5G wireless network slicing power allocation method based on deep learning

Abstract

The present invention specifically relates to a 5G wireless network slice power allocation method based on deep learning, including: obtaining the actual value of the allocated power of the corresponding network slice of the 5G wireless network in the previous allocated time period of the time period to be allocated as the initial value of the allocated power; calculating the predicted value of the allocated power of the corresponding network slice in the time period to be allocated based on the initial value of the allocated power; with the goal of ensuring that each network slice is isolated from each other, calculating the initial user power allocation value of each user in the corresponding network slice in the time period to be allocated in combination with the predicted value of the allocated power; with the goal of maximizing the power efficiency in the network slice, calculating the optimal user power allocation value of each user in the corresponding network slice in the time period to be allocated in combination with the initial user power allocation value of each user; allocating power to each user in the corresponding network slice based on the optimal user power allocation value. The present invention can ensure the network quality of each network slice and user, and can improve the power efficiency of the 5G wireless network.

Description

Technical Field

[0001] The present invention relates to the field of mobile communication network technology, and in particular to a 5G wireless network slicing power allocation method based on deep learning. Background Art

[0002] 5G wireless networks will support a large number of diversified business scenarios from vertical industries, such as smart security, high-definition video, smart home, autonomous driving, and augmented reality. These business scenarios usually have different communication requirements. If a dedicated physical network is built for each business scenario, it will inevitably lead to problems such as complex network operation and maintenance, high costs, and poor scalability. Therefore, in order to support multiple business scenarios with different performance requirements on a single physical network and meet the different network requirements of differentiated services, network slicing technology has emerged.

[0003] Each network slice is logically an independent end-to-end network, consisting of a set of network functions and corresponding resources, optimized for specific business scenarios, and providing end-to-end on-demand customized services. In order to ensure normal end-to-end communication, wireless resources must be allocated on the access side to enable more users to access the network while meeting the user QoS.

[0004] Among them, the Chinese patent with publication number CN106851705A discloses "A wireless network slicing method based on slicing flow table", including: extending the Openflow slicing flow table in the SDN architecture to the wireless access network, the matching fields of the slicing flow table include the remote transmitter RRH field and the baseband processing unit BBU field; generating at least one slicing network according to different slicing trigger conditions; wherein the slicing network includes a static slicing network, a dynamic slicing network and a semi-static slicing network.

[0005] The wireless network slicing method in the above existing scheme can realize the separation of data forwarding and control plane. However, there are many users in the network slices of 5G wireless network and their network requirements are different. The limited power of 5G wireless network needs to be allocated to users with different power requirements, and it is necessary to improve the power efficiency of 5G wireless network as much as possible while ensuring the quality of user network. This problem has not yet been solved in a targeted manner. Therefore, how to design a method that can ensure the quality of each network slice and user network and improve the power efficiency of 5G wireless network is a technical problem that needs to be solved urgently. Summary of the invention

[0006] In view of the above-mentioned deficiencies in the prior art, the technical problem to be solved by the present invention is: how to provide a 5G wireless network slicing power allocation method based on deep learning, so as to ensure the network quality of each network slice and user, and improve the power efficiency of the 5G wireless network, thereby improving the effect of 5G wireless network slicing power allocation.

[0007] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0008] A 5G wireless network slicing power allocation method based on deep learning includes:

[0009] S1: Obtaining an actual value of the allocated power of the corresponding network slice of the 5G wireless network in the previous allocated time period of the time period to be allocated as an initial value of the allocated power;

[0010] S2: Calculate the predicted value of the allocated power of the corresponding network slice in the time period to be allocated based on the initial value of the allocated power;

[0011] S3: To ensure that each network slice is isolated from each other, the initial user power allocation value of each user in the corresponding network slice in the time period to be allocated is calculated in combination with the allocation power prediction value;

[0012] S4: Taking the maximization of power efficiency within the network slice as the goal, the optimal user power allocation value of each user in the corresponding network slice in the time period to be allocated is calculated in combination with the initial user power allocation value of each user;

[0013] S5: Allocate power to each user in the corresponding network slice based on the optimal user power allocation value.

[0014] Preferably, in step S1, a time axis for representing the passage of time and divided into time steps is constructed, and a prediction window is established on the time axis, which includes a plurality of time steps and can be moved along the time passage direction of the time axis; the starting point and the end point of the prediction window along the time axis are the start and end times of the corresponding time period to be allocated, that is, the length of the prediction window along the time axis is equal to the time length of the time period to be allocated.

[0015] Preferably, in step S1, when the prediction window moves along the time axis, the step length of each movement is the length of the prediction window along the time axis, that is, the time length of the time period to be allocated.

[0016] Preferably, in step S2, the allocation power prediction value is calculated by using the power mean square error objective function and the slice power total constraint;

[0017] The power mean square error objective function is expressed by the following formula:

[0018]

[0019] Where: represents the predicted value of the allocated power of the mth network slice at time t; T Δ Indicates the length of the time period to be allocated; r m (t) represents the actual value of the allocated power of the mth network slice at time t; M represents the network slice set;

[0020] The slice power allocation constraint is expressed by the following formula:

[0021]

[0022] Where: represents the predicted value of the allocated power of the mth network slice at time t; Θ represents the total power in the 5G wireless network.

[0023] Preferably, the predicted value of the allocated power of the network slice is The actual value of the allocated power r m (t) The following circumstances exist:

[0024] 1) If Then the corresponding network slice needs to use the remaining power in the 5G wireless network;

[0025] 2) If The allocated power of the corresponding network slice remains unchanged during the allocated time period.

[0026] Preferably, the initial value of the allocated power is used as input data, and the stacked bidirectional long short-term memory algorithm in the recurrent neural network is combined with the slice power allocation constraint to solve the power mean square error objective function to calculate the predicted value of the allocated power of the corresponding network slice in the time period to be allocated.

[0027] Preferably, in step S3, the user power initial allocation value is calculated by using the user power allocation objective function and the user power total amount constraint;

[0028] The user power objective function is expressed by the following formula:

[0029]

[0030] Where: r m,u (t) represents the initial user power allocation value allocated to user u by the mth network slice at time t; M represents the network slice set;

[0031] The user power allocation constraint is expressed by the following formula:

[0032]

[0033] r m,u (t)≥h m,u (t);

[0034]

[0035] Ω m (D m (t)=0)≥T m ;

[0036] D m (t) = |r m,u (t+τ)-r m,u (t)|;

[0037] Where: Θ m (t) represents the total power allocated to the mth network slice; N represents the total number of users in the mth network slice; U m represents the user set of the mth network slice; h m,u (t) represents the power required by user u in the mth network slice at time t; R m Indicates the communication rate required by the user; B w represents the channel bandwidth; P m,u (t) represents the downlink transmission power between the base station and the user equipment; G m,u (t) represents the time-varying Rayleigh fading channel gain; T m represents the minimum threshold to ensure the isolation of network slice m; N0 represents the variance of additive white Gaussian noise; Ω m Indicates condition counter; D m (t) represents the power difference allocated to user u at time t+τ and time t in the prediction window; τ represents the time step length.

[0038] Preferably, the user power objective function is solved by combining the asynchronous advantage executor evaluator deep learning algorithm with the user power allocation constraint to calculate the initial user power allocation value of each user in the corresponding network slice in the time period to be allocated.

[0039] Preferably, in step S4, the optimal power allocation value of the user is calculated by using the power optimal objective function and the power optimal allocation constraint;

[0040] The power optimization objective function is expressed by the following formula:

[0041]

[0042] Where: P m,u (t) represents the downlink transmission power between the base station and the user equipment; η EE Represents the power efficiency index; G m,u (t) represents the time-varying Rayleigh fading channel gain; P c Indicates the power consumption of the circuit;

[0043] The optimal power allocation constraint is expressed by the following formula:

[0044]

[0045] 0≤P m,u (t)≤P max ;

[0046] 0≤h m,u (t)≤Θ m (t)+Θ s (t);

[0047] Where: h m,u 9t) represents the power required by user u in the mth network slice at time t; B w represents the channel bandwidth; P m,u (t) represents the downlink transmission power between the base station and the user equipment; G m,u (t) represents the time-varying Rayleigh fading channel gain; N0 represents the additive white Gaussian noise variance; R m Indicates the communication rate required by the user; Θ m (t) represents the total power allocated to the mth network slice; P max represents the maximum downlink transmission power between the base station and the user equipment; Θ s (t) represents the remaining power to be allocated in the 5G wireless network.

[0048] Preferably, the power optimal objective function is solved by combining the Dinkelbach iterative algorithm with the power optimal allocation constraint to calculate the optimal user power allocation value for each user in the corresponding network slice in the time period to be allocated.

[0049] The 5G wireless network slicing power allocation method based on deep learning in the present invention has the following beneficial effects:

[0050] The present invention calculates the predicted value of the allocated power of each network slice in the current time period to be allocated through the actual value of the allocated power in the previous allocated time period, so that the allocated power of each network slice can be effectively predicted before allocating power, thereby avoiding the problem of isolation destruction between network slices caused by frequent power reconfiguration between network slices (power load changes in one network slice will affect other network slices, and frequent resource allocation in each network slice will bring additional overhead to the 5G system and reduce the performance of the network slice), that is, the present invention can ensure the network quality of each network slice and user, thereby improving the effect of power allocation of 5G wireless network slices.

[0051] The present invention calculates the predicted value of the allocated power of each network slice in the current time period to be allocated through the actual value of the allocated power in the previous allocated time period, so that the allocated power of each network slice can be effectively predicted before allocating power, thereby avoiding the problem of isolation destruction between network slices caused by frequent power reconfiguration between network slices (power load changes in one network slice will affect other network slices, and frequent resource allocation in each network slice will bring additional overhead to the 5G system and reduce the performance of the network slice), that is, the present invention can ensure the network quality of each network slice and user, thereby improving the effect of power allocation of 5G wireless network slices.

[0052] The present invention aims to ensure that each network slice is isolated from each other. It calculates the initial user power allocation value of each user in each network slice in combination with the allocation power prediction value, so that while ensuring the isolation between each network slice, it can allocate the minimum power required for each user in the network slice to ensure that the 5G wireless network has enough residual power. Under the premise of ensuring the quality of each network slice and user network, the power efficiency of the 5G wireless network can be improved, thereby further improving the effect of power allocation of 5G wireless network slices.

[0053] The present invention aims to maximize the power efficiency within a network slice, and calculates the optimal user power allocation value for each user in each network slice in combination with the initial user power allocation value of each user, so that power can be redistributed on the basis of the initial user power allocation value, and while meeting the user's communication rate requirements, the power efficiency of the entire network slice can be maximized, that is, the network quality of each network slice and user can be guaranteed and the power efficiency of the 5G wireless network can be improved, thereby further improving the effect of power allocation of 5G wireless network slices. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to make the purpose, technical solution and advantages of the invention more clear, the present invention will be further described in detail below with reference to the accompanying drawings, in which:

[0055] Figure 1 The logical block diagram of the 5G wireless network slicing power allocation method based on deep learning;

[0056] Figure 2 A schematic diagram of the prediction window on the time axis. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but only represents selected embodiments of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work belong to the scope of protection of the present invention.

[0058] It should be noted that similar numbers and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings. In the description of the present invention, it should be noted that the orientation or position relationship indicated by the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inside", "outside", etc. is based on the orientation or position relationship shown in the drawings, or the orientation or position relationship in which the invention product is usually placed when used, which is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation of the present invention. In addition, the terms "first", "second", "third", etc. are only used to distinguish the description, and cannot be understood as indicating or implying relative importance. In addition, the terms "horizontal", "vertical", etc. do not mean that the components are absolutely horizontal or suspended, but can be slightly tilted. For example, "horizontal" only means that its direction is more horizontal than "vertical", and does not mean that the structure must be completely horizontal, but can be slightly tilted. In the description of the present invention, it is also necessary to explain that, unless otherwise clearly specified and limited, the terms "set", "install", "connect", and "connect" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0059] The following is a further detailed description through specific implementation methods:

[0060] Example:

[0061] This embodiment discloses a 5G wireless network slicing power allocation method based on deep learning.

[0062] like Figure 1 As shown, the 5G wireless network slice power allocation method based on deep learning includes:

[0063] S1: Obtaining an actual value of the allocated power of the corresponding network slice of the 5G wireless network in the previous allocated time period of the time period to be allocated as an initial value of the allocated power;

[0064] S2: Calculate the predicted value of the allocated power of the corresponding network slice in the time period to be allocated based on the initial value of the allocated power;

[0065] S3: To ensure that each network slice is isolated from each other, the initial user power allocation value of each user in the corresponding network slice in the time period to be allocated is calculated in combination with the allocation power prediction value;

[0066] S4: Taking the maximization of power efficiency within the network slice as the goal, the optimal user power allocation value of each user in the corresponding network slice in the time period to be allocated is calculated in combination with the initial user power allocation value of each user;

[0067] S5: Allocate power to each user in the corresponding network slice based on the optimal user power allocation value.

[0068] The present invention calculates the predicted value of the allocated power of each network slice in the current time period to be allocated through the actual value of the allocated power in the previous allocated time period, so that the allocated power of each network slice can be effectively predicted before allocating power, thereby avoiding the problem of isolation destruction between network slices caused by frequent power reconfiguration between network slices (power load changes in one network slice will affect other network slices, and frequent resource allocation in each network slice will bring additional overhead to the 5G system and reduce the performance of the network slice), that is, the present invention can ensure the network quality of each network slice and user, thereby improving the effect of power allocation of 5G wireless network slices.

[0069] The present invention aims to ensure that each network slice is isolated from each other. It calculates the initial user power allocation value of each user in each network slice in combination with the allocation power prediction value, so that while ensuring the isolation between each network slice, it can allocate the minimum power required for each user in the network slice to ensure that the 5G wireless network has enough residual power. Under the premise of ensuring the quality of each network slice and user network, the power efficiency of the 5G wireless network can be improved, thereby further improving the effect of power allocation of 5G wireless network slices.

[0070] The present invention aims to maximize the power efficiency within a network slice, and calculates the optimal user power allocation value for each user in each network slice in combination with the initial user power allocation value of each user, so that power can be redistributed on the basis of the initial user power allocation value, and while meeting the user's communication rate requirements, the power efficiency of the entire network slice can be maximized, that is, the network quality of each network slice and user can be guaranteed and the power efficiency of the 5G wireless network can be improved, thereby further improving the effect of power allocation of 5G wireless network slices.

[0071] In the specific implementation process, Figure 2 As shown, a time axis for representing the passage of time and divided into time steps is constructed, and a prediction window is established on the time axis, which includes several time steps and can move along the time passage direction of the time axis; the starting point and the end point of the prediction window along the time axis are the start and end times of the corresponding time period to be allocated, that is, the length of the prediction window along the time axis is equal to the time length of the time period to be allocated.

[0072] Among them, the time step length of the time axis and the length of the prediction window along the time axis can be dynamically adjusted based on the specific requirements of each network slice.

[0073] When the prediction window moves along the time axis, the step length of each movement is the length of the prediction window along the time axis, that is, the time length of the time period to be allocated.

[0074] The present invention constructs a time axis and establishes a prediction window on the time axis, so that the prediction window can be used as the carrier and boundary of each time period to be allocated, and the length of the time period in the prediction result is determined by the size of the prediction window, thereby better realizing the power allocation prediction of each time period to be allocated.

[0075] In the specific implementation process, the predicted value of the allocated power is calculated through the slice power objective function and the slice power allocation constraint;

[0076] The power mean square error objective function is expressed by the following formula:

[0077]

[0078] Where: represents the predicted value of the allocated power of the mth network slice at time t; T Δ Indicates the length of the time period to be allocated, that is, the length of the prediction window along the time axis (the number of time steps included); r m (t) represents the actual value of the allocated power of the mth network slice at time t; M represents the network slice set;

[0079] The slice power allocation constraint is expressed by the following formula:

[0080]

[0081] Where: represents the predicted value of the allocated power of the mth network slice at time t; Θ represents the total power in the 5G wireless network.

[0082] Prediction of allocated power for network slices The actual value of the allocated power r m (t) The following circumstances exist:

[0083] 1) If Then the corresponding network slice needs to use the remaining power in the 5G wireless network;

[0084] 2) If The allocated power of the corresponding network slice remains unchanged within the prediction window to ensure isolation between network slices.

[0085] In this embodiment, the initial value of the allocated power is used as input data, and the stacked bidirectional long short-term memory algorithm in the recurrent neural network is combined with the slice power allocation constraint to solve the power mean square error objective function to calculate the predicted value of the allocated power of the corresponding network slice in the time period to be allocated.

[0086] It should be noted that the stacked bidirectional long short-term memory algorithm is an existing mature neural network algorithm, and applying it to solve various objective functions is also an existing mature means. The present invention does not make any improvements to the existing stacked bidirectional long short-term memory algorithm, but only applies it to solve the power mean square error objective function designed in the present invention (only the corresponding parameters need to be adjusted according to existing means, and no technical improvements are involved).

[0087] In other preferred embodiments, other existing methods may also be used to solve the power mean square error objective function.

[0088] The present invention calculates the predicted value of the allocated power of each network slice in the current time period to be allocated through the actual value of the allocated power in the previous allocated time period, so that the allocated power of each network slice can be effectively predicted before allocating power, thereby avoiding the problem of isolation destruction between network slices caused by frequent power reconfiguration between network slices (power load changes in one network slice will affect other network slices, and frequent resource allocation in each network slice will bring additional overhead to the 5G system and reduce the performance of the network slice), that is, the present invention can ensure the network quality of each network slice and user, thereby improving the effect of power allocation of 5G wireless network slices.

[0089] In the specific implementation process, the user power initial allocation value is calculated through the user power objective function and the user power allocation constraint;

[0090] The user power objective function is expressed by the following formula:

[0091]

[0092] Where: r m,u (t) represents the initial user power allocation value allocated to user u by the mth network slice at time t; M represents the network slice set;

[0093] The user power allocation constraint is expressed by the following formula:

[0094]

[0095] r m,u (t)≥h m,u (t);

[0096]

[0097] Ω m (D m (t)=0)≥T m ;

[0098] D m (t) = |r m,u (t+τ)-r m,u (t)|;

[0099] Where: Θ m (t) represents the total power allocated to the mth network slice; N represents the total number of users in the mth network slice; U m represents the user set of the mth network slice; h m,u (t) represents the power required by user u in the mth network slice at time t; R m Indicates the communication rate required by the user; B w represents the channel bandwidth; P m,u (t) represents the downlink transmission power between the base station and the user equipment; G m,u (t) represents the time-varying Rayleigh fading channel gain; T m represents the minimum threshold to ensure the isolation of network slice m; N0 represents the variance of additive white Gaussian noise; Ω m Indicates condition counter; D m (t) represents the power difference allocated to user u at time t+τ and time t in the prediction window; τ represents the time step length.

[0100] In this embodiment, the user power objective function is solved by combining the asynchronous advantage executor evaluator deep learning algorithm with the user power allocation constraint to calculate the initial user power allocation value of each user in the corresponding network slice in the time period to be allocated.

[0101] It should be noted that the asynchronous advantage actuator evaluator deep learning algorithm is an existing mature deep reinforcement learning algorithm, and applying it to solve various objective functions is also an existing mature means. The present invention does not make any improvements to the existing asynchronous advantage actuator evaluator deep learning algorithm, but only applies it to solve the user power objective function designed in the present invention (only the corresponding parameters need to be adjusted according to the existing means, and no technical improvement is involved). For the relevant applications of the asynchronous advantage actuator evaluator deep learning algorithm, please refer to the corresponding content disclosed in the document "Adaptive PID Control Design Based on Asynchronous Advantage Actuator Evaluator Learning".

[0102] In other preferred embodiments, other existing methods may also be used to solve the user power objective function.

[0103] The present invention aims to ensure that each network slice is isolated from each other. It calculates the initial user power allocation value of each user in each network slice in combination with the allocation power prediction value, so that while ensuring the isolation between each network slice, it can allocate the minimum power required for each user in the network slice to ensure that the 5G wireless network has enough residual power. Under the premise of ensuring the quality of each network slice and user network, the power efficiency of the 5G wireless network can be improved, thereby further improving the effect of power allocation of 5G wireless network slices.

[0104] In the specific implementation process, the optimal power allocation value of the user is calculated through the power optimal objective function and the power optimal allocation constraint;

[0105] The power optimization objective function is expressed by the following formula:

[0106]

[0107] Where: P m,u (t) represents the downlink transmission power between the base station and the user equipment; η EE Represents the power efficiency index; G m,u (t) represents the time-varying Rayleigh fading channel gain; P c Indicates the power consumption of the circuit;

[0108] The optimal power allocation constraint is expressed by the following formula:

[0109]

[0110] 0≤P m,u (t)≤P max ;

[0111] 0≤h m,u (t)≤Θ m (t)+Θ s (t);

[0112] Where: h m,u (t) represents the power required by user u in the mth network slice at time t; B w represents the channel bandwidth; P m,u (t) represents the downlink transmission power between the base station and the user equipment; G m,u (t) represents the time-varying Rayleigh fading channel gain; N0 represents the additive white Gaussian noise variance; R m Indicates the communication rate required by the user; Θ m (t) represents the total power allocated to the mth network slice; P max represents the maximum downlink transmission power between the base station and the user equipment; Θ s (t) represents the remaining power to be allocated in the 5G wireless network.

[0113] In this embodiment, the power optimal objective function is solved by combining the Dinkelbach iterative algorithm with the power optimal allocation constraint to calculate the optimal user power allocation value for each user in the corresponding network slice in the time period to be allocated.

[0114] It should be noted that the Dinkelbach iteration algorithm is an existing mature optimization algorithm, and applying it to solve various objective functions is also an existing mature means. The present invention does not make any improvements to the existing Dinkelbach iteration algorithm, but only applies it to solve the power optimal objective function designed in the present invention (only the corresponding parameters need to be adjusted according to the existing means, and no technical improvement is involved).

[0115] In other preferred embodiments, other existing methods may also be used to solve the power optimal objective function.

[0116] The present invention adopts the generalized Dinkelbach's algorithm (Generalized Dinkelbach's Algorithm, GDA). For its application, reference may be made to the corresponding contents disclosed in the document "Research on Key Technologies of Resource Allocation and Interference Management in Mobile Communication Heterogeneous Networks".

[0117] The present invention aims to maximize the power efficiency within a network slice, and calculates the optimal user power allocation value for each user in each network slice in combination with the initial user power allocation value of each user, so that power can be redistributed on the basis of the initial user power allocation value, and while meeting the user's communication rate requirements, the power efficiency of the entire network slice can be maximized, that is, the network quality of each network slice and user can be guaranteed and the power efficiency of the 5G wireless network can be improved, thereby further improving the effect of power allocation of 5G wireless network slices.

[0118] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit the technical solution. Those skilled in the art should understand that those modifications or equivalent substitutions of the technical solution of the present invention that do not depart from the purpose and scope of the technical solution should be included in the scope of the claims of the present invention.

Claims

1. A 5G wireless network slice power allocation method based on deep learning, characterized in that: include: S1: Obtaining an actual value of the allocated power of the corresponding network slice of the 5G wireless network in the previous allocated time period of the time period to be allocated as an initial value of the allocated power; S2: Calculate the predicted value of the allocated power of the corresponding network slice in the time period to be allocated based on the initial value of the allocated power; In step S2, the allocation power prediction value is calculated by using the power mean square error objective function and the slice power total constraint; The power mean square error objective function is expressed by the following formula: Where: represents the predicted value of the allocated power of the mth network slice at time t; T Δ Indicates the length of the time period to be allocated; r m (t) represents the actual value of the allocated power of the mth network slice at time t; M represents the network slice set; The slice power allocation constraint is expressed by the following formula: Where: represents the predicted value of the allocated power of the mth network slice at time t; Θ represents the total power in the 5G wireless network; Taking the initial value of the allocated power as input data, the stacked bidirectional long short-term memory algorithm in the recurrent neural network is used in combination with the slice power allocation constraint to solve the power mean square error objective function, so as to calculate the allocated power prediction value of the corresponding network slice in the time period to be allocated; S3: To ensure that each network slice is isolated from each other, the initial user power allocation value of each user in the corresponding network slice in the time period to be allocated is calculated in combination with the allocation power prediction value; S4: Taking the maximization of power efficiency within the network slice as the goal, the optimal user power allocation value of each user in the corresponding network slice in the time period to be allocated is calculated in combination with the initial user power allocation value of each user; S5: Allocate power to each user in the corresponding network slice based on the optimal user power allocation value.

2. The 5G wireless network slice power allocation method based on deep learning according to claim 1, characterized in that: In step S1, a time axis for representing the passage of time and divided into time steps is constructed, and a prediction window including a plurality of time steps and capable of moving along the time passage direction of the time axis is established on the time axis; The starting point and the end point of the prediction window along the time axis are the start and end time of the corresponding time period to be allocated, that is, the length of the prediction window along the time axis is equal to the time length of the time period to be allocated.

3. The 5G wireless network slice power allocation method based on deep learning as claimed in claim 2, characterized in that: When the prediction window moves along the time axis, the step length of each movement is the length of the prediction window along the time axis, that is, the time length of the time period to be allocated.

4. The 5G wireless network slice power allocation method based on deep learning according to claim 1, characterized in that: Prediction of allocated power for network slices The actual value of the allocated power r m (t) The following circumstances exist: 1) If Then the corresponding network slice needs to use the remaining power in the 5G wireless network; 2) If The allocated power of the corresponding network slice remains unchanged during the allocated time period.

5. The 5G wireless network slice power allocation method based on deep learning according to claim 1, characterized in that: In step S3, the user power initial allocation value is calculated by using the user power allocation objective function and the user power total amount constraint; The user power objective function is expressed by the following formula: Where: r m,u (t) represents the initial user power allocation value allocated to user u by the mth network slice at time t; M represents the network slice set; The user power allocation constraint is expressed by the following formula: r m,u (t)≥h m,u (t); Ω m (D m (t)=0)≥T m ; D m (t)=|r m,u (t+τ)-r m,u (t)|; Where: Θ m (t) represents the total power allocated to the mth network slice; N represents the total number of users in the mth network slice; U m represents the user set of the mth network slice; h m,u (t) represents the power required by user u in the mth network slice at time t; R m Indicates the communication rate required by the user; B w represents the channel bandwidth; P m,u (t) represents the downlink transmission power between the base station and the user equipment; G m,u (t) represents the time-varying Rayleigh fading channel gain; T m represents the minimum threshold to ensure the isolation of network slice m; N0 represents the variance of additive Gaussian white noise; Ω m Indicates condition counter; D m (t) represents the power difference allocated to user u at time t+τ and time t in the prediction window; τ represents the time step length.

6. The 5G wireless network slice power allocation method based on deep learning as claimed in claim 5, characterized in that: The user power objective function is solved by combining the asynchronous advantage executor evaluator deep learning algorithm with the user power allocation constraint to calculate the initial user power allocation value for each user in the corresponding network slice in the time period to be allocated.

7. The 5G wireless network slice power allocation method based on deep learning according to claim 1, characterized in that: In step S4, the optimal power allocation value of the user is calculated by using the power optimal objective function and the power optimal allocation constraint; The power optimization objective function is expressed by the following formula: Where: P m,u (t) represents the downlink transmission power between the base station and the user equipment; η EE Represents the power efficiency index; G m,u (t) represents the time-varying Rayleigh fading channel gain; P c Indicates the power consumption of the circuit; The optimal power allocation constraint is expressed by the following formula: 0≤P m,u (t)≤P max ; 0≤h m,u (t)≤Θ m (t)+Θ s (t); Where: h m,u (t) represents the power required by user u in the mth network slice at time t; B w represents the channel bandwidth; P m,u (t) represents the downlink transmission power between the base station and the user equipment; G m,u (t) represents the time-varying Rayleigh fading channel gain; N0 represents the additive white Gaussian noise variance; R m Indicates the communication rate required by the user; Θ m (t) represents the total power allocated to the mth network slice; P max represents the maximum downlink transmission power between the base station and the user equipment; Θ s (t) represents the remaining power to be allocated in the 5G wireless network.

8. The 5G wireless network slice power allocation method based on deep learning according to claim 7, characterized in that: The power optimal objective function is solved by combining the Dinkelbach iterative algorithm with the power optimal allocation constraint to calculate the optimal user power allocation value for each user in the corresponding network slice in the time period to be allocated.

Images