Method and apparatus for controlling cell transmit power, electronic device, and storage medium

Abstract

The application provides a cell transmission power control method and device, electronic equipment and storage medium, and relates to the technical field of communication. The method comprises: obtaining a cell network state data set of a first period; wherein the cell network state data set comprises neighbor cell relationship information between each cell, first network state information of each cell at each time in the first period, connection relationship information between each cell and a user terminal, and second network state information of each connected user terminal at each time; and determining a transmission power prediction result of each cell according to the cell network state data set, so as to control the transmission power of each cell. The application fully utilizes the neighbor cell relationship between each cell, the connection relationship between each cell and the user terminal, and the network state information of each cell and each connected user terminal, to predict the transmission power of each cell.

Description

Technical Field

[0001] This invention relates to the field of communication technology, and in particular to a method, apparatus, electronic device, and storage medium for controlling cell transmit power. Background Technology

[0002] Downlink power control is a critical and fundamental issue in 5G systems. On the one hand, if the base station's transmit power is too low, path loss and shadow fading cannot be adequately compensated, resulting in weak coverage or even coverage blind spots. On the other hand, if the base station's signal transmit power is too high, it will cause over-coverage and consume more energy, leading to strong interference and high energy consumption. In current networks, the status of base station cells and users changes over time, requiring dynamic control of downlink power to adjust the base station's transmit power to an appropriate value according to network changes.

[0003] In related technologies, downlink power is mainly dynamically controlled based on the parameter settings of the base station cell side and the uplink feedback from users. For example, the transmit power of the edge user's dedicated PDCCH is adaptively adjusted based on the aggregation level of PDCCH (Physical Downlink Control Channel) CCE (Control Channel Element) and the feedback of PUCCH (Physical Uplink Control Channel). The power spectral density of PDSCH is also adaptively adjusted based on the scheduling MCS (Modulation and Coding Scheme), bandwidth, and estimated user received signal strength of PDSCH (Physical Downlink Shared Channel). However, the power control effect is poor, and the problem of mismatch between the downlink power of the base station and the actual demand still exists. Summary of the Invention

[0004] This invention provides a method, apparatus, electronic device, and storage medium for controlling cell transmit power, in order to solve the problem that the downlink power of the base station does not match the actual demand in the power control strategy of the prior art.

[0005] In a first aspect, embodiments of the present invention provide a method for controlling cell transmit power, comprising: acquiring a cell network status dataset for a first time period; wherein the cell network status dataset includes neighbor cell relationship information between cells, first network status information of each cell at each time point in the first time period and connection relationship information between each cell and a user terminal, and second network status information of each connected user terminal at each time point; and determining a transmit power prediction result for each cell based on the cell network status dataset, for use in controlling the transmit power of each cell.

[0006] In some embodiments, determining the transmit power prediction result for each cell based on the cell network state dataset includes: constructing a first cell user map based on the cell network state dataset; the first cell user map is an information representation of the network state relationships between the cells and between the cells and the connected user terminals at each time point; inputting the first cell user map into a first power prediction model to obtain the transmit power prediction result for each cell; wherein the first power prediction model is trained based on the first cell user map samples and the cell transmit power labels corresponding to the first cell user map samples.

[0007] As one possible implementation, constructing a first cell user graph based on the cell network state dataset includes: determining each cell as a cell node; determining first edge feature information between cell nodes based on neighbor cell relationship information between the cells; for a first moment within the first time period, determining each connected user terminal as a user node based on the connection relationship at the first moment, and determining second edge feature information between each user node and each cell node; determining initial feature information of each cell node based on the first network state information of each cell at the first moment, and determining initial feature information of each user node based on the second network state information of each connected user terminal at the first moment; determining a first cell user graph slice at the first moment based on each cell node, each user node, the first edge feature information, the second edge feature information, the initial feature information of each cell node, and the initial feature information of each user node; and determining the cell user graph based on the first cell user graph slices at each moment within the first time period.

[0008] In some embodiments, the first power prediction model includes a spatial fusion layer, a temporal fusion layer, and a linear mapping layer; the step of inputting the first cell user map into the first power prediction model to obtain the transmit power prediction result of each cell includes: sending the first cell user map to the spatial fusion layer to obtain a spatial fusion feature set of each cell node output by the spatial fusion layer; wherein, each spatial fusion feature set includes spatial fusion feature information of the corresponding cell node at each time step; the spatial fusion layer is used to fuse the initial feature information of each cell node with the initial feature information of its neighboring nodes based on the first cell user map; inputting the spatial fusion feature set of each cell node into the temporal fusion layer to obtain the spatiotemporal fusion feature information of each cell node output by the temporal fusion layer; the temporal fusion layer is used to fuse the spatial fusion feature information of each cell node at each time step within the first time period based on the spatial fusion feature set of each cell node; and inputting the spatiotemporal fusion feature information of each cell node into the linear mapping layer to obtain the transmit power prediction result of each cell output by the linear mapping layer.

[0009] In some embodiments, determining the predicted transmit power of each cell based on the cell network state dataset includes: constructing multiple second cell user maps based on multiple pre-divided cell sub-maps, the first network state information, the connectivity information, and the second network state information; wherein the multiple cell sub-maps are divided based on the neighbor cell relationship information of each cell, and the multiple second cell user maps are related to the multiple cell sub-maps. Figure One One-to-one correspondence; each second cell user map is an information expression of the network state relationship between each cell in the corresponding cell sub-map, and between each cell in the corresponding cell sub-map and each connected user terminal at each time point; the multiple second cell user maps are input into the second power prediction model to obtain the transmission power prediction result of each cell; wherein, the second power prediction model is trained based on the second cell user map samples and the cell transmission power labels corresponding to the second cell user map samples.

[0010] As an example, the multiple cell subgraphs are divided as follows: a cell topology map is constructed based on the neighbor relationships of each cell; the cell topology map is segmented based on a spectral clustering algorithm to obtain the various cell subgraphs.

[0011] Secondly, the present invention provides a cell transmission power control device, comprising: an acquisition module, configured to acquire a cell network status dataset for a first time period; wherein the cell network status dataset includes neighbor cell relationship information between cells, first network status information of each cell at each moment in the first time period and connection relationship information between each cell and a user terminal, and second network status information of each connected user terminal at each moment; and a prediction module, configured to determine a transmission power prediction result for each cell based on the cell network status dataset, for controlling the transmission power of each cell.

[0012] Thirdly, embodiments of the present invention provide an electronic device, including a processor and a memory storing a computer program, wherein the processor executes the program to implement the cell transmit power control method described in the first aspect.

[0013] Fourthly, embodiments of the present invention provide a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the cell transmit power control method described in the first aspect.

[0014] Fifthly, embodiments of the present invention provide a computer program product, including a computer program that, when executed by a processor, implements the cell transmit power control method described in the first aspect.

[0015] The present invention provides a method, apparatus, electronic device, and storage medium for controlling cell transmit power. It acquires a cell network status dataset for a first time period. This dataset includes neighbor cell relationship information between cells, first network status information of each cell at each moment within the first time period, connection relationship information between each cell and user terminals, and second network status information of each connected user terminal at each moment. Based on the cell network status dataset, a predicted transmit power for each cell is determined for controlling its transmit power. This invention fully utilizes the neighbor cell relationships between cells, the connection relationships between each cell and user terminals, and the network status information of each cell and each connected user terminal to predict the transmit power of each cell. The predicted transmit power is then used as the basis for controlling the cell transmit power, thereby improving the accuracy of power control and enhancing the signal coverage and strength of the cells covered by the base station, thus improving network coverage performance. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 This is one of the flowcharts illustrating the cell transmit power control method provided in an embodiment of the present invention;

[0018] Figure 2 This is a second schematic flowchart of a cell transmission power control method provided in an embodiment of the present invention.

[0019] Figure 3 The third flowchart illustrates the cell transmission power control method provided in this embodiment of the invention.

[0020] Figure 4 This is a schematic diagram of the user map of the first cell in an embodiment of the present invention;

[0021] Figure 5 The fourth flowchart illustrates the cell transmit power control method provided in this embodiment of the invention.

[0022] Figure 6 Fifth flowchart illustrating the cell transmit power control method provided in this embodiment of the invention;

[0023] Figure 7 This is a schematic diagram of cell submap division provided in an embodiment of the present invention;

[0024] Figure 8 This is a schematic diagram of the structure of a cell transmission power control device provided in an embodiment of the present invention;

[0025] Figure 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0027] Downlink power control is a critical and fundamental issue in 5G systems. On the one hand, if the base station's transmit power is too low, path loss and shadow fading cannot be adequately compensated, resulting in weak coverage or even coverage blind spots. On the other hand, if the base station's signal transmit power is too high, it will cause over-coverage and consume more energy, leading to strong interference and high energy consumption. In current networks, the status of base station cells and users changes over time, requiring dynamic control of downlink power to adjust the base station's transmit power to an appropriate value according to network changes.

[0028] In related technologies, downlink power is mainly dynamically controlled based on the parameter settings of the base station cell side and the uplink feedback from users. For example, the transmit power of the edge user-specific PDCCH is adaptively adjusted based on the PDCCH CCE aggregation level and PUCCH feedback, and the PDSCH power spectral density is adaptively adjusted based on the PDSCH scheduling MCS, bandwidth, and estimated user received signal strength. However, the power control effect is poor, and there is still a problem that the downlink power of the base station does not match the actual demand.

[0029] To address the aforementioned problems, this invention provides a method, apparatus, electronic device, and storage medium for controlling cell transmission power.

[0030] Figure 1 This is one of the flowcharts illustrating a cell transmit power control method provided in an embodiment of the present invention. Figure 1 As shown, the method may include the following steps.

[0031] Step 101: Obtain the cell network status dataset for the first time period; wherein, the cell network status dataset includes neighbor cell relationship information between cells, first network status information of each cell at each moment in the first time period and connection relationship information between each cell and user terminal, and second network status information of each connected user terminal at each moment.

[0032] In some embodiments, for scenarios where the transmit power of each cell is predicted at the current moment, the first time period can be a continuous time period including the current moment and a previous period, or a continuous time period in the past that is close to the current network state. For scenarios where the transmit power of cells is predicted in a future time period, the first time period can be a past time period that is close to the network state in the future time period.

[0033] In some embodiments of the present invention, the transmit power of all cells in the entire network can be predicted, where each cell refers to all cells in the entire network; alternatively, the transmit power of cells in a target area can be predicted, where each cell refers to all cells in the target area.

[0034] In some embodiments, neighbor cell relationship information for each cell can be obtained from the network management data platform. The first network state information for each cell at each moment within the first time period may include the maximum number of users, downlink average PRB utilization, downlink data throughput, interference intensity, etc., and this data can also be obtained from the network management data platform. The connection relationship information between each cell and user terminals at each moment within the first time period refers to the information of user terminals connected to each cell at each time period. For example, if a cell and a user terminal are in an RRC (Radio Resource Control) connection state, then a connection relationship is confirmed between the cell and the user terminal. Each connected user terminal at any given time refers to the user terminal that has accessed the serving cell at that time. The second network status information may include the measurement information of each connected user terminal to the serving cell, such as RSRP (Reference Signal Receiving Power), RSRQ (Reference Signal Receiving Quality), SINR (Signal to Interference plus Noise Ratio), etc. It may also include the measurement information of each connected user terminal to neighboring cells, such as RSRP, RSRQ, SINR, etc. All of the above data can be obtained from the network management data platform.

[0035] Step 102: Based on the cell network status dataset, determine the predicted transmit power for each cell to control the transmit power of each cell.

[0036] It is understandable that the cell network status data includes the neighbor cell relationships between cells in the first time period, the network status information of each cell and each connected user terminal at each moment in the first time period, and the data on the mutual connections between each cell and each user terminal. All of these data are mutually influential with the cell transmission power, so the interrelationships between the above data can be fully considered to predict the transmission power of each cell.

[0037] In some embodiments, the transmit power prediction result for each cell may include the cell transmit power prediction result for each cell at each moment within a second time period. The second time period may be a future time period close to the network state of the first time period, and the duration of the second time period may be the same as the duration of the first time period or shorter than the duration of the first time period.

[0038] In other embodiments, if the application scenario is to predict the transmit power of the current cell, and the first time period is a continuous time period including the current time and a previous time period, then the transmit power prediction result of each cell can include the transmit power prediction value of the corresponding cell at the current time.

[0039] In some embodiments, the process of determining the predicted transmit power of each cell based on a cell network state dataset may include: inputting the cell network state dataset into a trained neural network model to obtain the predicted transmit power of each cell. This neural network model can be trained based on a large number of past cell network state data samples and their corresponding cell transmit power labels.

[0040] In other embodiments, based on the neighboring cell relationships of each cell, for each cell, the first network state information of the cell can be fused with the first network state information of its neighboring cells to obtain the first fused state information of the cell; the second network state information of connected user terminals can be fused with the second network state information of all connected user terminals connected to the cell to obtain the second fused state information of the cell; the first fused state information and the second fused state information of the cell can be fused to obtain the third fused state information of the cell; and the information at each time point in the third fused state information can be fused to obtain the fourth fused state information; based on the preset mapping relationship between the fourth fused state information and the cell's transmit power, the transmit power prediction result of the cell can be determined. The fusion weight value in each step of the fusion process can be obtained based on the respective network state information, or it can be implemented by constructing a multi-layer attention network to achieve each step of the fusion process.

[0041] In some embodiments, the predicted transmit power of each cell can be directly used to control the transmit power of the cell, or it can be corrected and the transmit power of each cell can be controlled based on the corrected result.

[0042] As one implementation, the cell transmit power control method of the present invention can be applied to the network management platform side, executed by the network management platform side, and sends cell transmit power control commands to the base station side based on the transmit power prediction results of each cell.

[0043] As another implementation, the execution subject of the cell transmission power control method of the present invention can be any electronic device. The electronic device establishes a communication connection with the network management platform and sends the obtained transmission power prediction results of each cell to the network management platform, so that the network management platform can issue a cell transmission power control command to the base station based on the cell transmission power prediction results.

[0044] As another implementation, the execution subject of the cell transmission power control method of the present invention can be any electronic device. The electronic device establishes a communication connection with the base station and sends the obtained transmission power prediction results of each cell to the corresponding base station, so as to the network management platform side, so that the base station side can control the transmission power of the cell based on the cell transmission power prediction results.

[0045] The cell transmit power control method provided in this invention obtains a cell network status dataset for a first time period. This dataset includes neighbor cell relationship information between cells, first network status information of each cell at each moment within the first time period, connection relationship information between each cell and user terminals, and second network status information of each connected user terminal at each moment. Based on the cell network status dataset, a predicted transmit power for each cell is determined for controlling the transmit power of each cell. This invention fully utilizes the neighbor cell relationships between cells, the connection relationships between each cell and user terminals, and the network status information of each cell and each connected user terminal to predict the transmit power of each cell. The predicted transmit power of each cell is used as the decision basis for controlling the transmit power of the cell, thereby improving the accuracy of power control and enhancing the signal coverage and strength of the cells covered by the base station to meet requirements, thus improving network coverage performance.

[0046] Figure 2 This is a second schematic flowchart illustrating a method for controlling cell transmit power according to an embodiment of the present invention. Figure 2 As shown, based on the above embodiments, Figure 1 The implementation process of step 102 may include the following steps.

[0047] Step 201: Construct a first cell user graph based on the cell network status dataset; the first cell user graph is an information representation of the network status relationships between cells and between each cell and each connected user terminal at each time point.

[0048] In some embodiments, the first cell user graph is graph data constructed based on a cell network state dataset. The first cell user graph is a dynamic heterogeneous graph, comprising slices of the first cell user graph at various times within a first time period. Since the cell network state dataset contains a large amount of relational data, using the cell user graph to represent information provides a more intuitive representation of the relationships between nodes.

[0049] In some embodiments, for each time moment, each cell can be regarded as a cell node, each connected user terminal as a user node, the neighbor relationship between cells is expressed by the edges connecting each cell node, the connection relationship between each cell and each user terminal is expressed by the edges connecting each cell node and each user terminal, the first network state information of each cell is regarded as the first network state information of the corresponding cell node, and the second state information of each connected user terminal is regarded as the second network state information of the corresponding user node.

[0050] As an example, the user graph of the first cell is G = {G} t} t＝1，T1 , of which G t G represents the first cell user map slice at time t within time T1, where T1 is the first time period. t It includes multiple cell nodes, multiple user nodes, the first network status information of each cell node, the second network status information of each user node, the edges between cell nodes, and the edges between cell nodes and user nodes.

[0051] Step 202: Input the user map of the first cell into the first power prediction model to obtain the transmission power prediction result of each cell; wherein, the first power prediction model is trained based on the user map samples of the first cell and the cell transmission power labels corresponding to the user map samples of the first cell.

[0052] In some embodiments, the first power prediction model may be a graph neural network model that has been learned to predict the transmit power of a cell based on a first cell user map.

[0053] In some embodiments, the first cell user map sample can be constructed based on a large amount of historical cell network state datasets. For example, cell network state datasets from multiple past time periods are obtained, along with the transmit power labels of each cell corresponding to each time period. The duration of each time period is consistent with the duration of the first time period. The first cell user map is constructed based on the cell network state datasets of each time period, resulting in the first cell user map sample. The first cell user map sample and its corresponding cell transmit power label values ​​are then input into an initial first power prediction model for training, resulting in the trained first power prediction model.

[0054] It should be noted that the cell transmit power label corresponding to the first cell user map sample can be determined based on actual needs. For example, if it is necessary to predict the transmit power of each cell at the current moment based on a continuous time period consisting of the cell network state dataset of the current moment and a previous period, the transmit power setting value of the last moment in each historical time period can be used as the corresponding cell transmit power label value. As another example, if it is necessary to predict the transmit power of each cell in the future time period based on the cell network state dataset of past time periods, the cell transmit power setting values ​​of each moment in each historical time period can be used as the corresponding cell transmit power label values ​​during the training process.

[0055] According to the cell transmit power control method of the present invention, a first cell user map is constructed based on a cell network state dataset. The first cell user map represents the information of network state relationships between cells and between each cell and each connected user terminal at each time step. The first cell user map is input into a first power prediction model to obtain the transmit power prediction result for each cell. The first power prediction model is trained based on samples of the first cell user map and the corresponding cell transmit power labels. By constructing the first cell user map, the network state relationships between cells and between cells and users are fully utilized, and the cell transmit power is predicted using the first power prediction model as a decision-making basis for controlling cell transmit power, thereby improving the accuracy of power control.

[0056] Figure 3 This is the third flowchart illustrating the cell transmit power control method provided in this embodiment of the invention. Figure 3 As shown, the process of constructing the first cell user map based on the cell network status dataset may include the following steps.

[0057] Step 301: Each cell is identified as a cell node, and the first edge feature information between cell nodes is determined based on the neighbor cell relationship information between each cell.

[0058] In other words, each cell is treated as a node in the first cell user graph, and the neighboring relationships between cells are treated as edges. If two cells are neighbors, the edge weight between these two cell nodes is 1; if two cells are not neighbors, the edge weight between these two cell nodes is 0. The first edge feature information between each cell node is the edge weight.

[0059] Step 302: For the first moment within the first time period, based on the connection relationship information of the first moment, each connected user terminal of the first moment is identified as a user node, and the second-side feature information between each user node and each cell node is determined.

[0060] In some embodiments, the connected user terminals at the first time can be determined based on the connection relationship information between each cell and the user terminal at the first time, and each connected user terminal at the first time is also treated as a node in the user graph of the first cell, i.e., a user node. The connection relationship between each cell and the user terminal is used as the edge between the cell node and the user node. If the connected user terminal A and cell B are in a connected state at the first time, the edge weight between the user node corresponding to the connected user terminal A and the cell node corresponding to cell B at the first time is set to 1; otherwise, it is set to 0. The second edge feature information between each user node and each cell node is the edge weight between each user node and each cell node.

[0061] Step 303: Based on the first network state information of each cell at the first moment, determine the initial characteristic information of each cell node; based on the second signal state information of each connected user terminal at the first moment, determine the initial characteristic information of each user node.

[0062] In some embodiments, for each cell, the initial characteristic information of the cell node corresponding to the cell can be determined based on the first network state information of the cell at the first moment.

[0063] As an example, the initial feature information of a cell node can be defined as d. A A column vector of dimensions is used. For each cell, the first network state information of the cell at the first time step is encoded and transformed to obtain the initial feature information of the corresponding cell node. For example, if the first network state information includes the maximum number of users, the downlink average PRB (Physical Resource Block) utilization, downlink data throughput, etc., the above information can be concatenated, and the concatenated data can be encoded and transformed to obtain the initial feature information of the corresponding cell node.

[0064] Step 304: Based on each cell node, each user node, the first side feature information, the second side feature information, the initial feature information of each cell node, and the initial feature information of each user node, determine the first cell user map slice at the first moment.

[0065] In some embodiments, each cell node, first-side feature information, multiple user nodes determined at the first time step, second-side feature information, initial feature information of each cell node, and initial feature information of each user node are combined to form a first cell user map slice at the first time step.

[0066] Step 305: Determine the first cell user map based on the first cell user map slices at each time point within the first time period.

[0067] In some embodiments, the first cell user map includes slices of the first cell user map at each time point within the first time period. Therefore, the slices of the first cell user map at each time point within the first time period are constructed sequentially according to steps 302 to 304, and the first cell user map of the first time period is finally obtained.

[0068] like Figure 4 As shown, the first cell user graph can be a dynamic heterogeneous graph. The graph includes slices of the first cell user graph at various times within the first time period. Each slice of the first cell user graph includes cell nodes, user nodes, initial feature information of cell nodes, initial feature information of user nodes, edges between cell nodes, and edges between cell nodes and user nodes.

[0069] According to the cell transmit power control method of the present invention, by sequentially constructing first cell user map slices for each time period within a first time period, a first cell user map for the first time period is obtained, so as to intuitively express the network state relationship between cells and between cells and user terminals. This helps to make full use of this information when predicting cell transmit power, thereby improving the accuracy of cell transmit power prediction and further improving the effect of cell power control.

[0070] Figure 5 This is the fourth flowchart illustrating the cell transmit power control method provided in this embodiment of the invention. Figure 5 As shown, based on the above embodiments, the process of inputting the user map of the first cell into the first power prediction model to obtain the transmission power prediction results of each cell may include the following steps.

[0071] Step 501: Send the first cell user map to the spatial fusion layer to obtain the spatial fusion feature set of each cell node output by the spatial fusion layer; wherein, each spatial fusion feature set includes the spatial fusion feature information of the corresponding cell node at each time step; the spatial fusion layer is used to fuse the initial feature information of each cell node with the initial feature information of its neighboring nodes based on the first cell user map.

[0072] In some embodiments, the first power model can be a hierarchical attention network model, including a spatial fusion layer, a temporal fusion layer, and a linear mapping layer. The spatial fusion layer is used to determine the neighboring nodes of each cell node at each time step based on the first cell user graph. The neighboring nodes can include neighboring cell nodes of the corresponding cell node, and can also include user nodes connected to the corresponding cell node. The spatial fusion layer can also fuse the initial feature information of each cell node, the initial feature information of its neighboring cells, and the initial feature information of the connected user nodes to obtain a combined spatial fusion feature of that cell.

[0073] In some embodiments, the process of the spatial fusion layer fusing the initial feature information of each cell node with the initial feature information of its neighboring nodes may include: for time t, for the i-th cell node Ai, determining the set of neighboring cell nodes and the combination of neighboring user nodes of Ai according to the first cell user map; determining a first fusion weight coefficient for each neighboring cell node to Ai based on the initial feature information of Ai and the initial feature information of each neighboring cell node; determining a second fusion weight coefficient for each neighboring user node to Ai based on the initial feature information of Ai and the initial feature information of each neighboring user node; determining the fusion features of Ai's neighboring cell nodes based on the first fusion weight coefficient and the initial feature information of each neighboring cell; determining the fusion features of Ai's neighboring user nodes based on the second fusion weight coefficient and the initial feature information of each neighboring user; and further fusing the initial feature information of Ai, the fusion features of Ai's neighboring cell nodes, and the fusion features of Ai's neighboring user nodes to obtain the spatial fusion feature information of Ai at time t. Wherein, both the first fusion weight coefficient and the second fusion weight coefficient represent the degree of feature contribution of the neighboring nodes to node Ai.

[0074] As an example, the process of determining the first fusion weight coefficient of each neighboring cell node for Ai based on the initial feature information of Ai and the initial feature information of each neighboring cell node can be achieved by the following formula (1).

[0075]

[0076] in, Let Aj be the first fusion weight coefficient of the j-th neighboring cell node Aj to the i-th cell node Ai at time t. This represents the set of neighboring cell nodes of cell node Ai at time t. Let be the initial feature information of cell node Ai at time t; σ be the activation function; a1 be the attention factor, which can be a trainable column vector with a dimension of 2l1; and W1 be a trainable weight matrix with a dimension of l1×d. A d A is the dimension of the initial feature information of the Ai node; || indicates that the column vectors are concatenated row by row.

[0077] Based on the first fusion weight coefficient and the initial feature information of each neighboring cell, the method for determining the fusion features of Ai neighboring cell nodes can be as shown in formula (2).

[0078]

[0079] in, This refers to the fusion characteristics of Ai neighboring cell nodes.

[0080] Based on the initial feature information of Ai and the initial feature information of each neighboring user node, the process of determining the second fusion weight coefficient of each neighboring user node for Ai is shown in the following formula (3).

[0081]

[0082] in, Let Bj be the second fusion weight coefficient of the j-th user cell node Bj to the i-th cell node Ai at time t. This represents the set of neighboring user nodes of cell node Ai at time t. Let be the initial feature information of cell node Ai at time t; σ be the activation function; a2 be the attention factor, which can be a trainable column vector with a dimension of 2l1; W2 and W3 be trainable weight matrices, with W2 having a dimension of l2×d. A The dimension of W3 is l2×d B d B The dimension of the initial feature information of the user node.

[0083] Based on the second fusion weight coefficient and the initial feature information of each neighboring user, the process of determining the fusion features of Ai neighboring user nodes can be shown in the following formula (4).

[0084]

[0085] in, AI Neighbors' integrated features of community users.

[0086] Finally, the initial feature information of Ai, the fusion features of Ai's neighboring cell nodes, and the fusion features of Ai's neighboring user nodes are fused together to obtain the spatial fusion feature information of Ai at time t. The implementation method is shown in the following formula (5).

[0087]

[0088] Where W4, W5, and W6 represent the corresponding weighting coefficient matrices, which can be obtained through training; W4 has a dimension of l3×d. A The dimensions of W5 are l2×l1, and the dimensions of W6 are l3×l2. It is a column vector of dimension l3.

[0089] Step 502: Input the spatial fusion feature set of each cell node into the temporal fusion layer to obtain the spatiotemporal fusion feature information of each cell node output by the temporal fusion layer; the temporal fusion layer is used to fuse the spatial fusion feature information of each cell node at each time in the first time period based on the spatial fusion feature set of each cell node.

[0090] In some embodiments, the temporal fusion layer can perform fusion processing on the spatial fusion feature information of each cell node at each time point in the first time period based on the scaling dot product attention mechanism.

[0091] In some embodiments, for A i The set of spatial fusion features of nodes is This set can be combined into a matrix H. i H i The dimension is l3×l4. Calculate Q=H i W q K = H i W k V = H i W v W q W k and W v The dimensions of each matrix are l3×l4, and M is a mask matrix of dimension T1×T1. Its purpose is to ensure that when a cell node at time t aggregates in the temporal domain, it only aggregates spatial aggregation features from time t and earlier. Therefore, the elements m in M... i,j It is defined as follows (6).

[0092]

[0093] The final spatiotemporal fusion feature information of node Ai can be obtained based on the following formula (7). It should be noted that the final spatiotemporal fusion feature information of Ai can be a set of spatiotemporal fusion feature information at each time point, or it can be the spatiotemporal fusion feature information at the last time point in the first time period. This can be set based on actual needs.

[0094]

[0095] Among them, Z i Z is the set of spatiotemporal fusion feature information of node Ai; i It is a matrix of dimension T1×l4, Z i The t-th row vector That is, A i The spatiotemporal fusion feature information at time t.

[0096] Step 503: Input the spatiotemporal fusion feature information of each cell node into the linear mapping layer to obtain the transmit power prediction result of each cell output by the linear mapping layer.

[0097] According to the cell transmit power control method of the present invention, a first power prediction model including a spatial fusion layer, a temporal fusion layer and a linear mapping layer is used to deeply fuse the spatial and temporal information contained in the user map of the first cell to perform deep feature extraction on the user map of the first cell, thereby further improving the accuracy of transmit power prediction for each cell, and thus improving the accuracy and effect of cell power control.

[0098] To further improve computational efficiency, the present invention also provides another embodiment.

[0099] Figure 6 This is the fifth flowchart illustrating the cell transmit power control method provided in this embodiment of the invention. Figure 6 As shown, based on the above embodiments, the process of determining the transmission power prediction result of each cell according to the cell network status dataset may include the following steps.

[0100] Step 601: Construct multiple second cell user maps based on the previously divided cell sub-maps, first network status information, connectivity information, and second network status information; wherein the multiple cell sub-maps are divided based on the neighbor cell relationship information of each cell, and the multiple second cell user maps are connected to the multiple cell sub-maps. Figure One One-to-one correspondence; each second cell user map is an information representation of the network state relationships between cells within the corresponding cell sub-map, and between each cell within the corresponding cell sub-map and each connected user terminal at each time point.

[0101] It is understandable that the number of cells and user terminals in the entire network is enormous. The first cell user map constructed based on the cell network status dataset of the first time period involves a large amount of data, requiring significant computational resources and time-consuming model prediction and training. To improve computational efficiency, this embodiment of the invention constructs a second cell user map based on multiple pre-divided cell subgraphs. Each second cell user map thus only involves information about the cell nodes in the corresponding cell subgraph and the information about the user nodes connected to those cell nodes. Distributed parallel computation can be used for both constructing the second cell user map and predicting cell transmit power, thereby greatly improving computational efficiency.

[0102] In some embodiments, the cell subgraph is divided based on the neighbor cell relationship information between each cell node. Each cell subgraph may include multiple closely related cells, while the relationship between cells in different subgraphs is relatively sparse. This enables efficient information exchange between cells in the same cell subgraph and also improves computational efficiency.

[0103] In some embodiments, each cell subgraph may include multiple cell nodes and edges between cell nodes, using the neighboring cell relationships between cells as edges. If two cells are neighboring, the edge weight between the corresponding cell nodes is set to 1; otherwise, it is set to 0. That is, the cell subgraph may contain the first edge feature information between cell nodes.

[0104] In some embodiments, the process of constructing multiple second cell user maps based on the divided cell subgraphs, first network state information, connection relationship information, and second network state information is consistent with the process of constructing the first cell user subgraph. The only difference is that the second cell user map is constructed based on the cell nodes contained in each cell subgraph, not based on all cell nodes. For example, if the cell nodes and the first edge feature information between cell nodes in the second cell user map are consistent with those in cell subgraph A when constructing the second cell user map corresponding to cell subgraph A, for each time point, based on the cell nodes in cell subgraph A, the connected user terminals at that time point are obtained and identified as user nodes, and the second edge feature information between user nodes is determined. Based on the first network state information at that time point, the initial feature information of the cell nodes in cell subgraph A is determined. Based on the second network state information at that time point, the initial feature information of the user nodes is obtained. Based on the above information, the second cell user map slice in the second cell user map corresponding to cell subgraph A is determined.

[0105] In some embodiments, cell subgraphs can be divided in the following ways: a cell topology map is constructed based on the neighboring cell relationships of each cell; the cell topology map is segmented based on a spectral clustering algorithm to obtain each cell subgraph.

[0106] Specifically, when constructing the cell topology graph, cells are treated as nodes, and neighboring relationships between cells are treated as edges. That is, if two cells are neighboring, the corresponding edge weight is set to 1; otherwise, it is set to 0, resulting in an undirected weighted graph. Assume there are M nodes in the cell topology graph, and the corresponding adjacency matrix is ​​set as A = [a...]. ij ] M×M Then, based on the adjacency matrix, the degree matrix D = [d ij ] M×M D is a diagonal matrix, and the elements d on the diagonal are... ii =∑ j a ij Therefore, the corresponding Laplace matrix is ​​L = DA.

[0107] Based on the spectral clustering algorithm, neighboring cells that are closely related and may have overlapping coverage and mutual influence are grouped into the same subgraph. First, a representation matrix is ​​calculated. Then, by calculating the k eigenvalues ​​and corresponding eigenvectors of S, the matrix composed of the k eigenvectors is standardized by row to obtain the feature matrix F. Each row of F is taken as a sample, and there are a total of M samples. Then, clustering algorithms such as k-means are used to cluster these M samples. Let the clustering dimension be N, and finally the cell topology map is divided into N cell sub-maps. Figure 7 This is a schematic diagram of dividing a cell topology map into multiple cell sub-maps.

[0108] Therefore, by segmenting the cell topology map in this step, the cell topology map consisting of all the cells in the network is divided into different cell sub-maps, and subsequent processing is performed on each cell sub-map separately, thereby reducing information exchange costs and improving data processing efficiency.

[0109] Step 602: Input multiple second cell user maps into the second power prediction model to obtain the transmit power prediction result for each cell; wherein, the second power prediction model is trained based on the second cell user map samples and the cell transmit power labels corresponding to the second cell user map samples.

[0110] It should be noted that the difference between the second power prediction model and the first power prediction model is that the second power prediction model is trained based on the second cell user map samples and their corresponding cell transmit power labels, while the first power prediction model is trained based on the first cell user map samples. The second cell user map samples are constructed based on multiple pre-divided cell sub-maps and cell network state datasets from multiple historical time periods.

[0111] In some embodiments, the second power prediction model may also include a spatial fusion layer, a temporal fusion layer, and a linear mapping layer, and its implementation process and principle are consistent with the first power prediction model. However, the second cell user map contains less data, so the second power prediction model needs to make predictions based on the data in the second cell user map. Cell transmit power prediction for different second cell user maps can be accomplished through distributed parallel computing.

[0112] According to the cell transmit power control method of this invention, multiple second cell user maps are constructed based on multiple pre-divided cell sub-maps, first network state information, connection relationship information, and second network state information. These multiple second cell user maps are then input into a second power prediction model to obtain the transmit power prediction result for each cell. This invention divides the entire network into multiple cell sub-maps, enabling distributed parallel computation in both cell user map construction and cell transmit power prediction, thereby significantly improving computational efficiency.

[0113] The control device for cell transmission power provided in the embodiments of the present invention is described below. The control device for cell transmission power described below and the control method for cell transmission power described above can be referred to in correspondence.

[0114] Figure 8 This is a schematic diagram of the structure of a cell transmission power control device provided in an embodiment of the present invention. Figure 8 As shown, the device may include an acquisition module 810 and a prediction module 820.

[0115] The acquisition module 810 is used to acquire the cell network status dataset for the first time period; wherein, the cell network status dataset includes neighbor cell relationship information between each cell, the first network status information of each cell at each time in the first time period and the connection relationship information between each cell and the user terminal, and the second network status information of each connected user terminal at each time.

[0116] The prediction module 820 is used to determine the predicted transmit power of each cell based on the cell network status dataset, so as to control the transmit power of each cell.

[0117] In some embodiments, the prediction module 820 includes a construction unit 821 and a prediction unit 822; wherein, the construction unit 821 is used to construct a first cell user map based on a cell network state dataset; the first cell user map is an information representation of the network state relationships between cells and between each cell and each connected user terminal at each time step; the prediction unit 822 is used to input the first cell user map into a first power prediction model to obtain the transmission power prediction result for each cell; wherein, the first power prediction model is trained based on the first cell user map samples and the cell transmission power labels corresponding to the first cell user map samples.

[0118] In some embodiments, the construction unit 821 is specifically configured to: determine each cell as a cell node, and determine the first side feature information between each cell node based on the neighbor cell relationship information between each cell; for the first moment within the first time period, determine each connected user terminal at the first moment as a user node based on the connection relationship information at the first moment, and determine the second side feature information between each user node and each cell node; determine the initial feature information of each cell node based on the first network state information of each cell at the first moment, and determine the initial feature information of each user node based on the second signal state information of each connected user terminal at the first moment; determine the first cell user map slice at the first moment based on each cell node, each user node, the first side feature information, the second side feature information, the initial feature information of each cell node, and the initial feature information of each user node; and determine the first cell user map based on the first cell user map slices at each moment within the first time period.

[0119] In some embodiments, the first power prediction model includes a spatial fusion layer, a temporal fusion layer, and a linear mapping layer; the prediction unit 822 is further configured to: send a first cell user map to the spatial fusion layer to obtain a spatial fusion feature set of each cell node output by the spatial fusion layer; wherein, each spatial fusion feature set includes the spatial fusion feature information of the corresponding cell node at each time step; the spatial fusion layer is configured to fuse the initial feature information of each cell node with the initial feature information of its neighboring nodes based on the first cell user map; input the spatial fusion feature set of each cell node to the temporal fusion layer to obtain the spatiotemporal fusion feature information of each cell node output by the temporal fusion layer; the temporal fusion layer is configured to fuse the spatial fusion feature information of each cell node at each time step within a first time period based on the spatial fusion feature set of each cell node; input the spatiotemporal fusion feature information of each cell node to the linear mapping layer to obtain the transmit power prediction result of each cell output by the linear mapping layer.

[0120] In some embodiments, the construction unit 821 is further configured to: construct multiple second cell user maps based on the divided multiple cell sub-maps, first network state information, connection relationship information, and second network state information; wherein the multiple cell sub-maps are divided based on the neighbor cell relationship information of each cell, and the multiple second cell user maps are connected to the multiple cell sub-maps. Figure OneOne-to-one correspondence; each second cell user map is an information expression of the network state relationship between each cell in the corresponding cell submap, and between each cell in the corresponding cell submap and each connected user terminal at each time point; the prediction unit 822 is also used to: input multiple second cell user maps into the second power prediction model to obtain the transmission power prediction result of each cell; wherein, the second power prediction model is trained based on the cell transmission power labels corresponding to the second cell user map samples and the second cell user map samples.

[0121] In some embodiments, the device further includes a segmentation module 830 for constructing a cell topology map based on the neighboring cell relationships of each cell; and for segmenting the cell topology map based on a spectral clustering algorithm to obtain each cell sub-map.

[0122] It should be noted that the explanations and descriptions in the above embodiments regarding the control method for cell transmission power can also be applied to the control device for cell transmission power in the embodiments of the present invention, and will not be repeated here.

[0123] Figure 9 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 9 As shown, the electronic device may include a processor 910, a communication interface 920, a memory 930, and a communication bus 940, wherein the processor 910, the communication interface 920, and the memory 930 communicate with each other via the communication bus 940. The processor 910 can call the computer program in the memory 930 to execute the steps of the cell transmit power control method.

[0124] For example, the method includes: acquiring a cell network status dataset for a first time period; wherein the cell network status dataset includes neighbor cell relationship information between cells, first network status information of each cell at each time point in the first time period and connection relationship information between each cell and user terminal, and second network status information of each connected user terminal at each time point; and determining the transmission power prediction result of each cell based on the cell network status dataset for use in controlling the transmission power of each cell.

[0125] Furthermore, the logical instructions in the aforementioned memory 930 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0126] On the other hand, embodiments of the present invention also provide a computer program product, the computer program product including a computer program, the computer program being stored on a non-transitory computer-readable storage medium, and when the computer program is executed by a processor, the computer is able to execute the steps of the cell transmit power control method provided in the above embodiments.

[0127] The method includes: acquiring a cell network status dataset for a first time period; wherein the cell network status dataset includes neighbor cell relationship information between cells, first network status information of each cell at each moment in the first time period and connection relationship information between each cell and user terminal, and second network status information of each connected user terminal at each moment; and determining the transmission power prediction result of each cell based on the cell network status dataset, so as to control the transmission power of each cell.

[0128] On the other hand, embodiments of the present invention also provide a processor-readable storage medium storing a computer program for causing a processor to execute the steps of the cell transmit power control method provided in the above embodiments.

[0129] For example, the method includes: acquiring a cell network status dataset for a first time period; wherein the cell network status dataset includes neighbor cell relationship information between cells, first network status information of each cell at each time point in the first time period and connection relationship information between each cell and user terminal, and second network status information of each connected user terminal at each time point; and determining the transmission power prediction result of each cell based on the cell network status dataset for use in controlling the transmission power of each cell.

[0130] The processor-readable storage medium can be any available medium or data storage device that the processor can access, including but not limited to magnetic memory (e.g., floppy disk, hard disk, magnetic tape, magneto-optical disk (MO)), optical memory (e.g., CD, DVD, BD, HVD), and semiconductor memory (e.g., ROM, EPROM, EEPROM, non-volatile memory (NAND FLASH), solid-state drive (SSD)).

[0131] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0132] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0133] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for controlling the transmission power of a cell, characterized in that, include: Obtain the cell network status dataset for the first time period; wherein the cell network status dataset includes neighbor cell relationship information between cells, first network status information of each cell at each moment in the first time period and connection relationship information between each cell and user terminal, and second network status information of each connected user terminal at each moment. Based on the cell network status dataset, the predicted transmit power of each cell is determined for use in controlling the transmit power of each cell. The step of determining the predicted transmit power for each cell based on the cell network state dataset includes: Based on the cell network status dataset, a first cell user graph is constructed; the first cell user graph is an information representation of the network status relationships between the cells and between the cells and the connected user terminals at each time point. The first cell user map is input into the first power prediction model to obtain the transmission power prediction result of each cell; wherein, the first power prediction model is trained based on the first cell user map sample and the cell transmission power label corresponding to the first cell user map sample.

2. The method for controlling cell transmission power according to claim 1, characterized in that, The step of constructing a first cell user map based on the cell network state dataset includes: Each of the aforementioned cells is defined as a cell node, and the first edge feature information between the cell nodes is determined based on the neighbor cell relationship information between the cells. For the first moment within the first time period, based on the connection relationship information of the first moment, each connected user terminal at the first moment is determined as a user node, and the second side feature information between each user node and each cell node is determined. Based on the first network state information of each cell at the first time, the initial characteristic information of each cell node is determined; based on the second signal state information of each connected user terminal at the first time, the initial characteristic information of each user node is determined. Based on each cell node, each user node, the first side feature information, the second side feature information, the initial feature information of each cell node, and the initial feature information of each user node, the first cell user map slice at the first time moment is determined. The first cell user map is determined based on the first cell user map slices at each time point within the first time period.

3. The method for controlling cell transmission power according to claim 2, characterized in that, The first power prediction model includes a spatial fusion layer, a temporal fusion layer, and a linear mapping layer; the step of inputting the first cell user map into the first power prediction model to obtain the transmit power prediction result for each cell includes: The first cell user map is sent to the spatial fusion layer to obtain the spatial fusion feature set of each cell node output by the spatial fusion layer; wherein, each spatial fusion feature set includes the spatial fusion feature information of the corresponding cell node at each time; the spatial fusion layer is used to fuse the initial feature information of each cell node with the initial feature information of its neighboring nodes based on the first cell user map. The spatial fusion feature set of each cell node is input to the temporal fusion layer to obtain the spatiotemporal fusion feature information of each cell node output by the temporal fusion layer; the temporal fusion layer is used to fuse the spatial fusion feature information of each cell node at each time in the first time period based on the spatial fusion feature set of each cell node. The spatiotemporal fusion feature information of each cell node is input into the linear mapping layer to obtain the transmit power prediction result of each cell output by the linear mapping layer.

4. The method for controlling cell transmission power according to claim 1, characterized in that, The step of determining the predicted transmit power for each cell based on the cell network state dataset includes: Based on the multiple cell sub-graphs, the first network status information, the connection relationship information, and the second network status information, multiple second cell user graphs are constructed; wherein, the multiple cell sub-graphs are divided based on the neighbor cell relationship information of each cell, and the multiple second cell user graphs correspond one-to-one with the multiple cell sub-graphs; each second cell user graph is an information expression of the network status relationship between each cell in the corresponding cell sub-graph, and between each cell in the corresponding cell sub-graph and each connected user terminal at each time point; The multiple second cell user maps are input into the second power prediction model to obtain the transmit power prediction result for each cell; wherein, the second power prediction model is trained based on the second cell user map samples and the cell transmit power labels corresponding to the second cell user map samples.

5. The method for controlling cell transmission power according to claim 4, characterized in that, The multiple cell sub-maps are divided in the following manner: Based on the neighboring cell relationships of each cell, a cell topology map is constructed; Based on the spectral clustering algorithm, the cell topology map is segmented to obtain the cell sub-maps.

6. A control device for cell transmission power, characterized in that, include: The acquisition module is used to acquire the cell network status dataset for a first time period; wherein the cell network status dataset includes neighbor cell relationship information between cells, first network status information of each cell at each moment in the first time period and connection relationship information between each cell and user terminal, and second network status information of each connected user terminal at each moment. The prediction module is used to determine the predicted transmit power of each cell based on the cell network status dataset, so as to control the transmit power of each cell. The step of determining the predicted transmit power for each cell based on the cell network state dataset includes: Based on the cell network status dataset, a first cell user graph is constructed; the first cell user graph is an information representation of the network status relationships between the cells and between the cells and the connected user terminals at each time point. The first cell user map is input into the first power prediction model to obtain the transmission power prediction result of each cell; wherein, the first power prediction model is trained based on the first cell user map sample and the cell transmission power label corresponding to the first cell user map sample.

7. An electronic device comprising a processor and a memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the cell transmission power control method according to any one of claims 1 to 5.

8. A non-transitory 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 cell transmit power control method as described in any one of claims 1 to 5.

9. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the cell transmission power control method according to any one of claims 1 to 5.

Images