A smart microgrid distributed dynamic tracking method with privacy protection

Abstract

This invention relates to the field of smart microgrid energy dispatching technology, and in particular to a distributed dynamic tracking method for smart microgrids with privacy protection. The method includes the following steps: S1, determining the total number of smart microgrid entity nodes as N, and the real-time power consumption data of the i-th node as φ. i (t), and then combined with the weight coefficients w of each node i The process involves: S1) obtaining the objective function model for solving power consumption; S2) designing a dynamic tracking algorithm based on a random number perturbation mechanism; S3) detailing the execution process of the algorithm in S2; and S4) determining and verifying the algorithm parameters. The distributed dynamic tracking algorithm based on a random number perturbation mechanism proposed in this invention can ensure that sensitive information of nodes in the microgrid is not leaked, while enabling each node to accurately track the weighted average of the time-varying power consumption data of all nodes in the entire network, thereby obtaining the total power consumption data and providing a decision-making reference for supply and demand load balance control.

Description

Technical Field

[0001] This invention relates to the field of smart microgrid energy dispatching technology, and in particular to a distributed dynamic tracking method for smart microgrids with privacy protection features. Background Technology

[0002] With the development of renewable energy generation technologies such as photovoltaics and wind power, distributed generation has gradually become an effective way to meet the growing demand for loads, improve the overall efficiency of energy utilization, and enhance the reliability of power supply, and has been widely applied in distribution networks. To integrate the advantages of distributed generation, researchers have proposed the concept of a smart microgrid, defined as a system unit consisting of a group of micro-power sources, loads, energy storage systems, and control devices. It is an autonomous system capable of self-control, protection, and management, and can operate either connected to the external power grid or in isolation.

[0003] Smart microgrids, as a new type of network composed of multiple distributed power sources and their related components arranged in a certain topology, are more robust and cost-effective than traditional centralized power grids. However, effective management of power supply and demand loads within the microgrid requires significant attention. Because smart microgrids are a distributed network structure, traditional centralized control methods are no longer applicable. Therefore, a technique called dynamic average tracking (MAST) has been introduced into the research field of smart microgrids. MAST aims to design a distributed cooperative algorithm that allows each node in the network to track the average value of the time-varying reference signal of all nodes simply by exchanging information with its neighbors. In traditional MAST cooperative algorithms, each node must honestly communicate with its neighbors to reach a consensus on a variable of interest. However, in practical applications, this direct information exchange may lead to the leakage of sensitive information; therefore, malicious or curious attackers can easily obtain such data by listening to the channel, and then use the accessed power consumption data to infer detailed household or commercial activities of the target node. Summary of the Invention

[0004] The purpose of this invention is to address the supply and demand load balancing problem in smart microgrids by developing a distributed dynamic tracking method with privacy protection based on a random number perturbation mechanism, so as to rationally plan the power production, storage and dispatch processes.

[0005] A distributed dynamic tracking method for intelligent microgrids with privacy protection features, the method comprising the following steps:

[0006] S1. Determine the total number of intelligent microgrid entity nodes as N, and the time-varying reference signal of the i-th node as φ. i (t), and then combined with the weight coefficients w of each node i The objective function model for solving the time-varying reference signal is obtained.

[0007] S2. Based on the objective function model in S1, a dynamic tracking algorithm based on a random number perturbation mechanism is designed as follows:

[0008]

[0009] It uses virtual states that have been scrambled by random numbers. By exchanging information with each other, the leakage of the node's real information is avoided;

[0010] S3. Describe in detail the execution process of the algorithm in S2;

[0011] S4. Determination and verification of algorithm parameters;

[0012] Furthermore, in S2, z i (t), x i (t) and u i (t) represent the node's internal state, estimated state, and control input, respectively; parameter 0 < ω i <1, α is the regularization weight of agent i, α is the control gain, and γ > 0 are design parameters; This represents a random number generated by node i and passed to its neighbor node j. This represents the random number that node i receives from its neighbor node j. This can be viewed as estimating state x i (t) The virtual state after being disturbed by random numbers;

[0013] Furthermore, each node i generates a set of random numbers based on the number of its neighboring nodes j. The number of generated random numbers is equal to the number of its neighboring nodes j. Then, a random number is sent to each of its neighboring nodes. Each node i receives a random number sent by its neighboring node j. Then, estimate x based on its own state. i (t) The set of random numbers generated by itself and the received set of random numbers Calculate the virtual state used for information transmission

[0014] 2. Furthermore, S3 describes the specific algorithm flow as follows:

[0015] Step 1: Algorithm initialization, specifically including:

[0016] (1-1) Each node i generates N i random numbers Where N i It is the number of its neighboring nodes;

[0017] (1-2) Node i distributes random numbers to its neighboring nodes (e.g., ...). The random number is passed to the neighboring node i1 of i, and receives the random number generated and passed from its neighboring node j.

[0018] Step 2: Initialize the algorithm iteration time t = 0, and repeat the following steps;

[0019] Step 3: Node i calculates its state based on the set of random numbers it generates.

[0020] Step 4: Node i calculates the virtual state based on the set of random numbers passed from its neighboring nodes.

[0021] Step 5: Adjacent nodes exchange information using their respective virtual states. Each node i updates its control input as follows:

[0022]

[0023] Step 6: Node i updates its internal state z i (t+1) and estimator state x i (t+1) is as follows:

[0024] z i (t+1)=z i (t)+h(-γz i (t)+u i (t)),

[0025]

[0026] Where h is the step size used for iteration in the algorithm.

[0027] Step 7: Update the algorithm iteration time and determine whether the iteration has ended. This includes:

[0028] (7-1) The update iteration time is t = t + 1;

[0029] (7-2) Determine if the iteration has ended. Calculate the error between the two most recent state estimates: ||ε(t)||=||x(t)-x(t-1)||. If the error is less than the given threshold, i.e. ||ε(t)||<δ, then the algorithm ends; otherwise, continue to execute steps 3-7 until the algorithm ends.

[0030] The beneficial effects of this invention are:

[0031] Using this invention, the proposed algorithm makes information exchange between nodes no longer dependent on the actual state value x. iInstead of (t), it uses a virtual state that has been perturbed by random numbers. By exchanging information with each other, the leakage of the node's real information is avoided, thus protecting the privacy of the node's sensitive information. At the same time, the random number perturbation mechanism used in this invention is carefully designed to ensure that even if the information exchange between nodes is perturbed by random numbers, accurate dynamic target tracking can still be achieved. The distributed dynamic tracking algorithm based on the random number perturbation mechanism can ensure that the sensitive information of nodes in the microgrid is not leaked, while enabling each node to accurately track the weighted average of the time-varying power consumption data of all nodes in the entire network, thereby obtaining the total power consumption data and providing decision-making reference for supply and demand load balance control. Attached Figure Description

[0032] Figure 1 This is a flowchart of the steps of the present invention;

[0033] Figure 2 These are the specific execution steps of the algorithm in S3 of this invention. Detailed Implementation

[0034] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0035] refer to Figure 1-2 A distributed dynamic tracking method for intelligent microgrids with privacy protection features, comprising the following steps:

[0036] S1. The intelligent microgrid consists of N physical nodes, and the time-varying reference signal of the i-th node is φ. i (t), then the dynamic averaging tracking problem addresses how to ensure that each node can track all time-varying reference signals φ. i The average value of (t), i.e., the target being tracked, can be expressed as

[0037]

[0038] The problem described above is also called the absolute average tracking problem, because the tracking target is all φ. i The absolute average of (t) is used; however, in actual microgrid scenarios, since the importance of each entity node is different, the target value to be tracked is not necessarily the absolute average. In this case, in order to reflect the difference in importance of each node, a weight coefficient w is introduced for each node i. i Then the tracking target becomes all φ i The weighted average sum of (t), i.e.

[0039]

[0040] It should be noted that if all weighting coefficients are set to w i If =1 / N, then the dynamic weighted average tracking problem will degenerate into the absolute average tracking problem. This shows that the absolute average tracking is only a special case of the weighted average tracking, and also shows that the dynamic weighted average tracking has better universality.

[0041] S2. To solve the absolute average tracking problem, existing work has proposed the following dynamic tracking algorithm:

[0042]

[0043] x i (t)=z i (t)+φ i (t)

[0044] Where z i (t), x i (t) and u i (t) represent the node's internal state, estimated state, and control input, respectively. Although this algorithm exhibits good performance in terms of convergence speed and accuracy, its ultimate goal is to track all time-varying reference signals φ. i The absolute average of (t). On the other hand, the algorithm requires all nodes in the network to communicate directly with their neighboring nodes, meaning that the real state of each node is transmitted directly through the channel, which poses a significant risk of leakage of sensitive node information.

[0045] To overcome the shortcomings of the above algorithms, this invention proposes a privacy-preserving dynamic weighted average tracking algorithm based on a random number perturbation mechanism:

[0046]

[0047] Where z i (t), x i (t) and u i (t) represents the node's internal state, estimated state, and control input, respectively. Parameter 0 < ω i <1, α is the regularization weight of agent i, α is the control gain, and γ > 0 are design parameters. This represents a random number generated by node i and passed to its neighbor node j. This represents the random number that node i receives from its neighbor node j. This can be viewed as estimating state x i (t) The virtual state after being disturbed by random numbers.

[0048] Specifically, each node i first generates a set of random numbers based on the number of its neighbors. The number of random numbers generated is equal to the number of its neighboring nodes. Then, one random number is sent to each of its neighboring nodes. Each node i receives random numbers from its neighboring nodes. Then, estimate x based on its own state. i (t) The set of random numbers generated by itself and the received set of random numbers Calculate the virtual state used for information transmission Once all nodes have completed the calculation of their respective virtual states, the virtual states will be exchanged between neighboring nodes in the network.

[0049] From the perspective of the communication network, the combination of random numbers generated by the nodes themselves and those generated by their corresponding neighboring nodes plays a crucial role in ensuring accurate target tracking, as they cancel each other out when calculating their respective virtual states and performing state fusion. It is worth emphasizing that the introduced random number perturbation mechanism adds very little communication load to the algorithm execution because the generation and exchange of random numbers only occur during the initialization phase of the algorithm and are a one-time operation. However, it should be noted that once the communication network topology changes, the perturbation mechanism, including random number generation and exchange, must be re-executed and re-initialized.

[0050] S3, Step 1: Algorithm initialization, specifically including:

[0051] (1-1) Each node i generates N i random numbers Where N i It is the number of its neighboring nodes;

[0052] (1-2) Node i distributes random numbers to its neighboring nodes (e.g., ...). The random number is passed to the neighboring node i1 of i, and receives the random number generated and passed from its neighboring node j.

[0053] Step 2: Initialize the algorithm iteration time t = 0, and repeat the following steps;

[0054] Step 3: Node i calculates its state based on the set of random numbers it generates.

[0055] Step 4: Node i calculates the virtual state based on the set of random numbers passed from its neighboring nodes.

[0056] Step 5: Adjacent nodes exchange information using their respective virtual states. Each node i updates its control input as follows:

[0057]

[0058] Step 6: Node i updates its internal state z i (t+1) and estimator state x i (t+1) is as follows:

[0059] z i (t+1)=z i (t)+h(-γz i (t)+u i (t)),

[0060]

[0061] Where h is the step size used for iteration in the algorithm.

[0062] Step 7: Update the algorithm iteration time and determine whether the iteration has ended. This includes:

[0063] (7-1) The update iteration time is t = t + 1;

[0064] (7-2) Determine if the iteration has ended. Calculate the error between the two most recent state estimates: ||ε(t)||=||x(t)-x(t-1)||. If the error is less than the given threshold, i.e. ||ε(t)||<δ, then the algorithm ends; otherwise, continue to execute steps 3-7 until the algorithm ends.

[0065] S4. Determination and verification of algorithm parameters;

[0066] Assumption 1 (Connectivity): Assume that the network topology consisting of N nodes is bidirectional and connected.

[0067] Assumption 2 (Boundedness): For any node i∈{1,...,N} in the network topology, its time-varying reference signal φ i (t) and its derivative All are bounded, meaning they have positive constants. And σ makes:

[0068]

[0069] Theorem 1: Based on assumptions 1 and 2, if the control gain α satisfies:

[0070]

[0071] in If λ2 is the second smallest eigenvalue (i.e., the smallest non-zero eigenvalue) of the Lapacian matrix corresponding to the communication topology graph, then the algorithm proposed in this invention can guarantee the virtual state x after random number perturbation.+ (t) To achieve the weighted average signal within a finite time. Precise tracking, namely:

[0072]

[0073] Its corresponding lower time bound t * for:

[0074]

[0075] Where e(t0) is the initial steady-state error.

[0076] Theorem 1 above shows that the convergence time of the algorithm is bounded and exponential, with a corresponding lower time bound t. * It is mainly determined by the design parameter γ, the weight matrix W, and the initial steady-state error e(t0). Furthermore, the lower bound of the control gain α depends on the reference signal φ(t) and its time derivative. Global information. In practical applications, the global information of the reference signal and its upper bound of time derivative can be obtained by running the maximum consensus algorithm during the algorithm initialization phase.

[0077] Theorem 2: In the algorithm proposed in this invention, node i can protect its estimator state x. i (t), time-varying reference signal φ i The privacy of (t) is not disclosed unless all the neighbors of node i cooperate to infer the relevant state information of node i.

[0078] If all neighboring nodes of the target node are willing to cooperate in attacking the node, then the algorithm proposed in this invention can no longer protect the sensitive information of the target node. On the contrary, as long as at least one neighboring node is non-cooperative, the proposed algorithm can guarantee the privacy of the sensitive state of the target node. In addition, if the attacker does not have the network topology information of the system, it will not be able to launch an effective attack on the target, because it cannot determine whether a certain node belongs to the neighboring node of its attack target, and therefore cannot collect the random numbers generated by all the neighboring nodes of the target and further calculate and infer the true sensitive information of the attack target.

[0079] In the specific implementation process of this invention, (1) by introducing a weight coefficient to each node in the system, the proposed algorithm enables each node to accurately track the weighted average value of the time-varying reference signal instead of the simple absolute average value; (2) based on the random number perturbation mechanism, a privacy protection scheme is carefully designed for the dynamic tracking algorithm, so that the algorithm can accurately track the target point while avoiding the leakage of sensitive information of participating nodes; the distributed dynamic tracking algorithm based on the random number perturbation mechanism proposed in this invention can ensure that the sensitive information of nodes in the microgrid is not leaked, while enabling each node to accurately track the weighted average value of the time-varying power consumption data of all nodes in the entire network, thereby obtaining the total power consumption data and providing decision reference for supply and demand load balance control.

[0080] The specific embodiments described herein are merely illustrative of the spirit of the invention; those skilled in the art can make various modifications or additions to the described specific embodiments or use similar methods to replace them, without departing from the scope defined by the spirit of the invention.

Claims

1. A distributed dynamic tracking method for intelligent microgrids with privacy protection features, characterized in that, The method includes the following steps: S1. Determine the total number of intelligent microgrid entity nodes as N, and the time-varying reference signal of the i-th node as follows: Then, combined with the weight coefficients of each node The objective function model for solving the time-varying reference signal is obtained. ; S2. Based on the objective function model in S1, a distributed dynamic tracking algorithm based on a random number perturbation mechanism is designed as follows: ； In S2 , and These represent the node's internal state, estimated state, and control input, respectively; parameters , It is an intelligent agent Regularization weights, It is about controlling the gain. These are design parameters; Indicates that by node Generate and pass it on to neighboring nodes random numbers, Represents a node From neighboring nodes The received random number, It is an estimated state The virtual state after being disturbed by random numbers; It uses virtual states that have been scrambled by random numbers. They exchanged information with each other, thus avoiding the leakage of the nodes' real information; S3. Describe in detail the execution process of the algorithm in S2; S3 describes the specific algorithm flow as follows: Step 1: Algorithm initialization, specifically including: (1-1) Each node generate random numbers ,in It is the number of its neighboring nodes; (1-2) nodes The random number is passed to its neighboring nodes, and the data is received from its neighboring nodes. Generated and transmitted random numbers ; Step 2: Initialize the algorithm iteration time And repeat the following steps; Step 3: Nodes Calculate the state based on the set of random numbers generated by itself. ; Step 4: Nodes The virtual state is calculated based on the set of random numbers passed from neighboring nodes. ; Step 5: Adjacent nodes exchange information using their respective virtual states. The control input is updated as follows: Step 6: Node Update internal status and estimator state as follows: in It is the step size used for iteration in the algorithm; Step 7: Update the algorithm iteration time and determine whether the iteration has ended. This includes: (7-1) Update iteration time ; (7-2) Determine whether the iteration has ended and calculate the error between the two most recent state estimates. If the error is less than a given threshold, If the algorithm succeeds, then the algorithm ends; otherwise, continue executing steps 3-7 until the algorithm ends. S4. Determination and verification of algorithm parameters; Assumption 1: Connectivity: Assume that the network topology consisting of N nodes is bidirectional and connected; Assumption 2: Boundedness: For any node in the network topology Its time-varying reference signal and its derivative They are all bounded and contain positive constants. and Make: Theorem 1: Based on assumptions 1 and 2, if the control gain... satisfy: in , It is the second smallest eigenvalue of the Lapacian matrix corresponding to the communication topology, and the virtual state after random number perturbation. Achieve weighted average signal within a limited time. Precise tracking, Its corresponding lower time bound for: in This represents the initial steady-state error. Each node i generates a set of random numbers based on the number of its neighboring nodes j. The number of random numbers generated is equal to the number of its neighboring nodes j, and then a random number is sent to each of its neighboring nodes; each node i receives a random number sent from its neighboring node j. Then, estimate based on its own state. The set of random numbers generated by itself and the received set of random numbers Calculate the virtual state used for information transmission .

Images