Network load prediction method and device, electronic equipment and storage medium

Abstract

The invention discloses a network load prediction method and device, electronic equipment and a storage medium, and belongs to the technical field of wireless communication, and the method comprises the steps: determining a space-time undirected graph of cells in a network coverage area, and constructing an overlapping coverage relation matrix of the cells according to the overlapping coverage relation between the cells in the space-time undirected graph; and inputting the historical load data of each cell in the overlapping coverage relation matrix and the space-time undirected graph into a space-time diagram prediction model for load prediction, and outputting a load prediction result of each cell.

Description

Technical Field

[0001] This application belongs to the field of wireless communication technology, and in particular relates to a network load prediction method, apparatus, electronic device, storage medium and computer program product. Background Technology

[0002] With the continuous development of internet technology, wireless network traffic is experiencing explosive growth, and users' demands for wireless network service quality are constantly increasing. The energy consumption problem brought about by wireless network development is also becoming increasingly serious. Accurate prediction of wireless network load helps operators optimize wireless network resource allocation and energy consumption. For example, by predicting the load during peak hours, the transmission power of base stations can be dynamically adjusted, or base stations can be put into sleep mode during low-load periods, minimizing base station energy consumption while ensuring wireless network service quality.

[0003] In some scenarios, wireless network load forecasting primarily focuses on the load forecasting of individual cells. This involves predicting the future load of the wireless network based on load data from a single cell over several consecutive days. However, because cells coordinate with each other to provide network services to users, the load data of a single cell may not be very stable, exhibiting fluctuations and even extremely low loads, resulting in poor predictability. Therefore, predicting the future load of the wireless network based on the load data of a single cell leads to relatively low accuracy. Summary of the Invention

[0004] The purpose of this application is to provide a network load forecasting method, apparatus, electronic device, storage medium, and computer program product that can solve the problem of low accuracy in future load forecasting of wireless networks.

[0005] Firstly, a network load forecasting method is provided, executed by a terminal. The method includes: determining a spatiotemporal undirected graph of cells in the network coverage area; constructing an overlap coverage relationship matrix for each cell based on the overlap coverage relationship between cells in the spatiotemporal undirected graph; inputting the overlap coverage relationship matrix and the historical load data of each cell in the spatiotemporal undirected graph into a spatiotemporal graph forecasting model for load forecasting; and outputting the load forecasting results for each cell.

[0006] Secondly, a network load forecasting device is provided, comprising: a determination module, used to determine the spatiotemporal undirected graph of cells in the network coverage area, and construct an overlapping coverage relationship matrix of each cell based on the overlapping coverage relationship between each cell in the spatiotemporal undirected graph; and a forecasting module, used to input the overlapping coverage relationship matrix and the historical load data of each cell in the spatiotemporal undirected graph into the spatiotemporal graph forecasting model for load forecasting, and output the load forecasting results of each cell.

[0007] Thirdly, embodiments of this application provide an electronic device, including: a memory, a processor, and computer-executable instructions stored in the memory and executable on the processor, wherein the computer-executable instructions, when executed by the processor, implement the steps of performing the method of the first aspect.

[0008] Fourthly, embodiments of this application provide a computer-readable storage medium for storing computer-executable instructions that, when executed by a processor, implement the steps of the method of the first aspect.

[0009] Fifthly, embodiments of this application provide a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions that, when executed by a computer, cause the computer to perform the load forecasting method steps of the network as described in the first aspect.

[0010] In this embodiment, by determining the spatiotemporal undirected graph of cells in the network coverage area, an overlap coverage relationship matrix for each cell is constructed based on the overlap coverage relationships between cells in the spatiotemporal undirected graph. Then, the overlap coverage relationship matrix and the historical load data of each cell in the spatiotemporal undirected graph are input into the spatiotemporal graph prediction model for load prediction, and the load prediction results for each cell are output. Thus, this application uses the overlap coverage relationships between cells in the network coverage area and historical load data to predict the future network load of all cells in the network coverage area using a spatiotemporal graph prediction model. This achieves accurate prediction of the future load of each cell in the network coverage area, thereby ensuring user experience and improving network service quality within the network coverage area. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of this application 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 only some embodiments recorded in this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a flowchart illustrating the network load prediction method provided in the embodiments of this application.

[0013] Figure 2 This is a schematic diagram of a spacetime undirected graph provided in an embodiment of this application.

[0014] Figure 3 This is a schematic diagram of the model structure of the spatiotemporal graph prediction model provided in the embodiments of this application.

[0015] Figure 4This is a schematic diagram illustrating the process of selecting sample data provided in the embodiments of this application.

[0016] Figure 5 This is a schematic diagram of the overlapping coverage area of ​​the main cell and neighboring cells provided in the embodiments of this application.

[0017] Figure 6 This is a schematic diagram of the overlapping coverage area of ​​six cells in the network coverage area provided in the embodiments of this application.

[0018] Figure 7 This is a schematic diagram of the overlapping coverage relationship matrix in the network coverage area provided in the embodiments of this application.

[0019] Figure 8 This is a schematic diagram of the network load prediction device provided in the embodiments of this application.

[0020] Figure 9 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0021] To enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this application.

[0022] The technical solutions provided in this application will be described in detail below with reference to the accompanying drawings and through some embodiments and application scenarios.

[0023] Figure 1 This diagram illustrates a flowchart of a network load prediction method 100 provided in an embodiment of this application. This method can be executed by an electronic device, such as a terminal device or a server device. In other words, the method can be executed by software or hardware installed on the terminal device or server device. Servers include, but are not limited to, single servers, server clusters, cloud servers, or cloud server clusters. Figure 1 As shown, the method may include the following steps.

[0024] Step S101: Determine the spatiotemporal undirected graph of cells in the network coverage area, and construct the overlapping coverage relationship matrix of each cell based on the overlapping coverage relationship between cells in the spatiotemporal undirected graph.

[0025] Specifically, a cell typically corresponds to a geographical area. It transmits and receives wireless signals through a base station (BS) or cell site to provide network communication services to users within that geographical area. The network coverage area refers to the geographical area covered by the wireless communication network provided by each cell. A spatiotemporal undirected graph is a graph structure built with cells as nodes and overlapping coverage relationships between cells as edges. Cells are numbered from 1 to n, where n is the total number of cells. This spatiotemporal undirected graph can be stored using a matrix data structure. Furthermore, the overlapping coverage relationship matrix can be an N×N matrix, where N is the number of nodes in the spatiotemporal undirected graph. The element A in the matrix... ij This represents the overlapping relationship between node i and node j. In a spatiotemporal undirected graph, if node i and node j are connected and have an overlapping relationship, then A... ij The value is between 0 and 1, otherwise it is 0. This encoding method allows the overlap matrix to intuitively reflect the cooperative relationships between nodes in a spatiotemporal undirected graph.

[0026] Furthermore, in some areas, multiple cells may share network coverage, resulting in overlapping coverage areas. Therefore, the overlapping coverage relationship is used to represent the overlap of wireless network coverage areas between cells.

[0027] Furthermore, in this embodiment, the degree of overlap between cells is measured by the overlap coverage degree between cells within the network coverage area. Overlap coverage degree describes the number and distribution of other neighboring cells with signal strengths similar to the main cell within the network coverage area. It is defined as the area or proportion affected when the number of neighboring cells whose signal strength difference with the main cell is less than a certain threshold (including the main cell) exceeds a predetermined number. This threshold and predetermined number can be adjusted according to different network environments and evaluation needs; for example, in this embodiment, the threshold can be 6 dBm. Specifically, overlap coverage degree can be calculated using measurement report (MR) data. MR data includes, but is not limited to, the actual reference signal received power and the reference signal received power threshold value.

[0028] Furthermore, the network coverage area can be a user-specified area, such as a train station, airport, or shopping mall, or the user can specify a network frequency range as the network coverage area.

[0029] Furthermore, the following is... Figure 2 Taking an example, the spatiotemporal undirected graph of a cell in the network coverage area of ​​this application embodiment will be explained. Figure 2 A schematic diagram of a spacetime undirected graph provided in an embodiment of this application, as shown below. Figure 2As shown, there are five cells within a network coverage area, denoted as N0 to N5. These five cells collectively provide network services to users within the coverage area. The five cells within the network coverage area are represented by five vertices in a spatiotemporal undirected graph. In this graph, N0 to N5 represent the five cells. Overlapping coverage refers to the overlapping portions of the network coverage areas between adjacent cells. For example, nodes N0 and N2 represent cells with overlapping coverage areas; therefore, edge E02 represents the overlapping coverage relationship between cells N0 and N2. Each element in the overlap relationship matrix refers to the overlapping coverage relationship between a cell and its neighboring cells, as well as the overlapping coverage relationship within the cell itself. Figure 2 In the matrix, node N0 has overlapping coverage relationships with nodes N1, N2, and N3. The elements in the overlapping coverage relationship matrix of node N0 are the overlapping coverage relationships of node N0 itself, as well as the overlapping coverage relationships of node N0 with nodes N1, N2, and N3 respectively.

[0030] Step S102: Input the historical load data of each cell in the overlapping coverage relationship matrix and the spatiotemporal undirected graph into the spatiotemporal graph prediction model to perform load prediction, and output the load prediction results of each cell.

[0031] Specifically, historical load data includes call quality metrics for users in each cell under network coverage, such as Radio Resource Control (RRC) connections and Physical Resource Block (PRB) utilization. Historical load data can be data from a continuous period of 28 days in 15-minute increments, primarily reflecting the continuous changes in load over a historical period.

[0032] Furthermore, load data has three main characteristics: First, cyclicality, exhibiting periodic fluctuations within a predetermined time period. The length of this predetermined time period can be determined based on actual conditions, such as 28 days or 7 days. Second, dependency, with data from a predetermined time period after any given point in time showing a certain correlation with data from a predetermined time period before that point. For example, from 9:00 AM to 5:00 PM each day, the network experiences higher traffic and load, resulting in higher load data values ​​during this period. While a daily cycle can reflect the short-term cyclicality of the data, there are significant differences between weekend and weekday load data. Therefore, this embodiment chooses a seven-day cycle to predict the load situation for the next day. The seven-day period of load data can balance the short-term cyclicality at the daily level with the differences between weekend and weekday load data. Third, load data may have some gaps. Due to the long data collection period (e.g., 28 days), unforeseen circumstances may cause the loss of data for some time periods during load data collection.

[0033] Furthermore, the long data acquisition cycle can lead to the loss of data for certain time periods during load data collection. Therefore, it is necessary to complete the load data. The methods for completing the load data dataset (Data) are as follows:

[0034] The first method involves filling in the missing load data at position k, while the load data at positions k-1 (Data(k-1)) and k+1 (Data(k+1)) exist. The formula for completing the load data at position k (Data(k)) is as follows:

[0035]

[0036] The second approach is as follows: If the load data at position k is continuously missing (i.e., the load data at positions k-1, k, and k+1 are all missing, but the load data Data(k-2) at position k-2 and the load data Data(k+2) at position k+2 exist), then the formula for completing the load data at positions k-1, k, and k+1 is as follows:

[0037]

[0038] Furthermore, since the load data distribution of each cell may be in different intervals, it is necessary to normalize the load data and normalize the load data as a whole to the interval [0, 1]. In this embodiment, the maximum and minimum value normalization and decimal scaling method are used to normalize the load data.

[0039] Furthermore, the spatiotemporal graph prediction model can be a combination of a temporal series model and a Graph Convolutional Network (GCN) model. In GCN, the aggregation of node features is typically performed through a weighted summation, overlapping and covering elements A in the relation matrix. ij This can be viewed as the weights of node j when aggregating features of node i. This weight allocation method enables GCN to capture the structural information of spatiotemporal undirected graphs and generate node representations with rich semantic information. In this embodiment, the GCN model includes a convolutional graph filtering model, calculated as: X't = σ(A*Xt*W), where Xt is the historical load data of each node at time t, A is the overlap matrix between nodes, W is the weight parameter of the convolutional graph filtering model, and σ is the sigmoid activation function. Furthermore, the temporal models in this embodiment include Long Short-Term Memory (LSTM) networks and Gated Recurrent Unit (GRU) models, etc.

[0040] Furthermore, as Figure 2 Taking the spatiotemporal undirected graph shown as an example, the correlation between nodes extracted by the GCN model in this embodiment of the application is explained. Node N0 has an overlapping relationship with nodes N1, N2, and N3, and their corresponding historical load data at time t is X. N0 t, X N1 t, X N2 t, X N3 The historical load data at each time point and the overlap matrix between nodes N0 and N1, N2, and N3 in the spatiotemporal undirected graph are fed into the spatiotemporal graph prediction model. The GCN model of the spatiotemporal graph prediction model can extract the association information between nodes N0 and N1, N2, and N3 in the overlap matrix, such as the association X' of node N0 generated by the GCN model. N0 t integrates the load data information of nodes N1, N2, and N3 at time t.

[0041] Furthermore, as Figure 2 Using the undirected spatiotemporal graph shown as an example, this application illustrates the variation pattern of historical load data at different times for node N0 extracted by the time series model in this embodiment. Node N0 overlaps with nodes N1, N2, and N3. The historical load data X' of node N0 at time t+1... N0 Historical load data X' for t+1 and t-1 N0 Historical load data X' at times t-1 and t N0 The historical load data of node N0 at each time step and the overlap matrix between node N0 and nodes N1, N2, and N3 in the spatiotemporal undirected graph are fed into the spatiotemporal graph prediction model. The variation pattern X” of the load data of node N0 at time t is output by the time series model of the spatiotemporal graph prediction model. N0 t, which combines the historical load data X' of node N0 at time t-1. N0 Historical load data X' at times t-1 and t+1 N0 The variation pattern X of the load data of node N0 at time t+1. N0 t+1 also incorporates the historical load data X' of node N0 at time t-1. N0 Historical load data X' of node t-1 and node N0 at time t N0 Information about t.

[0042] Furthermore, the predetermined time period for selecting the historical load data of each cell can be selected based on the duration of the predetermined time period for predicting the load situation in the future. For example, wireless networks usually need to predict the load situation in the next 24 hours, so the predetermined time period can be selected as the historical 7 days. The historical load data of the past 7 days of each cell can be input into the spatiotemporal map prediction model to predict the load of each cell in the next 24 hours.

[0043] The following is based on Figure 3 The spatiotemporal graph prediction model provided in this application will be illustrated using an example. Figure 3 This is a schematic diagram of the model structure of the spatiotemporal graph prediction model provided in the embodiments of this application, as shown below. Figure 3 As shown, the spatiotemporal graph prediction model consists of two spatiotemporal convolutional modules and a fully connected output layer. Each spatiotemporal convolutional module contains two temporally gated convolutional layers and an intermediate spatial graph convolutional layer. Each spatiotemporal convolutional module applies residual connections and a bottleneck strategy, which enables the spatiotemporal graph prediction model to converge quickly while preventing overfitting. Corresponding to... Figure 3 The provided spatiotemporal graph prediction model, as an optional embodiment of this application, inputs the historical load data of each cell in the overlapping coverage relationship matrix and the spatiotemporal undirected graph into the spatiotemporal graph prediction model for load prediction. This includes: using the time-gated convolutional layer of the spatiotemporal graph prediction model to extract time features from the historical load data of each cell to obtain the variation patterns of the historical load data of each cell; using the spatial graph convolutional layer of the spatiotemporal graph prediction model to extract features from the historical load data of each cell within a preset time period and the overlapping coverage relationship matrix of each cell to obtain the correlation between the load data of each cell; and using the fully connected layer of the spatiotemporal graph prediction model to correlate the variation patterns and correlations to obtain the load prediction results of each cell within the network coverage area.

[0044] Specifically, the periodicity and trend of historical load data in each cell are obtained through a time-gated convolutional layer. The historical load data first undergoes a one-dimensional causal convolution in the time-gated convolutional layer, and then passes through a gated linear unit (h(X)) of the time-gated convolutional layer. The gated linear unit h(X) is defined as follows:

[0045] h(X)=(X*W+b)·σ(X*V+c)

[0046] Where X is the input historical load data, W and V are two different convolution kernels, both with a kernel size of 3 and 16 channels, σ is the sigmoid activation function, "·" represents multiplication, and b and c are weight parameters.

[0047] Furthermore, by integrating the correlation of load data between a cell and its neighboring cells through a spatial graph convolutional layer, and inputting the overlap coverage matrix into the spatial graph convolutional layer, the spatial graph convolutional layer can learn the correlation between the load data of the current node and its neighboring nodes after fusion. The operator h of the spatial graph convolutional layer... i The formula for (X) is defined as:

[0048]

[0049] Where, N iLet X be the set of neighboring cells of cell i, and let j be the set of neighboring cells of cell i. j It is the eigenvector of the load data and overlapping coverage relationship matrix of cell j in the previous layer, A ij Let σ represent the overlapping coverage relationship between cell i and cell j in the overlapping coverage relationship matrix, where σ is the ReLU activation function and W is the convolution kernel.

[0050] Furthermore, the spatiotemporal convolution module uses normalization and fully connected layer output layers to provide load prediction results based on the variation patterns of the temporally gated convolutional layer output and the correlation between the output of the spatial graph convolutional layer. These results are used as the output of the spatiotemporal graph prediction model, where the load prediction results include the future load changes for each cell.

[0051] In this embodiment, by determining the spatiotemporal undirected graph of cells in the network coverage area, an overlap coverage relationship matrix for each cell is constructed based on the overlap coverage relationships between cells in the spatiotemporal undirected graph. Then, the overlap coverage relationship matrix and the historical load data of each cell in the spatiotemporal undirected graph are input into the spatiotemporal graph prediction model for load prediction, and the load prediction results for each cell are output. Thus, this application uses the overlap coverage relationships between cells in the network coverage area and historical load data to predict the future network load of all cells in the network coverage area using a spatiotemporal graph prediction model. This achieves accurate prediction of the future load of each cell in the network coverage area, thereby ensuring user experience and improving network service quality within the network coverage area.

[0052] Furthermore, the spatiotemporal graph prediction model can be trained using the cross-entropy loss function and the Adam optimizer. The training batch size can be set to 16, and the number of training iterations can be set to 100. The sample data for training the spatiotemporal graph prediction model can be divided into input data and output data, with the output data serving as the true value for the spatiotemporal graph prediction model to be trained. In this embodiment, the historical load data can be in a 7-day cycle. Therefore, the historical load data for 7 consecutive days is used as the input data X for the spatiotemporal graph prediction model to be trained, and the historical load data for the following 8th day is used as the output data Y. The following uses... Figure 4 The process of selecting sample data will be explained using an example. Figure 4 This is a schematic diagram illustrating the sample data selection process provided in the embodiments of this application, such as... Figure 4 As shown, it contains 14 days of historical load data. Each square, 1-14, represents the load data for each day. Since the data interval is 15 minutes, there are a total of 96 data points with a 15-minute time granularity in a 24-hour day. Taking Sample 1 as an example, the data from days 1-7 is used as the input data, and the data from the following day 8 is used as the output data, and so on. Therefore, the input data for the sample is a matrix of (7, 96), and the output data for the sample is a matrix of (1, 96).

[0053] Furthermore, to obtain the above sample data, historical load data can be extracted using methods such as sliding windows. The window size is eight days, the sliding step is one day, the historical load data of the first seven days in the window is used as input data, and the historical load data of the last day is used as output data.

[0054] After obtaining the sample data, the sample data is integrated according to the cell number order. Assuming there are n cells and m days of historical load data, the final integrated sample data is a sample matrix of size (n, m-7), each sample matrix is ​​(8, 96), and the overall sample matrix size is (n, m-7, 8, 96).

[0055] When training the spatiotemporal map prediction model, the model predicts the load data for the 8th day using the historical load data for 7 consecutive days. Then, the predicted load data for the 8th day is compared with the historical load data for the 8th day. The model is iteratively trained using the error between the two data until the error between them meets the preset accuracy.

[0056] Furthermore, as an optional embodiment of this application, determining the spatiotemporal undirected graph of cells in the network coverage area includes: determining the overlapping coverage relationship between the main cell and neighboring cells based on the actual reference signal received power, reference signal received power threshold value, and historical load data of each cell in the spatiotemporal undirected graph of the main cell and neighboring cells in each group of cells in the network coverage area, wherein each group includes one main cell and at least one neighboring cell adjacent to the main cell; and determining the spatiotemporal undirected graph with each cell as a node and the overlapping coverage relationship between cells as an edge.

[0057] Specifically, the following are examples: Figure 5 The primary cell and neighboring cells within the network coverage area of ​​this application embodiment will be described using an example. Figure 5 This is a schematic diagram of the overlapping coverage area of ​​the primary cell and neighboring cells provided in an embodiment of this application. For example... Figure 5 As shown, there is a primary cell A and a neighboring cell B. Both primary cell A and neighboring cell B have their own coverage edges. The area inside the coverage edge is the area that their signals can cover. The overlapping coverage area 501 between the two cells is the area of ​​overlapping coverage between the two cells. For primary cell A and neighboring cell B, MR data measurement and reporting can be performed by the base station or network equipment. Each record represents the MR data measurement record of a mobile device. The MR data includes the Reference Signal Received Power (RSRP) values ​​of the primary cell providing network services to the mobile device and its neighboring cells.

[0058] As an optional embodiment of this application, the overlapping coverage relationship between the main cell and neighboring cells is determined based on the actual reference signal received power, reference signal received power threshold value, and historical load data of each cell in the spatiotemporal undirected graph. This includes: selecting a first number of data entries and a second number of data entries from the historical load data according to a first condition and a second condition, wherein the first condition is that the actual reference signal received power of the main cell is greater than its own reference signal received power threshold value, and the second condition is that the actual reference signal received power of the neighboring cell is greater than its own reference signal received power threshold value; selecting a first number of data entries that simultaneously satisfies the first condition and a second number of data entries that satisfies the first condition from the historical load data; determining the overlapping coverage degree between the main cell and each neighboring cell in each group based on the first number of data entries and the second number of data entries; and determining the overlapping coverage relationship between the main cell and neighboring cells based on the overlapping coverage degree.

[0059] Specifically, the ratio of the first data entry number to the second data entry number can be used to determine the overlap coverage between the primary cell and its neighboring cells. The overlap coverage is calculated using the following formula:

[0060]

[0061] Among them, Cov ij The overlap coverage of the neighboring cell numbered j relative to the primary cell numbered i, where i is not equal to j.

[0062] The first condition is that the RSRP of the primary cell numbered i is greater than its own coverage edge RSRP threshold value.

[0063] The second condition is that the RSRP of the neighboring cell j is greater than the coverage edge RSRP threshold value of the neighboring cell j.

[0064] Furthermore, the overlap coverage meter for the primary cell numbered i with respect to itself can be Cov. ii =1.

[0065] Furthermore, with Figure 5 Taking the primary cell A and neighboring cell B as an example, the overlap coverage can also be determined using any of the following methods:

[0066] (1) The overlap coverage of the main cell A with its neighboring cell B = the area of ​​overlap coverage between the main cell A and the neighboring cell B / the coverage area of ​​cell A.

[0067] (2) The overlap coverage of the main cell A with its neighboring cell B = the number of mobile devices in the overlapping area of ​​the main cell A and the neighboring cell B / the number of mobile devices in the coverage area of ​​cell A.

[0068] After determining the overlap coverage, an overlap relationship matrix is ​​determined based on the overlap coverage. As an optional embodiment of this application, determining the overlap coverage relationship between the primary cell and neighboring cells based on the overlap coverage includes: sorting the overlap coverage between the primary cell and each neighboring cell in each group in descending order; selecting the top N neighboring cells in the overlap coverage ranking as the basic coverage cells of the primary cell, where N is the upper limit of the number of basic coverage cells; when the neighboring cell is the basic coverage cell of the primary cell, determining the overlap coverage relationship between the primary cell and the neighboring cell using the overlap coverage of the neighboring cell and the maximum overlap coverage among the basic coverage cells; when the neighboring cell is not the basic coverage cell of the primary cell, determining the overlap coverage relationship between the primary cell and the neighboring cell as a preset value.

[0069] Specifically, the overlap coverage between the main cell and each neighboring cell is sorted from highest to lowest. Then, the neighboring cells with the highest "maximum number of basic coverage cells" are selected. If the number of neighboring cells of the main cell is less than the maximum number of basic coverage cells, all neighboring cells are selected, and these selected neighboring cells are used as the basic coverage cells of the main cell. The maximum number of basic coverage cells can be set by the user or configured in the cell's parameter information. In this embodiment, the maximum number of basic coverage cells is configured to 6.

[0070] Specifically, the overlap coverage relationship between the main cell and the neighboring cell can be determined based on whether the neighboring cell is a basic coverage cell of the main cell. If the neighboring cell is a basic coverage cell of the main cell, the overlap coverage relationship can be determined as the ratio of the overlap coverage degree of the neighboring cell to the maximum overlap coverage degree. If the neighboring cell is not a basic coverage cell of the main cell, the overlap coverage relationship between the main cell and the neighboring cell is determined to be a preset value.

[0071] Specifically, the overlapping relationship can be represented by the following formula:

[0072]

[0073] In the above formula, Cov ij Cov represents the overlap coverage of the neighboring cell j relative to the primary cell i, where i is not equal to j. max The maximum overlap coverage of all basic coverage cells of primary cell i. A ij This is the overlap coverage matrix from primary cell i to neighboring cell j. The overlap coverage from primary cell i to itself is denoted as A. ii =1. "0" is the default value.

[0074] The following is based on Figure 6 and Figure 7 The overlap relationship matrix provided in the embodiments of this application will be explained using an example. Figure 6This is a schematic diagram of the overlapping coverage area of ​​six cells in the network coverage area provided in this application embodiment. Figure 7 This is a schematic diagram of the overlapping coverage relationship matrix in the network coverage area provided in an embodiment of this application. For example... Figure 6 As shown, assuming there are 6 cells in the network coverage area, Figure 6 The system uses numbers 0-5 to represent six cells. Numbers 0 and 4 represent primary cells, and numbers 1, 2, 3, and 5 represent adjacent cells. Primary cell 0 overlaps with adjacent cells 1, 2, and 3, and primary cell 4 overlaps with adjacent cells 3 and 5. For example... Figure 7 As shown, the overlapping coverage matrix of each cell in the network coverage area is 5*5. Here, element A in the overlapping coverage matrix... 01 A 00 And A 05 Let's take the numerical values ​​of the elements in the matrix as an example to illustrate, A 01 The overlapping coverage relationship between cell 0 and cell 1 can be calculated as described in the above embodiment. A 00 Indicates the overlapping coverage relationship of cell 0 itself, A 00 =1, A 05 This indicates the overlapping coverage relationship between cell 0 and cell 5, such as... Figure 6 As shown, since there is no overlapping coverage area between cell 0 and cell 5, therefore A 05 The value of is 0. The meanings of the remaining elements in this overlap matrix are analogous, and will not be repeated here in the embodiments of this application.

[0075] Furthermore, as an optional embodiment of this application, after selecting the top N neighboring cells in the overlap coverage ranking as the basic coverage cells of the main cell, the method further includes: when the main cell is in a carrier shutdown energy-saving state or the main cell is a poor quality cell, determining the target cell from the basic coverage cells to provide network services to the area covered by the main cell.

[0076] Specifically, in scenarios where multiple cells collaborate to provide network services within a coverage area, such as during carrier shutdown for energy conservation, energy-saving cells cease service and enter a shutdown state. In scenarios where the primary cell is a low-quality cell, the network provided by that cell is poor. To address these scenarios, cell handover is necessary. A high-quality cell can be selected from the basic coverage cells, thus switching the energy-saving or low-quality cell to a high-quality cell to continue providing network service to users. Because the basic coverage cells have a high degree of overlap with the primary cell, switching an energy-saving or low-quality cell to a basic coverage cell has a high success rate and ensures uninterrupted network service, improving network reliability.

[0077] Figure 8 The diagram shows a structural schematic of a network load forecasting device provided in an embodiment of this application. The device 800 includes: a determination module 801, used to determine the spatiotemporal undirected graph of cells in the network coverage area, and construct an overlapping coverage relationship matrix of each cell based on the overlapping coverage relationship between each cell in the spatiotemporal undirected graph; and a forecasting module 802, used to input the overlapping coverage relationship matrix and the historical load data of each cell in the spatiotemporal undirected graph into the spatiotemporal graph forecasting model for load forecasting, and output the load forecasting results of each cell.

[0078] In this embodiment, by determining the spatiotemporal undirected graph of cells in the network coverage area, an overlap coverage relationship matrix for each cell is constructed based on the overlap coverage relationships between cells in the spatiotemporal undirected graph. Then, the overlap coverage relationship matrix and the historical load data of each cell in the spatiotemporal undirected graph are input into the spatiotemporal graph prediction model for load prediction, and the load prediction results for each cell are output. Thus, this application uses the overlap coverage relationships between cells in the network coverage area and historical load data to predict the future network load of all cells in the network coverage area using a spatiotemporal graph prediction model. This achieves accurate prediction of the future load of each cell in the network coverage area, thereby ensuring user experience and improving network service quality within the network coverage area.

[0079] Optionally, the prediction module 802 is further configured to use the time-gated convolutional layer of the spatiotemporal graph prediction model to extract time features from the historical load data of each cell, thereby obtaining the variation pattern of the historical load data of each cell; use the spatial graph convolutional layer of the spatiotemporal graph prediction model to extract features from the historical load data of each cell within a preset time period and the overlapping coverage matrix of each cell, thereby obtaining the correlation between the load data of each cell; and use the fully connected layer of the spatiotemporal graph prediction model to correlate the variation pattern and correlation, thereby obtaining the load prediction results of each cell within the network coverage area.

[0080] Optionally, the determining module 801 is further configured to determine the overlapping coverage relationship between the main cell and the neighboring cells based on the actual reference signal received power, the reference signal received power threshold value, and the historical load data of each cell in the spatiotemporal undirected graph of the main cell and the neighboring cells in each group of cells within the network coverage area. Each group contains one main cell and at least one neighboring cell adjacent to the main cell. The spatiotemporal undirected graph is determined with each cell as a node and the overlapping coverage relationship between cells as an edge.

[0081] Optionally, the determining module 801 is further configured to select a first number of data entries and a second number of data entries from historical load data based on a first condition and a second condition, wherein the first condition is that the actual reference signal received power of the primary cell is greater than its own reference signal received power threshold value, and the second condition is that the actual reference signal received power of the neighboring cell is greater than its own reference signal received power threshold value; select a first number of data entries that simultaneously satisfy the first condition and the second number of data entries that satisfy the first condition from historical load data; determine the overlap coverage between the primary cell and each neighboring cell in each group based on the first number of data entries and the second number of data entries; and determine the overlap coverage relationship between the primary cell and the neighboring cells based on the overlap coverage.

[0082] Optionally, the determining module 801 is further configured to sort the overlap coverage between the main cell and each neighboring cell in each group in descending order; select the top N neighboring cells in the overlap coverage sort as the basic coverage cells of the main cell, where N is the upper limit of the number of basic coverage cells; when the neighboring cell is the basic coverage cell of the main cell, determine the overlap coverage relationship between the main cell and the neighboring cell using the overlap coverage of the neighboring cell and the maximum overlap coverage among the basic coverage cells; when the neighboring cell is not the basic coverage cell of the main cell, determine the overlap coverage relationship between the main cell and the neighboring cell as a preset value.

[0083] Optionally, the determining module 801 is also used to determine the ratio of the overlapping coverage of adjacent cells to the maximum overlapping coverage as the overlapping coverage relationship.

[0084] Optionally, the determining module 801 is also used to determine the target cell from the basic coverage cells to provide network services to the area covered by the main cell when the main cell is in a carrier shutdown energy-saving state or the main cell is a poor quality cell.

[0085] Optionally, the determining module 801 is also used to determine the ratio of the first data entry number and the second data entry number to determine the overlap coverage between the primary cell and each neighboring cell.

[0086] The device 800 provided in this application embodiment can execute the methods in the preceding method embodiments and realize the functions and beneficial effects of the methods in the preceding method embodiments, which will not be repeated here.

[0087] Figure 9This diagram illustrates the hardware structure of an electronic device implementing the embodiments of this application. Referring to the diagram, at the hardware level, the electronic device includes a processor and optionally, an internal bus, a network interface, and a memory. The memory may include RAM, such as high-speed random-access memory (RAM), or non-volatile memory, such as at least one disk storage device. Of course, the electronic device may also include other hardware required for other services.

[0088] The processor, network interface, and memory can be interconnected via an internal bus, which can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, this diagram uses only a single bidirectional arrow, but it does not imply that there is only one bus or one type of bus.

[0089] Memory is used to store programs. Specifically, programs may include program code, which includes computer operation instructions. Memory may include main memory and non-volatile memory, and provides instructions and data to the processor.

[0090] The processor reads the corresponding computer program from non-volatile memory into main memory and then executes it, forming a device at the logical level that is assigned to a specific user. The processor executes the program stored in memory and specifically performs the following: Figure 1 The methods disclosed in the embodiments shown achieve the functions and beneficial effects of the methods described in the preceding method embodiments, and will not be repeated here.

[0091] The above is as stated in this application. Figure 1The methods disclosed in the illustrated embodiments can be applied to or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above methods can be completed by integrated logic circuits in the processor's hardware or by instructions in software form. The processor can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software module can reside in a mature storage medium in the field, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method.

[0092] The electronic device can also execute the methods described in the preceding method embodiments and achieve the functions and beneficial effects of the methods described in the preceding method embodiments, which will not be repeated here.

[0093] Of course, in addition to software implementation, the electronic device of this application does not exclude other implementation methods, such as logic devices or a combination of hardware and software, etc. In other words, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0094] This application also proposes a computer-readable storage medium that stores one or more programs, which, when executed by an electronic device including multiple applications, cause the electronic device to perform... Figure 1 The methods disclosed in the embodiments shown achieve the functions and beneficial effects of the methods described in the preceding method embodiments, and will not be repeated here.

[0095] The computer-readable storage medium includes read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks, etc.

[0096] Furthermore, embodiments of this application also provide a computer program product, the computer program product including a computer program stored on a non-transitory computer-readable storage medium, the computer program including program instructions, which, when executed by a computer, implement the following process: Figures 1-3 The methods disclosed in the embodiments shown achieve the functions and beneficial effects of the methods described in the preceding method embodiments, and will not be repeated here.

[0097] In summary, the above description is merely a preferred embodiment of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

[0098] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, a computer can be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smartphone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or any combination of these devices.

[0099] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0100] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0101] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

Claims

1. A method for predicting network load, characterized in that, The load forecasting method for the network includes: Determine the spatiotemporal undirected graph of cells in the network coverage area, and construct the overlapping coverage relationship matrix of each cell based on the overlapping coverage relationship between each cell in the spatiotemporal undirected graph; The overlapping coverage relationship matrix and the historical load data of each cell in the spatiotemporal undirected graph are input into the spatiotemporal graph prediction model to perform load prediction, and the load prediction results of each cell are output.

2. The network load forecasting method as described in claim 1, characterized in that, The step of inputting the overlapping coverage relationship matrix and the historical load data of each cell in the spatiotemporal undirected graph into the spatiotemporal graph prediction model for load prediction includes: The temporal gating convolutional layer of the spatiotemporal graph prediction model is used to extract the temporal features of the historical load data of each cell to obtain the variation pattern of the historical load data of each cell. The spatial graph convolutional layer of the spatiotemporal graph prediction model is used to extract features from the historical load data of each cell within a preset time period and the overlapping coverage matrix of each cell to obtain the correlation between the load data of each cell. By using the fully connected layer of the spatiotemporal graph prediction model to correlate the change patterns and correlations, the load prediction results of each cell within the network coverage area are obtained.

3. The network load forecasting method as described in claim 1 or 2, characterized in that, The determination of the spatiotemporal undirected graph of cells in the network coverage area includes: Based on the actual reference signal received power, reference signal received power threshold value, and historical load data of each cell in the spatiotemporal undirected graph of the main cell and the adjacent cells in each group of cells within the network coverage area, the overlapping coverage relationship between the main cell and the adjacent cells is determined, wherein each group includes one main cell and at least one adjacent cell adjacent to the main cell. The spatiotemporal undirected graph is determined by using each of the aforementioned cells as nodes and the overlapping coverage relationships between the aforementioned cells as edges.

4. The network load forecasting method as described in claim 3, characterized in that, The step of determining the overlap coverage relationship between the primary cell and the neighboring cells based on the actual reference signal received power, the reference signal received power threshold value, and the historical load data of each cell in the spatiotemporal undirected graph includes: The first number of data entries and the second number of data entries are selected from the historical load data according to the first condition and the second condition, wherein the first condition is that the actual reference signal received power of the primary cell is greater than its own reference signal received power threshold value, and the second condition is that the actual reference signal received power of the neighboring cell is greater than its own reference signal received power threshold value. From the historical load data, select a first number of data entries that simultaneously satisfy both the first and second conditions, and a second number of data entries that simultaneously satisfy the first condition. The overlap coverage between the main cell and each of the neighboring cells in each group is determined based on the first data count and the second data count. The overlapping coverage relationship between the primary cell and the neighboring cells is determined based on the overlapping coverage.

5. The network load forecasting method as described in claim 4, characterized in that, Determining the overlapping coverage relationship between the primary cell and the neighboring cells based on the overlapping coverage includes: Sort the overlap coverage between the main cell and each of the neighboring cells in each group in descending order; The top N adjacent cells in the overlap coverage ranking are selected as the basic coverage cells of the main cell, where N is the upper limit of the number of basic coverage cells. When the adjacent cell is the base coverage cell of the main cell, the overlapping coverage relationship between the main cell and the adjacent cell is determined by using the overlapping coverage of the adjacent cell and the maximum overlapping coverage in the base coverage cell; If the adjacent cell is not a base coverage cell of the main cell, the overlap coverage relationship between the main cell and the adjacent cell is determined to be a preset value.

6. The network load forecasting method as described in claim 5, characterized in that, The step of determining the overlap coverage relationship between the primary cell and the neighboring cells using the overlap coverage of the adjacent cells and the maximum overlap coverage among the basic coverage cells includes: The ratio of the overlap coverage of the adjacent cells to the maximum overlap coverage is determined as the overlap coverage relationship.

7. The network load forecasting method as described in claim 5, characterized in that, After selecting the top N neighboring cells in the overlap coverage ranking as the base coverage cells of the primary cell, the method further includes: When the primary cell is in a carrier-off energy-saving state or the primary cell is a poor-quality cell, a target cell is determined from the basic coverage cells to provide network services to the area covered by the primary cell.

8. The network load forecasting method as described in claim 4, characterized in that, The step of determining the overlap coverage between the primary cell and each of the neighboring cells in each group based on the first data count and the second data count includes: The ratio of the first number of data entries to the second number of data entries is determined as the overlap coverage between the main cell and each of the neighboring cells.

9. A network load forecasting device, characterized in that, include: The determination module is used to determine the spatiotemporal undirected graph of cells in the network coverage area, and construct the overlapping coverage relationship matrix of each cell based on the overlapping coverage relationship between each cell in the spatiotemporal undirected graph; The prediction module is used to input the overlapping coverage relationship matrix and the historical load data of each cell in the spatiotemporal undirected graph into the spatiotemporal graph prediction model to perform load prediction, and output the load prediction results of each cell.

10. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the load forecasting method for the network as described in any one of claims 1-8.

11. A computer-readable medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the network load forecasting method as described in any one of claims 1-8.

12. A computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program including program instructions that, when executed by a computer, cause the computer to perform the steps of the network load forecasting method as described in any one of claims 1-8.

Images