Power distribution network cooperative interaction method based on integration of communication, inductance and calculation
By adopting synesthesia computing integrated collaborative interaction method in the distribution network, channel selection and computing resource allocation are optimized, and the problems of high latency and imbalance in terminal data upload are solved, thereby achieving low latency and efficient data Transmission and resource allocation.
Patent Information
- Application Number
- CN202411286179.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-13
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2044-09-13
AI Technical Summary
In the process of uploading terminal data to edge servers, multiple terminals select the same channel, resulting in poor learning performance, slow convergence speed, unreasonable resource allocation, high data transmission queue delay, and failure to balance delay and The relationship between security.
A synergistic interaction method based on synesthesia computing is proposed to optimize the edge data transmission process through channel selection principle and computing resource allocation principle. The channel selection principle is to use the signal-to-interference noise ratio according to the edge-end signal-to-interference noise ratio when the channel is connected to the terminal. The terminal access channel is selected in descending order. The principle of computing resource allocation is to use the difference between the total computing resources and data to calculate the allocation resources instead of block authentication, build a joint optimization problem model and fight the DQN network and Markov decision-making process. Perform the solution.
Realize edge-end collaborative optimization of channel selection and efficient transmission of low-latency and low-power data, decoupling data calculation and block authentication calculation, balance data processing delay and blockchain consensus security, and improve the overall system Performance and efficiency.
Smart Images

Figure CN119094530A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of wireless communications, and in particular relates to a distribution network collaborative interaction method based on integrated telepathy and computing. Background Art
[0002] With the widespread application of technologies such as 5G (fifth-generation mobile communication technology), cognitive IoT (cognitive Internet of Things), and Web3.0 (third-generation Internet) in distribution networks, revolutionary changes have taken place in their operation modes and service innovations, which also put forward new requirements for the integration and collaborative interaction of multi-dimensional heterogeneous resources such as perception, transmission, and computing in distribution networks. Through the integrated resource security allocation technology of inter-sensing computing, distribution networks can achieve rapid transmission and processing of large-capacity data, monitor real-time operating status data of electrical equipment, ensure data transaction security, and support low-latency, high-security data transmission and interaction in distribution networks, which is a major boost to the digitalization and intelligence of power grids.
[0003] However, the prior art ignores the situation where multiple terminals select the same channel during the process of uploading terminal data to the edge server, and cannot effectively mine the state information of the access confrontation from the edge server feedback, resulting in poor learning performance, slow convergence, unreasonable resource allocation, and high data transmission queuing delay. The prior art also ignores the coupling of data calculation delay and block authentication calculation delay involved in computing resource allocation, and does not consider the relationship between dynamic balance delay and security, resulting in the inability to ensure a balanced compromise between delay and security. Summary of the invention
[0004] In view of this, the present invention aims to propose a distribution network collaborative interaction method based on integrated synergy and computing, so as to solve the above-mentioned problems existing in the prior art.
[0005] In order to solve the above technical problems, the present invention provides the following technical solutions:
[0006] A distribution network collaborative interaction method based on integrated synergy and computing includes the following steps:
[0007] Get the end-side data queue, which is the total amount of data that all terminals are ready to transmit to the edge server;
[0008] Transmit data through channel-connected terminals;
[0009] Generate a data calculation queue for the edge server, where the data calculation queue is the amount of data to be processed transmitted from the terminal to the edge server;
[0010] Obtain various delay information in the edge-end collaboration process, where each delay information includes at least transmission-related delay information generated by the total data volume of the end-side data queue during the data transmission process and calculation-related delay information generated by the amount of data to be processed in the data calculation queue after the calculation resources are allocated;
[0011] Taking the set latency requirement as the optimization goal, the edge data transmission process and computing resource allocation process are continuously optimized based on the channel selection principle and computing resource allocation principle, so as to realize the coordinated interaction of the distribution network based on the integration of tele-sensing and computing;
[0012] The channel selection principle is that when a terminal access confrontation occurs on a channel, the terminal access channel is selected in descending order of the edge signal-to-interference-noise ratio.
[0013] The computing resource allocation principle is to use the difference between the total computing resources and the data computing allocation resources to replace the resources allocated for block authentication computing.
[0014] Furthermore, taking the set delay requirement as the optimization goal, the edge data transmission process and the computing resource allocation process are continuously optimized based on the channel selection principle and the computing resource allocation principle, which are constructed as the first joint optimization problem model for solution. The first joint optimization problem model is as follows:
[0015]
[0016] In the formula, For Terminal Channel sensing time allocation ratio, For the Time Slot Terminal The channel selection indicator variable, For the Time Slot Terminal Through the channel The transmission power of the transmitted data, The edge server is used to process data from the terminal. The computing resources allocated to the data, is the set of time slots, For terminal collection, The queuing delay of the end-side queue transmission. Calculate the queue delay weight for the edge side. For the The calculation queue delay of the data calculation queue, Calculate the delay weight for blockchain authentication, Calculate latency for blockchain authentication; is the time allocation ratio constraint, For terminal collection; is the channel availability constraint, For the Time Slot Terminal The channel availability indicator variable; , is the channel selection constraint, which means that each terminal can only select at most one channel for transmission in each time slot, and each channel can be used at most Terminal multiplexing, is the total number of channels, For the channels, is a channel set; is the power allocation constraint, which means that the terminal transmission power is discretized from arrive of levels, among which The level is , and are the minimum and maximum values of the transmission power respectively; Assign constraints to computing resources, indicating the processing The total computing resources of the terminal data must be less than or equal to the available computing resources of the edge server ; Represents the terminal energy consumption constraint, that is, the terminal exist The sum of the sensing and transmission energy consumption of each time slot must be less than or equal to the maximum energy consumption constraint. ; represents the channel selection reliability constraint, i.e., the terminal Select Channel The SINR must be greater than or equal to the threshold ; Indicates The amount of data in the end-side data queue per time slot and the amount of data in the data calculation queue The average rate is stable.
[0017] Furthermore, based on Lyapunov optimization, the first joint optimization problem model is transformed into the second joint optimization problem model, as follows:
[0018]
[0019] In the formula, Represents the weights of data transmission delay, data calculation delay and blockchain authentication calculation delay, The amount of data collected for the client-side data queue. The amount of data transmitted by the client-side data queue. The amount of data processed by the edge server, For the period up to time slot, the terminal The amount of energy consumption that exceeds the expected amount, To sense energy consumption, The transmission energy consumption.
[0020] Furthermore, the second joint optimization problem model is transformed into the first sub-optimization problem model for solving, so as to optimize the edge data transmission process. The first sub-optimization problem model is specifically as follows:
[0021]
[0022] In the formula, Represents the optimization objective of the first sub-optimization problem model.
[0023] Furthermore, the first sub-optimization problem model is modeled as a Markov decision process for solution. The Markov decision process includes the following elements:
[0024] State Space: , For Terminal No. The status of each time slot;
[0025] Action Space: , For Terminal No. The action taken in each time slot;
[0026] award: .
[0027] Furthermore, the first joint optimization method based on the adversarial DQN network is used to solve the Erkov decision process. The first joint optimization method includes a main network and a target network, and the corresponding parameter vectors are and ;
[0028] The main network is used to utilize the status of the current time slot and main network parameters Generate strategy , according to the strategy adopted Algorithms determine actions and strategies Parameterized as ,Right now , used to indicate that when the system is in state When the parameter Take action under the strategy When terminal access confrontation occurs, the channel selection principle is used to solve the confrontation problem;
[0029] The target network is used for rewards based on environmental feedback and combined with the terminal Current Status and the next state Calculate the TD error and loss function, and the main network adjusts the strategy using the gradient descent method based on the loss function.
[0030] Furthermore, the second joint optimization problem model is transformed into a second sub-optimization problem model for solving, so as to optimize the computing resource allocation process. The second sub-optimization problem model is specifically as follows:
[0031]
[0032] In the formula, For Terminal The computational complexity of the data is The time slot is long.
[0033] Furthermore, based on the result of the first sub-optimization problem model, the second sub-optimization problem model is transformed into a convex optimization problem, and the optimality condition for computing resource allocation is obtained based on the dual decomposition and the Carlo-Kuhn-Tucker condition, as follows:
[0034] use Represents the computing resources used for blockchain authentication calculations. The second sub-optimization problem model is transformed into a convex optimization problem as follows:
[0035]
[0036] In the formula, represents the optimization objective of a convex optimization problem;
[0037] structure The Lagrangian function is:
[0038]
[0039] In the formula, and Respectively and The Lagrange multiplier of ;
[0040] Using the dual decomposition method, Translates to:
[0041]
[0042] According to the Carlo-Kuhn-Tucker condition, the necessary condition for the Lagrangian function to take an extreme value is The first-order partial derivative of is equal to zero, expressed as:
[0043]
[0044] Solve the above formula to get the optimal computing resource allocation decision.
[0045] Furthermore, the optimal computing resource allocation decision solution process is obtained, which specifically includes the following steps:
[0046] Initialize the number of iterations , iterative step length and And the convergence accuracy ;
[0047] Perform iterative condition judgment, if or Update The optimal value of for
[0048]
[0049] Update Lagrange multipliers and for
[0050]
[0051]
[0052] make Repeat the above steps until the convergence accuracy is met. , the income This is the result of the optimal allocation of computing resources, that is, .
[0053] Furthermore, the delay information includes at least:
[0054] The calculation queue delay of the data calculation queue:
[0055]
[0056] Blockchain authentication calculation delay:
[0057] .
[0058] In summary, the present invention provides a distribution network collaborative interaction method based on the integration of communication and calculation, the method comprising obtaining a data queue on the end side; connecting the terminal to transmit data through a channel; generating a data calculation queue of an edge server; obtaining various delay information in the edge-end collaborative process; taking the set delay requirement as the optimization goal, the edge data transmission process and the computing resource allocation process are continuously optimized based on the channel selection principle and the computing resource allocation principle, thereby realizing the distribution network collaborative interaction based on the integration of communication and calculation; wherein, the channel selection principle is that when the channel has terminal access confrontation, the terminal access channel is selected in descending order according to the edge signal to noise ratio; the computing resource allocation principle is to replace the resources allocated by the block authentication calculation with the difference between the total computing resources and the data calculation allocation resources. In the present invention, when a terminal access confrontation occurs in a certain channel, the terminal is arranged in descending order based on SINR, and the terminal with a larger SINR is preferentially allowed to access the channel, thereby solving the confrontation and realizing the edge-end collaborative optimization of channel selection and efficient transmission of low-latency and low-power data. At the same time, the difference between the total computing resources and the data calculation allocation resources is used to replace the resources allocated by the block authentication calculation, thereby realizing the decoupling between data calculation and block authentication calculation. BRIEF DESCRIPTION OF THE DRAWINGS
[0059] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0060] Figure 1 A flow chart of a distribution network collaborative interaction method based on integrated synergy and computing provided by an embodiment of the present invention;
[0061] Figure 2 A flow chart of a distribution network collaborative interaction method based on integrated synergy and computing provided in an embodiment of the present invention;
[0062] Figure 3 A diagram of a distribution network collaborative interaction system based on integrated synergy and computing provided by an embodiment of the present invention;
[0063] Figure 4 A weighted delay performance diagram provided by an embodiment of the present invention;
[0064] Figure 5 A schematic diagram of energy consumption performance of a 5G cognitive IoT terminal device provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0065] In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, 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 embodiments described below are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0066] See also Figure 1 The embodiment of the present invention provides a distribution network collaborative interaction method based on integrated synergy and computing, comprising the following steps:
[0067] S1: Obtain the end-side data queue, which is the total amount of data that all terminals are ready to transmit to the edge server;
[0068] S2: Transmit data through the channel connection terminal;
[0069] S3: Generate a data computing queue for the edge server, where the data computing queue is the amount of data to be processed transmitted from the terminal to the edge server;
[0070] S4: Acquire various delay information in the edge-end collaboration process, where each delay information includes at least transmission-related delay information generated by the total data volume of the end-side data queue during the data transmission process and computing-related delay information generated by the to-be-processed data volume of the data computing queue after computing resources are allocated;
[0071] S5: Taking the set latency requirement as the optimization goal, the edge data transmission process and computing resource allocation process are continuously optimized based on the channel selection principle and computing resource allocation principle, so as to realize the coordinated interaction of the distribution network based on the integration of telepathy and computing;
[0072] The channel selection principle is that when a terminal access confrontation occurs on a channel, the terminal access channel is selected in descending order of the edge signal-to-interference-noise ratio.
[0073] The computing resource allocation principle is to use the difference between the total computing resources and the data computing allocation resources to replace the resources allocated for block authentication computing.
[0074] It should be noted that edge computing is an information transmission and processing technology derived from cloud computing, and is an extension of cloud computing on the user side. Edge computing is to sink data computing and data storage partially or completely to the edge of the network close to the user, thereby reducing data transmission delay, shortening service response time, reducing network bandwidth, and improving computing efficiency. Distributed edge computing meets the requirements of a large number of energy terminals in the energy Internet for response speed and Q0S service quality. Blockchain technology based on consensus algorithm is essentially a distributed accounting technology. Commonly used consensus algorithms include proof of work PoW (Proof of Work) and Byzantine fault tolerance PBFT (Practical Byzantine fault Tolerance) algorithm. Blockchain technology and edge computing have a wide range of applications in the communication architecture and application scenarios of the energy Internet. Integrating blockchain and edge computing into a system can achieve reliable access and control of the network, and storage and computing are distributed at the edge, thereby providing large-scale network servers, data storage, and validity computing close to the end in a secure manner.
[0075] In the distribution network collaborative interaction system based on blockchain technology and edge computing technology, in order to solve the problems of terminal access confrontation and unreasonable allocation of computing resources in the channel, this embodiment proposes a distribution network collaborative interaction method based on channel selection principle and computing resource allocation principle.
[0076] In this method, the channel selection principle is used to solve the problem of terminal access confrontation in the channel during the coordinated interaction of the distribution network. When terminal access confrontation occurs in the channel, the terminal access channel is selected in descending order according to the edge signal to noise ratio. That is, when a terminal access confrontation occurs in a channel, that is, the number of terminals that select the same channel for data transmission exceeds the maximum allowed number of terminals to access the channel, the edge server will collect the edge signal to interference plus noise ratio (SINR) information, and arrange the terminals in descending order based on SINR, giving priority to allowing terminals with larger SINR to access the channel, thereby solving the confrontation.
[0077] The computing resource allocation principle is adopted to solve the problem that the computing resource allocation in the collaborative interaction process of the distribution network involves the coupling of data computing delay and block authentication computing delay, without considering the relationship between dynamic balance delay and security. The difference between the total computing resources and the data computing allocation resources is used to replace the resources allocated for block authentication calculation, thereby realizing the decoupling between data computing and block authentication calculation, and optimizing the entire interactive collaborative process in combination with the delay information in the collaborative interaction process, thereby achieving a balanced compromise between data processing delay and blockchain consensus security.
[0078] The edge server obtains the end-side data queue, and generates a data calculation queue after the end-side data queue starts transmitting data according to the channel selection principle. The execution optimization of the collaborative interaction method proposed in this embodiment can be reflected according to the amount of data in the queue, such as latency.
[0079] This embodiment provides a distribution network collaborative interaction method based on integrated communication and computing. When a terminal access confrontation occurs on a certain channel, the method arranges the terminals in descending order based on SINR, and preferentially allows terminals with larger SINR to access the channel, thereby resolving the confrontation and achieving edge-end collaborative optimization of channel selection and efficient transmission of low-latency and low-power data. At the same time, the difference between the total computing resources and the data computing allocation resources is used to replace the resources allocated by the block authentication calculation, thereby achieving decoupling between data calculation and block authentication calculation.
[0080] In a preferred embodiment of the present invention, the edge data transmission process and the computing resource allocation process are continuously optimized based on the channel selection principle and the computing resource allocation principle, and are constructed as a first joint optimization problem model for solving. The first joint optimization problem model is specifically as follows:
[0081] (1)
[0082] In the formula, For Terminal Channel sensing time allocation ratio, For the Time Slot Terminal The channel selection indicator variable, For the Time Slot Terminal Through the channel The transmission power of the transmitted data, The edge server is used to process data from the terminal. The computing resources allocated to the data, is the set of time slots, For terminal collection, The queuing delay of the end-side queue transmission. Calculate the queue delay weight for the edge side. For the The calculation queue delay of the data calculation queue, Calculate the delay weight for blockchain authentication, Calculate latency for blockchain authentication; is the time allocation ratio constraint, For terminal collection; is the channel availability constraint, For the Time Slot Terminal The channel availability indicator variable; , is the channel selection constraint, which means that each terminal can only select at most one channel for transmission in each time slot, and each channel can be used at most Terminal multiplexing, is the total number of channels, For the channels, is a channel set; is the power allocation constraint, which means that the terminal transmission power is discretized from arrive of levels, among which The level is , and are the minimum and maximum values of the transmission power respectively; Assign constraints to computing resources, indicating the processing The total computing resources of the terminal data must be less than or equal to the available computing resources of the edge server ; Represents the terminal energy consumption constraint, that is, the terminal exist The sum of the sensing and transmission energy consumption of each time slot must be less than or equal to the maximum energy consumption constraint. ; represents the channel selection reliability constraint, i.e., the terminal Select Channel The SINR must be greater than or equal to the threshold ; Indicates The amount of data in the end-side data queue per time slot and the amount of data in the data calculation queue The average rate is stable.
[0083] In the process of uploading terminal data to the edge server, there are situations where multiple terminals choose the same channel, which is prone to channel conflicts and causes serious interference, thereby affecting the data transmission quality and system performance. Existing methods cannot effectively mine the state information of the access confrontation from the edge server feedback, resulting in poor learning performance, slow convergence, unreasonable resource allocation, and high data transmission queuing delay. Therefore, this embodiment constructs a joint optimization problem of multi-dimensional resource allocation for communication and computing. The optimization goal is to minimize the weighted sum of the end-side transmission queue queuing delay, the edge-side computing queue queuing delay, and the block authentication computing delay by jointly optimizing the time allocation ratio, channel selection, power allocation, and computing resource allocation.
[0084] Specifically, considering the low latency and security requirements of distribution network services for data collection, transmission, and processing, a joint optimization problem of multi-dimensional resource allocation of synaesthesia is constructed. The optimization goal is to minimize the weighted sum of the end-side transmission queue queuing delay, the edge-side computing queue queuing delay, and the block authentication computing delay by jointly optimizing the time allocation ratio, channel selection, power allocation, and computing resource allocation. The joint optimization problem is constructed as shown in Equation 1.
[0085] It is necessary to combine the multi-dimensional resource optimization decision of single time slot with The long-term terminal energy consumption constraint of the time slot ensures the decoupling between time slots. To solve this problem, in a preferred embodiment of the present invention, a virtual queue in Lyapunov optimization is introduced to constrain the long-term terminal energy consumption. Transformed into queue stability constraints to achieve decoupling between time slots. The corresponding virtual energy deficit queue is , the queue backlog evolution formula is:
[0086] (2)
[0087] In the formula, Representatives up to time slot, the terminal The amount of energy consumption that exceeds the expected amount is accumulated. The larger the terminal The more energy consumption exceeds expectations, the more difficult it is to meet long-term terminal energy consumption constraints. Otherwise, it means the terminal The less energy consumption than expected.
[0088] definition Represents the connection vector between the data queue and the virtual queue. The Lyapunov function is defined as
[0089] (3)
[0090] Lyapunov drift is defined as The conditional expected change in two consecutive time slots can effectively ensure a low queue backlog by minimizing the Lyapunov drift plus penalty upper bound. Therefore, the optimization problem is converted into
[0091] (4)
[0092] In the formula, Represents the weights of data transmission delay, data calculation delay and blockchain authentication calculation delay, The amount of data collected for the client-side data queue. The amount of data transmitted by the client-side data queue. The amount of data processed by the edge server, To sense energy consumption, The transmission energy consumption.
[0093] In a specific implementation of this embodiment, a distribution network collaborative interaction model based on integrated synergy and calculation is proposed, and the joint optimization problems P1 and P2 are solved according to the model. The model provides a feasible calculation method for each parameter in the above joint optimization problem, so as to facilitate understanding of how the constructed joint optimization problem specifically realizes the distribution network collaborative interaction method in conjunction with other embodiments.
[0094] The model consists of two parts: channel collaborative interaction perception and data transmission model and edge collaborative interaction computing model. These two parts are introduced in detail below.
[0095] (1) Channel collaborative interaction perception and data transmission model
[0096] Assume that there is a 5G cognitive IoT terminals, the collection of which is The perception layer contains channels, whose set is The time slot set is , each time slot is In the Each terminal senses the channel availability and transmits data based on cognitive radio technology. The channel sensing time allocation ratio is ,in For channel sensing, For data transmission. Assume that the energy consumption required to sense a channel is , the time required is , then the terminal The perceived energy consumption is
[0097] (5)
[0098] in, Indicates rounding down. Indicates the number of channels that the terminal can perceive.
[0099] Definition Time Slot Terminal The channel selection indicator variable is , When indicates Time Slot Terminal Select access channel . Define Time Slot Terminal The channel availability indicator variable is . Based on channel cooperative interaction perception, channel availability is obtained, where Indicates channel For Terminal Available, otherwise, Indicates unavailable, only available channels can be selected for data transmission, therefore, .
[0100] terminal The collected data is stored in the local cache, which can be modeled as a data queue on the end. The input of the queue is the amount of collected data. , the queue output is the amount of data transmitted Therefore, the backlog of data queue on the client side evolves to
[0101] (6)
[0102] In the formula, Indicates taking the maximum value.
[0103] Based on Little's Law, the queuing delay of the end-side queue is:
[0104] (7)
[0105] In the formula, To The average data arrival rate up to time slots is calculated as: .
[0106] No. Time slot, terminal Through the channel The transmission data rate is
[0107] (8)
[0108] In the formula, Indicates terminal Select Channel The signal-to-interference-noise ratio is Indicates terminal Through the channel The co-channel interference of the transmitted data, represents the Gaussian white noise power, Indicates Time Slot Terminal Selected Channel bandwidth, Indicates Time Slot Terminal Through the channel The transmission power of the transmitted data, Indicates Time Slot Terminal Selected Channel The channel gain.
[0109] terminal The amount of data transmitted to the edge server through the 5G channel is the minimum value of the queue backlog and the theoretical transmission capacity, expressed as
[0110] (9)
[0111] terminal In the The transmission energy consumption of a time slot is
[0112] (10)
[0113] (2) Edge collaborative interactive computing model
[0114] In each time slot, the edge server optimizes the allocation of computing resources to process data from 5G cognitive IoT terminals and complete the blockchain authentication calculations required for consensus. The edge server is used to process data from the terminal. The computing resources allocated to the data. is the total available computing resources of the edge server, then the computing resources allocated to complete the block authentication calculation are .
[0115] On the edge server there is data queues, corresponding to Terminals. Among them, the terminal The data calculation queue at the edge server is
[0116] (11)
[0117] In the formula, For Terminal In the The amount of data transmitted to the edge server in each time slot, For the The edge server processes the data from the terminal The amount of data is expressed as
[0118] (12)
[0119] In the formula, For Terminal The computational complexity of the data.
[0120] On the edge server, The calculation queuing delay of the data calculation queue is
[0121] (13)
[0122] In the formula, To The average data arrival rate at the edge side up to time slots is calculated as: .
[0123] Consider the consensus of the alliance blockchain based on the Practical Byzantine Fault Tolerance (PBFT) algorithm. In the consensus process, there are master nodes and slave nodes. The master node is responsible for proposing new transactions and blocks, and broadcasting transactions and blocks to other slave nodes, playing the role of interactively coordinating the operations of other nodes. The slave nodes are responsible for receiving transactions and blocks broadcast by the master node and verifying their legitimacy. PBFT includes five steps: request, pre-preparation, preparation, submission and reply.
[0124] Define the computing resources required to complete the consistency certification calculation as , then the blockchain authentication calculation delay is
[0125] (14)
[0126] In the formula, Computing resources used for blockchain authentication.
[0127] The present invention proposes a distribution network collaborative interaction method based on adversarial learning and integrated communication, transmission and computing. This method is divided into two stages. The first stage is a method proposed for terminal access confrontation in the channel, and the second stage is a method proposed for the coupling of data computing delay and block authentication computing delay involved in computing resource allocation.
[0128] In a preferred embodiment of the present invention, for the first stage, a joint optimization method for time allocation, channel allocation and energy control based on a deep adversarial Q-network (DQN) (i.e., the first joint optimization method) is proposed. When terminal access confrontation occurs, confrontation perception is performed based on edge SINR information, and terminals with larger SINR are preferentially allowed to access the channel, thereby solving the terminal confrontation problem. The terminal confrontation perception information is fed back to the adversarial DQN network in the form of rewards. With the assistance of the target network, the adversarial DQN main network realizes efficient learning of the confrontation perception information, effectively improves the adversarial DQN learning rate, and reduces terminal access confrontation conflicts. The specific scheme of this method is as follows:
[0129] (1) Joint optimization method of time allocation, channel allocation and energy control based on adversarial DQN
[0130] The transformed problem Further decomposed into Sub-question, The subproblem is the joint optimization of time allocation, channel allocation and power control.
[0131] The sub-problem is stated as
[0132] (15)
[0133] In a further embodiment of the present invention, The sub-problems are modeled as Markov decision processes (MDPs), which are described as follows.
[0134] State Space: Time Slots Middle Terminal The state space of Required information, including terminal The client-side queue is backlogged , the terminal calculation queue backlog at the edge server , Virtual Energy Deficit Queue , weight as well as Therefore, the state space can be expressed as .
[0135] Action space: The action space is determined by the channel-aware time allocation ratio , channel selection indicator variable and control power Composition, expressed as .
[0136] Reward: Due to To minimize the problem, the reward of the MDP is defined as The negative value of the optimization objective, i.e. .
[0137] In a further embodiment of the present invention, the first-stage method uses an adversarial DQN network to solve the constructed MDP problem. The adversarial DQN network includes a main network and a target network, and their parameter vectors are respectively and Among them, the main network uses the state of the current time slot and main network parameters Generate strategy , according to the strategy adopted The algorithm determines the action. Strategy Parameterized as ,Right now , used to indicate that when the system is in state When the parameter Take action under the strategy When a terminal access confrontation occurs, the edge server perceives the confrontation based on the collected edge SINR information and preferentially allows terminals with larger SINR to access the channel, thereby solving the terminal confrontation problem. The target network rewards based on environmental feedback and combines the terminal Current Status and the next state Calculate the TD error and loss function, and the main network adjusts the strategy using the gradient descent method based on the loss function.
[0138] The proposed joint optimization method of time allocation, channel allocation and energy control based on adversarial DQN includes five stages: network initialization, action selection, adversarial perception, action execution and network parameter update. The execution steps are described as follows:
[0139] 1) Initialize the main network parameters Initialize the target network parameters as random values .
[0140] 2) In the time slots, terminal Strategy obtained based on adversarial DQN network ,use Strategy selection action . Strategy For exploration, Select actions for utilization. As training progresses, exploration Will continue to decay to a very small value, and finally the model with high probability Execute the current best strategy.
[0141] 3) Define a collection Indicates channel selection The terminal set, for the set Terminal in ,calculate use of ,when When Select Channel , and order , on the contrary, when Terminals are not allowed Select Channel , and order , from the collection Remove . Determine whether any channel has terminal access confrontation. If the channel Terminal access confrontation occurs, that is, , then for the set The terminal in To sort in descending order, select Terminal Access Channel , and remove the remaining terminals from the collection Remove from.
[0142] 4) Each terminal, such as Decision-making based on actions The time allocation ratio , Power Control And channel selection strategy , perform channel sensing and data transmission, and observe rewards And further update the queue according to equations (2) and (11) , .
[0143] The TD error is calculated based on the current state and the next state, expressed as
[0144] (16)
[0145] In the formula, is the discount factor, which indicates the extent to which the target network considers the value of future rewards; for The Q function of the main network and the target network, through different network parameters Calculate the network Q value. Introducing the Q value of the target network can improve the stability of learning. is the penalty function, expressed as
[0146] (17)
[0147] In the formula, For Channel The number of times selected.
[0148] TD error represents the prediction error of the environment in the current state, thereby guiding the agent to learn and adjust strategies. The loss function is calculated based on the TD error and is expressed as
[0149] (18)
[0150] Main network update based on loss function
[0151] (19)
[0152] in, Indicates the update step size.
[0153] 6) ,until The iteration ends when Time slots for update .
[0154] In a preferred embodiment of the present invention, for the second stage, a computing resource allocation method based on latency and security compromise (i.e., the second joint optimization method) is proposed to replace the resources allocated for block authentication calculations with the difference between the total computing resources and the data computing allocation resources, thereby achieving decoupling between data calculations and block authentication calculations, and using the optimal time allocation, channel allocation, and power control joint optimization results to transform the original problem into a convex optimization problem. Then, based on the dual decomposition and Karush-Kuhn-Tucker (KKT) conditions, the optimality conditions for computing resource allocation are derived, and they are iteratively solved to achieve a balanced compromise between data processing latency and blockchain consensus security. The specific scheme of this method is as follows:
[0155] (2) Computing resource allocation method based on latency and security tradeoff
[0156] The transformed problem Further decomposed into Sub-question, The sub-problem is the side computing resource allocation optimization sub-problem.
[0157] based on Optimal time allocation, channel allocation and power control results of sub-problems The sub-problem is expressed as
[0158] (20)
[0159] In a further embodiment of the present invention, data calculation and blockchain authentication calculation are decoupled, and the Represents the computing resources used for blockchain authentication calculations, and further , , , Substitute Transformed into a convex optimization problem
[0160] (twenty one)
[0161] structure The Lagrangian function is
[0162] (twenty two)
[0163] In the formula, and Respectively and The Lagrange multiplier of .
[0164] Using the dual decomposition method, Convert to
[0165] (twenty three)
[0166] According to the KKT condition, the necessary condition for the Lagrangian function to take an extreme value is The first-order partial derivative of is equal to zero, expressed as
[0167] (twenty four)
[0168] By solving equation (24), the optimal computing resource allocation decision can be obtained.
[0169] In a further embodiment of the present invention, a computing resource allocation method based on latency and security trade-off is proposed to solve equation (24), and the specific steps are as follows:
[0170] 1) Initialize the number of iterations , iterative step length and And the convergence accuracy .
[0171] 2) Perform iterative condition judgment, if or Update The optimal value of for
[0172] (25)
[0173] 3) Update the Lagrange multiplier and for
[0174] (26)
[0175] (27)
[0176] 4) Order Repeat steps 2)-3) until the convergence accuracy is met , the income This is the result of the optimal allocation of computing resources, that is, .
[0177] Compared with the existing technology, it has the following advantages:
[0178] 1. The present invention proposes a joint optimization method of time planning, channel allocation and energy control based on adversarial DQN. When a terminal access confrontation occurs on a channel, that is, the number of terminals that select the same channel for data transmission exceeds the maximum threshold of the number of terminals allowed to access the channel, the edge server will collect the edge SINR information and sort the terminals in descending order based on SINR, giving priority to terminals with larger SINR to access the channel, thereby resolving the confrontation and achieving edge collaborative optimization of channel selection and efficient transmission of low-latency, low-power data.
[0179] 2. The present invention proposes a computing resource allocation method based on latency and security compromise. The difference between the total computing resources and the data computing allocation resources is used to replace the resources allocated for block authentication calculation, thereby realizing the decoupling between data calculation and block authentication calculation, and converting the original problem into a convex optimization problem. Then, based on the dual decomposition and KKT conditions, the optimality conditions for computing resource allocation are derived, and a computing resource allocation method based on latency and security compromise is proposed for iterative solution, so as to achieve a balanced compromise between data processing latency and blockchain consensus security.
[0180] The above is a detailed introduction to an embodiment of a distribution network collaborative interaction method based on integrated communication and computing of the present invention. The following is a detailed introduction to an embodiment of a distribution network collaborative interaction system based on integrated communication and computing of the present invention.
[0181] See also Figure 2 This embodiment provides a distribution network collaborative interaction system based on integrated sensing, transmission and computing, which can be divided into two parts: a sensing, transmission and computing integrated edge computing device and a 5G cognitive IoT terminal device.
[0182] 1) Sensing, transmission and computing integrated edge computing device. The sensing, transmission and computing integrated edge computing device proposed in the present invention includes: a joint optimization module of time allocation, channel allocation and energy control based on adversarial DQN, a computing resource allocation module based on latency and security compromise, a communication module, a data storage module and a power supply module.
[0183] Joint optimization module of time allocation, channel allocation and energy control based on adversarial DQN: Based on the edge SINR information collected by the SINR information collection module, five steps of network initialization, action selection, adversarial perception, action execution and network parameter update are performed, thereby realizing the joint optimization of time allocation, channel allocation and energy control for edge collaboration and solving terminal access adversarial problems.
[0184] Computing resource allocation module based on latency and security compromise: The computing resource allocation module based on latency and security compromise uses the optimal strategy generated by the joint optimization module of time allocation, channel allocation and energy control based on adversarial DQN to decouple data calculation from blockchain authentication calculation, transform the edge server computing resource allocation optimization problem into a convex optimization problem, and obtain the optimal result of computing resource allocation based on dual decomposition and KKT conditions.
[0185] Communication module: The communication module is responsible for communicating with the 5G cognitive IoT terminal, receiving the electrical equipment data collected by the terminal and issuing resource allocation strategies. At the same time, when a terminal access conflict occurs, it collects edge SINR information.
[0186] Data storage module: The data storage module is mainly composed of flash memory unit (Flash) and random access memory unit (RAM). The RAM unit, as a memory module of the portable electronic positioning tag device, provides a basis for the high-speed operation of the integrated edge computing device; while the flash memory unit, as a long-term storage module for device data, provides sufficient storage space for the received electrical equipment data.
[0187] Power module: The power module is responsible for providing a stable voltage level to each module in the integrated sensing, transmission and computing edge computing device and ensuring the normal operation of the system circuit.
[0188] 2) 5G cognitive IoT terminal device. The 5G cognitive IoT terminal device proposed in the present invention includes: a data acquisition module, a channel sensing module, a communication module, a data storage module and a power module.
[0189] Data acquisition module: The data acquisition module is responsible for collecting data from the electrical equipment connected to it.
[0190] Channel sensing module: The channel sensing module senses the availability of the channel based on cognitive radio technology.
[0191] Communication module: The communication module is responsible for communicating with the integrated sensing, transmission and computing edge computing device, uploading the collected electrical equipment data and receiving the optimal resource allocation strategy issued by the integrated sensing, transmission and computing edge computing device.
[0192] Data storage module: The data storage module is mainly composed of a flash memory unit (Flash) and a random access memory unit (RAM). The RAM unit, as a memory module for portable electronic positioning tag devices, provides a basis for the high-speed operation of 5G cognitive IoT terminal devices; while the flash memory unit, as a long-term storage module for device data, provides sufficient storage space for the received electrical equipment data.
[0193] Power module: The power module is responsible for providing a stable voltage level to each module in the 5G cognitive IoT terminal device and ensuring the normal operation of the system circuit.
[0194] The present invention proposes a distribution network collaborative interaction system based on integrated communication, sensing and computing, including an integrated edge computing device for sensing, transmission and computing and a 5G cognitive IoT terminal device. Among them, the integrated edge computing device for sensing, transmission and computing includes a joint optimization module for time allocation, channel allocation and energy control based on adversarial DQN, a computing resource allocation module based on latency and security compromise, a communication module, a data storage module and a power module; the 5G cognitive IoT terminal device includes a data acquisition module, a channel sensing module, a communication module, a data storage module and a power module. The distribution network collaborative interaction system based on integrated communication, sensing and computing can, firstly, in terms of perception, optimize the time allocation ratio of channel perception and data transmission based on cognitive radio technology to achieve intelligent perception; secondly, in terms of transmission, jointly optimize channel selection and power allocation, solve the access confrontation problem through edge-end collaborative interaction, and achieve low-latency and low-power data transmission; finally, in terms of computing, optimize the edge computing resource allocation to achieve low-latency data processing and blockchain consensus security, ensure the balance between latency and security, and provide a basis for efficient and safe collaborative interaction of distribution networks.
[0195] The following is a simulation analysis example to illustrate the advantages of the present invention compared to the prior art.
[0196] The simulation experiment of the method and system proposed in the present invention uses MATLAB software and is based on Intel Core i7-6900K CPU. The test site area is 100m×100m, including 8 channels, 25 5G cognitive IoT terminal devices and 1 integrated sensing, transmission and computing edge computing device. The 5G cognitive IoT terminal devices and the integrated sensing, transmission and computing edge computing devices are randomly distributed in the simulation area. In this section, the proposed method is compared with two different comparison methods. Comparison method 1 uses simulated anneal arithmetic (SAA), which is a random optimization method based on Monte Carlo iteration strategy. This method can realize random global optimal solution search in the feasible domain, and can get rid of the local optimal solution in a probabilistic jump manner, and finally converge to the global optimal solution. However, comparison method 1 cannot perform channel confrontation sensing, and a random access strategy is adopted when terminal access confrontation occurs. Comparison method 2 uses the traditional DQN algorithm to allocate resources based on queue backlog perception to ensure minimum queue delay. However, it lacks channel confrontation perception, resulting in excessive delay in blockchain authentication calculation and greatly reduced security.
[0197] Depend on Figure 4It can be seen that compared with the two comparison methods, the weighted delay of the proposed method is reduced by 21.23% and 14.39% respectively. This is because the proposed method is based on adversarial perception, which enhances the learning ability of DQN, avoids terminal access conflicts, improves resource utilization efficiency, and reduces data transmission delay. At the same time, the proposed algorithm achieves a compromise allocation of data computing resources and blockchain authentication computing resources.
[0198] Depend on Figure 5 It can be seen that compared with the two comparison methods, the energy consumption of the 5G cognitive IoT terminal device of the proposed method is reduced by 5.48% and 3.28% respectively. As can be seen from the figure, the average energy consumption fluctuation of the proposed method is relatively small and always remains within the energy consumption constraint range. The fundamental reason for the reduction in energy consumption is that the proposed method optimizes the channel perception time allocation ratio of the channel allocation of the 5G cognitive IoT terminal device. This optimization realizes comprehensive channel perception, reduces terminal access confrontation, and ensures reliable and stable data transmission.
[0199] Based on the same inventive concept, the embodiment of the present application also provides a distribution network collaborative interaction system based on integrated communication and computing for realizing the distribution network collaborative interaction method based on integrated communication and computing involved above. The implementation scheme for solving the problem provided by the system is similar to the implementation scheme recorded in the above method, so the specific limitations in the distribution network collaborative interaction system embodiment based on integrated communication and computing provided below can be referred to the limitations of the distribution network collaborative interaction method based on integrated communication and computing above, and will not be repeated here.
[0200] The embodiment of the present invention provides a distribution network collaborative interaction system based on integrated synergy and computing, including:
[0201] Edge servers and terminals;
[0202] The edge server and the terminal interact with each other in the distribution network. The steps of the collaborative interaction based on the edge server are as follows:
[0203] Get the end-side data queue, which is the total amount of data that all terminals are ready to transmit to the edge server;
[0204] Transmit data through channel-connected terminals;
[0205] Generate a data calculation queue for the edge server, where the data calculation queue is the amount of data to be processed transmitted from the terminal to the edge server;
[0206] Obtain various delay information in the edge-end collaboration process, where each delay information includes at least transmission-related delay information generated by the total data volume of the end-side data queue during the data transmission process and calculation-related delay information generated by the amount of data to be processed in the data calculation queue after the calculation resources are allocated;
[0207] Taking the set latency requirement as the optimization goal, the edge data transmission process and computing resource allocation process are continuously optimized based on the channel selection principle and computing resource allocation principle, so as to realize the coordinated interaction of the distribution network based on the integration of tele-sensing and computing;
[0208] The channel selection principle is that when a terminal access confrontation occurs on a channel, the terminal access channel is selected in descending order of the edge signal-to-interference-noise ratio.
[0209] The computing resource allocation principle is to use the difference between the total computing resources and the data computing allocation resources to replace the resources allocated for block authentication computing.
[0210] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. 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 invention.
Claims
1. A distribution network collaborative interaction method based on integrated synergy and computing, characterized in that: The steps include: Obtaining a terminal-side data queue, where the terminal-side data queue is a total amount of data that all terminals are ready to transmit to the edge server; Transmit data through channel-connected terminals; Generate a data calculation queue for the edge server, where the data calculation queue is the amount of data to be processed transmitted from the terminal to the edge server; Acquire various delay information in the edge-end collaboration process, wherein the various delay information includes at least transmission-related delay information generated by the total data volume of the end-side data queue during the data transmission process and calculation-related delay information generated by the amount of data to be processed in the data calculation queue after the calculation resources are allocated; Taking the set latency requirement as the optimization goal, the edge data transmission process and computing resource allocation process are continuously optimized based on the channel selection principle and computing resource allocation principle, so as to realize the coordinated interaction of the distribution network based on the integration of tele-sensing and computing; The channel selection principle is that when a terminal access confrontation occurs in a channel, the terminal access channel is selected in descending order of the edge signal-to-interference-noise ratio; The computing resource allocation principle is to use the difference between the total computing resources and the data computing allocation resources to replace the resources allocated for block authentication computing.
2. The distribution network collaborative interaction method based on integrated synergy and calculation according to claim 1 is characterized in that: Taking the set delay requirement as the optimization goal, the edge data transmission process and the computing resource allocation process are continuously optimized based on the channel selection principle and the computing resource allocation principle, which are constructed as the first joint optimization problem model for solution. The first joint optimization problem model is specifically as follows: In the formula, For Terminal Channel sensing time allocation ratio, For the Time Slot Terminal The channel selection indicator variable, For the Time Slot Terminal Through the channel The transmission power of the transmitted data, The edge server is used to process data from the terminal. The computing resources allocated to the data, is the set of time slots, For terminal collection, The queuing delay of the end-side queue transmission. Calculate the queue delay weight for the edge side. For the The calculation queue delay of the data calculation queue, Calculate the delay weight for blockchain authentication, Calculate latency for blockchain authentication; is the time allocation ratio constraint, For terminal collection; is the channel availability constraint, For the Time Slot Terminal The channel availability indicator variable; , is the channel selection constraint, which means that each terminal can only select at most one channel for transmission in each time slot, and each channel can be used at most Terminal multiplexing, is the total number of channels, For the channels, is a channel set; is the power allocation constraint, which means that the terminal transmission power is discretized from arrive of levels, among which The level is , and are the minimum and maximum values of the transmission power respectively; Assign constraints to computing resources, indicating the processing The total computing resources of the terminal data must be less than or equal to the available computing resources of the edge server ; Represents the terminal energy consumption constraint, that is, the terminal exist The sum of the sensing and transmission energy consumption of each time slot must be less than or equal to the maximum energy consumption constraint. ; represents the channel selection reliability constraint, i.e., the terminal Select Channel The SINR must be greater than or equal to the threshold ; Indicates The amount of data in the end-side data queue per time slot and the amount of data in the data calculation queue The average rate is stable.
3. The distribution network collaborative interaction method based on integrated synergy and calculation according to claim 2 is characterized in that: Based on Lyapunov optimization, the first joint optimization problem model is converted into the second joint optimization problem model, as follows: In the formula, Represents the weights of data transmission delay, data calculation delay and blockchain authentication calculation delay, the amount of data collected for the end-side data queue, is the amount of data transmitted by the end-side data queue, The amount of data processed by the edge server, For the period up to time slot, the terminal The amount of energy consumption that exceeds the expected amount, To sense energy consumption, The transmission energy consumption.
4. The distribution network collaborative interaction method based on integrated synergy and calculation according to claim 3 is characterized in that: The second joint optimization problem model is transformed into a first sub-optimization problem model for solving, so as to optimize the edge data transmission process. The first sub-optimization problem model is specifically as follows: In the formula, Represents the optimization objective of the first sub-optimization problem model.
5. The distribution network collaborative interaction method based on integrated communication and calculation according to claim 4 is characterized in that: The first sub-optimization problem model is modeled as a Markov decision process for solving, and the Markov decision process includes the following elements: State Space: , For Terminal No. The status of each time slot; Action Space: , For Terminal No. The action taken in each time slot; award: .
6. The distribution network collaborative interaction method based on integrated synergy and calculation according to claim 5 is characterized in that: The first joint optimization method based on the adversarial DQN network is used to solve the Erkov decision process. The first joint optimization method includes a main network and a target network, and the corresponding parameter vectors are and ; The master network is used to utilize the status of the current time slot and main network parameters Generate strategy , according to the strategy adopted Algorithms determine actions and strategies Parameterized as ,Right now , used to indicate that when the system is in state When the parameter Take action under the strategy When terminal access confrontation occurs, the channel selection principle is used to solve the confrontation problem; The target network is used for rewards based on environmental feedback and combined with the terminal Current Status and the next state The TD error and the loss function are calculated, and the main network adjusts the strategy using the gradient descent method based on the loss function.
7. The distribution network collaborative interaction method based on integrated synergy and calculation according to claim 4 is characterized in that: The second joint optimization problem model is transformed into a second sub-optimization problem model for solving, so as to optimize the computing resource allocation process. The second sub-optimization problem model is specifically as follows: In the formula, For Terminal The computational complexity of the data is The time slot is long.
8. The distribution network collaborative interaction method based on integrated communication and calculation according to claim 7 is characterized in that: Based on the result of the first sub-optimization problem model, the second sub-optimization problem model is transformed into a convex optimization problem, and the optimality condition of computing resource allocation is obtained based on the dual decomposition and the Carlo-Kuhn-Tucker condition, which is as follows: use Represents the computing resources used for blockchain authentication calculation. The second sub-optimization problem model is transformed into a convex optimization problem as follows: In the formula, represents the optimization objective of the convex optimization problem; structure The Lagrangian function is: In the formula, and Respectively and The Lagrange multiplier of ; Using the dual decomposition method, Translates to: According to the Carlo-Kuhn-Tucker condition, the necessary condition for the Lagrangian function to take an extreme value is The first-order partial derivative of is equal to zero, expressed as: Solve the above formula to get the optimal computing resource allocation decision.
9. The distribution network collaborative interaction method based on integrated communication and calculation according to claim 8 is characterized in that: The solution process for obtaining the optimal computing resource allocation decision specifically includes the following steps: Initialize the number of iterations , iterative step length and And the convergence accuracy ; Perform iterative condition judgment, if or Update The optimal value of for Update Lagrange multipliers and for make Repeat the above steps until the convergence accuracy is met. , the income This is the result of the optimal allocation of computing resources, that is, .
10. The distribution network collaborative interaction method based on integrated communication and calculation according to claim 1 is characterized in that: The delay information at least includes: The calculation queuing delay of the data calculation queue is: Blockchain authentication calculation delay: 。
Citation Information
Patent Citations
DNP protocol based online communication method of power system simulator and power system
CN101707586A
Mobile key enrollment and use
US20230234537A1
Cited By
A cloud computing-based dual-network collaborative data sharing method and system
CN123098014A
