrier shared network according to claim 1, characterized in that, When constructing the heterogeneous multi-relationship graph, the method further includes: If multiple physical cells are affected by the same environmental factor, a common virtual node corresponding to the environmental factor is introduced into the heterogeneous relationship graph. Construct environmental impact edges between the public virtual node and the graph nodes corresponding to the cells affected by environmental factors.

4. The fault handling method for a cross-carrier shared network according to claim 1, characterized in that, The initialization of information for the graph nodes includes: Configure node attribute information for the graph nodes, the node attribute information including the cell identifier of the corresponding physical cell and the operator identity identifier to which it belongs; The network operation parameters of each physical cell are collected and preprocessed to obtain structured feature data, and node characteristic vectors of corresponding graph nodes are generated based on the structured feature data. The node attribute information and the node feature vector are bound to the corresponding graph node to complete the information initialization of the graph node.

5. The fault handling method for a cross-carrier shared network according to claim 1, characterized in that, For the cross-carrier coupling edge, determining the edge weight of each of the said relational edges includes: Obtain the number of used physical resource blocks and transmit power of the coupled adjacent physical cells corresponding to the cross-operator coupling edge; Obtain the total number of physical resource blocks and the maximum allowed transmit power of the coupled adjacent physical cells; The resource load percentage is determined based on the number of used physical resource blocks and the total number of physical resource blocks in the coupled adjacent physical cells. The transmission power ratio is determined based on the transmission power of the coupled adjacent physical cells and the maximum allowed transmission power. The edge weight of the cross-carrier coupling edge is dynamically calculated based on the resource load ratio and the transmit power ratio.

6. The fault handling method for a cross-carrier shared network according to claim 4, characterized in that, The construction of the heterogeneous multi-relation graph neural network including relation graph convolutional layers includes: Using the relation graph convolutional layer as the basic operation layer of the network, the heterogeneous multi-relation graph neural network is built based on the heterogeneous multi-relation graph; The node feature vectors are used as the initial input features of each graph node. Based on the type of the relation edges and the edge weights, a weighted graph convolution aggregation operation is performed on the graph nodes to complete the construction of the heterogeneous multi-relation graph neural network.

7. The fault handling method for a cross-carrier shared network according to claim 1, characterized in that, The generation of a digital twin of a cross-carrier shared network based on the heterogeneous multi-relationship graph neural network includes: The network operation parameters of the real shared network across operators are collected in real time, and the data is synchronized with the heterogeneous multi-relationship graph neural network. The digital twin that is consistent with the state of the real shared network is obtained through network simulation.

8. The fault handling method for a cross-carrier shared network according to claim 1, characterized in that, The process of performing factual scenario simulation and counterfactual scenario simulation in the digital twin for the suspicious operation, respectively, to obtain corresponding factual performance prediction values ​​and counterfactual performance prediction values, includes: Based on all current graph node features and current edge weights, network forward propagation is performed in the digital twin to obtain the actual performance prediction value of the degraded cell; In the digital twin, the network parameters corresponding to the suspicious operation are rolled back to before the operation was executed. Based on the rolled-back network parameters, the edge weights of each cross-carrier coupling edge are recalculated, and network forward propagation is performed to obtain the counterfactual performance prediction value of the degraded cell.

9. The fault handling method for a cross-carrier shared network according to claim 1, characterized in that, The causal effect analysis of the suspicious operation based on the factual performance prediction value and the counterfactual performance prediction value includes: The difference between the factual performance prediction value and the counterfactual performance prediction value is calculated and denoted as the causal effect value; If the causal effect value is greater than the preset degradation threshold, the corresponding suspicious operation is determined to be the target causal operation.

10. The fault handling method for a cross-carrier shared network according to claim 9, characterized in that, The step of generating fault handling instructions based on the target causal operation and sending them to the corresponding network includes: A target optimization function is constructed based on the performance loss terms of the degraded cell and the interfering cell; the target optimization function includes a first performance loss term of the degraded cell and a second performance loss term of the interfering cell performing the target causal operation; The gradient of the objective optimization function is backpropagated in the digital twin to obtain the parameter adjustment amount, which is the equilibrium optimization value that minimizes the performance loss of the degraded cell and the impact on the interfering cell. The fault handling instruction is generated based on the parameter adjustment amount and sent to the corresponding network.

11. The fault handling method for a cross-carrier shared network according to claim 1, characterized in that, The process of generating fault handling instructions based on causal effect analysis results and sending them to the corresponding network includes: The fault handling instructions are generated based on the causal effect analysis results. The fault handling instructions are executed in the digital twin, and the execution results are verified. If the verification is successful, the fault handling instruction will be sent to the corresponding real network to repair the performance of the degraded cell.

12. A fault handling device for a cross-carrier shared network, characterized in that, include: The heterogeneous graph construction module is used to take the physical cells of each operator as graph nodes and construct the relationship edges between adjacent graph nodes based on the operator type of adjacent physical cells to obtain a heterogeneous multi-relationship graph. The relationship edges include cooperative edges within the same operator and coupling edges across operators. The twin construction module is used to initialize the information of the graph nodes, determine the edge weights of each relation edge, and construct a heterogeneous multi-relation graph neural network containing relation graph convolutional layers, and generate a digital twin of a cross-carrier shared network based on the heterogeneous multi-relation graph neural network; The simulation and deduction module is used to filter out suspicious operations that may interfere with the degraded cell when the performance degradation of any of the physical cells is detected. In the digital twin, the module performs factual scenario simulation and deduction and counterfactual scenario simulation and deduction for the suspicious operations respectively to obtain the corresponding factual performance prediction value and counterfactual performance prediction value. The causal analysis module is used to perform causal effect analysis on the suspicious operation based on the factual performance prediction value and the counterfactual performance prediction value. The fault handling module is used to generate fault handling instructions based on the causal effect analysis results and send them to the corresponding network to repair the performance of the degraded cell.

13. An electronic device, characterized in that, include: processor; Memory for storing the executable instructions of the processor; The processor is configured to execute the instructions to implement the fault handling method for a cross-carrier shared network as described in any one of claims 1 to 11.

14. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the fault handling method for the cross-carrier shared network as described in any one of claims 1 to 11.

Images