WO2024040960A1 - 动态负荷预测方法、装置、基站和存储介质 - Google Patents

动态负荷预测方法、装置、基站和存储介质 Download PDF

Info

Publication number
WO2024040960A1
WO2024040960A1 PCT/CN2023/083168 CN2023083168W WO2024040960A1 WO 2024040960 A1 WO2024040960 A1 WO 2024040960A1 CN 2023083168 W CN2023083168 W CN 2023083168W WO 2024040960 A1 WO2024040960 A1 WO 2024040960A1
Authority
WO
WIPO (PCT)
Prior art keywords
load
base station
cell
neighbor
layer
Prior art date
Application number
PCT/CN2023/083168
Other languages
English (en)
French (fr)
Inventor
刘学斌
Original Assignee
中兴通讯股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 中兴通讯股份有限公司 filed Critical 中兴通讯股份有限公司
Publication of WO2024040960A1 publication Critical patent/WO2024040960A1/zh

Links

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W16/00—Network planning, e.g. coverage or traffic planning tools; Network deployment, e.g. resource partitioning or cells structures
    • H04W16/22—Traffic simulation tools or models
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W24/00—Supervisory, monitoring or testing arrangements
    • H04W24/06—Testing, supervising or monitoring using simulated traffic
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W28/00—Network traffic management; Network resource management
    • H04W28/02—Traffic management, e.g. flow control or congestion control
    • H04W28/10—Flow control between communication endpoints
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W72/00—Local resource management
    • H04W72/12—Wireless traffic scheduling
    • H04W72/1263—Mapping of traffic onto schedule, e.g. scheduled allocation or multiplexing of flows

Definitions

  • the present disclosure relates to the field of base station systems, and in particular, to a dynamic load prediction method, device, base station and storage medium.
  • the load of a mobile communication network usually refers to the number of users (number of terminals) accessing the network base station cell, or the various resource occupancy rates of the network base station cell.
  • Network load is a data indicator that has attracted much attention in the operation and maintenance of mobile communication networks. It is a key indicator for real-time network monitoring and is directly related to the smooth operation of the network, user perception and energy saving. Therefore, it is a research hotspot and application of mobile communication networks. focus.
  • network base stations can be assisted and guided to implement various energy-saving measures, wake up in time, implement load balancing in advance, and admit and control special guarantees for high traffic, ensuring the smooth operation of base stations and user service perception.
  • the differential integrated autoregressive moving average (ARIMA) model is commonly used.
  • the historical load data of the network base station cells is used to train, learn and regress the model, so as to realize the prediction of each base station cell. Forecasting of future loads; deep neural network models are also used, such as: Recurrent Neural Network (RNN) model and improved Long-short term Memory (LSTM) to achieve load forecasting; but it is common knowledge
  • RNN Recurrent Neural Network
  • LSTM Long-short term Memory
  • the prediction method usually requires a long enough time and enough historical data to achieve effective training and learning, and the prediction can be more accurate.
  • the load prediction also requires that the load changes of the base station cell have periodic rules.
  • the present disclosure provides a dynamic load prediction method, device, base station and storage medium.
  • the present disclosure provides a dynamic load prediction method, which is applied to target base station cells under a high-speed private network, including: determining cross-layer co-directional groups among multiple base station cells located on preset load migration lines, Among them, the base station cells located at both ends of the cross-layer co-directional group are the Kth layer main neighbor cells corresponding to the target base station cell, and K is a preset value; the real-time load at the current moment and the history corresponding to the previous moment are obtained.
  • the historical predicted load is used to represent the load migration changes that will occur at the current moment predicted at the previous moment; receive the left neighbor zone in-direction group and the right neighbor zone in-direction group of the cross-layer in-direction group and transmit them respectively cross-layer neighboring cell load data, wherein the corresponding cross-layer neighboring cell load data includes historical load data and real-time load data, and the historical load data is used to represent the K layer of the corresponding neighboring cell same-direction group.
  • the real-time load data is used to represent the actual load of the K-layer primary neighbor cell in the corresponding neighboring cell same-direction group at the current moment; using the preset load prediction model, Load prediction is performed based on the real-time load, the historical predicted load and the cross-layer neighboring cell load data to obtain the forward predicted load and reverse predicted load corresponding to the target base station cell, and based on the forward prediction
  • the load and reverse forecast load determine the corresponding load forecast results.
  • the present disclosure provides a dynamic load prediction device, applied to a target base station cell under a high-speed private network, including: a determination module configured to determine, among multiple base station cells located on a preset load migration line, Cross-layer in-directional group, in which the base station cells located at both ends of the cross-layer in-directional group are the K-th layer primary neighbor cells corresponding to the target base station cell, and K is a preset value; the acquisition module is configured to obtain the current The real-time load at the moment and the historical predicted load corresponding to the previous moment, the historical predicted load is used to represent the load migration changes that will occur at the current moment predicted at the previous moment; the receiving module is configured to receive the cross-layer synchronization The cross-layer neighboring cell load data transmitted to the left neighboring cell in-directional group and the right neighboring cell in-directional group respectively, wherein the corresponding cross-layer neighboring cell load data includes historical load data and actual load data.
  • the historical load data is used to represent the predicted load corresponding to the K-layer main neighbor cells of the corresponding neighboring cell group in the same direction at the previous moment.
  • the real-time load data is used to represent the corresponding neighboring cells in the same direction.
  • the actual load of the K-layer primary neighbor cells owned by the group at the current moment; the prediction module is configured to use a preset load prediction model, based on the real-time load, the historical predicted load and the cross-layer neighbor load data Perform load prediction to obtain the forward predicted load and reverse predicted load corresponding to the target base station cell, and determine the corresponding load prediction result based on the forward predicted load and reverse predicted load.
  • a base station including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory complete communication with each other through the communication bus; the memory stores computer programs; the processor, When executing the program stored in the memory, the dynamic load prediction method described in the first aspect is implemented.
  • a fourth aspect provides a computer-readable storage medium on which a computer program is stored.
  • the computer program is executed by a processor, the dynamic load prediction method as described in the first aspect is implemented.
  • Figure 1 is a schematic flowchart of a dynamic load prediction method provided by an embodiment of the present disclosure
  • Figure 2 is a schematic structural diagram of a cross-layer co-directional group in an embodiment of the present disclosure
  • Figure 3 is a structural block diagram of a dynamic load prediction device provided by an embodiment of the present disclosure.
  • Figure 4 is a schematic structural diagram of a base station according to an embodiment of the present disclosure.
  • Base Station is the basic unit of mobile networks.
  • Cell the English abbreviation of Cell, is the basic unit of a base station. Under the cloverleaf structure of a conventional cellular base station, one base station has three cells with 120° sector coverage. Under the high-speed rail private network, one base station has two or more cells.
  • User Equipment User Equipment
  • New Radio (NEW Ratio, NR for short).
  • LTE Long Term Evolution
  • RNN Recurrent Neural Network
  • LSTM Long Short-term Memory
  • FIG 1 is a schematic flowchart of a dynamic load prediction method provided by an embodiment of the present disclosure.
  • an embodiment of the present disclosure provides a dynamic load prediction method, which is applied to a target base station cell under a high-speed private network.
  • the method includes the following steps S101 to S104.
  • Step S101 Determine a cross-layer co-directional group among multiple base station cells located on the preset load migration line, where the base station cells located at both ends of the cross-layer co-directional group are the Kth layer primary neighbor cells corresponding to the target base station cell.
  • K is the default value.
  • the dynamic load prediction method is to use the corresponding base station cell to predict the load that is coming at the next moment based on the relevant load at the current moment and the relevant load at the previous moment, and learn through the load prediction.
  • the goal is to take prediction accuracy as the goal, but to consider the predicted load and the actual load together, and to obtain the corresponding load change trend and change value of the base station cell in the future, so as to achieve energy saving, load control, traffic guarantee, etc. of the base station cell.
  • the operation provides effective reference and guidance; in this embodiment, the target base station cell is used for load prediction, that is, one of the multiple base station cells in the load migration direction.
  • the high-speed private network under consideration is the high-speed private network corresponding to the straight-line high-speed railway line. Therefore, for each base station cell of multiple base station cells under the high-speed private network, the corresponding cross-layer same-direction There is only one group; there are two cross-layer same-direction groups corresponding to the base station cell where the central intersection point of the T-shaped or cross-shaped high-speed rail line is located. For the base station cell where the central intersection point of the " ⁇ "-shaped high-speed rail line is located, there are There are three cross-layer co-directional groups. For the base station cell where the central intersection point of the M-shaped high-speed rail line is located, there are four cross-layer co-directional groups.
  • each base station cell in N determines the primary neighbor cell, that is, in the left and right neighbor cells corresponding to the base station cell, the primary neighbor cell of each base station cell is screened and determined based on the statistical data of "neighbor cell pair" switching times.
  • Neighboring cells for example, hourly granular statistics of the number of handovers of base station cell C i based on its own period of time (such as 168 hours corresponding to 1 week) and its neighboring cells, including handovers to neighboring cells and neighboring cells.
  • the number of handovers to the base station cell C i is divided into two types of neighbor cells according to the clustering algorithm, and the neighbor cell with a greater number of handovers is used as the main neighbor cell of the base station cell C i ; of course, in some optional implementations,
  • the network planning and optimization personnel can use the network engineering parameter information of the base station cell C i and neighboring cells, such as installation location, coverage direction, etc., or refer to the handover statistical data to manually filter out the base station cell C i along the high-speed rail line.
  • the nearest neighboring cells on the left and right sides of the base station cell C i are set as the main neighboring cells of the base station cell C i; then, the nearest neighboring cells on the left and right sides of the base station cell C i are used as the same direction group corresponding to the base station cell C i , and then based on The correlation between the co-directional groups corresponding to multiple base station cells is to determine the cross-layer co-directional group corresponding to each base station cell. For example: if C 2 is the right neighbor of the co-directional group of C 1 , C 3 is the co-directional group of C 2 To the right neighbor of the group, it can be confirmed that C 3 is the right neighbor of C 1 spanning 1 layer. By analogy, the right neighbor C k of C 1 spanning K layers can be obtained. ⁇ C 1 , C 2 ,...C K ⁇ form the cross-layer same-direction group of C 1 across K layers.
  • the determined cross-layer codirectional group corresponding to the target base station cell includes two cross-layer codirectional groups, the left and the right, for example: the left neighbor cell codirectional group and the right neighbor cell codirectional group.
  • the left neighbor cell same-direction group and the right neighbor cell same-direction group both have K-layer main neighbor cells, and K is a preset value.
  • the target base station cell C i hopes to predict the possible future loads earlier in advance, and the induction changes
  • K can be set very large, but this will increase the cost and difficulty of transmitting load information across layers, and the random uncertainty will also increase. Therefore, it is necessary to take into account the target requirements and overhead costs of load forecasting. Wait to set the K value, for example: K ⁇ 9.
  • Step S102 Obtain the real-time load at the current moment and the historical predicted load corresponding to the previous moment.
  • the historical predicted load is used to represent the load migration changes that will occur at the current moment predicted at the previous moment.
  • the interval is a set time period T.
  • the predicted load refers to the load that will occur at the current moment that has been predicted at the previous moment; considering the dimension of base station cell objects, the base station cells participating in load prediction including items There are two dimensions: a base station cell and a cross-layer co-directional group.
  • the load corresponding to the target base station cell includes the real-time load at the current moment and the predicted load corresponding to the previous moment, because the predicted load is a prediction of the load changes that may occur in the base station cell in the future.
  • the predicted load corresponding to a moment is used to determine the difference between the real-time load and the predicted load at the current moment to represent whether the load change trend of the target base station cell itself matches the predicted change trend.
  • the time period T needs to be set taking into account the time situation. For example: if T is too small, the overhead of load prediction and load information transmission will be very large, and if T is too large, the high-speed The train is likely to pass through the target base station cell and some of its associated neighboring cells, causing the corresponding base station cell to be unable to sample the real actual load, and the target base station cell to be unable to achieve effective load prediction; when considering the real-time load, the corresponding current load of each base station cell is The real-time load corresponding to the moment needs to be considered. The real-time load of the base station cell located in the base station cell corresponding to the forward load migration direction is reversely migrated into the base station cell. Therefore, the real-time load of the base station cell is reduced based on the detected real-time load. In addition to the load migrated from the corresponding base station cell at the previous moment, the real-time load at the current moment matches the load migration situation.
  • Step S103 Receive the cross-layer neighboring cell load data transmitted respectively by the left neighbor cell co-directional group and the right neighbor cell co-directional group of the cross-layer co-directional group.
  • the corresponding cross-layer neighbor cell load data includes historical load data and real-time load data. Load data, historical load data is used to represent the predicted load of the K-layer main neighbor cells of the corresponding neighboring cell same-direction group at the previous moment, and real-time load data is used to represent the K of the corresponding neighboring cell same-direction group. The actual load of the layer primary neighbor at the current moment.
  • the cross-layer codirectional groups corresponding to the target base station cell C i are only the left neighbor cell codirectional group and the right neighbor cell codirectional group. One of them, at this time, the target base station cell C i only performs forward load prediction or reverse load prediction.
  • Step S104 Use the preset load prediction model to perform load prediction based on real-time load, historical predicted load and cross-layer neighboring cell load data, and obtain the forward predicted load and reverse predicted load corresponding to the target base station cell, and based on the forward predicted load The forecast load and reverse forecast load determine the corresponding load forecast results.
  • the preset load prediction model sets the corresponding operation function, and based on the operation function, the real-time load corresponding to the target base station cell, the historical predicted load and the corresponding cross-layer neighboring cell load data are respectively corrected.
  • forward load forecasting and reverse load forecasting and after obtaining the corresponding forward forecast load and reverse forecast load, calculate the total forecast load, and then determine the corresponding forecast result based on the total forecast load.
  • a cross-layer co-directional group is determined among multiple base station cells located on the preset load migration line.
  • the base station cells located at both ends of the cross-layer co-directional group are all the K-th base station cells corresponding to the target base station cell.
  • Layer main neighbor area, K is the preset value; obtain the real-time load at the current moment and the historical predicted load corresponding to the previous moment.
  • the historical predicted load is used to represent the load migration changes that will occur at the current moment predicted at the previous moment; receive
  • the corresponding cross-layer neighbor load data includes historical load data and real-time load data.
  • Historical load data It is used to represent the predicted load corresponding to the K-layer main neighbor cells of the corresponding neighbor group in the same direction at the previous moment.
  • the real-time load data is used to represent the K-layer main neighbor cells of the corresponding neighbor group in the same direction at the current time.
  • the actual load at the moment use the preset load prediction model to perform load prediction based on real-time load, historical predicted load and cross-layer neighboring cell load data to obtain the forward predicted load and reverse predicted load corresponding to the target base station cell, and based on Forward load prediction and reverse load prediction determine the corresponding load prediction results, which solves the problem in related technologies that the load prediction method cannot be applied to high-speed mobile communication scenarios where users move at high speed, load bursts, and there is no precise time cycle rule, and realizes the base station
  • the cell can detect and sense the user load and changing trend of the more distant neighbor cells in advance at a precise time point, and combine the predicted load and actual load of the base station cell to provide energy saving, load control, and traffic of the base station cell. Ensure the beneficial effects of providing effective reference and guidance for other operations.
  • a preset load prediction model is used to perform load prediction based on real-time load, historical predicted load and cross-layer neighboring cell load data to obtain the forward predicted load and reverse predicted load corresponding to the target base station cell.
  • Step 21 Use the load prediction model to perform forward load calculation on the historical load data and real-time load data corresponding to the real-time load, historical predicted load and the left neighbor group in the same direction, and obtain the forward predicted load.
  • the corresponding load migration line that from left to right is the forward direction and from right to left is the reverse direction.
  • the forward load it is also necessary to refer to the right adjacent area in the same direction.
  • the load transmitted by the first-layer primary neighbor cell in the group to the target base station cell at the last moment is used to determine the accurate real-time load of the target base station cell, so that the predicted load is closer to the load change trend, that is, it can better reflect the target The changing trend of base station cells.
  • Step 22 Use the load prediction model to perform reverse load calculation on the historical load data and real-time load data corresponding to the real-time load, historical predicted load and the right neighbor group in the same direction to obtain the reverse predicted load.
  • the following steps 31 to 34 are also implemented.
  • Step 31 Obtain the first actual load of each base station cell located on the preset load migration line at the current moment.
  • the first actual load includes a part of the load that migrated to the target base station cell at the last moment and a part of the load used for corresponding load prediction; in order to make the load prediction more consistent with the expected load prediction, the predicted load The changing trend is more accurate.
  • Step 32 Determine the nearest left neighbor cell and the nearest right neighbor cell corresponding to each base station cell, and obtain the actual load corresponding to the nearest right neighbor cell and the nearest left neighbor cell corresponding to the corresponding base station cell at the previous moment, where , set the forward direction from the left end to the right end of the preset load migration line.
  • the nearest left neighbor is used to represent the first base station cell on the left side of the corresponding base station cell in the forward direction.
  • the nearest right neighbor is Yu represents the first base station cell located on the right side of the corresponding base station cell in the forward direction.
  • the nearest left neighbor cell is the first base station cell located to the left (behind) of the corresponding base station cell in the corresponding set forward direction
  • the nearest right neighbor cell is the first base station cell located in the corresponding set forward direction.
  • the first base station cell located on the right side of the base station cell (that is, in front) receives the load of the reverse migration of the base station cell
  • the nearest right neighbor cell receives the load of the forward migration of the base station cell.
  • Step 33 Based on the actual load corresponding to the nearest right neighbor cell at the previous moment, determine the first migration load of the nearest right neighbor cell that migrates to the base station cell according to the first preset ratio at the previous moment, and based on the load of the nearest left neighbor cell at the previous moment.
  • the actual load corresponding to a moment determines the second migration load of the nearest left neighbor cell that migrated to the base station cell according to the second preset ratio at the previous moment.
  • Step 34 Determine the real-time load corresponding to the base station cell when performing forward prediction at the current time based on the difference between the first actual load and the first migration load, and determine the real-time load corresponding to the base station cell based on the difference between the first actual load and the second migration load.
  • the actual load corresponding to the forward prediction or the reverse prediction of the base station cell at the current moment is determined based on the difference between the first actual load and the first migration load and the second migration load respectively. In this way, the actual load is determined in this way.
  • the load at the current moment can make the predicted load obtained by the load prediction operation more reflective of the load change trend of the base station cell, for example, more accurately predict the arrival time of the corresponding load of the corresponding base station cell.
  • the first actual load of each base station cell located on the preset load migration line at the current moment through the above steps determine the nearest left neighbor and nearest right neighbor corresponding to each base station cell, and obtain the corresponding The actual load corresponding to the nearest right neighbor cell and the nearest left neighbor cell corresponding to the base station cell at the previous moment; based on the actual load corresponding to the nearest right neighbor cell at the previous moment, determine the first predetermined load of the nearest right neighbor cell at the previous moment.
  • the first migration load is proportionally migrated to the base station cell, and based on the actual load corresponding to the nearest left neighbor cell at the previous moment, determine the second migration load of the nearest left neighbor cell at the last moment according to the second preset proportion.
  • the difference between the actual load and the second migration load is used to determine the real-time load corresponding to the base station cell when performing reverse prediction at the current moment, achieving the acquisition of accurate real-time load of each base station cell, and further providing energy saving, load control, and Provide effective reference and guidance for operations such as call traffic assurance.
  • the load forecasting model includes a forward forecasting load calculation formula and reverse forecast load calculation formula
  • m represents the current moment
  • m-1 represents the previous moment
  • ⁇ represents the forward direction
  • ⁇ represents the reverse direction
  • K represents the number of layers in the cross-layer same-direction group
  • p is the P spans corresponding to the target base station cell C i.
  • the label of a cross-layer codirectional group within a layer codirectional group is the actual load of target base station cell C i at the current moment
  • the actual load at the last moment is the first layer main neighbor cell in the same direction group of the right neighbor cell corresponding to the target base station cell C i
  • the actual load at the last moment is the first preset ratio, is the jith-level main neighbor in the left neighbor group.
  • the actual load at the current moment is the jith-level main neighbor in the left neighbor group.
  • the corresponding nearest right neighbor is the jith-level main neighbor in the left neighbor group.
  • the corresponding forward forecast load at the previous moment Represents the reverse prediction load of the target base station cell C i at the current time and the previous time, respectively.
  • the actual load at the last moment is the jith-level main neighbor in the right neighbor group.
  • the actual load at the current moment It is the ji-th main neighbor cell in the same direction group of right neighbor cells.
  • the corresponding nearest left neighbor is the jith-level main neighbor in the right neighbor group.
  • ⁇ , ⁇ , and ⁇ are weight factors between [0, 1]
  • ⁇ is the weight of the target base station cell C i’s own load
  • (1- ⁇ ) is the neighboring cell
  • ⁇ is the weight of the real-time load of each base station cell
  • (1- ⁇ ) is the weight of the predicted load of each base station cell
  • ⁇ is the first preset ratio
  • ⁇ is the forward load
  • 1- ⁇ is the second preset ratio
  • 1- ⁇ is the load ratio of the load of the corresponding nearest left neighbor cell that is migrated to the corresponding base station cell at the current moment to the actual load of the corresponding nearest left neighbor cell at the previous moment in reverse load prediction.
  • the corresponding forward load prediction direction and reverse load prediction direction are represented by " ⁇ " and " ⁇ " respectively. Therefore, for the left adjacent area of the forward load prediction and reverse load prediction respectively,
  • the neighbor cell numbers corresponding to the primary neighbor cells in the same direction group and the right neighbor cell same direction group also have directionality.
  • the time interval between the current moment and the previous moment is the cycle T of load prediction and load information transmission.
  • the purpose of load prediction is not mainly to expect the predicted load and the actual load to be as close as possible, that is, it is not to improve the prediction accuracy.
  • the core goal it is necessary to combine the predicted load and the actual load into a comprehensive consideration to provide effective reference and guidance for operations such as energy saving, load control, and traffic guarantee in the base station community.
  • the corresponding load prediction result is determined based on the forward predicted load and the reverse predicted load, which is implemented through the following steps 41 to 42.
  • Step 41 Sum the forward predicted load and the reverse predicted load to obtain the first predicted total load corresponding to the target base station on the preset load migration line.
  • p is the label of the cross-layer in-directional group among the P cross-layer in-directional groups corresponding to the target base station cell, represents the first predicted total load, represents the forward forecast load, Represents reverse forecast load.
  • Step 42 Accumulate the first predicted total load corresponding to the plurality of preset load migration lines to generate a total predicted load, where the load prediction result includes the total predicted load.
  • p is the label of one cross-layer codirectional group among the P cross-layer codirectional groups corresponding to the target base station cell, is the total forecast load, Represents the first predicted total load.
  • the total predicted load is equal to the first predicted total load.
  • the target base station community located at the intersection of multiple load transfer lines The corresponding total forecast load is the sum of the total loads of multiple first forecasts.
  • the first predicted total load corresponding to the preset load transfer line of the target base station is obtained; the first predicted total load corresponding to multiple preset load transfer lines is accumulated. Predict the total load and generate the total predicted load, where the load prediction result includes the total predicted load, and implement the summary calculation of the total load arriving at the next time predicted by the target base station cell at the current moment.
  • the following steps 51 to 53 are also performed.
  • Step 51 Obtain the total predicted load corresponding to the current moment and the real-time load corresponding to the current moment, and obtain the previous load respectively.
  • the forward predicted load and the reverse predicted load are summarized and the predicted total load corresponding to multiple load migration lines is summed to obtain the predicted load corresponding to the target base station cell at the current moment.
  • Step 52 Generate a load forecast based on at least two of the actual load corresponding to the current moment, the total predicted load corresponding to the current moment, the historical total predicted load corresponding to the previous moment, and the historical actual load corresponding to the previous moment. difference function, and determine the calculated value corresponding to the corresponding difference function, where the difference function is used to characterize the prediction performance of load forecasting.
  • the expression manner of different load prediction sensitivity parameters is determined through a preset calculation function that optimizes the parameters of the load prediction model. For example, the real-time load corresponding to the current moment and the total load corresponding to the current moment can be used. Difference in predicted loads as a function of difference.
  • the so-called differential function for generating load prediction refers to using the actual load corresponding to the current time, the total predicted load corresponding to the current time, the historical total predicted load corresponding to the previous time, and the total predicted load corresponding to the previous time. At least two of the historical actual loads are corresponding parameters, and then a preset difference function is defined; and determining the calculated value corresponding to the difference function means bringing the above four parameters into the corresponding defined difference function The function value obtained in is the corresponding calculated value.
  • Step 53 Determine whether the calculated value corresponding to the difference function is greater than the preset threshold, and adjust the weight factor corresponding to the load prediction model according to the judgment result, where the weight factor corresponding to the load prediction model includes the target base station cell C i The weight ⁇ of the own load, the weight ⁇ of the actual load of each base station cell, and the first preset proportion ⁇ .
  • the load prediction model is sensitive to the load prediction.
  • the preset threshold that is, the corresponding threshold
  • the load prediction model is sensitive to the load prediction.
  • Setting a threshold indicates that the prediction is fast and sensitive. Increase ⁇ and ⁇ according to the preset step size to reduce the weight of the neighbor load and the predicted load.
  • the difference function value is negative And is lower than the preset threshold (the set lower limit of the threshold), indicating that the prediction is too slow and too slow, then ⁇ and ⁇ are reduced according to the preset step size to increase the weight of the neighboring cell load and the predicted load; in this embodiment , the default setting value of the preset step size is 0.05.
  • the total predicted load corresponding to the current moment and the real-time load corresponding to the current moment are respectively obtained, and the historical total predicted load corresponding to the previous moment and the historical actual load corresponding to the previous moment are respectively obtained; based on the current At least two of the actual load corresponding to the moment, the total predicted load corresponding to the current moment, the historical total predicted load corresponding to the previous moment, and the historical actual load corresponding to the previous moment are used to generate a difference function of the load forecast, and Determine the calculated value corresponding to the difference function, where the difference function is used to characterize the prediction performance of load forecasting; determine whether the calculated value corresponding to the difference function is greater than the preset threshold, and calculate the corresponding value of the load forecast model based on the judgment result.
  • the weighting factor is adjusted to realize real-time adjustment of the model parameters of the load prediction model based on the prediction results, thereby taking into account the accuracy of the load prediction and extending the detection sensing distance corresponding to the load prediction.
  • the difference function includes one of the following: the difference between the actual load corresponding to the current moment and the historical total predicted load corresponding to the previous moment, the difference between the actual load corresponding to the current moment and the previous moment The ratio of the corresponding difference between the historical total predicted load and the actual load corresponding to the current moment, the difference between the historical total predicted load corresponding to the previous moment and the total predicted load corresponding to the current moment, the actual load corresponding to the current moment. The difference between the load and the historical actual load corresponding to the previous moment and the difference between the total predicted load corresponding to the current moment and the historical total predicted load corresponding to the previous moment, the difference between the actual load corresponding to the current moment and the previous moment The load difference is generated by weighting the difference between the historical actual loads corresponding to the time.
  • each base station cell supports optimizing the corresponding weight factors after determining the corresponding load prediction results, that is, maintaining and optimizing the model parameters of the load prediction model to form a personality that adapts to itself.
  • the corresponding model parameters are optimized in the following manner, that is, the first difference function is defined and generated in the following manner.
  • L i (m) is the actual load corresponding to the current moment of the target base station cell C i , is the historical total predicted load corresponding to the previous moment.
  • Li (m) is the actual load corresponding to the current moment
  • Li (m-1) represents the historical actual load corresponding to the previous moment.
  • model parameters are adjusted in real time according to changes in actual load and predicted load, taking into account prediction accuracy and longer detection sensing distance; and, when it is learned that the predicted load is increasing, that is, when the train load is approaching, ⁇ is changed and ⁇ are increased according to the preset step size, further focusing on the target base station cell's own load and actual load to improve the prediction accuracy.
  • the predicted load is decreasing, that is, when the train load is moving away, ⁇ and ⁇ are increased according to the preset step size.
  • the step size is reduced to further focus on the neighboring cell load and predicted load to extend the distance for detecting the neighboring cell load of the sensing base station.
  • Method 3 Calculate the center of gravity CoL(m, p) of the preset overall load of the high-speed private network section according to the following formula:
  • the high-speed private network section is composed of the target base station cell C i and its corresponding 2K layer primary neighbor cell in a cross-K layer in-directional group.
  • CoL(m, p) means that at time m, the target base station cell C i corresponds to The center of gravity of the overall load of the high-speed private network section, represents the real-time load of each base station cell C j in the high-speed private network section at time m, and p is the label of one of the P cross-layer in-directional groups corresponding to the target base station cell C i ; then, The center of gravity of the overall load as a function of the difference.
  • each base station cell determines the center of gravity of the overall load of the high-speed private network section based on its own and the corresponding actual load of each base station neighbor cell in the corresponding cross-layer cross-layer in-directional group.
  • the change in the position of the center of gravity of the load is used to determine the moving direction of the train on the high-speed rail line corresponding to the high-speed private network and the corresponding direction of load movement.
  • ⁇ decreases according to the preset step size, otherwise it increases.
  • the preset step size can be 0.05, and the adjustment of ⁇ cannot exceed the preset value.
  • receiving the cross-layer neighbor cell load data respectively transmitted by the left neighbor cell isotropic group and the right neighbor cell isotropic group of the cross-layer codirectional group is implemented through the following steps 61 to 62.
  • Step 61 After detecting the first target load data from the first neighbor cell group load data received by the nearest left neighbor cell, it combines the first target load data with its own actual load corresponding to the current time and the previous time.
  • the corresponding historical predicted load is synthesized into the first cross-layer neighbor cell load data according to the preset format and transmitted to the corresponding target base station cell.
  • the load data of the first neighbor cell group is in the same direction as the nearest left neighbor cell's own left neighbor cell.
  • Cross-layer neighbor load data corresponding to the group, the first target load data includes the nearest left neighbor's own left neighbor and is excluded from the same direction group
  • the target base station cell is set to C 8
  • the left neighbor cell co-directional group corresponding to the target base station cell C 8 is set to ⁇ C 2 , C 3 , C 4 , C 5 , C 6 , C 7 ⁇
  • the nearest left neighbor cell corresponding to the target base station cell C 8 is C 7
  • Step 62 After detecting the second target load data from the second neighbor cell group load data received by the nearest right neighbor cell, it combines the second target load data with its own actual load corresponding to the current time and the previous time.
  • the corresponding historical predicted load is synthesized into the second cross-layer neighbor cell load data in a preset format and transmitted to the corresponding target base station cell.
  • the second neighbor cell group load data includes the right neighbor cell of the nearest right neighbor cell itself.
  • the cross-layer neighbor load data corresponding to the direction group, the second target load data includes the cross-layer corresponding to all the primary neighbor cells in the same direction group of the nearest right neighbor cell's own right neighbor cell after excluding its Kth layer primary neighbor cell. Neighboring cell load data;
  • the target base station cell is set to C 8
  • the right neighbor cell co-directional group corresponding to the target base station cell C 8 is set to ⁇ C 9 , C 10 , C 11 , C 12 , C 13 , C 14 ⁇
  • the nearest right neighbor cell corresponding to the target base station cell C 8 is C 9.
  • the nearest right neighbor cell and the nearest left neighbor cell constitute the left and right main neighbor cells of the corresponding base station cell; in this embodiment, the load information transmitted by the base station cell is real-time based on its own current moment. load, the corresponding historical predicted load at the previous moment, and the corresponding load information of the primary neighbor in the cross-layer same-direction group.
  • the nearest left neighbor cell before preparing to transmit the corresponding load data to the corresponding target base station cell, the nearest left neighbor cell will first receive the corresponding cross-layer neighbor cell load data transmitted by its own corresponding left neighbor cell in the same direction group. At this time , the main neighbor cell of the Kth layer in the same-direction group of the corresponding left neighbor cell of the nearest left neighbor cell does not belong to the main neighbor cell in the same-direction group of the left neighbor cell corresponding to the target base station cell.
  • the nearest left neighbor cell transmits
  • the load data does not include the load data of the Kth layer main neighbor cell of the same direction group of the left neighbor cell corresponding to the nearest left neighbor cell; before preparing to transmit the corresponding load data to the corresponding target base station cell, the nearest right neighbor cell will first receive its own The corresponding cross-layer neighbor cell load data transmitted by the corresponding right neighbor cell in the same direction group.
  • the main neighbor cell of the Kth layer in the right neighbor cell in the same direction group corresponding to the nearest right neighbor cell does not belong to the target base station cell.
  • the right neighbor cell is the primary neighbor cell in the same direction group. Therefore, the load data transmitted by the target base station cell does not include the load data of the Kth layer primary neighbor cell of the right neighbor cell in the same direction group corresponding to the nearest right neighbor cell.
  • the load of the K-th layer main neighbor cell is changed from the neighbor cell group stored in a preset format to the left neighbor cell in-directional group corresponding to the nearest left neighbor cell/right neighbor cell codirectional group corresponding to the nearest right neighbor cell.
  • the real-time load of the nearest left neighbor/nearest right neighbor at the current moment and the historical predicted load corresponding to the previous moment are supplemented, and then the corresponding load data in a preset format is generated and transmitted to the target base station. community.
  • the base station cell can transmit load information in a directional relay manner.
  • determining a cross-layer codirectional group in multiple base station cells located on a preset load migration line is implemented through the following steps 71 to 73.
  • Step 71 Obtain the handover frequency granularity data corresponding to each base station cell on the preset load migration line, and select the neighbor cell with the largest handover frequency granularity data as the main neighbor cell of the base station cell; wherein, the handover frequency granularity data is used for It means that within the preset time, the base station is small The number of user handovers between neighboring cells corresponding to the cell and the base station cell.
  • Step 72 Use the left main neighbor cell and the right main neighbor cell corresponding to each selected base station cell as the base station cell co-directional group corresponding to the base station cell.
  • Step 73 Taking the target base station cell as the starting point, select the same-direction group of K consecutive base station cells corresponding to the forward and reverse directions of the preset load migration line, and transfer the selected base station cells to the same direction.
  • the base station cells in the group undergo deduplication processing to obtain a cross-layer co-directional group.
  • the base station cell C i is based on hourly granular statistical data of the number of handovers of the "neighbor cell pairs" of its neighboring cells over a period of time (such as 168 hours corresponding to one week), including handovers to neighboring cells and neighboring cells.
  • the number of handovers to the base station cell C i is divided into two types of neighbor cells according to the clustering algorithm, and the neighbor cell with more handover times is used as the main neighbor cell of the base station cell C i ; of course, optimization by network planning can also be used personnel, based on the network engineering parameter information of the base station cell C i and neighboring cells, such as installation location, coverage direction, etc., or by referring to handover statistical data, manually select the closest neighbors on the left and right sides of the base station cell C i along the high-speed rail line.
  • Neighboring cells are set as the main neighboring cells of the base station cell C i ; then, the nearest neighboring cells on the left and right sides of the base station cell C i are used as the same direction group corresponding to the base station cell C i , and then based on the corresponding group of multiple base station cells
  • the association between co-directional groups is to determine the cross-layer co-directional group corresponding to each base station cell. For example: if C 2 is the right neighbor cell of the C 1 co-directional group, and C 3 is the right neighbor cell of the C 2 co-directional group , then it can be correlatively confirmed that C 3 is the right neighbor area of C 1 spanning 1 layer, and by analogy, the right neighbor area C k of C 1 spanning K layers can be obtained. ⁇ C 1 , C 2 ,...C K ⁇ form the cross-layer same-direction group of C 1 across K layers.
  • each main neighbor area of the base station cell is matched and filtered according to the load migration spatio-temporal characteristic pattern on the same direction line, and the left main neighbor area and the right main neighbor area are selected, and divided into the same direction route group, that is, the same direction route group. to the group.
  • the target base station cell C i and its main neighboring cells select the periods with the highest number of handovers from the hourly granular handover statistics of 1 week. For example, select 2 busy hours with the most number of handovers every day, then there are During the 14-hour period, obtain the statistical data D2 of the number of users in these periods with a granularity of 10 to 60 seconds (such as 30 seconds), divide D2 into two categories through the clustering algorithm, and set the load with a large number of users to 1, and the load with a small number of users. The load is set to 0, and then, the main neighbor area 1 and the main neighbor area 2 that meet the load migration spatio-temporal characteristics as shown in Table 1 and Table 2 are searched and filtered out (corresponding to one of them). In this way, the filtered out main neighbor area 1 and main neighbor cell 2 serve as the left neighbor cell and right neighbor cell of the target base station cell C i .
  • the cross-layer codirectional group of each base station cell is determined based on multiple associated codirectional groups; if C 2 is the right neighbor of the C 1 codirectional group, C 3 is the right neighbor of the C 2 codirectional group. If the right neighbor area is the right neighbor area, it can be confirmed that C 3 is the right neighbor area of C 1 spanning 1 layer.
  • the right neighbor area C k of C 1 spanning K layers can be obtained, ⁇ C 1 , C 2 ,...C K ⁇ constitutes the cross-layer codirectional group of C 1 across the K layer; in the embodiment of the present disclosure, the determined cross-layer codirectional group corresponding to the target base station cell includes two cross-layer codirectional groups, the left and the right.
  • K is the default value, for example: target base station Cell C i hopes to predict possible future loads earlier in advance and sense farther-distance loads, so K can be set very large, but this will increase the cost and difficulty of transmitting load information across layers and lead to random uncertainty. will also increase.
  • the K value needs to be set taking into account the target requirements and overhead costs of load forecasting, for example: K ⁇ 9; as shown in Figure 2, ⁇ C 1 , C 2 , C 3 , C 4 , C 5 , C 6 ⁇ is the cross-layer same-direction left neighbor area of C 7 spanning 6 layers. ⁇ C 8 , C 9 , C 10 , C 11 , C 12 , C 13 ⁇ is the cross-layer same-direction area of C 7 spanning 6 layers. In the right adjacent area, load information can be passed left or right through the 6th floor based on relay transfer.
  • Embodiments 1 to 6 are given below to illustrate the load prediction of the embodiments of the present disclosure as follows:
  • the base station cell can prepare for operations such as energy saving or load control in advance.
  • the prediction model can detect and sense the upcoming load and changing trend in advance. Combining the actual load and predicted load, the base station cell can prepare in advance. Operations such as energy saving or load control.
  • the prediction model can detect and sense the upcoming load and changing trend in advance. Combining the actual load and predicted load, the base station cell can prepare for energy saving in advance. or load control operations.
  • the prediction model can detect and sense the upcoming load and changing trend in advance. Combining the actual load and predicted load, the community can prepare for energy saving or load control operations in advance.
  • the embodiments of the present disclosure also provide a dynamic load prediction device, which is used to implement the above embodiments and optional implementations. What has been described will not be described again.
  • the terms “module”, “unit”, “subunit”, etc. used below may be a combination of software and/or hardware that implements predetermined functions.
  • the apparatus described in the following embodiments is preferably implemented in software, implementation in hardware, or a combination of software and hardware, is also possible and contemplated.
  • Figure 3 is a structural block diagram of a dynamic load prediction device provided by an embodiment of the present disclosure. As shown in Figure 3, the device includes: an acquisition module 32, an acquisition module 32, a receiving module 33 and a prediction module 34.
  • the determination module 31 is configured to determine a cross-layer co-directional group among multiple base station cells located on the preset load migration line, where the base station cells located at both ends of the cross-layer co-directional group are the K-th base station cells corresponding to the target base station cell.
  • the main neighbor area of the layer, K is the default value.
  • the acquisition module 32 is coupled to the determination module 31 and is configured to obtain the real-time load at the current moment and the historical predicted load corresponding to the previous moment.
  • the historical predicted load is used to represent the load migration that will occur at the current moment predicted at the previous moment. Variety.
  • the receiving module 33 is coupled to the acquisition module 32 and is configured to receive the cross-layer neighbor load data respectively transmitted by the left neighbor group and the right neighbor group of the cross-layer group, wherein the corresponding cross-layer Neighboring cell load data includes historical load data and real-time load data.
  • Historical load data is used to represent the predicted load corresponding to the K-layer main neighbor cell in the corresponding neighboring cell same-direction group at the previous moment
  • real-time load data is used to represent The actual load of the K-layer primary neighbor cell of the corresponding neighboring cell same-direction group at the current moment.
  • the prediction module 34 is coupled to the receiving module 33 and is configured to use a preset load prediction model to perform load prediction based on real-time load, historical predicted load and cross-layer neighboring cell load data, and obtain the forward prediction corresponding to the target base station cell. load and reverse forecast load, and determine the corresponding load forecast results based on the forward forecast load and reverse forecast load.
  • a cross-layer co-directional group is determined among multiple base station cells located on the preset load migration line.
  • the base station cells located at both ends of the cross-layer co-directional group are all corresponding to the target base station cell.
  • the Kth layer main neighbor area of , K is the preset value; obtain the real-time load at the current moment and the historical predicted load corresponding to the previous moment.
  • the historical predicted load is used to represent the load migration that will occur at the current moment predicted at the previous moment. Change; receive the cross-layer neighboring cell load data transmitted by the left neighbor cell co-directional group and the right neighbor cell co-directional group of the cross-layer co-directional group.
  • the corresponding cross-layer neighbor cell load data includes historical load data and real-time load data.
  • the historical load data is used to represent the predicted load corresponding to the K-layer primary neighbors of the corresponding neighbor group in the same direction at the previous moment
  • the real-time load data is used to represent the K-layer primary neighbors of the corresponding neighbor group in the same direction.
  • the actual load of the area at the current moment use the preset load prediction model to perform load prediction based on real-time load, historical predicted load and cross-layer neighboring cell load data, and obtain the forward predicted load and reverse predicted load corresponding to the target base station cell. , and determine the corresponding load prediction results based on forward load prediction and reverse load prediction, solving the problem in related technologies that the load prediction method cannot be applied to high-speed mobile communication scenarios with high-speed user movement, load bursts, and no precise time cycle rules.
  • the base station cell can detect and sense the user load and changing trend of the more distant neighbor cells in advance at a precise time point, and combines the predicted load and actual load of the base station cell to provide energy saving and load control for the base station cell. It provides effective reference and guidance for operations such as , traffic protection and other operations.
  • the prediction module 34 is also configured to use the load prediction model to perform forward load calculation on the historical load data and real-time load data corresponding to the real-time load, the historical predicted load and the left neighbor group in the same direction, Obtain the forward forecast load; and use the load forecast model to perform reverse load calculation on the historical load data and real-time load data corresponding to the real-time load, the historical forecast load and the right neighbor group in the same direction, and obtain the reverse forecast load.
  • the device is also used to obtain the first actual value of each base station cell located on the preset load migration line at the current moment.
  • Actual load determine the nearest left neighbor and nearest right neighbor corresponding to each base station cell, and obtain the actual load corresponding to the nearest right neighbor and nearest left neighbor of the corresponding base station cell at the previous moment, where , set the forward direction from the left end to the right end of the preset load migration line.
  • the nearest left neighbor is used to represent the first base station cell on the left side of the corresponding base station cell in the forward direction.
  • the nearest right neighbor is It represents the first base station cell located on the right side of the corresponding base station cell in the forward direction; based on the actual load corresponding to the nearest right neighbor cell at the previous moment, it is determined that the nearest right neighbor cell presses the first preset at the previous moment The first migration load proportionally migrated to the base station cell, and based on the actual load corresponding to the nearest left neighbor cell at the previous moment, determine the second migration load of the nearest left neighbor cell migrated to the base station cell according to the second preset proportion at the previous moment ; According to the difference between the first actual load and the first migration load, determine the real-time load corresponding to the base station cell when performing forward prediction at the current moment, and according to the difference between the first actual load and the second migration load, determine the base station cell The real-time load corresponding to the reverse prediction at the current moment.
  • the load forecasting model includes a forward forecasting load calculation formula and reverse forecast load calculation formula
  • m represents the current moment
  • m-1 represents the previous moment
  • ⁇ represents the forward direction
  • ⁇ represents the reverse direction
  • K represents the number of layers in the cross-layer same-direction group
  • p is the P spans corresponding to the target base station cell C i.
  • L i (m) is the actual load of the target base station cell C i at the current moment
  • the actual load at the last moment is the first layer main neighbor cell in the same direction group of the right neighbor cell corresponding to the target base station cell C i
  • the actual load at the last moment is the first preset ratio, is the jith-level main neighbor in the left neighbor group.
  • the actual load at the current moment is the jith-level main neighbor in the left neighbor group.
  • the corresponding nearest right neighbor is the jith-level main neighbor in the left neighbor group.
  • the corresponding forward forecast load at the previous moment Represents the reverse prediction load of the target base station cell C i at the current time and the previous time, respectively.
  • the actual load at the last moment is the jith-level main neighbor in the right neighbor group.
  • the actual load at the current moment It is the ji-th main neighbor cell in the same direction group of right neighbor cells.
  • the corresponding nearest left neighbor is the jith-level main neighbor in the right neighbor group.
  • ⁇ , ⁇ , and ⁇ are weight factors between [0, 1]
  • ⁇ is the weight of the target base station cell C i’s own load
  • (1- ⁇ ) is the neighboring cell
  • ⁇ is the weight of the real-time load of each base station cell
  • (1- ⁇ ) is the weight of the predicted load of each base station cell
  • ⁇ is the first preset ratio
  • 1- ⁇ is the second Default ratio.
  • the prediction module 34 is also configured to sum the forward predicted load and the reverse predicted load to obtain the first predicted total load corresponding to the target base station on the preset load migration line; accumulate multiple predicted loads. Assume the first predicted total load corresponding to the load migration line to generate a total predicted load, where the load prediction result includes the total predicted load.
  • the device is further configured to obtain the total predicted load corresponding to the current moment and the real-time load corresponding to the current moment, and to obtain the historical total corresponding to the previous moment.
  • the predicted load and the historical actual load corresponding to the previous moment based on the actual load corresponding to the current moment, the total predicted load corresponding to the current moment, the historical total predicted load corresponding to the previous moment, and the historical actual load corresponding to the previous moment. At least two of the loads are used to generate the difference function of the load forecast, and the calculated value corresponding to the corresponding difference function is determined.
  • the difference function is used to characterize the prediction performance of the load forecast; it is judged whether the calculated value of the difference function is greater than
  • the threshold is preset, and the weight factor corresponding to the load prediction model is adjusted according to the judgment result.
  • the weight factor corresponding to the load prediction model includes the weight ⁇ of the target base station cell C i's own load, and the actual load of each base station cell. Weight ⁇ , first preset ratio ⁇ .
  • the difference function includes one of the following: the difference between the actual load corresponding to the current moment and the historical total predicted load corresponding to the previous moment, the difference between the actual load corresponding to the current moment and the historical total predicted load corresponding to the previous moment.
  • the load difference is generated by weighting the difference between the corresponding historical actual loads.
  • the acquisition module 31 is further configured to obtain handover frequency granularity data corresponding to each base station cell on the preset load migration line, and select the neighboring cell with the largest handover frequency granularity data as the main base station cell.
  • Neighboring cells where the handover frequency granularity data is used to represent the number of user handovers between the base station cell and the neighboring cells corresponding to the base station cell within a preset time; the left primary neighbor cell and the right primary neighbor cell corresponding to each selected base station cell are The neighboring cells are used as the co-directional group of base station cells corresponding to the base station cell; starting from the target base station cell, select the co-directional group of base station cells corresponding to K consecutive base station cells along the forward and reverse directions of the preset load migration line. , and perform deduplication processing on the base station cells in the selected base station cell co-directional group to obtain a cross-layer co-directional group.
  • FIG 4 is a schematic structural diagram of a base station according to an embodiment of the present disclosure.
  • an embodiment of the present disclosure provides a base station, including a processor 41, a communication interface 42, a memory 43 and a communication bus 44, where the processor 41 , communication interface 42, memory 43 completes mutual communication through communication bus 44; memory 43 stores computer programs; processor 41 implements the method in Figure 1 when executing the program stored in memory 43.
  • the processing in the base station implements the method steps in Figure 1, and the technical effects brought about are consistent with the technical effects of the dynamic load prediction method in Figure 1 in the above embodiment, and will not be described again here.
  • the communication bus mentioned in the above base station can be the Peripheral Component Interconnect (PCI) bus or the Extended Industry Standard Architecture (EISA) bus, etc.
  • PCI Peripheral Component Interconnect
  • EISA Extended Industry Standard Architecture
  • the communication bus can be divided into address bus, data bus, control bus, etc. For ease of presentation, only one thick line is used in Figure 4, but it does not mean that there is only one bus or one type of bus.
  • the communication interface is used for communication between the above terminal and other devices.
  • the memory may include Random Access Memory (RAM) or non-volatile memory (non-volatile memory), such as at least one disk memory.
  • RAM Random Access Memory
  • non-volatile memory non-volatile memory
  • the memory may also be at least one storage device located far away from the aforementioned processor.
  • the above-mentioned processor can be a general-purpose processor, including a central processing unit (Central Processing Unit, referred to as CPU), a network processor (Network Processor, referred to as NP), etc.; it can also be a digital signal processor (Digital Signal Processing, referred to as DSP) , Application Specific Integrated Circuit (ASIC for short), Field-Programmable Gate Array (FPGA for short) or other programmable logic devices, discrete gate or transistor logic devices, and discrete hardware components.
  • CPU Central Processing Unit
  • NP Network Processor
  • DSP Digital Signal Processing
  • ASIC Application Specific Integrated Circuit
  • FPGA Field-Programmable Gate Array
  • Embodiments of the present disclosure also provide a computer-readable storage medium on which a computer program is stored.
  • the computer program is executed by a processor, the dynamic load prediction method as provided in any of the foregoing method embodiments is implemented.
  • a computer program product containing instructions is also provided, which when run on a computer causes the computer to execute the dynamic load prediction method described in any of the above embodiments.
  • embodiments of the present disclosure provide dynamic load prediction methods, devices, base stations and storage media, by determining cross-layer co-directional groups in multiple base station cells located on preset load migration lines, where, The base station cells located at both ends of the cross-layer co-directional group are the Kth layer primary neighbor cells corresponding to the target base station cell, and K is the default value; obtain the real-time load at the current moment and the historical predicted load corresponding to the previous moment, and the historical predicted load Used to represent the load migration changes that will occur at the current moment predicted at the previous moment; receive the cross-layer neighbor load data transmitted by the left neighbor group and the right neighbor group respectively, where, The corresponding cross-layer neighbor load data includes historical load data and real-time load data.
  • the historical load data is used to represent the predicted load and real-time load corresponding to the K-layer main neighbor zone of the corresponding neighbor group in the same direction at the previous moment.
  • the data is used to represent the actual load of the K-layer main neighbor cells of the corresponding neighbor group in the same direction at the current moment; the preset load prediction model is used to calculate the load based on real-time load, historical predicted load and cross-layer neighbor load data.
  • the base station cell can detect and sense the user load and changing trend of the arrival of more distant neighboring cells in advance at a precise time point, and pass Combining the predicted load and actual load of the base station cell, it can provide effective reference and guidance for operations such as energy saving, load control, and traffic guarantee of the base station cell.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

本公开涉及一种动态负荷预测方法、装置、基站和存储介质，该方法包括：在位于预设负荷迁移线路上的多个基站小区中，确定目标基站小区的跨层同向组，获取目标基站小区当前时刻的实时负荷和上一时刻所对应的历史预测负荷；接收目标基站小区跨层同向组的左邻区同向组和右邻区同向组分别传递的跨层邻区负荷数据，对应的跨层邻区负荷数据均包括历史负荷数据和实时负荷数据；利用预设的负荷预测模型，基于实时负荷、历史预测负荷以及跨层邻区负荷数据进行负荷预测，得到目标基站小区所对应的正向预测负荷和反向预测负荷，并基于正向预测负荷和反向预测负荷确定对应的负荷预测结果。

Description

动态负荷预测方法、装置、基站和存储介质
相关申请的交叉引用
本公开要求享有2022年08月25日提交的名称为“动态负荷预测方法、装置、基站和存储介质”的中国专利申请CN202211028169.9的优先权，其全部内容通过引用并入本公开中。
技术领域
本公开涉及基站系统领域，尤其涉及一种动态负荷预测方法、装置、基站和存储介质。
背景技术
相关技术中，移动通信网络的负荷通常指接入网络基站小区的用户数(终端数)，或者网络基站小区的各种资源占用率。网络负荷是移动通信网络运维保障中一个备受关注的数据指标，是网络实时监控的关键指标，直接关系到网络的平稳运行、用户感知及是否节能，故属于移动通信网络的研究热点及应用重点。相关技术中，通过负荷预测，可以协助和指导网络基站实现各种节能措施、及时唤醒、以及提前实施负载均衡、接纳控制高话务特殊保障，保证基站的平稳运行和用户的业务感知。
相关技术中，对负荷预测，常用差分整合自回归滑动平均(Auto Regressive Integrated Moving Average，简称ARIMA)模型，利用网络基站小区自身的历史负荷数据，对模型进行训练学习和回归，实现对各基站小区未来负荷的预测；还有采用深度神经网络模型，例如：循环神经网络(Recurrent Neural Network，简称RNN)模型和改进的长短期记忆(Long-short term Memory，简称LSTM)实现负荷预测；但习知的预测方法通常需要足够长时间、足够多的历史数据，才能实现有效的训练学习，才能预测的比较准，同时，进行的负荷预测还需要要求基站小区的负荷变化具有周期规律的，对于没有较明显时间周期规律、网络拓扑结构和用户移动轨迹复杂、或者网络拓扑结构相对简单但是用户高速移动、负荷突发时间不定的移动通信场景，相关负荷预测方法无法实现精准有效的负荷预测。
针对相关技术中负荷预测方法无法适用用户高速移动、负荷突发、没有精确时间周期规律的高速移动通信场景的问题，尚未存在有效的解决方案。
发明内容
本公开提供了一种动态负荷预测方法、装置、基站和存储介质。
第一方面，本公开提供了一种动态负荷预测方法，应用于高速专网下的目标基站小区，包括：在位于预设负荷迁移线路上的多个基站小区中，确定跨层同向组，其中，位于所述跨层同向组两端的基站小区均为所述目标基站小区对应的第K层主邻区，K为预设值；获取当前时刻的实时负荷和上一时刻所对应的历史预测负荷，所述历史预测负荷用于表征在上一时刻预测的当前时刻将产生的负荷迁移变化；接收所述跨层同向组的左邻区同向组和右邻区同向组分别传递的跨层邻区负荷数据，其中，对应的所述跨层邻区负荷数据均包括历史负荷数据和实时负荷数据，所述历史负荷数据用于表征对应的邻区同向组所具有的K层主邻区在上一时刻对应的预测负荷，所述实时负荷数据用于表征对应的邻区同向组所具有的K层主邻区在当前时刻的实际负荷；利用预设的负荷预测模型，基于所述实时负荷、所述历史预测负荷以及所述跨层邻区负荷数据进行负荷预测，得到所述目标基站小区所对应的正向预测负荷和反向预测负荷，并基于所述正向预测负荷和反向预测负荷确定对应的负荷预测结果。
第二方面，本公开提供了一种动态负荷预测装置，应用于高速专网下的目标基站小区，包括：确定模块，被配置为在位于预设负荷迁移线路上的多个基站小区中，确定跨层同向组，其中，位于所述跨层同向组两端的基站小区均为所述目标基站小区对应的第K层主邻区，K为预设值；获取模块，被配置为获取当前时刻的实时负荷和上一时刻所对应的历史预测负荷，所述历史预测负荷用于表征在上一时刻预测的当前时刻将产生的负荷迁移变化；接收模块，被配置为接收所述跨层同向组的左邻区同向组和右邻区同向组分别传递的跨层邻区负荷数据，其中，对应的所述跨层邻区负荷数据均包括历史负荷数据和实 时负荷数据，所述历史负荷数据用于表征对应的邻区同向组所具有的K层主邻区在上一时刻对应的预测负荷，所述实时负荷数据用于表征对应的邻区同向组所具有的K层主邻区在当前时刻的实际负荷；预测模块，被配置为利用预设的负荷预测模型，基于所述实时负荷、所述历史预测负荷以及所述跨层邻区负荷数据进行负荷预测，得到所述目标基站小区所对应的正向预测负荷和反向预测负荷，并基于所述正向预测负荷和反向预测负荷确定对应的负荷预测结果。
第三方面，提供了一种基站，包括处理器、通信接口、存储器和通信总线，其中，处理器、通信接口、存储器通过通信总线完成相互间的通信；存储器，其存放计算机程序；处理器，其执行存储器上所存放的程序时，实现第一方面所述的动态负荷预测方法。
第四方面，提供了一种计算机可读存储介质，其上存储有计算机程序，所述计算机程序被处理器执行时实现如第一方面所述的动态负荷预测方法。
本公开的一个或多个实施例的细节在以下附图和描述中提出，以使本公开的其他特征、目的和优点更加简明易懂。
附图说明
此处的附图被并入说明书中并构成本说明书的一部分，示出了符合本公开的实施例，并与说明书一起用于解释本公开的原理。
为了更清楚地说明本公开实施例或相关技术中的技术方案，下面将对实施例或相关技术描述中所需要使用的附图作简单地介绍，显而易见地，对于本领域普通技术人员而言，在不付出创造性劳动性的前提下，还可以根据这些附图获得其他的附图。
图1是本公开实施例提供的动态负荷预测方法的流程示意图；
图2是本公开实施例中跨层同向组的结构示意图；
图3是本公开实施例提供的动态负荷预测装置的结构框图；以及
图4是本公开实施例的基站的结构示意图。
具体实施方式
为使本公开实施例的目的、技术方案和优点更加清楚，下面将结合本公开实施例中的附图，对本公开实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例是本公开的一部分实施例，而不是全部的实施例。基于本公开中的实施例，本领域普通技术人员在没有做出创造性劳动的前提下所获得的所有其他实施例，都属于本公开保护的范围。
在对本公开实施例进行描述之前，对本公开所涉及的相关技术名称进行如下说明。
基站(Base Station，简称BS)，属于移动网络的基本单元。
小区，英文缩写Cell，基站的基本单元，常规蜂窝基站三叶草结构下，一个基站有3个120°扇形覆盖的小区，高铁专网下则1个基站有两个或多个的小区。
用户设备(User Equipment，简称UE)。
第五代移动通信(the Fifth-Generation mobile communication，简称5G)。
第四代移动通信(the Fourth-Generation mobile communication，简称4G)。
新空口(NEW Ratio，简称NR)。
长期演进(Long Term Evolution，简称LTE)。
差分整合自回归滑动平均(Auto Regressive Integrated Moving Average，简称ARIMA)，基于ARMA改进的时间序列模型。
循环神经网络(Recurrent Neural Network，简称RNN)。
长短期记忆(Long Short-term Memory，简称LSTM)，是一种基于RNN的改进神经网络模型。
下面将结合本公开实施例中的附图，对本公开实施例中的技术方案进行描述。
图1为本公开实施例提供的动态负荷预测方法的流程示意图。如图1所示，本公开实施例提供了一种动态负荷预测方法，应用于高速专网下的目标基站小区，该方法包括如下步骤S101至步骤S104。
步骤S101，在位于预设负荷迁移线路上的多个基站小区中，确定跨层同向组，其中，位于跨层同向组两端的基站小区均为目标基站小区对应的第K层主邻区，K为预设值。
在本公开实施例中，动态负荷预测方法是由对应的基站小区根据当前时刻的相关负荷和上一时刻的相关负荷，对下一个时刻将要到来的负荷进行预测，并通过负荷的预测，以获知基站小区在未来对应的负荷将要发生的变化、负荷变化的趋势以及负荷变化的变化量；在本实施例中，进行的负荷预测并不以预测的负荷和实际负荷接近为主要目标，也就是不以预测准确度为目标，而是以将预测负荷和实际负荷结合考虑，以获取基站小区在未来对应的负荷变化趋势及变化值为目标，籍以为基站小区的节能、负荷控制、话务保障等操作提供有效的参考和指导；在本实施例中，进行负荷预测的是目标基站小区，也就是处于负荷迁移方向上的多个基站小区其中的一个。
在本实施例中，考量的高速专网是一字形高铁线路所对应的高速专网，因此，对于在高速专网下的多个基站小区的每个基站小区而言，对应的跨层同向组就只要一个；在T字形或十字形高铁线路的中心相交点所处的基站小区对应有两个跨层同向组，对于“Ж”字形高铁线路的中心相交点所在的基站小区，则存在三个跨层同向组，对于米字形高铁线路的中心相交点所在的基站小区，则存在四个跨层同向组。
在本实施例中，在确定目标基站小区对应的跨层同向组之前，会对设置在高铁线路对应的高速专网一段线路上的多个基站小区Ci(i＝1，2，……，N)中的每个基站小区进行主邻区的确定，也就是在基站小区对应的左、右邻区中，根据“邻区对”切换次数统计数据，筛选和确定每个基站小区的主邻区，例如，基站小区Ci根据自身一段时间(如1周对应的168个小时)与其各个邻区的“邻区对”切换次数的小时粒度统计数据，包括切出到邻区和邻区切入到基站小区Ci的次数，按照聚类算法分成两类邻区，将切换次数属于更多一类的邻区作为基站小区Ci的主邻区；当然，在一些可选实施方式中，可以采用由网络规划优化人员，基于基站小区Ci及邻区的网络工程参数信息，如：安装位置、覆盖方向等，或再参考切换统计数据，手工筛选出基站小区Ci沿高铁线路方向上左右两边最近的相邻邻区，并设定为基站小区Ci的主邻区；然后，将基站小区Ci左右两边最近的相邻邻区作为基站小区Ci对应的同向组，然后基于多个基站小区对应的同向组之间的关联关系，以确定各基站小区对应的跨层同向组，例如：假如C2是C1同向组的右邻区，C3是C2同向组的右邻区，则可以关联确认C3是C1跨越1层的右邻区，以此类推，可以获得C1跨越K层的右邻区Ck。{C1，C2，...CK}就组成了C1跨越K层的跨层同向组。
在本公开实施例中，确定出的目标基站小区对应的跨层同向组是包括左、右两个跨层同向组的，例如：左邻区同向组和右邻区同向组，其中，左邻区同向组和右邻区同向组均具有K层主邻区，K为预设值，例如：目标基站小区Ci期望提前更早预测出未来可能到来的负荷，感应更远距离的负荷，那么K可以设定很大，但会使的跨层传递负荷信息的开销和难度增大，随机不确定性也会增加，因此，需要兼顾考虑负荷预测的目标要求和开销代价等来设定K值，例如：K≤9。
步骤S102，获取当前时刻的实时负荷和上一时刻所对应的历史预测负荷，历史预测负荷用于表征在上一时刻预测的当前时刻将产生的负荷迁移变化。
在本公开实施例中，在对下一时刻的负荷进行预测时，从时间维度考虑，则需要考虑是当前时刻和上一时刻两个维度，其中，上一时刻、当前时刻及下一时刻之间间隔一个设定的时间周期T，时间周期T是根据高铁专网下典型的基站小区平均覆盖距离以及高铁列车平均车速来确定，取值范围可为10～60秒，可选T＝30秒；从负荷信息维度考虑，则需要考虑实时负荷和预测负荷两个维度，预测负荷是指上一时刻已预测的当前时刻会出现的负荷；从基站小区对象维度考虑，则参与负荷预测的基站小区包括目 标基站小区、一个跨层同向组两个维度。
在本实施例中，对于涉及到目标基站小区对应的负荷包括当前时刻的实时负荷和上一时刻对应预测的预测负荷，因为预测负荷是基站小区在未来可能发生的负荷变化的预测，通过在上一时刻对应的预测负荷，籍以确定当前时刻的实时负荷与已预测的负荷的差异，以表征目标基站小区本身的负荷变化趋势是否与预测的变化趋势匹配。
需要说明的是，在考虑时间维度时，对于时间周期T需要考虑时间情况进行设定，例如：如果T太小，负荷预测和负荷信息传递的开销会很大，而如果T太大，则高速列车很可能穿越出目标基站小区及其关联的一些邻区，导致对应的基站小区无法采样到真实的实际负荷，目标基站小区无法实现有效的负荷预测；在考虑实时负荷时，对应各个基站小区当前时刻对应的实时负荷，需要考虑的位于基站小区对应于负荷迁移方向前向的基站小区反向迁移入基站小区的实时负荷，因此，基站小区的实时负荷是在检查到的实时负荷的基础上减除上一时刻由对应的基站小区迁移入的负荷，如此，当前时刻的实时负荷才与负荷迁移情况匹配。
步骤S103，接收跨层同向组的左邻区同向组和右邻区同向组分别传递的跨层邻区负荷数据，其中，对应的跨层邻区负荷数据均包括历史负荷数据和实时负荷数据，历史负荷数据用于表征对应的邻区同向组所具有的K层主邻区在上一时刻对应的预测负荷，实时负荷数据用于表征对应的邻区同向组所具有的K层主邻区在当前时刻的实际负荷。
在本实施例中，当目标基站小区Ci位于高速专网的首、末段位置时，目标基站小区Ci对应的跨层同向组只有左邻区同向组和右邻区同向组其中一个，此时，目标基站小区Ci仅进行正向负荷预测或反向负荷预测。
步骤S104，利用预设的负荷预测模型，基于实时负荷、历史预测负荷以及跨层邻区负荷数据进行负荷预测，得到目标基站小区所对应的正向预测负荷和反向预测负荷，并基于正向预测负荷和反向预测负荷确定对应的负荷预测结果。
在本公开实施例中，预设的负荷预测模型通过设定对应的运算函数，并基于运算函数，将目标基站小区对应的实时负荷、历史预测负荷以及对应的跨层邻区负荷数据分别进行正向负荷预测和反向负荷预测，并在得到对应的正向预测负荷和反向预测负荷，进行总预测负荷的计算，然后根据总预测负荷确定对应的预测结果。
通过上述步骤S101至步骤S104，采用在位于预设负荷迁移线路上的多个基站小区中，确定跨层同向组，位于跨层同向组两端的基站小区均为目标基站小区对应的第K层主邻区，K为预设值；获取当前时刻的实时负荷和上一时刻所对应的历史预测负荷，历史预测负荷用于表征在上一时刻预测的当前时刻将产生的负荷迁移变化；接收跨层同向组的左邻区同向组和右邻区同向组分别传递的跨层邻区负荷数据，对应的跨层邻区负荷数据均包括历史负荷数据和实时负荷数据，历史负荷数据用于表征对应的邻区同向组所具有的K层主邻区在上一时刻对应的预测负荷，实时负荷数据用于表征对应的邻区同向组所具有的K层主邻区在当前时刻的实际负荷；利用预设的负荷预测模型，基于实时负荷、历史预测负荷以及跨层邻区负荷数据进行负荷预测，得到目标基站小区所对应的正向预测负荷和反向预测负荷，并基于正向预测负荷和反向预测负荷确定对应的负荷预测结果，解决了相关技术中负荷预测方法无法适用用户高速移动、负荷突发、没有精确时间周期规律的高速移动通信场景的问题，实现了基站小区能够在精确的时间点上提前探测和感应到更远距离的邻区到来的用户负荷及变化趋势，并通过结合基站小区的预测负荷和实际负荷，为基站小区的节能、负荷控制、话务保障等操作提供有效的参考和指导的有益效果。
在一些实施例中，利用预设的负荷预测模型，基于实时负荷、历史预测负荷以及跨层邻区负荷数据进行负荷预测，得到目标基站小区所对应的正向预测负荷和反向预测负荷，通过如下步骤21至步骤22 实现。
步骤21、利用负荷预测模型，对实时负荷、历史预测负荷和左邻区同向组所对应的历史负荷数据与实时负荷数据，进行正向负荷计算，得到正向预测负荷。
在本实施例中，定义在对应的负荷迁移线路上，从左到右为正向，从右到左为反向，同时，在进行正向负荷计算时，还需要参考处于右邻区同向组中的第一层主邻区在上一时刻向目标基站小区传动的负荷，以确定目标基站小区精准的实时负荷，从而使预测出的负荷更接近负荷变化的趋势，也就是更能体现目标基站小区的变化趋势。
步骤22、利用负荷预测模型，对实时负荷、历史预测负荷和右邻区同向组所对应的历史负荷数据与实时负荷数据，进行反向负荷计算，得到反向预测负荷。
在本实施例中，在进行反向负荷计算时，还需要参考处于左邻区同向组中的第一层主邻区在上一时刻向目标基站小区传动的负荷，以确定目标基站小区精准的实时负荷，从而使预测出的负荷更接近负荷变化的趋势，也就是更能体现目标基站小区的变化趋势。
为了获取每个基站小区精准的实时负荷，在一些实施例中，还实施如下步骤31至步骤34。
步骤31、获取位于预设负荷迁移线路上的每个基站小区在当前时刻的第一实际负荷。
在本实施例中，第一实际负荷包括上一时刻迁移至目标基站小区的一部分负荷和用于对应负荷预测的一部分负荷；为使的负荷的预测与预期的负荷预测更匹配，预测出的负荷变化趋势更精准，在本实施例中，需要排除上一时刻迁移至目标基站小区的一部分负荷后，再进行对应的预测。步骤32、确定每个基站小区所对应的最近左邻区和最近右邻区，并分别获取对应的基站小区所对应的最近右邻区和最近左邻区在上一时刻对应的实际负荷，其中，设定从预设负荷迁移线路的左端至右端为正向方向，最近左邻区用于表征在正向方向上，位于对应的基站小区左侧的第一个基站小区，最近右邻区用于表征在正向方向上，位于对应的基站小区右侧的第一个基站小区。
在本实施例中，最近左邻区为在对应设定的正向方向上，位于对应的基站小区左侧(后方)的第一个基站小区，最近右邻区为在对应设定的正向方向上，位于基站小区右侧(也就是前方)的第一个基站小区，最近左邻区接收基站小区反向迁移的负荷，最近右邻区接收由基站小区正向迁移的负荷。
步骤33、基于最近右邻区在上一时刻对应的实际负荷，确定最近右邻区在上一时刻按第一预设比例迁移至基站小区的第一迁移负荷，以及基于最近左邻区在上一时刻对应的实际负荷，确定最近左邻区在上一时刻按第二预设比例迁移至基站小区的第二迁移负荷。
步骤34、根据第一实际负荷和第一迁移负荷的差值，确定基站小区在当前时刻进行正向预测时所对应的实时负荷，以及根据第一实际负荷和第二迁移负荷的差值，确定基站小区在当前时刻进行反向预测时所对应的实时负荷。
在本实施例中，基于第一实际负荷分别与第一迁移负荷和第二迁移负荷的差值确定当前时刻基站小区正向预测或反向预测所对应的真实的负荷，如此，采用此方式确定的当前时刻的负荷，能使的负荷预测运算得到的预测负荷更加能反映基站小区的负荷的变化趋势，例如：更加准确的预测对应基站小区的对应负荷的到来的时刻。
通过上述步骤中的获取位于预设负荷迁移线路上的每个基站小区在当前时刻的第一实际负荷；确定每个基站小区所对应的最近左邻区和最近右邻区，并分别获取对应的基站小区所对应的最近右邻区和最近左邻区在上一时刻对应的实际负荷；基于最近右邻区在上一时刻对应的实际负荷，确定最近右邻区在上一时刻按第一预设比例迁移至基站小区的第一迁移负荷，以及基于最近左邻区在上一时刻对应的实际负荷，确定最近左邻区在上一时刻按第二预设比例迁移至基站小区的第二迁移负荷；根据第一实际负荷和第一迁移负荷的差值，确定基站小区在当前时刻进行正向预测时所对应的实时负荷，以及根据第一实 际负荷和第二迁移负荷的差值，确定基站小区在当前时刻进行反向预测时所对应的实时负荷，实现了获取每个基站小区精准的实时负荷，进一步为基站小区的节能、负荷控制、话务保障等操作提供有效的参考和指导。
在一些实施例中，负荷预测模型包括正向预测负荷计算公式和反向预测负荷计算公式

其中，m表示当前时刻，m-1表示上一时刻，→表示正向方向，←表示反向方向，K表示跨层同向组的层数，p为目标基站小区Ci对应的P个跨层同向组中的一个跨层同向组的标号，为目标基站小区Ci在当前时刻的实际负荷，分别表示目标基站小区Ci在当前时刻和上一时刻对应的正向预测负荷，为目标基站小区Ci对应的右邻区同向组中的第一层主邻区在上一时刻的实际负荷，γ为第一预设比例，为左邻区同向组中的第j-i层主邻区在当前时刻的实际负荷，为左邻区同向组中的第j-i层主邻区对应的最近右邻区在上一时刻的实际负荷，为左邻区同向组中的第j-i层主邻区在上一时刻对应的正向预测负荷， 分别表示目标基站小区Ci在当前时刻和上一时刻的反向预测负荷，为目标基站小区Ci对应的左邻区同向组中的第一层主邻区在上一时刻的实际负荷，为右邻区同向组中的第j-i层主邻区在当前时刻的实际负荷，为右邻区同向组中第j-i层主邻区对应的最近左邻区在上一时刻的实际负荷，为右邻区同向组中的第j-i层主邻区在上一时刻对应的反向预测负荷；α、β、γ为介于[0，1]之间的权重因子，α为目标基站小区Ci自身负荷的权重，(1-α)为邻区同向组对应负荷的权重，β为每个基站小区的实时负荷的权重，(1-β)为每个基站小区的预测负荷的权重，γ为第一预设比例，且γ为正向负荷预测中，对应的最近右邻区在当前时刻迁移至对应的基站小区的负荷与对应的最近右邻区上一时刻的实际负荷的负荷比，1-γ为第二预设比例，且1-γ为在反向负荷预测中，对应的最近左邻区在当前时刻迁移至对应的基站小区的负荷与对应的最近左邻区上一时刻的实际负荷的负荷比。
在本实施例中，基站小区对应为一字形高铁网络对应的基站小区，对应的跨层同向组的个数P＝1；在本实施例中，α、β、γ的默认值为均0.5。
在本实施例中，对应的正向负荷预测方向和反向负荷预测方向分别用“→”和“←”所表示，因此，对于分别进行的正向负荷预测和反向负荷预测的左邻区同向组和右邻区同向组中主邻区所对应的邻区编号也均对应带有方向性，通过用带有方向数据的邻区编号籍以说明在对应的负荷预测时，跨层同向组的主邻区均是从第一层主邻区递增到第K层主邻区的，例如：当j＝i+2时，则表示目标基站小区Ci对 应的左邻区同向组中的第2层主邻区，又例如：当j＝i+K时，表示目标基站小区Ci对应的右邻区同向组中的第K层主邻区，j仅表示对应的基站小区的标号，并不代表参与负荷预测计算的参数大小。
需要说明的是，当前时刻和上一时刻之间的时间间隔为负荷预测和负荷信息传递的周期T，T根据高铁专网下典型的基站小区平均覆盖距离以及高铁列车平均车速来确定，T的取值范围大致为10～60秒，可选T＝30秒；需要说明的是，如果T太小，负荷预测和负荷信息传递的开销会很大，而如果T太大，则高速列车很可能穿越出目标基站小区及目标基站小区关联的部分主邻区，导致对应的基站小区无法采样到真实的实时负荷，目标基站小区无法实现有效的负荷预测。
需要进一步说明的是，在T设置为匹配基站小区覆盖距离和高铁列车速度的情况下，若α和β越大，也就是目标基站小区自身负荷权重和实际负荷权重越大，预测负荷的准确度越高，但是负荷的探测距离会越短，不利于感应更远距离的负荷，不利于提前更早预测负荷的到来；若α和β越小，即目标基站小区自身负荷权重和实际负荷权重越小，则负荷探测距离会越长，但预测负荷的准确度越低；在本公开实施例中，负荷预测的目的并非主要是期望预测负荷和实际负荷尽量接近，也就是并非以提高预测准确度为核心目标，而是要将预测负荷和实际负荷结合在一起综合考虑，为基站小区的节能、负荷控制、话务保障等操作提供有效的参考和指导。
在一些实施例中，基于正向预测负荷和反向预测负荷确定对应的负荷预测结果，通过如下步骤41至步骤42实现。
步骤41、对正向预测负荷和反向预测负荷求和，得到目标基站在预设负荷迁移线路上所对应的第一预测总负荷。
在本实施例中，采用如下计算公式计算第一预测总负荷：
其中，p为目标基站小区对应的P个跨层同向组中跨层同向组的标号，表示第一预测总负荷，表示正向预测负荷，表示反向预测负荷。
步骤42、累加多条预设负荷迁移线路所对应的第一预测总负荷，生成总预测负荷，其中，负荷预测结果包括总预测负荷。
在本实施例中，采用如下公式计算总预测负荷
其中，p为目标基站小区对应的P个跨层同向组中一个跨层同向组的标号，为总预测负荷，表示第一预测总负荷。
在本实施例中，对于一字形高铁网络而言，总预测负荷等于第一预测总负荷，对应十字形、米字形的高铁网络而言，位于多条负荷迁移线路交汇点出的目标基站小区所对应的总预测负荷则为多项第一预测总负荷之和。
通过上述步骤中的对正向预测负荷和反向预测负荷求和，得到目标基站在预设负荷迁移线路上所对应的第一预测总负荷；累加多条预设负荷迁移线路所对应的第一预测总负荷，生成总预测负荷，其中，负荷预测结果包括总预测负荷，实现在当前时刻，目标基站小区预测的下一时刻到来的总负荷的汇总运算。
在一些实施例中，在确定对应的负荷预测结果之后，还实施如下步骤51至步骤53。
步骤51、分别获取当前时刻所对应的总预测负荷和当前时刻所对应的实时负荷，以及分别获取上一 时刻所对应的历史总预测负荷和上一时刻所对应的历史实际负荷。
在本实施例中，获取对正向预测负荷和反向预测负荷进行汇总以及对多条负荷迁移线路对应的预测总负荷求和，从而得到当前时刻目标基站小区所对应的预测负荷。
步骤52、基于当前时刻所对应的实际负荷、当前时刻所对应的总预测负荷、上一时刻所对应的历史总预测负荷、上一时刻所对应的历史实际负荷其中至少两项，生成负荷预测的差变函数，并确定对应的差变函数所对应的计算值，其中，差变函数用于表征负荷预测的预测性能。
在本实施例中，通过预设的对负荷预测模型的参数进行优化的计算函数确定不同的负荷预测灵敏度参数的表现方式，例如：可以采用当前时刻所对应的实时负荷与当前时刻所对应的总预测负荷的差值作为差变函数。
在本实施例中，所谓生成负荷预测的差变函数是指以当前时刻所对应的实际负荷、当前时刻所对应的总预测负荷、上一时刻所对应的历史总预测负荷、上一时刻所对应的历史实际负荷其中至少两项为对应的参数，进而定义出预设的差变函数；而确定差变函数所对应的计算值是指将上述四项参数中的带入对应定义的差变函数中所得到的函数值，即为对应的计算值。
步骤53、判断差变函数所对应的计算值是否大于预设阈值，并根据判断结果对负荷预测模型所对应的权重因子进行调整，其中，负荷预测模型所对应的权重因子包括目标基站小区Ci自身负荷的权重α、每个基站小区的实际负荷的权重β、第一预设比例γ。
在本实施例中，通过判断差变函数所对应的计算值是否大于预设阈值(也就是对应的门限)，从而确定负荷预测模型对负荷的预测是否灵敏，当差变函数值为正且超过预设阈值(设定的阈值上限)，表明预测偏快且灵敏，将α和β，按预设步长进行增大，以降低邻区负荷和预测负荷的权重，反之，当差变函数值为负且低于预设阈值(设定的阈值下限)，表明预测偏慢偏迟钝，则将α和β按预设步长进行减小，以提高邻区负荷和预测负荷的权重；在本实施例中，预设步长默认的设定值为0.05。
通过上述步骤中的分别获取当前时刻所对应的总预测负荷和当前时刻所对应的实时负荷，以及分别获取上一时刻所对应的历史总预测负荷和上一时刻所对应的历史实际负荷；基于当前时刻所对应的实际负荷、当前时刻所对应的总预测负荷、上一时刻所对应的历史总预测负荷、上一时刻所对应的历史实际负荷其中至少两项，生成负荷预测的差变函数，并确定差变函数所对应的计算值，其中，差变函数用于表征负荷预测的预测性能；判断差变函数所对应的计算值是否大于预设阈值，并根据判断结果对负荷预测模型所对应的权重因子进行调整，实现了根据预测结果对负荷预测模型的模型参数进行实时调整，从而实现兼顾负荷预测的准确度，并使负荷预测对应的探测感应距离延长。
在一些可选实施方式中，差变函数包括以下其中一种：当前时刻所对应的实际负荷与上一时刻所对应的历史总预测负荷的差值、当前时刻所对应的实际负荷与上一时刻所对应的历史总预测负荷的差值和当前时刻所对应的实际负荷的比值、上一时刻所对应的历史总预测负荷和当前时刻所对应的总预测负荷的差值、当前时刻所对应的实际负荷与上一时刻所对应的历史实际负荷的差值和由对当前时刻所对应的总预测负荷与上一时刻所对应的历史总预测负荷的差值、当前时刻所对应的实际负荷与上一时刻所对应的历史实际负荷的差值进行加权所生成的负荷差值。
在本实施例中，每个基站小区都支持在确定对应的负荷预测结果后，对对应的权重因子进行优化，也就是对负荷预测模型的模型参数进行维护和优化调整，形成适配自身的个性化动态负荷预测模型，在一些可选实施方式中，采用如下方式进行对应的模型参数的优化，也即按如下方式定义及生成第一的差变函数。
方式一：定义如下公式的差变函数dif(m):
或定义如下公式的差变函数dif(m):
其中，Li(m)为目标基站小区Ci当前时刻所对应的实际负荷，为上一时刻所对应的历史总预测负荷。
方式二：定义如下公式的差变函数dif(m)
其中，表示当前时刻所对应的总预测负荷，表示上一时刻所对应的历史总预测负荷。
或定义如下公式计算差变函数dif(m)：
Dif(m)＝Li(m)-Li(m-1)
其中，Li(m)为当前时刻所对应的实际负荷，Li(m-1)表示上一时刻所对应的历史实际负荷。
或进一步采用如下公式定义差变函数dif(m)：
其中，表示当前时刻所对应的总预测负荷，表示上一时刻所对应的历史总预测负荷，Li(m)为当前时刻所对应的实际负荷，Li(m-1)表示上一时刻所对应的历史实际负荷，μ当前时刻和上一时刻的预测负荷的权重，μ的取值范围为[0，1]，默认取值为0.5。
在方式二中，根据实际负荷和预测负荷的变化实时调整模型参数，兼顾预测准确度和更长的探测感应距离；并且，在当获知到预测负荷在增加，即列车负荷在靠近时，将α和β按预设步长增大，进一步聚焦在目标基站小区自身负荷和实际负荷上，以提高预测准确度，当预测负荷在减少，即列车负荷在远离时，则将α和β按预设步长减小，进一步聚焦在邻区负荷和预测负荷上，以扩展探测感应基站邻区负荷的距离。
方式三：按如下公式计算预设的高速专网区段的总体负荷的重心CoL(m，p)：
其中，高速专网区段由目标基站小区Ci和其对应的一个跨K层同向组的2K层主邻区所组成，CoL(m，p)表示在m时刻，目标基站小区Ci对应的高速专网区段的总体负荷的重心，表示高速专网区段中每个基站小区Cj在m时刻的实时负荷，p为目标基站小区Ci对应的P个跨层同向组中的一个跨层同向组的标号；然后，将总体负荷的重心作为差变函数。
在本实施例中，每个基站小区，采用基于其自身和对应的跨K层的跨层同向组中每个基站邻区的实际负荷，判断高速专网区段总体负荷的重心，通过总体负荷的重心位置的变化来判断在高速专网对应的高铁线路上的列车的移动方向以及对应造成的负荷移动方向，当判断高速专网区段总体负荷的重心位置是从左到右正向移动，则γ按预设步长减小，反之则增大，在本实施例中，预设步长可以取值0.05，γ的调整不能超过预设的取值。
在一些实施例中，接收跨层同向组的左邻区同向组和右邻区同向组分别传递的跨层邻区负荷数据，通过如下步骤61至步骤62实现。
步骤61、最近左邻区在从其接收到的第一邻区组负荷数据中检测出第一目标负荷数据后，将第一目标负荷数据和其自身在当前时刻对应的实际负荷和上一时刻所对应的历史预测负荷，按预设格式合成第一跨层邻区负荷数据，并向对应的目标基站小区传递，其中，第一邻区组负荷数据最近左邻区自身的左邻区同向组所对应的跨层邻区负荷数据，第一目标负荷数据包括最近左邻区自身的左邻区同向组中排除 其第K层主邻区后的所有主邻区所对应的跨层邻区负荷数据。
在一些可选实施方式中，设定目标基站小区为C8，设定目标基站小区C8对应的左邻区同向组为{C2、C3、C4、C5、C6、C7}，目标基站小区C8对应的最近左邻区为C7，目标基站小区C8接收其左邻区同向组传递的跨层邻区负荷数据包括如下过程：最近左邻区C7所接收到其自身的左邻区同向组{C1、C2、C3、C4、C5、C6}的跨层邻区负荷数据M0＝{L(1)，L(2)，L(3)，L(4)，L(5)，L(6)}，最近左邻区C7在M0中删除其左邻区同向组中第K层主邻区C1所对应的负荷L(1)后，添加其自身当前时刻传递给目标基站小区C8对应的负荷L(7)，构成对应的第一跨层邻区负荷数据，也就是M1＝{L(2)，L(3)，L(4)，L(5)，L(6)，L(7)}，并传递给目标基站小区C8。
步骤62、最近右邻区在从其接收到的第二邻区组负荷数据中检测出第二目标负荷数据后，将第二目标负荷数据和其自身在当前时刻对应的实际负荷和上一时刻所对应的历史预测负荷，按预设格式合成第二跨层邻区负荷数据，并向对应的目标基站小区传递，其中，第二邻区组负荷数据包括最近右邻区自身的右邻区同向组所对应的跨层邻区负荷数据，第二目标负荷数据包括最近右邻区自身的右邻区同向组中排除其第K层主邻区后的所有主邻区所对应的跨层邻区负荷数据；
在一些可选实施方式中，设定目标基站小区为C8，设定目标基站小区C8对应的右邻区同向组为{C9、C10、C11、C12、C13、C14}，目标基站小区C8对应的最近右邻区为C9，目标基站小区C8接收其右邻区同向组传递的跨层邻区负荷数据包括如下过程：最近右邻区C9所接收到其自身的左邻区同向组{C10、C11、C12、C13、C14、C15}的跨层邻区负荷数据M2＝{L(10)，L(11)，L(12)，L(13)，L(14)，L(15)}，最近右邻区C9在M2中删除其右邻区同向组中第K层主邻区C15所对应的负荷L(15)后，添加其自身当前时刻传递给目标基站小区C8对应的负荷L(9)，构成对应的第二跨层邻区负荷数据，也就是M3＝{L(9)，L(10)，L(11)，L(12)，L(13)，L(14)}，并传递给目标基站小区C8。
在本实施例中，最近右邻区和最近左邻区组成了对应的基站小区的左、右主邻区；在本实施例中，基站小区传递的负荷信息均是为基于自身当前时刻的实时负荷、上一时刻对应的历史预测负荷以及对应的跨层同向组中的主邻区的负荷信息。
在本实施例中，准备向对应的目标基站小区传递对应的负荷数据之前，最近左邻区先会接收其自身对应的左邻区同向组传递的对应的跨层邻区负荷数据，此时，最近左邻区的对应的左邻区同向组中第K层的主邻区则不属于目标基站小区对应的左邻区同向组中的主邻区了，因此，最近左邻区传递的负荷数据不包括最近左邻区对应的左邻区同向组第K层主邻区的负荷数据；准备向对应的目标基站小区传递对应的负荷数据之前，最近右邻区先会接收其自身对应的右邻区同向组传递的对应的跨层邻区负荷数据，此时，最近右邻区对应的右邻区同向组中第K层的主邻区则不属于目标基站小区对应的右邻区同向组中的主邻区了，因此，目标基站小区传递的负荷数据不包括最近右邻区对应的右邻区同向组第K层主邻区的负荷数据。
在本实施例中，将最近左邻区对应的左邻区同向组/最近右邻区对应的右邻区同向组第K层主邻区的负荷从按预设格式存储的邻区组负荷数据中删除后，补充最近左邻区/最近右邻区当前时刻的实时负荷和上一时刻所对应的历史预测负荷，进而生成按预设格式存在的对应的负荷数据，并传递至目标基站小区。
通过上述步骤实现基站小区按定向接力的方式传递负荷信息。
在一些实施例中，在位于预设负荷迁移线路上的多个基站小区中，确定跨层同向组，通过如下步骤71至步骤73实现。
步骤71、获取在预设负荷迁移线路上的每个基站小区所对应的切换频次粒度数据，并选取切换频次粒度数据最大的邻区作为基站小区的主邻区；其中，切换频次粒度数据用于表征在预设时间内，基站小 区与基站小区对应的邻区之间用户切换的次数。
步骤72、将选取的每个基站小区对应的左主邻区和右主邻区作为基站小区对应的基站小区同向组。
步骤73、以目标基站小区为起点，沿预设负荷迁移线路的正向方向和反向方向分别选取K个连续的基站小区所对应的基站小区同向组，并将选取到的基站小区同向组中基站小区进行去重复处理，以得到跨层同向组。
在本实施例中，基站小区Ci根据自身一段时间(如1周对应的168个小时)与其各个邻区的“邻区对”切换次数的小时粒度统计数据，包括切出到邻区和邻区切入到基站小区Ci的次数，按照聚类算法分成两类邻区，将切换次数属于更多一类的邻区作为基站小区Ci的主邻区；当然，也可以采用由网络规划优化人员，基于基站小区Ci及邻区的网络工程参数信息，如：安装位置、覆盖方向等，或再参考切换统计数据，手工筛选出基站小区Ci沿高铁线路方向上左右两边最近的相邻邻区，并设定为基站小区Ci的主邻区；然后，将基站小区Ci左右两边最近的相邻邻区作为基站小区Ci对应的同向组，然后基于多个基站小区对应的同向组之间的关联关系，以确定各基站小区对应的跨层同向组，例如：假如C2是C1同向组的右邻区，C3是C2同向组的右邻区，则可以关联确认C3是C1跨越1层的右邻区，以此类推，可以获得C1跨越K层的右邻区Ck。{C1，C2，...CK}就组成了C1跨越K层的跨层同向组。
在本实施例中，将基站小区的各主邻区按同一方向线路上的负荷迁移时空特征图案，匹配筛选出左主邻区和右主邻区，并划分到同一方向路线组，也就是同向组。
针对目标基站小区Ci及其主邻区，从1周小时粒度的切换统计数据中选出切换次数最多的几个时段，例如，每天选取2个切换次数最多的忙时时段，则1周有14个小时时段，获取这些时段里10～60秒级(如30秒级)粒度的用户数统计数据D2，通过聚类算法将D2分成两类，将用户数多的负荷设置为1，少的负荷设置为0，然后，搜索筛选出符合如表1和表2的负荷迁移时空特征(对应满足其中之一即可)的主邻区1和主邻区2，如此，筛选出的主邻区1和主邻区2就作为目标基站小区Ci的左邻区和右邻区。
表1
表2
在本实施例中，基于多个关联关系的同向组，确定各基站小区的跨层同向组；假如C2是C1同向组的右邻区，C3是C2同向组的右邻区，则可以关联确认C3是C1跨越1层的右邻区，以此类推，可以获得C1跨越K层的右邻区Ck，{C1，C2，...CK}就组成了C1跨越K层的跨层同向组；在本公开实施例中，确定出的目标基站小区对应的跨层同向组是包括左、右两个跨层同向组的，例如：左邻区同向组和右邻区同向组，其中，左邻区同向组和右邻区同向组均具有K层主邻区，K为预设值，例如：目标基站小区Ci期望提前更早预测出未来可能到来的负荷，感应更远距离的负荷，那么K可以设定很大，但会使的跨层传递负荷信息的开销和难度增大，随机不确定性也会增加，因此，需要兼顾考虑负荷预测的目标要求和开销代价等来设定K值，例如：K≤9；参考图2所示，{C1，C2，C3，C4，C5，C6}就是C7跨越6层的跨层同向左邻区，{C8，C9，C10，C11，C12，C13}就是C7跨越6层的跨层同向右邻区，负荷信息可以基于接力传递向左或向右贯穿6层。
以下给出实施例1至6对本公开实施例的负荷预测进行说明如下：
在进行说明之前，给定一段包含有20个基站小区相连组成连续覆盖的一字形高铁专网，相邻的基站小区互为主邻区，互为左邻或右舍，并组成跨层同向组；实际负荷情况如各个实施例相应图表所示，数值代表了用户数的负荷；实施例设定了几种典型的负荷分布、迁移、变化场景；在本公开实施例1至6中，K＝6，α＝0.5，β＝0.5，γ＝0.5，负荷预测模型的模型参数的优化采用上述模型参数的优化方式二中的：并且μ＝0.5。
实施例1
模拟单列高铁从左到右正向移动，实际负荷和预测负荷分别如表3和表4所示，从预测负荷的结果看，预测模型可以提前探测和感应到即将到来的负荷及变化趋势，结合实际负荷和预测负荷，基站小区可以提前准备节能或负荷控制等操作。
表3
表4
实施例2
模拟单列高铁从右到左反向移动。实际负荷和预测负荷分别如表5和表6所示，从预测负荷的结果看，预测模型可以提前探测和感应到即将到来的负荷及变化趋势，结合实际负荷和预测负荷，基站小区可以提前准备节能或负荷控制等操作。

表5
表6
实施例3
模拟双列高铁分别从左到右和从右到左相向移动。实际负荷和预测负荷分别表7和表8所示，从预测负荷的结果看，预测模型可以提前探测和感应到即将到来的负荷及变化趋势，结合实际负荷和预测负荷，基站小区可以提前准备节能或负荷控制等操作。
表7

表8
实施例4
模拟双列高铁分别从两个起点同时从左到右正向移动。实际负荷和预测负荷分别表9和表10所示，从预测负荷的结果看，预测模型可以提前探测和感应到即将到来的负荷及变化趋势，结合实际负荷和预测负荷，基站小区可以提前准备节能或负荷控制等操作。
表9
表10
实施例5
模拟单列高铁从左到右正向移动过程中某时刻某小区随机增开反向单列高铁。实际负荷和预测负荷分别如表11和表12所示，从预测负荷的结果看，预测模型可以提前探测和感应到即将到来的负荷及变化趋势，结合实际负荷和预测负荷，基站小区可以提前准备节能或负荷控制等操作。

表11
表12
实施例6
模拟单列高铁从左到右正向移动过程中，到某些中间小区，随机增加或减少用户数，来模拟到达车站的乘客上下车带来的用户数负荷变化；实际负荷和预测负荷分别如表13和表14所示，从预测负荷的结果看，预测模型可以提前探测和感应到即将到来的负荷及变化趋势，结合实际负荷和预测负荷，小区可以提前准备节能或负荷控制等操作。
表13

表14
本公开实施例中还提供了一种动态负荷预测装置，该装置用于实现上述实施例及可选实施方式，已经进行过说明的不再赘述。以下所使用的术语“模块”、“单元”、“子单元”等可以实现预定功能的软件和/或硬件的组合。尽管在以下实施例中所描述的装置较佳地以软件来实现，但是硬件，或者软件和硬件的组合的实现也是可能并被构想的。
图3是本公开实施例提供的动态负荷预测装置的结构框图，如图3所示，该装置包括：获取模块32、获取模块32、接收模块33和预测模块34。
确定模块31，被配置为在位于预设负荷迁移线路上的多个基站小区中，确定跨层同向组，其中，位于跨层同向组两端的基站小区均为目标基站小区对应的第K层主邻区，K为预设值。
获取模块32，与确定模块31耦合连接，被配置为获取当前时刻的实时负荷和上一时刻所对应的历史预测负荷，历史预测负荷用于表征在上一时刻预测的当前时刻将产生的负荷迁移变化。
接收模块33，与获取模块32耦合连接，被配置为接收跨层同向组的左邻区同向组和右邻区同向组分别传递的跨层邻区负荷数据，其中，对应的跨层邻区负荷数据均包括历史负荷数据和实时负荷数据，历史负荷数据用于表征对应的邻区同向组所具有的K层主邻区在上一时刻对应的预测负荷，实时负荷数据用于表征对应的邻区同向组所具有的K层主邻区在当前时刻的实际负荷。
预测模块34，与接收模块33耦合连接，被配置为利用预设的负荷预测模型，基于实时负荷、历史预测负荷以及跨层邻区负荷数据进行负荷预测，得到目标基站小区所对应的正向预测负荷和反向预测负荷，并基于正向预测负荷和反向预测负荷确定对应的负荷预测结果。
通过本公开实施例的动态负荷预测装置，采用在位于预设负荷迁移线路上的多个基站小区中，确定跨层同向组，位于跨层同向组两端的基站小区均为目标基站小区对应的第K层主邻区，K为预设值；获取当前时刻的实时负荷和上一时刻所对应的历史预测负荷，历史预测负荷用于表征在上一时刻预测的当前时刻将产生的负荷迁移变化；接收跨层同向组的左邻区同向组和右邻区同向组分别传递的跨层邻区负荷数据，对应的跨层邻区负荷数据均包括历史负荷数据和实时负荷数据，历史负荷数据用于表征对应的邻区同向组所具有的K层主邻区在上一时刻对应的预测负荷，实时负荷数据用于表征对应的邻区同向组所具有的K层主邻区在当前时刻的实际负荷；利用预设的负荷预测模型，基于实时负荷、历史预测负荷以及跨层邻区负荷数据进行负荷预测，得到目标基站小区所对应的正向预测负荷和反向预测负荷，并基于正向预测负荷和反向预测负荷确定对应的负荷预测结果，解决了相关技术中负荷预测方法无法适用用户高速移动、负荷突发、没有精确时间周期规律的高速移动通信场景的问题，实现了基站小区能够在精确的时间点上提前探测和感应到更远距离的邻区到来的用户负荷及变化趋势，并通过结合基站小区的预测负荷和实际负荷，为基站小区的节能、负荷控制、话务保障等操作提供有效的参考和指导的有益效果。
在一些实施例中，该预测模块34还被配置为利用负荷预测模型，对实时负荷、历史预测负荷和左邻区同向组所对应的历史负荷数据与实时负荷数据，进行正向负荷计算，得到正向预测负荷；以及利用负荷预测模型，对实时负荷、历史预测负荷和右邻区同向组所对应的历史负荷数据与实时负荷数据，进行反向负荷计算，得到反向预测负荷。
在一些实施例中，该装置还用于获取位于预设负荷迁移线路上的每个基站小区在当前时刻的第一实 际负荷；确定每个基站小区所对应的最近左邻区和最近右邻区，并分别获取对应的基站小区所对应的最近右邻区和最近左邻区在上一时刻对应的实际负荷，其中，设定从预设负荷迁移线路的左端至右端为正向方向，最近左邻区用于表征在正向方向上，位于对应的基站小区左侧的第一个基站小区，最近右邻区用于表征在正向方向上，位于对应的基站小区右侧的第一个基站小区；基于最近右邻区在上一时刻对应的实际负荷，确定最近右邻区在上一时刻按第一预设比例迁移至基站小区的第一迁移负荷，以及基于最近左邻区在上一时刻对应的实际负荷，确定最近左邻区在上一时刻按第二预设比例迁移至基站小区的第二迁移负荷；根据第一实际负荷和第一迁移负荷的差值，确定基站小区在当前时刻进行正向预测时所对应的实时负荷，以及根据第一实际负荷和第二迁移负荷的差值，确定基站小区在当前时刻进行反向预测时所对应的实时负荷。
在一些实施例中，负荷预测模型包括正向预测负荷计算公式和反向预测负荷计算公式

其中，m表示当前时刻，m-1表示上一时刻，→表示正向方向，←表示反向方向，K表示跨层同向组的层数，p为目标基站小区Ci对应的P个跨层同向组中的一个跨层同向组的标号，Li(m)为目标基站小区Ci在当前时刻的实际负荷，分别表示目标基站小区Ci在当前时刻和上一时刻对应的正向预测负荷，为目标基站小区Ci对应的右邻区同向组中的第一层主邻区在上一时刻的实际负荷，γ为第一预设比例，为左邻区同向组中的第j-i层主邻区在当前时刻的实际负荷，为左邻区同向组中的第j-i层主邻区对应的最近右邻区在上一时刻的实际负荷，为左邻区同向组中的第j-i层主邻区在上一时刻对应的正向预测负荷， 分别表示目标基站小区Ci在当前时刻和上一时刻的反向预测负荷，为目标基站小区Ci对应的左邻区同向组中的第一层主邻区在上一时刻的实际负荷，为右邻区同向组中的第j-i层主邻区在当前时刻的实际负荷，为右邻区同向组中第j-i层主邻区对应的最近左邻区在上一时刻的实际负荷，为右邻区同向组中的第j-i层主邻区在上一时刻对应的反向预测负荷；α、β、γ为介于[0，1]之间的权重因子，α为目标基站小区Ci自身负荷的权重，(1-α)为邻区同向组对应负荷的权重，β为每个基站小区的实时负荷的权重，(1-β)为每个基站小区的预测负荷的权重，γ为第一预设比例，1-γ为第二预设比例。
在一些实施例中，该预测模块34还被配置为对正向预测负荷和反向预测负荷求和，得到目标基站在预设负荷迁移线路上所对应的第一预测总负荷；累加多条预设负荷迁移线路所对应的第一预测总负荷，生成总预测负荷，其中，负荷预测结果包括总预测负荷。
在一些实施例中，在确定对应的负荷预测结果之后，该装置还用于分别获取当前时刻所对应的总预测负荷和当前时刻所对应的实时负荷，以及分别获取上一时刻所对应的历史总预测负荷和上一时刻所对应的历史实际负荷；基于当前时刻所对应的实际负荷、当前时刻所对应的总预测负荷、上一时刻所对应的历史总预测负荷、上一时刻所对应的历史实际负荷其中至少两项，生成负荷预测的差变函数，并确定对应的差变函数所对应的计算值，其中，差变函数用于表征负荷预测的预测性能；判断差变函数的计算值是否大于预设阈值，并根据判断结果对负荷预测模型所对应的权重因子进行调整，其中，负荷预测模型所对应的权重因子包括目标基站小区Ci自身负荷的权重α、每个基站小区的实际负荷的权重β、第一预设比例γ。
在一些实施例中，差变函数包括以下其中一种：当前时刻所对应的实际负荷与上一时刻所对应的历史总预测负荷的差值、当前时刻所对应的实际负荷与上一时刻所对应的历史总预测负荷的差值和当前时刻所对应的实际负荷的比值、上一时刻所对应的历史总预测负荷和当前时刻所对应的总预测负荷的差值、当前时刻所对应的实际负荷与上一时刻所对应的历史实际负荷的差值和由对当前时刻所对应的总预测负荷与上一时刻所对应的历史总预测负荷的差值、当前时刻所对应的实际负荷与上一时刻所对应的历史实际负荷的差值进行加权所生成的负荷差值。
在一些实施例中，该获取模块31还被配置为获取在预设负荷迁移线路上的每个基站小区所对应的切换频次粒度数据，并选取切换频次粒度数据最大的邻区作为基站小区的主邻区；其中，切换频次粒度数据用于表征在预设时间内，基站小区与基站小区对应的邻区之间用户切换的次数；将选取的每个基站小区对应的左主邻区和右主邻区作为基站小区对应的基站小区同向组；以目标基站小区为起点，沿预设负荷迁移线路的正向方向和反向方向分别选取K个连续的基站小区所对应的基站小区同向组，并将选取到的基站小区同向组中基站小区进行去重复处理，以得到跨层同向组。
图4是本公开实施例的基站的结构示意图，如图4所示，本公开实施例提供了一种基站，包括处理器41、通信接口42、存储器43和通信总线44，其中，处理器41，通信接口42，存储器43通过通信总线44完成相互间的通信；存储器43，存放计算机程序；处理器41，执行存储器43上所存放的程序时，实现图1中的方法。
该基站中的处理实现图1中的方法步骤，所带来的技术效果与上述实施例执行图1中的动态负荷预测方法的技术效果一致，在此不再赘述。
上述基站提到的通信总线可以是外设部件互连标准(Peripheral Component Interconnect，简称PCI)总线或扩展工业标准结构(Extended Industry Standard Architecture，简称EISA)总线等。该通信总线可以分为地址总线、数据总线、控制总线等。为便于表示，图4中仅用一条粗线表示，但并不表示仅有一根总线或一种类型的总线。
通信接口用于上述终端与其他设备之间的通信。
存储器可以包括随机存取存储器(Random Access Memory，简称RAM)，也可以包括非易失性存储器(non-volatile memory)，例如至少一个磁盘存储器。可选的，存储器还可以是至少一个位于远离前述处理器的存储装置。
上述的处理器可以是通用处理器，包括中央处理器(Central Processing Unit，简称CPU)、网络处理器(Network Processor，简称NP)等；还可以是数字信号处理器(Digital Signal Processing，简称DSP)、专用集成电路(Application Specific Integrated Circuit，简称ASIC)、现场可编程门阵列(Field－Programmable Gate Array，简称FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。
本公开实施例还提供了一种计算机可读存储介质，其上存储有计算机程序，计算机程序被处理器执行时实现如前述任意一个方法实施例提供的动态负荷预测方法。
在本公开提供的又一实施例中，还提供了一种包含指令的计算机程序产品，当其在计算机上运行时，使得计算机执行上述实施例中任一所述的动态负荷预测方法。
与相关技术相比，本公开实施例中提供了动态负荷预测方法、装置、基站和存储介质，通过在位于预设负荷迁移线路上的多个基站小区中，确定跨层同向组，其中，位于跨层同向组两端的基站小区均为目标基站小区对应的第K层主邻区，K为预设值；获取当前时刻的实时负荷和上一时刻所对应的历史预测负荷，历史预测负荷用于表征在上一时刻预测的当前时刻将产生的负荷迁移变化；接收跨层同向组的左邻区同向组和右邻区同向组分别传递的跨层邻区负荷数据，其中，对应的跨层邻区负荷数据均包括历史负荷数据和实时负荷数据，历史负荷数据用于表征对应的邻区同向组所具有的K层主邻区在上一时刻对应的预测负荷，实时负荷数据用于表征对应的邻区同向组所具有的K层主邻区在当前时刻的实际负荷；利用预设的负荷预测模型，基于实时负荷、历史预测负荷以及跨层邻区负荷数据进行负荷预测，得到目标基站小区所对应的正向预测负荷和反向预测负荷，并基于正向预测负荷和反向预测负荷确定对应的负荷预测结果，解决了相关技术中负荷预测方法无法适用用户高速移动、负荷突发、没有精确时间周期规律的高速移动通信场景的问题，实现了基站小区能够在精确的时间点上提前探测和感应到更远距离的邻区到来的用户负荷及变化趋势，并通过结合基站小区的预测负荷和实际负荷，为基站小区的节能、负荷控制、话务保障等操作提供有效的参考和指导的有益效果。
需要说明的是，在本文中，诸如“第一”和“第二”等之类的关系术语仅仅用来将一个实体或者操作与另一个实体或操作区分开来，而不一定要求或者暗示这些实体或操作之间存在任何这种实际的关系或者顺序。而且，术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、方法、物品或者设备不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、方法、物品或者设备所固有的要素。在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括所述要素的过程、方法、物品或者设备中还存在另外的相同要素。
以上所述仅是本公开的具体实施方式，使本领域技术人员能够理解或实现本公开。对这些实施例的多种修改对本领域的技术人员来说将是显而易见的，本文中所定义的一般原理可以在不脱离本公开的精神或范围的情况下，在其它实施例中实现。因此，本公开将不会被限制于本文所示的这些实施例，而是要符合与本文所公开的原理和新颖特点相一致的最宽的范围。

Claims (13)

  1. 一种动态负荷预测方法，应用于高速专网下的目标基站小区，包括：
    在位于预设负荷迁移线路上的多个基站小区中，确定跨层同向组，其中，位于所述跨层同向组两端的基站小区均为所述目标基站小区对应的第K层主邻区，K为预设值；
    获取当前时刻的实时负荷和上一时刻所对应的历史预测负荷，所述历史预测负荷用于表征在上一时刻预测的当前时刻将产生的负荷迁移变化；
    接收所述跨层同向组的左邻区同向组和右邻区同向组分别传递的跨层邻区负荷数据，其中，对应的所述跨层邻区负荷数据均包括历史负荷数据和实时负荷数据，所述历史负荷数据用于表征对应的邻区同向组所具有的K层主邻区在上一时刻对应的预测负荷，所述实时负荷数据用于表征对应的邻区同向组所具有的K层主邻区在当前时刻的实际负荷；以及
    利用预设的负荷预测模型，基于所述实时负荷、所述历史预测负荷以及所述跨层邻区负荷数据进行负荷预测，得到所述目标基站小区所对应的正向预测负荷和反向预测负荷，并基于所述正向预测负荷和反向预测负荷确定对应的负荷预测结果。
  2. 根据权利要求1所述的方法，其中，利用预设的负荷预测模型，基于所述实时负荷、所述历史预测负荷以及所述跨层邻区负荷数据进行负荷预测，得到所述目标基站小区所对应的正向预测负荷和反向预测负荷，包括：
    利用所述负荷预测模型，对所述实时负荷、所述历史预测负荷和所述左邻区同向组所对应的所述历史负荷数据与所述实时负荷数据，进行正向负荷计算，得到所述正向预测负荷；以及
    利用所述负荷预测模型，对所述实时负荷、所述历史预测负荷和所述右邻区同向组所对应的所述历史负荷数据与所述实时负荷数据，进行反向负荷计算，得到所述反向预测负荷。
  3. 根据权利要求2所述的方法，其中，所述方法还包括：
    获取位于预设负荷迁移线路上的每个所述基站小区在当前时刻的第一实际负荷；
    确定每个所述基站小区所对应的最近左邻区和最近右邻区，并分别获取对应的所述基站小区所对应的所述最近右邻区和所述最近左邻区在上一时刻对应的实际负荷，其中，设定从所述预设负荷迁移线路的左端至右端为正向方向，所述最近左邻区用于表征在所述正向方向上，位于对应的所述基站小区左侧的第一个基站小区，所述最近右邻区用于表征在所述正向方向上，位于对应的所述基站小区右侧的第一个基站小区；
    基于所述最近右邻区在上一时刻对应的实际负荷，确定所述最近右邻区在上一时刻按第一预设比例迁移至所述基站小区的第一迁移负荷，以及基于所述最近左邻区在上一时刻对应的实际负荷，确定所述最近左邻区在上一时刻按第二预设比例迁移至所述基站小区的第二迁移负荷；以及
    根据所述第一实际负荷和所述第一迁移负荷的差值，确定所述基站小区在当前时刻进行正向预测的情况下所对应的所述实时负荷，以及根据所述第一实际负荷和所述第二迁移负荷的差值，确定所述基站小区在当前时刻进行反向预测的情况下所对应的所述实时负荷。
  4. 根据权利要求3所述的方法，其中，所述负荷预测模型包括正向预测负荷计算公式和反向预测负荷计算公式

    其中，m表示当前时刻，m-1表示上一时刻，→表示所述正向方向，←表示反向方向，K表示跨层同向组的层数，p为目标基站小区Ci对应的P个跨层同向组中的一个跨层同向组的标号，Li(m)为目标基站小区Ci在当前时刻的实际负荷，分别表示目标基站小区Ci在当前时刻和上一时刻对应的正向预测负荷，为目标基站小区Ci对应的所述右邻区同向组中的第一层主邻区在上一时刻的实际负荷，γ为所述第一预设比例，为所述左邻区同向组中的第j-i层主邻区在当前时刻的实际负荷，为所述左邻区同向组中的第j-i层主邻区对应的最近右邻区在上一时刻的实际负荷，为所述左邻区同向组中的第j-i层主邻区在上一时刻对应的正向预测负荷，分别表示目标基站小区Ci在当前时刻和上一时刻的反向预测负荷，为目标基站小区Ci对应的所述左邻区同向组中的第一层主邻区在上一时刻的实际负荷，为所述右邻区同向组中的第j-i层主邻区在当前时刻的实际负荷，为所述右邻区同向组中第j-i层主邻区对应的最近左邻区在上一时刻的实际负荷，为所述右邻区同向组中的第j-i层主邻区在上一时刻对应的反向预测负荷；α、β、γ为介于[0，1]之间的权重因子，α为目标基站小区Ci自身负荷的权重，(1-α)为邻区同向组对应负荷的权重，β为每个基站小区的实时负荷的权重，(1-β)为每个基站小区的预测负荷的权重，γ为所述第一预设比例，1-γ为所述第二预设比例。
  5. 根据权利要求4所述的方法，其中，基于所述正向预测负荷和反向预测负荷确定对应的负荷预测结果，包括：
    对所述正向预测负荷和所述反向预测负荷求和，得到所述目标基站在所述预设负荷迁移线路上所对应的第一预测总负荷；以及
    累加多条预设负荷迁移线路所对应的所述第一预测总负荷，生成总预测负荷，其中，所述负荷预测结果包括所述总预测负荷。
  6. 根据权利要求5所述的方法，其中，在确定对应的负荷预测结果之后，所述方法还包括：
    分别获取当前时刻所对应的所述总预测负荷和当前时刻所对应的实际负荷，以及分别获取上一时刻所对应的历史总预测负荷和上一时刻所对应的历史实际负荷；
    基于当前时刻所对应的实际负荷、当前时刻所对应的所述总预测负荷、上一时刻所对应的历史总预测负荷、上一时刻所对应的历史实际负荷其中至少两项，生成负荷预测的差变函数，并确定对应的所述差变函数所对应的计算值，其中，所述差变函数用于表征负荷预测的预测性能；
    判断所述差变函数所对应的计算值是否大于预设阈值，并根据判断结果对所述负荷预测模型所对应的权重因子进行调整，其中，所述负荷预测模型所对应的权重因子包括所述目标基站小区Ci自身负荷的权重α、每个基站小区的实际负荷的权重β、所述第一预设比例γ。
  7. 根据权利要求6所述的方法，其中，所述差变函数包括以下其中一种：当前时刻所对应的实际负荷与上一时刻所对应的历史总预测负荷的差值、当前时刻所对应的实际负荷与上一时刻所对应的历史总预测负荷的差值和当前时刻所对应的实际负荷的比值、上一时刻所对应的历史总预测负荷和当前时刻所 对应的所述总预测负荷的差值、当前时刻所对应的实际负荷与上一时刻所对应的历史实际负荷的差值和由对当前时刻所对应的所述总预测负荷与上一时刻所对应的历史总预测负荷的差值、当前时刻所对应的实际负荷与上一时刻所对应的历史实际负荷的差值进行加权所生成的负荷差值。
  8. 根据权利要求6所述的方法，还包括：
    按如下公式计算预设的高速专网区段的总体负荷的重心CoL(m，p)：
    其中，所述高速专网区段由所述目标基站小区Ci和其对应的一个所述跨K层同向组的2K层主邻区所组成，CoL(m，p)表示在m时刻，所述目标基站小区Ci对应的所述高速专网区段的总体负荷的重心，表示所述高速专网区段中每个基站小区Cj在m时刻的实际负荷，p为目标基站小区Ci对应的P个跨层同向组中的一个跨层同向组的标号；以及
    确定所述差变函数包括所述总体负荷的重心。
  9. 根据权利要求3所述的方法，其中，接收所述跨层同向组的左邻区同向组和右邻区同向组分别传递的跨层邻区负荷数据，包括：
    所述最近左邻区在从其接收到的第一邻区组负荷数据中检测出第一目标负荷数据后，将所述第一目标负荷数据和其自身在当前时刻对应的实际负荷和上一时刻所对应的历史预测负荷，按预设格式合成第一跨层邻区负荷数据，并向对应的所述目标基站小区传递，其中，所述第一邻区组负荷数据包括所述最近左邻区自身的左邻区同向组所对应的跨层邻区负荷数据，所述第一目标负荷数据包括所述最近左邻区自身的左邻区同向组中排除其第K层主邻区后的所有所述主邻区所对应的跨层邻区负荷数据；以及
    所述最近右邻区在从其接收到的第二邻区组负荷数据中检测出第二目标负荷数据后，将所述第二目标负荷数据和其自身在当前时刻对应的实际负荷和上一时刻所对应的历史预测负荷，按预设格式合成第二跨层邻区负荷数据，并向对应的所述目标基站小区传递，其中，所述第二邻区组负荷数据包括所述最近右邻区自身的右邻区同向组所对应的跨层邻区负荷数据，所述第二目标负荷数据包括所述最近右邻区自身的右邻区同向组中排除其第K层主邻区后的所有所述主邻区所对应的跨层邻区负荷数据。
  10. 根据权利要求1所述的方法，其中，在位于预设负荷迁移线路上的多个基站小区中，确定跨层同向组，包括：
    获取在预设负荷迁移线路上的每个所述基站小区所对应的切换频次粒度数据，并选取所述切换频次粒度数据最大的邻区作为所述基站小区的主邻区；其中，所述切换频次粒度数据用于表征在预设时间内，所述基站小区与所述基站小区对应的邻区之间用户切换的次数；
    将选取的每个所述基站小区对应的左主邻区和右主邻区作为所述基站小区对应的基站小区同向组；以及
    以所述目标基站小区为起点，沿所述预设负荷迁移线路的正向方向和反向方向分别选取K个连续的所述基站小区所对应的所述基站小区同向组，并将选取到的所述基站小区同向组中所述基站小区进行去重复处理，以得到所述跨层同向组。
  11. 一种动态负荷预测装置，应用于高速专网下的目标基站小区，包括：
    确定模块，被配置为在位于预设负荷迁移线路上的多个基站小区中，确定跨层同向组，其中，位于所述跨层同向组两端的基站小区均为所述目标基站小区对应的第K层主邻区，K为预设值；
    获取模块，被配置为获取当前时刻的实时负荷和上一时刻所对应的历史预测负荷，所述历史预测负 荷用于表征在上一时刻预测的当前时刻将产生的负荷迁移变化；
    接收模块，被配置为接收所述跨层同向组的左邻区同向组和右邻区同向组分别传递的跨层邻区负荷数据，其中，对应的所述跨层邻区负荷数据均包括历史负荷数据和实时负荷数据，所述历史负荷数据用于表征对应的邻区同向组所具有的K层主邻区在上一时刻对应的预测负荷，所述实时负荷数据用于表征对应的邻区同向组所具有的K层主邻区在当前时刻的实际负荷；以及
    预测模块，被配置为利用预设的负荷预测模型，基于所述实时负荷、所述历史预测负荷以及所述跨层邻区负荷数据进行负荷预测，得到所述目标基站小区所对应的正向预测负荷和反向预测负荷，并基于所述正向预测负荷和反向预测负荷确定对应的负荷预测结果。
  12. 一种基站，包括处理器、通信接口、存储器和通信总线，其中，处理器、通信接口、存储器通过通信总线完成相互间的通信；
    存储器，其存放计算机程序；
    处理器，其执行存储器上所存放的程序时，实现权利要求1-10任一项所述的动态负荷预测方法。
  13. 一种计算机可读存储介质，其上存储有计算机程序，其中，所述计算机程序被处理器执行时实现如权利要求1-10任一项所述的动态负荷预测方法。
PCT/CN2023/083168 2022-08-25 2023-03-22 动态负荷预测方法、装置、基站和存储介质 WO2024040960A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202211028169.9 2022-08-25
CN202211028169.9A CN117692941A (zh) 2022-08-25 2022-08-25 动态负荷预测方法、装置、基站和存储介质

Publications (1)

Publication Number Publication Date
WO2024040960A1 true WO2024040960A1 (zh) 2024-02-29

Family

ID=90012301

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/083168 WO2024040960A1 (zh) 2022-08-25 2023-03-22 动态负荷预测方法、装置、基站和存储介质

Country Status (2)

Country Link
CN (1) CN117692941A (zh)
WO (1) WO2024040960A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117951555A (zh) * 2024-03-27 2024-04-30 山东德源电力科技股份有限公司 一种用于负控安装测试装置的策略制定方法

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2014161131A (ja) * 2014-06-12 2014-09-04 Hitachi Ltd 無線通信システム、そのシステム用の収容先変更方法、並びに管理装置、及び無線基地局
CN112954707A (zh) * 2019-12-11 2021-06-11 中国电信股份有限公司 基站的节能方法、装置、基站和计算机可读存储介质
CN113132945A (zh) * 2019-12-30 2021-07-16 中国移动通信集团辽宁有限公司 一种铁路专网基站小区节能调度方法和系统
CN113891415A (zh) * 2021-11-17 2022-01-04 西藏先锋绿能环保科技股份有限公司 可节能小区快速筛选方法

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2014161131A (ja) * 2014-06-12 2014-09-04 Hitachi Ltd 無線通信システム、そのシステム用の収容先変更方法、並びに管理装置、及び無線基地局
CN112954707A (zh) * 2019-12-11 2021-06-11 中国电信股份有限公司 基站的节能方法、装置、基站和计算机可读存储介质
CN113132945A (zh) * 2019-12-30 2021-07-16 中国移动通信集团辽宁有限公司 一种铁路专网基站小区节能调度方法和系统
CN113891415A (zh) * 2021-11-17 2022-01-04 西藏先锋绿能环保科技股份有限公司 可节能小区快速筛选方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
WANG YONGGANG, ZHU SHIHUA, LU LING: "PREDICTION OF REVERSE TRAFFIC LOAD IN MULTIPLE SERVICES CDMA SYSTEMS", JOURNAL OF ELECTRONICS AND INFORMATION TECHNOLOGY., vol. 25, no. 2, 1 February 2003 (2003-02-01), pages 152 - 157, XP093142948 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117951555A (zh) * 2024-03-27 2024-04-30 山东德源电力科技股份有限公司 一种用于负控安装测试装置的策略制定方法

Also Published As

Publication number Publication date
CN117692941A (zh) 2024-03-12

Similar Documents

Publication Publication Date Title
US10395529B2 (en) Traffic signal control using multiple Q-learning categories
CN105009475B (zh) 考虑到用户设备(ue)移动性的用于准入控制和资源可用性预测的方法和系统
CN110505650B (zh) 随机异构分层网容量智能评估方法及装置
CN103200125B (zh) 电力数据网节点拥塞规避方法和系统
WO2024040960A1 (zh) 动态负荷预测方法、装置、基站和存储介质
Liu et al. Intelligent handover triggering mechanism in 5G ultra-dense networks via clustering-based reinforcement learning
CN102880975B (zh) Vanet中一种基于负载均衡的竞价博弈方法
CN106102099A (zh) 一种基于驻留时间的异构车联网切换方法
Chaurasia et al. MPMAC: Clustering based MAC protocol for VANETs
CN107396376A (zh) 小区预负荷均衡方法及装置
CN103686895B (zh) 切换控制方法、无线网络控制器和接入节点
CN112235804A (zh) 基站远端单元动态划归方法和装置、小区组网方法和系统
CN111586777B (zh) 室内环境下的网络切换方法、装置、电子设备及存储介质
CN102150453B (zh) 用于在移动台编组到达之前预先调整无线电基站的传输参数的方法、无线电信网络和节点
CN104956733A (zh) 无线通信网络的准入控制的方法及装置
CN103929752A (zh) 基站间动态协同覆盖方法
Midya et al. QoS aware distributed dynamic channel allocation for V2V communication in TVWS spectrum
Lu et al. Predictive contention window-based broadcast collision mitigation strategy for vanet
Cao et al. ETCS: An efficient traffic congestion scheduling scheme combined with edge computing
CN102857305A (zh) 一种多节点联合的频谱感知方法和系统
CN109068375B (zh) 一种基于以用户为中心的超密集网络uudn的动态ap分组方法
WO2024060523A1 (zh) 时域资源分配方法、装置、电子设备及存储介质
CN113992560B (zh) 一种活跃度感知的社交车辆分簇方法、装置及计算机设备
Moore et al. A hybrid (active-passive) clustering technique for VANETs
Singh et al. NWCA: A new weighted clustering algorithm to form stable cluster in VANET

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 23856056

Country of ref document: EP

Kind code of ref document: A1