A Time-Varying Nash Equilibrium Search Method for Secondary Control of Non-Uniform Decomposition DC Microgrid under a Hierarchical Framework
By employing a non-uniform decomposition method within a hierarchical framework and simplified reinforcement learning, the problems of time-varying Nash equilibrium search and privacy protection in DC microgrids are solved, achieving precise voltage control and privacy protection in dynamic environments, and improving the system's anti-interference capability and resilience.
Patent Information
- Application Number
- CN202611242638.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-17
- Publication Date
- 2026-11-17
AI Technical Summary
In DC microgrids, how to achieve time-varying Nash equilibrium search and protect privacy in dynamic environments, especially in multi-bus systems, how to balance voltage recovery and current distribution while avoiding sensitive information leakage and improving system resilience.
A non-uniform decomposition method under a hierarchical framework is adopted. By constructing a game model and value function for a multi-bus DC microgrid, the model is separated into an estimation layer and an execution layer. A Nash equilibrium estimator is used for real-time estimation, and simplified reinforcement learning and non-uniform decomposition communication topology are employed to protect privacy, thereby achieving Nash equilibrium control of the voltage of each bus.
It achieves accurate tracking of Nash equilibrium under unknown nonlinear dynamic environment, protects sensitive information in communication network, improves the system's anti-interference ability and privacy security, and ensures that the voltage of each bus reaches Nash equilibrium.
Smart Images

Figure CN123121490A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of industrial process control technology, specifically relating to a time-varying Nash equilibrium search method for secondary control of a non-uniformly decomposed DC microgrid under a hierarchical framework. Background Technology
[0002] DC microgrids utilizing renewable energy are becoming increasingly important in practical engineering applications due to their advantages in improving energy efficiency and reducing conversion losses. Their control schemes are typically designed in different layers, with secondary and tertiary control layers supporting microgrid operation and achieving optimization goals. Generally, in the implementation of distributed secondary control in multi-bus DC microgrids, the challenge lies in balancing voltage recovery and proportional current distribution.
[0003] The dynamic objectives of voltage recovery and current distribution conflicting in power systems can be viewed as a game theory problem under dynamic conditions, and thus transformed into a time-varying Nash equilibrium search problem. Therefore, research on distributed time-varying Nash equilibrium provides a convenient and effective method for the development of the aforementioned technical fields. In practical engineering, DC microgrid systems often exhibit unknown nonlinear disturbances, making it difficult to directly apply most single-integrator or dual-integrator Nash equilibrium estimators. Furthermore, if unknown dynamic characteristics are directly identified and a distributed NE estimator is introduced, the physical network will use neighbor information, leading to weak anti-interference capabilities and the risk of further privacy leaks.
[0004] In distributed systems, attackers can intercept control signals from the secondary control layer, thereby obtaining sensitive information about energy supply and demand. Leakage of voltage information transmitted in the secondary control communication topology weakens the consistency performance of local distributed power sources, further reducing the resilience of the microgrid. In this situation, privacy issues can lead to poor microgrid operation and even overload. Therefore, privacy protection in the secondary control of DC microgrids is of great significance in ensuring the secure transmission of sensitive information.
[0005] Therefore, how to solve the problem of time-varying Nash equilibrium search and privacy protection in microgrid game problem under dynamic environment, and how to protect privacy while completing the Nash equilibrium search, is the technical problem that this invention aims to solve. Summary of the Invention
[0006] The purpose of this invention is to provide a time-varying Nash equilibrium search method for secondary control of a non-uniformly decomposed DC microgrid under a hierarchical framework, so as to solve the problems mentioned in the background art.
[0007] The objective of this invention is achieved as follows: a time-varying Nash equilibrium search method for secondary control of a non-uniformly decomposed DC microgrid under a hierarchical framework, characterized by the following steps:
[0008] Step S1: Define the bus set of the multi-bus DC microgrid and construct a game model for the multi-bus DC microgrid;
[0009] Step S2: Construct the value function of a multi-bus DC microgrid using a game theory model;
[0010] Step S3: Based on the value function of the multi-bus DC microgrid, clarify the optimization control objectives of each bus, and divide the secondary control of the multi-bus DC microgrid into the multi-bus DC microgrid estimation layer and the multi-bus DC microgrid execution layer;
[0011] Step S4: Construct a multi-bus DC microgrid communication topology and a Nash equilibrium estimator based on non-uniform decomposition, and estimate the Nash equilibrium of other buses based on the Nash equilibrium estimator.
[0012] Step S5: Construct a Nash equilibrium estimator to estimate the Nash equilibrium of each bus voltage in real time and output the Nash equilibrium estimate to the physical layer;
[0013] Step S6: Determine the tracking error based on the Nash equilibrium estimate, obtain the estimate of the optimal trigger error using simplified reinforcement learning, update the trigger control input based on the game-theoretic trigger conditions, track the Nash equilibrium estimate output by the estimation layer, and achieve the optimal control objective for each bus.
[0014] Preferably, in step S1, defining the bus set of a multi-bus DC microgrid and constructing the value function of the multi-bus DC microgrid are specifically as follows:
[0015] definition busbar voltage, It is a busbar The reference voltage, It is a busbar The load current, It is a busbar The load current proportionality factor;
[0016] Based on the voltage recovery and current proportional distribution tasks in the secondary control of a multi-bus DC microgrid, the bus... voltage Define the bus set of a multi-bus DC microgrid. ; Indicates the number of buses in a multi-bus DC microgrid;
[0017] Based on the bus set of a bus-based DC microgrid, a game theory model for a multi-bus DC microgrid is constructed. The game theory model for a multi-bus DC microgrid is as follows: ;
[0018] A game theory model for multi-bus DC microgrids describes the game relationship between voltage recovery and current proportional distribution.
[0019] Preferably, in step S2, a multi-bus DC microgrid value function is constructed using a game model, specifically as follows:
[0020] Based on the game theory relationship between voltage recovery and current proportional distribution, the value function of a multi-bus DC microgrid is obtained. The value function of a multi-bus DC microgrid is as follows:
[0021] ;
[0022] Wherein, it represents the busbar, excluding the busbar. Voltage of other external busbars; It is a time-varying regularization parameter; It is a busbar With busbar The impedance between the wires; It is the load current distribution factor. ; It is the sum of the conductor admittances of each busbar and its adjacent busbars. .
[0023] Preferably, in step S3, the optimal control objective of each bus is defined based on the value function of the multi-bus DC microgrid, and the secondary control of the multi-bus DC microgrid is divided into a multi-bus DC microgrid estimation layer and a multi-bus DC microgrid execution layer, specifically:
[0024] The optimization control objective for each bus is to achieve Nash equilibrium in voltage values. The secondary control of the multi-bus DC microgrid is divided into a multi-bus DC microgrid estimation layer and a multi-bus DC microgrid execution layer.
[0025] The objective of the estimation layer for multi-bus DC microgrids is to calculate Nash equilibrium. The model for the estimation layer of multi-bus DC microgrids is as follows:
[0026] ;
[0027] in, It is a Nash equilibrium estimate and serves as a reference signal for the execution layer. It is a busbar The input to the estimation layer model enables the Nash equilibrium estimation. Converging to Nash equilibrium;
[0028] The objective of the execution layer of the multi-bus DC microgrid is to control the voltage of the buses. The model of the execution layer of the multi-bus DC microgrid is as follows: ;
[0029] in, It is an unknown nonlinear dynamic. It is a busbar The execution layer triggers the control input, causing the bus to... Voltage tracking Nash equilibrium estimation This allows the voltage to reach Nash equilibrium, minimizing the value function of the multi-bus DC microgrid. .
[0030] Preferably, in step S4, constructing a multi-bus DC microgrid communication topology and a Nash equilibrium estimator based on non-uniform decomposition, and estimating the Nash equilibrium of other buses based on the Nash equilibrium estimator, specifically involves:
[0031] The Nash equilibrium estimator observer is:
[0032] ;
[0033] in, busbar busbar Interactive observation of Nash equilibrium estimation busbar busbar The internal observations of the Nash equilibrium estimate; and These are the time-varying decomposition weights of the interaction estimate and the internal estimate of the Nash equilibrium estimate, respectively. , , Decompose the weight constants; It is a busbar According to the data received on the communication network and right The restoration; These are time-varying weighted combination coefficients; It is the observation gain; It is the decomposition coupling coefficient;
[0034] The Nash equilibrium estimator will use the bus. busbar The observation decomposition of the Nash equilibrium estimate is based on the busbar Nash equilibrium estimation interactive observation and busbar Nash equilibrium estimation internal observations ;and Transmitting data on the communication network and interacting with neighbors on the communication network. Only locally and Interactive; each bus transmits Nash equilibrium estimates and interactive observations only on the communication network. Combination coefficients of time-varying weights Time-varying decomposition weights and Selected as a non-zero bounded time-varying function, the weight combination coefficients are according to Real-time generation; busbar Received and Afterwards, through Complete locally The restoration; this process is performed locally and not publicly disclosed on the communication network; the Nash equilibrium estimate is an internal estimate. It is also not disclosed on communication networks.
[0035] Preferably, the Nash equilibrium estimator is:
[0036] ;
[0037] in, It is a busbar The estimate of the Nash equilibrium, It is to estimate the gain. It is a busbar value function The first Hessian matrix Line number List the blocks, Value function gradient with respect to time The partial derivatives;
[0038] Time-varying decomposition weights of interaction estimates and internal estimates of Nash equilibrium estimates and The time-varying weight combination coefficient must be strictly greater than 0. and observation gain Must meet Choose a larger decomposition coupling coefficient and estimated gain Accelerate the convergence speed.
[0039] Preferably, step S6, which determines the tracking error based on the Nash equilibrium estimate, specifically involves:
[0040] Step S6-1: Determine the tracking error:
[0041] Based on the Nash equilibrium estimation output from the communication layer And measure the busbar voltage The tracking error was determined to be... ;
[0042] Step S6-2: Define the continuous control input as , To account for the error between the trigger control input and the continuous control input; The trigger parameter; the output of the communication layer is a Nash equilibrium estimate. ;
[0043] Step S6-3: Use reinforcement learning to obtain the estimate of the optimal triggering error, and track the Nash equilibrium estimate of the estimation layer output based on the game triggering conditions.
[0044] Preferably, in step S6-3, reinforcement learning is used to obtain an estimate of the optimal triggering error, and the Nash equilibrium estimate output by the estimation layer is tracked based on the game's triggering conditions. Specifically:
[0045] Reinforcement learning includes an execution network and an evaluation network, with the input being the tracking error. The output is the trigger control input. ;
[0046] The optimal trigger error estimate obtained through reinforcement learning is: ;
[0047] in, To unknown dynamics The estimated weights, To unknown dynamics The estimated activation function, This is the weight of the execution network, with a layer number of 1 and a number of nodes. ; It is the activation function of the execution network. These are positive design parameters; Indicates the transpose operation;
[0048] The triggering condition for the game is: ;
[0049] Real-time recording at the execution layer and And determine the triggering condition. If the condition is met, then the control input is triggered, and reinforcement learning is used for updating; the update operation is as follows:
[0050] ;
[0051] in , It is a positive definite matrix;
[0052] If the condition is not met, the control input is triggered. Remains unchanged; Nash equilibrium estimation of the estimation layer output Tracking is underway.
[0053] Compared with the prior art, the present invention has the following improvements and advantages:
[0054] 1. By leveraging the dynamic characteristics of time-varying Nash equilibrium, the time derivative term of the gradient and the Hessian matrix of the value function are introduced as feedforwards to effectively compensate for fluctuations caused by TVNE. This is achieved by introducing compensation for the drift caused by the time-varying Nash equilibrium. Simultaneously, a hierarchical control scheme is adopted, dividing the control system into an estimation layer and an execution layer, isolating the estimation and implementation of Nash equilibrium. In the execution layer, a control method based on simplified reinforcement learning is used to track the Nash equilibrium estimate output by the estimation layer, achieving more accurate tracking of time-varying Nash equilibrium and enabling each bus voltage to reach Nash equilibrium in environments with unknown nonlinear dynamics.
[0055] 2. By introducing non-uniform decomposition, the Nash equilibrium estimation observations transmitted by the estimation layer on the communication topology are decomposed into two parts, interactive and internal, according to different time-varying weights. Furthermore, only the interactive Nash equilibrium estimation observations and time-varying weight combination coefficients are transmitted in the communication network, preventing eavesdroppers from obtaining the complete observations and thus protecting privacy. This method can protect sensitive information related to Nash equilibrium estimation information transmitted in the communication network. Attached Figure Description
[0056] Figure 1 This is an overall flowchart of the method of the present invention.
[0057] Figure 2 This is a schematic diagram illustrating the effectiveness of the Nash equilibrium estimator.
[0058] Figure 3 This is a schematic diagram showing the proportional distribution of current energy by a distributed power source.
[0059] Figure 4 This is a diagram illustrating the effectiveness of privacy protection.
[0060] Figure 5 For the execution layer voltage of each bus Nash equilibrium estimation relative to the output of the NE estimation layer A schematic diagram showing the results of tracking error. Detailed Implementation
[0061] The invention will be further summarized below with reference to the accompanying drawings.
[0062] like Figure 1 As shown, a time-varying Nash equilibrium search method for secondary control of a non-uniformly decomposed DC microgrid under a hierarchical framework is presented. This method includes the following steps:
[0063] Step S1 defines the bus set of a multi-bus DC microgrid and constructs a game model for the multi-bus DC microgrid, specifically as follows:
[0064] definition busbar voltage, busbar The reference voltage, busbar The load current, It is a busbar The load current proportionality factor;
[0065] Based on the voltage recovery and current proportional distribution tasks in the secondary control of a multi-bus DC microgrid, the bus... voltage Define the bus set of a multi-bus DC microgrid. .
[0066] Based on the bus set of a bus-based DC microgrid, a game theory model for multi-bus DC microgrids is constructed. A regularized game model for multi-bus DC microgrids describes the game relationship between voltage recovery and current proportional distribution.
[0067] In step S2, a game theory model for multi-bus DC microgrids is used to construct the value function of the multi-bus DC microgrid, which is specifically expressed as follows:
[0068] Based on the game-theoretic relationship between voltage recovery and current proportional distribution, the value function of a multi-bus DC microgrid is obtained:
[0069] ;
[0070] in, Including busbar Voltage of other buses It is a time-varying regularization parameter; It is a busbar With busbar The impedance between the wires, It is the load current distribution factor. ; It is the sum of the conductor admittances of each busbar and its adjacent busbars. ;
[0071] After constructing the value function of a multi-bus DC microgrid, a game theory model of the multi-bus DC microgrid is combined. The objective for each bus is clearly defined as achieving Nash equilibrium in the secondary control of a multi-bus DC microgrid, thereby minimizing the value function of the multi-bus DC microgrid. .
[0072] In step S3, the optimal control objective of each bus is clarified based on the value function of the multi-bus DC microgrid. The secondary control of the multi-bus DC microgrid is divided into a multi-bus DC microgrid estimation layer and a multi-bus DC microgrid execution layer, specifically as follows:
[0073] To achieve Nash equilibrium in voltage values, the secondary control process of the multi-bus DC microgrid is divided into a multi-bus DC microgrid estimation layer and a multi-bus DC microgrid execution layer.
[0074] The objective of the estimation layer for multi-bus DC microgrids is to calculate Nash equilibrium. The model for the estimation layer of multi-bus DC microgrids is as follows: ;
[0075] The objective of the execution layer in a multi-bus DC microgrid is to control the voltage of the buses. The model for the execution layer of a multi-bus DC microgrid is as follows: ;
[0076] in, It is an unknown nonlinear dynamic. It is a busbar The execution layer triggers the control input, causing the bus to... Voltage tracking Nash equilibrium estimation This allows the voltage to reach Nash equilibrium, minimizing the value function of the multi-bus DC microgrid. .
[0077] In step S4, a multi-bus DC microgrid communication topology and a Nash equilibrium estimator based on non-uniform decomposition are constructed, and the Nash equilibrium estimates of other buses are estimated based on the Nash equilibrium estimator. Specifically:
[0078] Define the communication topology of a multi-bus DC microgrid. ,in This represents the edge set of the communication topology. It is the adjacency matrix of the communication topology; if and only if hour In other cases . Indicates busbar With busbar They can obtain information uploaded by each other in the communication network;
[0079] Construct a Nash equilibrium estimator observer based on non-uniform decomposition:
[0080] ;
[0081] in, busbar busbar Interactive observation of Nash equilibrium estimation busbar busbar The internal observations of the Nash equilibrium estimate The interaction estimation time-varying decomposition weights of the Nash equilibrium estimate are... These are the time-varying decomposition weights of the internal estimates of the Nash equilibrium estimate; These are time-varying weighted combination coefficients. It is a busbar right The restoration; It is the observation gain; It is the decomposition coupling coefficient;
[0082] The Nash equilibrium estimator will use the bus... busbar The observation decomposition of the Nash equilibrium estimate is based on the busbar Nash equilibrium estimation interactive observation and busbar Nash equilibrium estimation internal observations ;and Transmitting data on the communication network and interacting with neighbors on the communication network. Only locally and Interactive; each bus transmits Nash equilibrium estimates and interactive observations only on the communication network. Combination coefficients of time-varying weights Time-varying decomposition weights and Selected as a non-zero bounded time-varying function, the weight combination coefficients are according to Real-time generation. Busbar Received and Afterwards, through Complete locally The restoration. This process is performed locally and not publicly disclosed over the communication network; the Nash equilibrium estimate is internally estimated. It is also not disclosed on communication networks. The attacker and eavesdropper model is... It cannot estimate interactive observations solely based on Nash equilibrium. Combination coefficients of time-varying weights Complete the estimation of convergence to the true Nash equilibrium value The acquisition of.
[0083] In step S5, a Nash equilibrium estimator is constructed to estimate the Nash equilibrium of each bus voltage in real time and output the Nash equilibrium estimate to the physical layer. Specifically:
[0084] The Nash equilibrium estimator is: ;
[0085] in, It is a busbar The estimate of the Nash equilibrium, It is to estimate the gain. It is a busbar value function The first Hessian matrix Line number List the blocks, Value function gradient with respect to time The partial derivatives;
[0086] Time-varying decomposition weights of interaction estimates and internal estimates of Nash equilibrium estimates and The time-varying weight combination coefficient must be strictly greater than 0. and observation gain Must meet Choose a larger decomposition coupling coefficient and estimated gain It can speed up the convergence process;
[0087] Choosing the Lyapunov function yields an upper bound on the Nash equilibrium estimation error:
[0088] ;
[0089] in, This indicates the error in the Nash equilibrium estimation. This represents the column vector of the stack of all Nash equilibrium estimates. For true time-varying Nash equilibrium; This represents the error in the Nash equilibrium estimation of observations. This represents the column vector of the Nash equilibrium estimation observation stack for all buses. Indicates busbar The Nash equilibrium estimate of the stacked column vectors of the observations. This represents a column vector representing the stacked reduction errors of all bus observations for the Nash equilibrium estimate. Indicates busbar The column vector of the stacked reconstructed errors of the Nash equilibrium estimation observations for all other buses. Indicates busbar busbar The reconstruction of observations for Nash equilibrium estimation. Indicates busbar busbar Nash equilibrium estimation interactive observation and bus busbar The error of the internal observations in the Nash equilibrium estimate; the upper bound of the error norm of the Nash equilibrium estimate characterizes the accuracy of the Nash equilibrium estimator in estimating the Nash equilibrium.
[0090] The upper bound of the error norm of the Nash equilibrium estimate is derived as follows: ;
[0091] in, For the estimated gain Observation gain Decompose coupling coefficients Combination coefficients of time-varying weights matrix The smallest eigenvalue.
[0092] In step S6, the tracking error is determined based on the Nash equilibrium estimate. Simplified reinforcement learning is used to obtain the estimate of the optimal trigger error. The trigger control input is updated based on the game's triggering conditions. The tracking is performed on the Nash equilibrium estimate output by the estimation layer. Specifically, this is manifested as follows:
[0093] Step S6-1: Determine the tracking error:
[0094] Based on the Nash equilibrium estimation output from the communication layer And measure the busbar voltage The tracking error was determined to be... Step S6-2: Define the continuous control input as... , To account for the error between the trigger control input and the continuous control input; For triggering parameters;
[0095] In step S6-3, reinforcement learning is used to obtain an estimate of the optimal triggering error. Based on the game's triggering conditions, the Nash equilibrium estimate of the estimation layer output is tracked, specifically as follows:
[0096] Reinforcement learning includes an execution network and an evaluation network, with the input being the tracking error. The output is the trigger control input. ;
[0097] The optimal trigger error estimate obtained through reinforcement learning is: ;
[0098] To unknown dynamics The estimated weights, To unknown dynamics The estimated activation function, This is the weight of the execution network, with a layer number of 1 and a number of nodes. ; It is the activation function of the execution network. It is a positive design parameter; and the weight is... , The update process satisfies:
[0099] ;
[0100] ;
[0101] in This is the weight of the evaluation network, with 1 layer and 1 node. . It is a positive definite matrix, activation function and All are selected as Gaussian functions; and the weights are... The update process satisfies ;
[0102] The initial values of the weights are all set to 0, and the activation function... , , These are positive design parameters;
[0103] The trigger condition for the game is the error between the trigger control input and the continuous control input. Does the inner product satisfy the inequality: ;
[0104] Real-time recording at the execution layer and And determine the triggering condition. Does it meet the requirements?
[0105] If the trigger condition If the condition is met, the control input is triggered, and reinforcement learning is used to update it; the update operation is as follows:
[0106] ;
[0107] in , It is a positive definite matrix;
[0108] If the condition is not met, the control input is triggered. Remains unchanged; Nash equilibrium estimation of the estimation layer output Tracking is underway.
[0109] A distributed time-varying Nash equilibrium control system for a non-uniformly decomposed DC microgrid under a hierarchical framework, the system comprising:
[0110] Multi-bus DC microgrid game model establishment module: used to determine the value function of each bus in the DC microgrid and clarify that the objective of each bus is to achieve Nash equilibrium in the secondary control of the multi-bus DC microgrid;
[0111] Time-varying non-uniform decomposition observation module: used for real-time observation of Nash equilibrium estimates for each bus at the estimation layer, while ensuring privacy. Each bus only transmits its Nash equilibrium estimate interactive observation over the communication network. Combination coefficients of time-varying weights Time-varying decomposition weights and Selected as a non-zero bounded time-varying function, the weight combination coefficients are according to Generate in real time. Busbar Received and Afterwards, through Complete locally The restoration. This process is performed locally and not publicly disclosed over the communication network; the Nash equilibrium estimate is internally estimated. It is also not disclosed on communication networks. The attacker and eavesdropper model is... It cannot estimate interactive observations solely based on Nash equilibrium. Combination coefficients of time-varying weights Complete the estimation of convergence to the true Nash equilibrium value Acquisition;
[0112] Time-varying Nash equilibrium estimation module: used to estimate the Nash equilibrium of a multi-bus DC microgrid in real time at the estimation layer based on the observed values of the Nash equilibrium estimates of other buses output by the time-varying non-uniform decomposition observation module, and output the Nash equilibrium estimate to the execution layer;
[0113] Multi-bus DC microgrid execution layer error construction module: used for real-time measurement and tracking of errors at the execution layer;
[0114] The execution layer control module of the multi-bus DC microgrid is used to track the Nash equilibrium estimation in real time and realize the Nash equilibrium of the voltage of each bus. At the same time, the module adopts a simplified reinforcement learning and game-based event triggering method to avoid unknown gradients and reduce the computational burden.
[0115] To verify the feasibility and effectiveness of the method of the present invention, the following experiments were conducted:
[0116] The simulation platform is Matlab 2023b, with a time step of 0.000005 and time-varying decomposition weights. , The attacker and eavesdropper models are set as follows:
[0117] ;
[0118] An attempt was made to eavesdrop on the estimation of bus 6 to bus 1 by intercepting information from the communication network; parameter selection for multi-bus DC microgrids: , V, V, V, V, V, V; A, A, A, A, A, A; , , , , , ;
[0119] , , , , , , , , , , .
[0120] busbar , , , , , , , , , The activation function is chosen as the Gaussian function. , , ;
[0121] where and Is Uniformly distributed random numbers ; .
[0122] like Figure 2 As shown, under the proposed Nash equilibrium estimator, the estimation of each bus within 15 seconds is performed. , The output Nash equilibrium estimation signal It can quickly converge to Nash equilibrium. Its effectiveness has been verified.
[0123] like Figure 3 As shown, under the proposed Nash equilibrium estimator, the output current of all buses within 15 seconds... Satisfy the demand , , .Right now It can be quickly allocated according to the proportion.
[0124] like Figure 4As shown, eavesdroppers cannot rely on publicly available information on communication networks. Obtain accurate information The information it stole Observations compared with the true Nash equilibrium estimate The difference is significant, which verifies the effectiveness of privacy protection.
[0125] like Figure 5 As shown, the execution layer voltage of each bus Nash equilibrium estimation relative to the output of the NE estimation layer Tracking error The norm rapidly decays to near 0, and the execution layer voltage... It can converge to Nash equilibrium quickly. Within the bounded neighborhood, the effectiveness of the execution layer was verified.
[0126] The above description is merely an embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the present invention should be included within the scope of the claims of the present invention.
Claims
1. A time-varying Nash equilibrium search method for secondary control of a non-uniformly decomposed DC microgrid under a hierarchical framework, characterized in that: The method includes the following steps: Step S1: Define the bus set of the multi-bus DC microgrid and construct a game model for the multi-bus DC microgrid; Step S2: Construct the value function of a multi-bus DC microgrid using a game theory model; Step S3: Based on the value function of the multi-bus DC microgrid, clarify the optimization control objectives of each bus, and divide the secondary control of the multi-bus DC microgrid into the multi-bus DC microgrid estimation layer and the multi-bus DC microgrid execution layer; Step S4: Construct a multi-bus DC microgrid communication topology and a Nash equilibrium estimator based on non-uniform decomposition, and estimate the Nash equilibrium of other buses based on the Nash equilibrium estimator. Step S5: Construct a Nash equilibrium estimator to estimate the Nash equilibrium of each bus voltage in real time and output the Nash equilibrium estimate to the physical layer; Step S6: Determine the tracking error based on the Nash equilibrium estimate, obtain the estimate of the optimal trigger error using simplified reinforcement learning, update the trigger control input based on the game-theoretic trigger conditions, track the Nash equilibrium estimate output by the estimation layer, and achieve the optimal control objective for each bus.
2. The time-varying Nash equilibrium search method for secondary control of a non-uniformly decomposed DC microgrid under a hierarchical framework according to claim 1, characterized in that: In step S1, the bus set of the multi-bus DC microgrid is defined, and the value function of the multi-bus DC microgrid is constructed, specifically as follows: definition busbar voltage, It is a busbar The reference voltage, It is a busbar The load current, It is a busbar The load current proportionality factor; Based on the voltage recovery and current proportional distribution tasks in the secondary control of a multi-bus DC microgrid, the bus... voltage Define the bus set of a multi-bus DC microgrid. ; Indicates the number of buses in a multi-bus DC microgrid; Based on the bus set of a bus-based DC microgrid, a game theory model for a multi-bus DC microgrid is constructed. The game theory model for a multi-bus DC microgrid is as follows: ; A game theory model for multi-bus DC microgrids describes the game relationship between voltage recovery and current proportional distribution.
3. The time-varying Nash equilibrium search method for secondary control of a non-uniformly decomposed DC microgrid under a hierarchical framework according to claim 1, characterized in that: In step S2, a multi-bus DC microgrid value function is constructed using a game theory model, specifically as follows: Based on the game theory relationship between voltage recovery and current proportional distribution, the value function of a multi-bus DC microgrid is obtained. The value function of a multi-bus DC microgrid is as follows: ; Wherein, it represents the busbar, excluding the busbar. Voltage of other external busbars; It is a time-varying regularization parameter; It is a busbar With busbar The impedance between the wires; It is the load current distribution factor. ; It is the sum of the conductor admittances of each busbar and its adjacent busbars. ; This indicates the number of buses in a multi-bus DC microgrid.
4. The time-varying Nash equilibrium search method for secondary control of a non-uniformly decomposed DC microgrid under a hierarchical framework as described in claim 1, characterized in that: In step S3, the optimal control objectives for each bus are defined based on the value function of the multi-bus DC microgrid. The secondary control of the multi-bus DC microgrid is divided into a multi-bus DC microgrid estimation layer and a multi-bus DC microgrid execution layer, specifically: The optimization control objective for each bus is to achieve Nash equilibrium in voltage values. The secondary control of the multi-bus DC microgrid is divided into a multi-bus DC microgrid estimation layer and a multi-bus DC microgrid execution layer. The objective of the estimation layer for multi-bus DC microgrids is to calculate Nash equilibrium. The model for the estimation layer of multi-bus DC microgrids is as follows: ; in, It is a Nash equilibrium estimate and serves as a reference signal for the execution layer. It is a busbar The input to the estimation layer model enables the Nash equilibrium estimation. Converging to Nash equilibrium; Indicates the number of buses in a multi-bus DC microgrid; The objective of the execution layer of the multi-bus DC microgrid is to control the voltage of the buses. The model of the execution layer of the multi-bus DC microgrid is as follows: ; in, It is an unknown nonlinear dynamic. It is a busbar The execution layer triggers the control input, causing the bus to... Voltage tracking Nash equilibrium estimation This allows the voltage value to reach Nash equilibrium, minimizing the value function of the multi-bus DC microgrid. .
5. The time-varying Nash equilibrium search method for secondary control of a non-uniformly decomposed DC microgrid under a hierarchical framework according to claim 1, characterized in that: In step S4, a multi-bus DC microgrid communication topology and a Nash equilibrium estimator based on non-uniform decomposition are constructed, and the Nash equilibrium estimates of other buses are estimated based on the Nash equilibrium estimator. Specifically: The Nash equilibrium estimator observer is: ; in, busbar busbar Interactive observation of Nash equilibrium estimation busbar busbar The internal observations of the Nash equilibrium estimate; and These are the time-varying decomposition weights of the interaction estimate and the internal estimate of the Nash equilibrium estimate, respectively. , , Decompose the weight constants; It is a busbar According to the data received on the communication network and right The restoration; These are time-varying weighted combination coefficients; It is the observation gain; It is the decomposition coupling coefficient; The Nash equilibrium estimator will use the bus. busbar The observation decomposition of the Nash equilibrium estimate is based on the busbar Nash equilibrium estimation interactive observation and busbar Nash equilibrium estimation internal observations ;and Transmitting data on the communication network and interacting with neighbors on the communication network. Only locally and Interactive; each bus transmits Nash equilibrium estimates and interactive observations only on the communication network. Combination coefficients of time-varying weights Time-varying decomposition weights and Selected as a non-zero bounded time-varying function, the weight combination coefficients are according to Real-time generation; busbar Received and Afterwards, through Complete locally The restoration; this process is performed locally and not publicly disclosed on the communication network; the Nash equilibrium estimate is an internal estimate. It is also not disclosed on communication networks.
6. The time-varying Nash equilibrium search method for secondary control of a non-uniformly decomposed DC microgrid under a hierarchical framework according to claim 1, characterized in that: The Nash equilibrium estimator is: ; in, It is a busbar The estimate of the Nash equilibrium, It is to estimate the gain. It is a busbar value function The first Hessian matrix Line 1 List the blocks, Value function gradient with respect to time The partial derivatives; Time-varying decomposition weights of interaction estimates and internal estimates of Nash equilibrium estimates and The time-varying weight combination coefficient must be strictly greater than 0. and observation gain Must meet Choose a larger decomposition coupling coefficient and estimated gain Accelerate the convergence speed.
7. The time-varying Nash equilibrium search method for secondary control of a non-uniformly decomposed DC microgrid under a hierarchical framework according to claim 1, characterized in that: Step S6, which determines the tracking error based on the Nash equilibrium estimation, specifically involves: Step S6-1: Determine the tracking error: Based on the Nash equilibrium estimation output from the communication layer And measure the busbar voltage The tracking error was determined to be... ; Step S6-2: Define the continuous control input as , To account for the error between the trigger control input and the continuous control input; The trigger parameter; the output of the communication layer is a Nash equilibrium estimate. ; Step S6-3: Use reinforcement learning to obtain the estimate of the optimal triggering error, and track the Nash equilibrium estimate of the estimation layer output based on the game triggering conditions.
8. The time-varying Nash equilibrium search method for secondary control of a non-uniform decomposition DC microgrid under a hierarchical framework according to claim 7, characterized in that: In step S6-3, reinforcement learning is used to obtain an estimate of the optimal triggering error. Based on the game's triggering conditions, the Nash equilibrium estimate output by the estimation layer is tracked, specifically as follows: Reinforcement learning includes an execution network and an evaluation network, with the input being the tracking error. The output is the trigger control input. ; The optimal trigger error estimate obtained through reinforcement learning is: ; in, To unknown dynamics The estimated weights, To unknown dynamics The estimated activation function, This is the weight of the execution network, with a layer number of 1 and a number of nodes. ; It is the activation function of the execution network. These are positive design parameters; Indicates the transpose operation; The triggering condition for the game is: ; Real-time recording at the execution layer and And determine the triggering condition. If the condition is met, then the control input is triggered, and reinforcement learning is used for updating; the update operation is as follows: ; in , It is a positive definite matrix; If the condition is not met, the control input is triggered. Remains unchanged; Nash equilibrium estimation of the estimation layer output Tracking is underway.
