Discontinuous reception parameter control energy-saving method for electric power virtual private network

Abstract

The invention discloses a discontinuous reception (DRX) parameter control energy-saving method for an electric power virtual private network, and designs a discontinuous reception (DRX) parameter control method based on reinforcement learning, which can effectively learn a downlink service mode from a base station, so that a monitoring terminal has the capability of dynamically adjusting DRX parameter configuration along with the change of the service mode, and the service performance of the monitoring terminal is improved. The high-power-consumption operation time of the monitoring terminal is reduced while the time delay requirement is met, and the terminal energy consumption can be actively and effectively reduced on the terminal side.

Description

Technical Field

[0001] The present application relates to the field of communications, and in particular to a discontinuous reception parameter control energy-saving method for a power virtual private network. Background Art

[0002] In the 5G power virtual private network, terminal energy-saving technology is particularly important. With the widespread application of 5G technology, a large number of terminal devices in the power system, such as smart meters, sensors, and distribution controllers, have dramatically increased the number of devices connected to the network. These terminal devices need to run continuously for a long time to ensure real-time monitoring and data transmission of the power system, but since these devices are usually deployed in remote areas or environments where frequent maintenance is difficult, how to extend the service life of the equipment through efficient energy-saving technology has become an urgent problem to be solved.

[0003] The traditional terminal communication mode adopts a continuous reception method, which causes the terminal to maintain high energy consumption when there is no data transmission, and cannot effectively adapt to the complex business scenarios and irregular data traffic in the 5G network. The diverse needs of 5G networks, such as remote fault detection, intelligent scheduling and automatic control, have put forward higher requirements on the latency and power consumption of terminals. In some mission-critical scenarios, excessive latency may cause system failures, while excessive power consumption will shorten the service life of the equipment and increase maintenance costs. Therefore, how to dynamically adjust the energy consumption of terminal equipment in different network scenarios has become one of the technical problems that need to be solved in 5G power virtual private networks.

[0004] To this end, the existing technology has gradually introduced discontinuous reception (DRX) technology, which aims to reduce the energy consumption of the terminal through a periodic sleep and wake-up mechanism. However, the existing DRX mechanism has many shortcomings. First, the DRX mechanism is usually designed based on a fixed cycle and lacks the ability to flexibly respond to the actual network status and traffic demand. In the 5G power virtual private network, business traffic is often irregular, and different application scenarios have different requirements for terminal latency and power consumption. For example, tasks such as remote fault detection and automated control are very sensitive to latency, while applications such as narrowband Internet of Things (NB-IoT) have more stringent requirements on energy consumption. The existing fixed DRX cycle cannot adapt to these complex application scenarios, which may result in excessive latency or excessive power consumption, making it difficult to flexibly balance the energy efficiency and latency of the terminal in different scenarios.

[0005] In addition, traditional energy-saving mechanisms usually do not fully consider the impact of multi-user access and channel changes. In the 5G power virtual private network, multiple terminal devices access the network at the same time, facing a complex wireless channel environment. Since the receiving environment and communication load of the terminal device change over time, the existing fixed DRX cycle may not be able to effectively respond to network changes in some scenarios. For example, when the network load is low, the terminal device still maintains a high activation state, thereby wasting energy; when the network load is high, the fixed sleep mechanism may delay the reception of critical data packets, affecting the real-time and reliability of the task.

[0006] In terms of the technical path to meet the energy-saving needs of terminals, most of the existing energy-saving solutions adopt static strategies, lack the ability to adapt to complex dynamic network environments, and are difficult to meet the increasingly diverse business needs in 5G power virtual private networks. Especially in smart grids, the distribution of equipment is wide, the business scenarios are complex and changeable, and the dynamic and diverse business models require terminal devices to be able to adjust DRX parameters according to actual conditions to better optimize energy efficiency. This makes the existing energy-saving technologies show great limitations in dealing with the balance between power consumption and latency of terminal devices. Summary of the invention

[0007] In view of the above-mentioned deficiencies in the prior art, the present invention provides a parameter control energy-saving method for a discontinuous reception system to solve the problem of how to effectively save energy.

[0008] In order to achieve the above-mentioned invention object, the technical solution adopted by the present invention is: a discontinuous reception parameter control energy-saving method for a power virtual private network, comprising:

[0009] S1. Initialize the recurrent neural network, the current neural network, the target neural network and the experience playback buffer through the power virtual private network terminal;

[0010] S2. Initialize and configure basic parameters and DRX parameters of the power virtual private network terminal;

[0011] S3, dynamically updating the discontinuous reception parameters of the power virtual private network based on the initialized recurrent neural network, the current neural network, the target neural network and the experience playback buffer, the basic parameters of the power virtual private network terminal and the DRX parameters;

[0012] S4. Use the updated discontinuous reception parameters to perform energy-saving control of the power virtual private network.

[0013] Further: S2 includes:

[0014] S21, initializing the power virtual private network terminal's demand for low power consumption λ and the terminal's acceptable maximum average delay T max ;

[0015] S22, send the DRX initial parameter DRX (T) to the monitoring terminal device through the base station of the power virtual private network. c ,T on ,T so ) and configure through the terminal;

[0016] Among them, T c To monitor the length of each DRX cycle after the terminal starts the DRX mode, T on To activate the timer, T so is the DRX start offset.

[0017] Further: S3 includes:

[0018] S31, obtaining the current state, and selecting the action configuration DRX parameters of the current neural network according to the ε-greedy strategy;

[0019] S32, the equipment to be monitored in the power virtual private network uploads information to the base station, the base station aggregates the data and sends it to the monitoring terminal in the form of data packets, and the monitoring terminal records important data within the decision interval;

[0020] S33, based on the important data in the decision interval, the feedback value in the previous decision cycle is calculated through the monitoring terminal, and stored in the experience playback buffer in combination with the current state, action and state of the next cycle;

[0021] S34, sampling a batch of data from the playback buffer, and calculating a target value using a target neural network;

[0022] S35, updating the parameters of the current neural network according to the target value;

[0023] S36. Repeat steps S31-S35 until the number of cycles reaches a preset number, completing the dynamic update of the discontinuous reception parameters of the power virtual private network.

[0024] Further: S32 includes:

[0025] S321, the equipment to be monitored in the power virtual private network uploads information to the base station, and the base station aggregates the data and sends it to the monitoring terminal in the form of a data packet;

[0026] S322, wait for the downlink data packet from the base station in the power virtual private network to arrive at the terminal through the unit decision window of the monitoring terminal, and record each decision interval time T d The number of complete DRX cycles in N c , and the time difference between the i-th data packet sent from the base station to the terminal in the DRX sleep state and the terminal state s when the i-th data packet arrives at the terminali As important data within the decision interval.

[0027] Further: S33 includes:

[0028] S331, calculating the feedback value r in the previous decision cycle through the monitoring terminal according to the important data in the decision interval;

[0029] S332, obtain the state s of the next cycle t+1 , and the current state s t , action a, feedback value r and the state of the next cycle are packaged and stored in the experience replay buffer.

[0030] Further: In S331, the expression of the feedback value r in the previous decision cycle is:

[0031] r = σf ED (λ,T max )+(1-σ)f E ' D (λ,T max )

[0032]

[0033] Among them, f ED (λ,T max ) is an indicator for measuring the energy efficiency delay matching of the monitoring terminal, α is the energy efficiency measurement indicator of the current decision cycle, T offset represents the length of the first DRX sleep phase in the decision interval, β is the average transmission delay of the current decision cycle, σ is the trust parameter, and f′ ED (λ,T max ) is the feedback value of the next decision cycle predicted by the recurrent neural network, α′ is the energy efficiency measurement index of the next cycle currently predicted, β′ is the average transmission delay of the next decision cycle predicted, and s i ∈{s0,s1,s2,s3} represents the state of the terminal when the i-th data packet arrives, s0 indicates that the terminal is receiving and processing downlink data, s1 indicates that the Inactivity timer is working, s2 indicates the active mode under DRX, and s3 indicates the sleep mode under DRX.

[0034] Further: In S34, the expression for calculating the target value y is:

[0035]

[0036] in, is the target neural network, r i Represents the feedback value of the i-th decision cycle.

[0037] The beneficial effects of the present invention are:

[0038] 1. The DRX parameter control energy-saving method for the power virtual private network provided by the present invention can effectively learn the downlink service mode of the UE in the power private network, and provide measurement indicators that can optimize energy efficiency and downlink transmission delay, which can reduce the high power consumption working time of the terminal and effectively reduce the terminal energy consumption of the power virtual private network.

[0039] 2. The present invention defines a measurement criterion that can reflect terminal energy efficiency and transmission delay. ED (λ,T max ), machine learning is based on maximizing the feedback value, which can meet the simulation of different requirements of machine learning for different types of services on latency and power consumption in actual service scenarios;

[0040] 3. The present invention is based on a deep Q network and uses a neural network to approximate the Q value function, avoiding the need to store the Q value of each state-action pair separately. It is more suitable for high-dimensional scenarios with complex data input in power virtual private networks, avoiding the explosion of state space, which makes the calculation and storage costs of power terminals become unaffordable;

[0041] 4. The energy efficiency and delay measurement criteria of the next decision window service flow based on the recurrent neural network in the present invention E ' D (λ,T max ), and combined with the DQN prediction based on the previous decision window f ED (λ,T max ) to achieve accurate prediction of the feedback value of future business traffic. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A schematic diagram of an embodiment of a method for controlling energy saving using discontinuous reception parameters in a power virtual private network.

[0043] Figure 2 Schematic diagram of the DRX mechanism workflow. DETAILED DESCRIPTION

[0044] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0045] Embodiment 1:

[0046] like Figure 1As shown, in this embodiment, a discontinuous reception parameter control energy saving method for a power virtual private network is provided, characterized in that it includes:

[0047] S1. Initialize the recurrent neural network, the current neural network, the target neural network and the experience playback buffer through the power virtual private network terminal;

[0048] S2. Initialize and configure basic parameters and DRX parameters of the power virtual private network terminal;

[0049] S3, dynamically updating the discontinuous reception parameters of the power virtual private network based on the initialized recurrent neural network, the current neural network, the target neural network and the experience playback buffer, the basic parameters of the power virtual private network terminal and the DRX parameters;

[0050] S4. Use the updated discontinuous reception parameters to perform energy-saving control of the power virtual private network.

[0051] Specifically, in S1, the power virtual private network terminal is initialized using random network parameters, which are randomly generated by the terminal or sent by the base station;

[0052] Specifically, S2 includes:

[0053] S21, initializing the power virtual private network terminal's demand for low power consumption λ and the terminal's acceptable maximum average delay T max ;

[0054] λ is used to measure the demand for low power consumption of this type of monitoring terminal and balance the terminal's preference between power consumption and latency: when the λ value is large, the monitoring terminal prefers high energy efficiency and can meet the requirements of longer standby and working time; when the λ value is small, the monitoring terminal prefers low latency and timely grasps the status of each device in the power grid;

[0055] When the average delay of the terminal exceeds T max When the delay is greater than 0.05, it means that the monitoring terminal of this type cannot meet the minimum delay requirement of the service, and the delay part of the measurement index has a negative impact on the overall index;

[0056] S22, send the DRX initial parameter DRX (T) to the monitoring terminal device through the base station of the power virtual private network. c ,T on ,T so ) and configure through the terminal;

[0057] Among them, T c To monitor the length of each DRX cycle after the terminal starts the DRX mode, T on The activation time timer is used to define the time during which the monitoring terminal is in the activation state in each DRX cycle.so DRX start offset, used to define the starting position of the activation time in the DRX cycle.

[0058] The DRX mechanism workflow is as follows Figure 2 shown.

[0059] Specifically, S3 includes:

[0060] S31, obtaining the current state, and selecting the action configuration DRX parameters of the current neural network according to the ε-greedy strategy;

[0061] Among them, the current state s i The running time of the DRX cycle in the current decision window is used to indicate the starting point and the ending point of each cycle, as well as the arrival time of each downlink service data packet and the working state of the terminal when each downlink data packet arrives at the terminal; the data stored by the terminal in the previous decision window is the default value of the terminal when it enters the cycle for the first time;

[0062] The DRX parameters are configured by selecting the action with the maximum value according to the Q network with a probability of 1-ε, and the DRX parameters are configured by randomly selecting the action with a probability of ε (T c ,T on ,T so ), thereby reducing the high power consumption working time of the monitoring terminal when receiving downlink service traffic from the base station, thereby achieving the purpose of active energy saving on the terminal side;

[0063] S32, the equipment to be monitored in the power virtual private network uploads information to the base station, the base station aggregates the data and sends it to the monitoring terminal in the form of data packets, and the monitoring terminal records important data within the decision interval;

[0064] S32 includes:

[0065] S321, the equipment to be monitored in the power virtual private network uploads information to the base station, and the base station aggregates the data and sends it to the monitoring terminal in the form of a data packet;

[0066] S322, wait for the downlink data packet from the base station in the power virtual private network to arrive at the terminal through the unit decision window of the monitoring terminal, and record each decision interval time T d The number of complete DRX cycles in N c , and the time difference between the i-th data packet sent from the base station to the terminal in the DRX sleep state and the terminal state s when the i-th data packet arrives at the terminal i As important data within the decision interval;

[0067] S33, based on the important data in the decision interval, the feedback value in the previous decision cycle is calculated through the monitoring terminal, and stored in the experience playback buffer in combination with the current state, action and state of the next cycle;

[0068] S33 includes:

[0069] S331, calculating the feedback value r in the previous decision cycle through the monitoring terminal according to the important data in the decision interval;

[0070] In S331, the expression of the feedback value r in the previous decision cycle is:

[0071] r = σf ED (λ,T max )+(1-σ)f E ' D (λ,T max )

[0072]

[0073] Among them, f ED (λ,T max ) is an indicator for measuring the energy efficiency-delay matching of the monitoring terminal. This indicator can comprehensively monitor the comprehensive requirements of the monitoring terminal for energy efficiency and delay according to the service type. α is the energy efficiency measurement indicator of the current decision cycle, which indicates the ratio of the time when the monitoring terminal enters the DRX sleep state during the decision interval to the total time of the decision cycle. T offset =0.1*L represents the length of the first DRX sleep phase in the decision interval, in milliseconds. This length is related to the frame number when the monitoring terminal enters the DRX mode. Subframe number T c and T so Related to, where L satisfies The smallest positive integer, β is the average transmission delay of the current decision cycle, σ is the trust parameter, the larger the σ is, the more the monitoring terminal trusts the parameters of the downlink service traffic from the base station actually measured by the terminal in the previous decision cycle, and the smaller the σ is, the more it trusts the downlink service traffic characteristic parameters in the next decision cycle predicted by the recurrent neural network trained by the terminal, f E ' D (λ,T max ) is the feedback value of the next decision cycle predicted by the recurrent neural network, α' is the energy efficiency measurement index of the current next cycle, β' is the average transmission delay of the next decision cycle, and is predicted by the recurrent neural network D, s i∈{s0,s1,s2,s3} represents the state of the terminal when the i-th data packet arrives, s0 indicates that the terminal is receiving and processing downlink data, s1 indicates that the Inactivity timer is working, s2 indicates the active mode under DRX, and s3 indicates the sleep mode under DRX.

[0074] S332, obtain the state s of the next cycle t+1 , and the current state s t , action a, feedback value r and the state of the next cycle are packaged and stored in the experience playback buffer;

[0075] S34, sampling a batch of data from the playback buffer, and calculating a target value using a target neural network;

[0076] In S34, the expression for calculating the target value y is:

[0077]

[0078] in, is the target neural network, r i represents the feedback value of the i-th decision cycle;

[0079] In S34, if the playback buffer is full when sampling is performed, a batch of old data is randomly discarded;

[0080] S35, updating the parameters of the current neural network according to the target value;

[0081] S36. Repeat steps S31-S35 until the number of cycles reaches a preset number, completing the dynamic update of the discontinuous reception parameters of the power virtual private network.

[0082] Embodiment 2:

[0083] This embodiment provides another example of a method for dynamically controlling energy saving of discontinuous reception parameters of a power virtual private network, which can perform dynamic control and repeatedly adjust discontinuous reception parameters of a power virtual private network, including:

[0084] S1. Initialize the recurrent neural network D, current neural network Q, and target neural network through the power virtual private network terminal and experience replay buffer;

[0085] S2. Initialize and configure basic parameters and DRX parameters of the power virtual private network terminal;

[0086] S3, dynamically updating the discontinuous reception parameters of the power virtual private network based on the initialized recurrent neural network, the current neural network, the target neural network and the experience playback buffer, the basic parameters of the power virtual private network terminal and the DRX parameters;

[0087] S4. Use the updated continuous reception parameters to perform energy-saving control of the power virtual private network.

[0088] Specifically, S2 includes:

[0089] S21, initializing the power virtual private network terminal's demand for low power consumption λ and the terminal's acceptable maximum average delay T max ;

[0090] S22, send the DRX initial parameter DRX (T) to the monitoring terminal device through the base station of the power virtual private network. c ,T on ,T so ) and configure through the terminal;

[0091] Among them, T c To monitor the length of each DRX cycle after the terminal starts the DRX mode, T on To activate the timer, T so is the DRX start offset.

[0092] In particular, S3 includes:

[0093] S31, obtain the current state, use the ε-greedy strategy for the current neural network, and select an action to configure DRX parameters;

[0094] S32, the equipment to be monitored in the power virtual private network uploads information to the base station, the base station aggregates the data and sends it to the monitoring terminal in the form of data packets, and records important data within the decision interval;

[0095] S32 includes:

[0096] S321, the equipment to be monitored in the power virtual private network uploads information to the base station, and the base station aggregates the data and sends it to the monitoring terminal in the form of a data packet;

[0097] S322, wait for the downlink data packet from the base station in the power virtual private network to arrive at the terminal through the unit decision window of the monitoring terminal, and record each decision interval time T d The number of complete DRX cycles in N c And the time difference between the i-th data packet sent from the base station to the terminal in the DRX sleep state As important data within the decision interval;

[0098] S33, based on the important data in the decision interval, the feedback value in the previous decision cycle is calculated through the monitoring terminal, and stored in the experience playback buffer in combination with the current state, action and state of the next cycle;

[0099] S33 includes:

[0100] S331, calculating the feedback value r in the previous decision cycle through the monitoring terminal according to the important data in the decision interval;

[0101] In S331, the expression of the feedback value r in the previous decision cycle is:

[0102] r = σf ED (λ,T max )+(1-σ)f E ' D (λ,T max )

[0103]

[0104] Among them, f ED (λ,T max ) is an indicator for measuring the energy efficiency-delay matching of the monitoring terminal. This indicator can comprehensively monitor the comprehensive requirements of the monitoring terminal for energy efficiency and delay according to the service type. α is the energy efficiency measurement indicator of the current decision cycle, which indicates the ratio of the time when the monitoring terminal enters the DRX sleep state during the decision interval to the total time of the decision cycle. T offset =0.1*L represents the length of the first DRX sleep phase in the decision interval, in milliseconds. This length is related to the frame number when the monitoring terminal enters the DRX mode. Subframe number T c and T so Related to, where L satisfies The smallest positive integer, β is the average transmission delay of the current decision cycle, σ is the trust parameter, the larger the σ is, the more the monitoring terminal trusts the parameters of the downlink service traffic from the base station actually measured by the terminal in the previous decision cycle, and the smaller the σ is, the more it trusts the downlink service traffic characteristic parameters in the next decision cycle predicted by the recurrent neural network trained by the terminal, f E ' D (λ,T max ) is the feedback value of the next decision cycle predicted by the recurrent neural network, α' is the energy efficiency measurement index of the current next cycle, β' is the average transmission delay of the next decision cycle, and is predicted by the recurrent neural network D, s i ∈{s0,s1,s2,s3} represents the state of the terminal when the i-th data packet arrives, s0 indicates that the terminal is receiving and processing downlink data, s1 indicates that the Inactivity timer is working, s2 indicates the active mode under DRX, and s3 indicates the sleep mode under DRX.

[0105] S332, obtain the state s of the next cycle t+1 , and the current state s t, action a, feedback value r and the state of the next cycle are packaged and stored in the experience playback buffer;

[0106] S34, sampling a batch of data from the playback buffer, and calculating a target value using a target neural network;

[0107] In S34, the expression for calculating the target value y is:

[0108]

[0109] in, is the target neural network, r i represents the feedback value of the i-th decision cycle;

[0110] In S34, if the playback buffer is full when sampling is performed, a batch of old data is randomly discarded;

[0111] S35, updating the parameters of the current neural network according to the target value;

[0112] S36, judging whether the number of cycles reaches the preset number:

[0113] If yes, then go to S37;

[0114] If not, return to S31;

[0115] S37, reset the target network parameters to the current network parameters Retrain the recurrent neural network D according to the stored data, adjust the trust parameter σ and return to S31.

[0116] In this embodiment, by continuously modifying the DRX parameters, the downlink service mode of the UE in the power private network can be effectively learned, and the energy efficiency and downlink transmission delay measurement index f can be optimized. ED (λ,T max )’s DRX parameter configuration scheme.

[0117] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A discontinuous reception parameter control energy saving method for a power virtual private network, characterized in that: include: S1. Initialize the recurrent neural network, the current neural network, the target neural network and the experience playback buffer through the power virtual private network terminal; S2. Initialize and configure basic parameters and DRX parameters of the power virtual private network terminal; S3, dynamically updating the discontinuous reception parameters of the power virtual private network based on the initialized recurrent neural network, the current neural network, the target neural network and the experience playback buffer, the basic parameters of the power virtual private network terminal and the DRX parameters; S4. Use the updated discontinuous reception parameters to perform energy-saving control of the power virtual private network.

2. The non-continuous reception parameter control energy saving method of the power virtual private network according to claim 1 is characterized in that S2 include: S21, initializing the power virtual private network terminal's demand for low power consumption λ and the terminal's acceptable maximum average delay T max ; S22, send the DRX initial parameter DRX (T) to the monitoring terminal device through the base station of the power virtual private network. c ,T on ,T so ) and configure through the terminal; Among them, T c To monitor the length of each DRX cycle after the terminal starts the DRX mode, T on To activate the timer, T so is the DRX start offset.

3. The energy-saving method for discontinuous reception parameter control of a power virtual private network according to claim 2, characterized in that S3 include: S31, obtaining the current state, and selecting the action configuration DRX parameters of the current neural network according to the ε-greedy strategy; S32, the equipment to be monitored in the power virtual private network uploads information to the base station, the base station aggregates the data and sends it to the monitoring terminal in the form of data packets, and the monitoring terminal records important data within the decision interval; S33, based on the important data in the decision interval, the feedback value in the previous decision cycle is calculated through the monitoring terminal, and the feedback value is stored in the experience playback buffer in combination with the current state, action and state of the next cycle; S34, sampling a batch of data from the playback buffer, and calculating a target value using a target neural network; S35, updating the parameters of the current neural network according to the target value; S36. Repeat steps S31-S35 until the number of cycles reaches a preset number, completing the dynamic update of the discontinuous reception parameters of the power virtual private network.

4. The energy-saving method for discontinuous reception parameter control of a power virtual private network according to claim 3 is characterized in that: S32 includes: S321, the equipment to be monitored in the power virtual private network uploads information to the base station, and the base station aggregates the data and sends it to the monitoring terminal in the form of a data packet; S322, by monitoring the unit decision window of the terminal, waiting for the downlink data packet from the base station in the power virtual private network to arrive at the terminal, and recording the number N of complete DRX cycles in each decision interval c , and the time difference between the i-th data packet sent from the base station to the terminal in the DRX sleep state and the terminal state s when the i-th data packet arrives at the terminal i As important data within the decision interval.

5. The energy-saving method for discontinuous reception parameter control of a power virtual private network according to claim 4, characterized in that: S33 includes: S331, calculating the feedback value r in the previous decision cycle through the monitoring terminal according to the important data in the decision interval; S332, obtain the state s of the next cycle t+1 , and the current state s t , action a, feedback value r and the state of the next cycle are packaged and stored in the experience replay buffer.

6. The energy-saving method for discontinuous reception parameter control of a power virtual private network according to claim 5, characterized in that: In S331, the expression of the feedback value r in the previous decision cycle is: r=σf ED (λ,T max )+(1-σ)f′ ED (λ,T max ) Among them, f ED (λ,T max ) is an indicator for measuring the energy efficiency delay matching of the monitoring terminal, α is the energy efficiency measurement indicator of the current decision cycle, T offset represents the length of the first DRX sleep phase in the decision interval, β is the average transmission delay of the current decision cycle, σ is the trust parameter, and f′ ED (λ,T max ) is the feedback value of the next decision cycle predicted by the recurrent neural network, α′ is the energy efficiency measurement index of the next cycle currently predicted, β′ is the average transmission delay of the next decision cycle predicted, and s i ∈{s0,s1,s2,s3} represents the state of the terminal when the i-th data packet arrives, s0 indicates that the terminal is receiving and processing downlink data, s1 indicates that the Inactivity timer is working, s2 indicates the active mode under DRX, and s3 indicates the sleep mode under DRX.

7. The energy-saving method for discontinuous reception parameter control of a power virtual private network according to claim 1, characterized in that: In S34, the expression for calculating the target value y is: in, is the target neural network, r i Represents the feedback value of the i-th decision cycle.

Images