Virtual power plant scheduling method and system with communication review mechanism

Abstract

The invention relates to a virtual power plant scheduling method and system under Byzantine attack, and the method comprises the following steps: collecting the real-time power of each distributed power generation node in a virtual power plant and the connection topology information between the distributed power generation nodes, and obtaining an adjacent matrix of the distributed power generation nodes according to the connection topology information; taking the minimum power generation cost of all distributed power generation nodes as a target, simulating abnormal distributed power generation nodes by Byzantine nodes, updating the real-time power of the distributed power generation nodes by adopting a Lagrange duality method on the premise of referring to the positions of all elements in a constraint set, and determining the real-time power of the distributed power generation nodes on the basis of a communication review mechanism. And the communication cost is reduced while scheduling control is performed on the virtual power plant. The method is effective and reliable, the Byzantine attack is judged through a distributed algorithm, power scheduling participated by all power generation nodes is provided on the premise of overcoming risks, and an optimal potential scheduling mode capable of saving communication cost is provided for a virtual power plant.

Description

Technical Field

[0001] The present invention relates to the field of power systems, and more specifically, to a virtual power plant dispatching method and system with a communication review mechanism under network attacks. Background Art

[0002] As society's demand for energy increases and environmental issues continue to increase, people are increasingly developing and utilizing renewable energy sources such as wind and solar energy. Developing and utilizing renewable energy has become a strategic choice for countries around the world. Renewable energy is usually connected to the large power grid in the form of distributed power sources. Distributed resources have the characteristics of small capacity and dispersed geographical locations, which brings management challenges to the power system. Therefore, the concept of virtual power plants came into being.

[0003] At present, the economic dispatch methods of power grids mainly include centralized and distributed methods. Traditional power systems mostly use centralized economic dispatch. Centralized economic dispatch coordinates and optimizes power generation resources and load demand through the dispatch center to maximize the overall economic benefits. However, once the dispatch center fails, it will lead to the failure of the entire economic dispatch task. At the same time, with the increasing complexity of the power grid structure, the amount of information that the dispatch center needs to process and calculate continues to grow, which puts higher standards on the dispatch center and increases the difficulty of power system control. Distributed economic dispatch strategies no longer require a dispatch center that centralizes global information, but instead communicate with neighboring units through units.

[0004] The economic dispatch of virtual power plants refers to coordinating the aggregated distributed power generation resources to achieve the goal of minimizing resource operating costs and maximizing benefits under the premise of satisfying constraints. The economic dispatch of virtual power plants is an important way to solve the scheduling of virtual power plants. How to economically implement the scheduling of virtual power plants is also one of the core issues of smart grid operation. Based on the conceptual nature of virtual power plants, virtual power plants often use a distributed approach for economic dispatch. Distributed economic dispatch strategies have many benefits. They can not only reduce the total cost of power generation, but also improve the flexibility and robustness of economic dispatch strategies. Among them, the dispatch method using decentralized distributed dispatch algorithms is widely used in economic dispatch in power grids.

[0005] However, the decentralized virtual power plant economic dispatch strategy still faces the following problems: (1) The distributed generation nodes in the virtual power plant need to communicate with neighboring nodes. When there are malicious nodes, faulty nodes and risky nodes in the network, it will lead to system instability, inconsistent decision-making and other problems, which greatly reduces the security of the system. (2) The communication resources in the virtual power plant are often limited. With the increase in the complexity of the virtual power plant topology and the number of nodes, there is a problem of communication congestion caused by excessive communication times, which directly affects the realization of the economic dispatch target. Improving communication efficiency means a significant increase in communication costs. (3) The decentralized virtual power plant economic dispatch problem needs to consider the power generation characteristics of each distributed generation node, which causes great trouble in the setting or solution process of the economic dispatch target. On the one hand, when the economic dispatch index comprehensively considers the characteristics of each power generation node, the objective function is often difficult to solve; on the other hand, simplifying the economic dispatch model means a significant reduction in the influencing factors, resulting in inaccurate solution results and low confidence in the dispatch basis.

[0006] In view of the above problems, a new virtual power plant distributed scheduling method and system is urgently needed. Summary of the invention

[0007] In order to address the deficiencies in the prior art, the present invention provides a virtual power plant scheduling method and system with a communication review mechanism under network attacks. The method adopts Byzantine attacks to simulate the failure risks or malicious nodes in the distributed scheduling network, takes the minimum power generation cost as the goal, allows the dual variables and local cost gradients to participate in the update of the real-time power of the nodes, and reduces the communication cost based on the communication review mechanism, providing a basis for virtual power plant scheduling.

[0008] The present invention adopts the following technical solution.

[0009] A first aspect of the present invention relates to a virtual power plant scheduling method with a communication review mechanism, the method comprising the following steps: collecting the real-time power of each distributed power generation node in the virtual power plant, and the connection topology information between the distributed power generation nodes, and obtaining the adjacency matrix of the distributed power generation nodes based on the connection topology information; taking the minimum power generation cost of all distributed power generation nodes as the goal, using Byzantine nodes to simulate abnormal distributed power generation nodes, using the Lagrangian duality method to update the real-time power of the distributed power generation nodes under the premise of the positions of all elements in the reference constraint set, and controlling the communication between nodes through the communication review mechanism to schedule and control the virtual power plant.

[0010] Preferably, the abnormal distributed power generation nodes are simulated with Byzantine nodes, including: using maximum attack or clipped Gaussian attack to attack some nodes in the distributed power generation nodes to modify the power of some nodes and modify some nodes into Byzantine nodes; the attack method is obtained by analyzing the abnormal distribution of distributed power generation nodes in the virtual power plant.

[0011] Preferably, the Lagrangian dual method is used to update the real-time power of the distributed generation nodes under the premise of referring to the positions of all elements in the constraint set, including: instructing each distributed generation node in the virtual power plant to send real-time node information to its neighboring nodes; calculating and reviewing the preparation node information of the current distributed generation node based on the real-time node information of all neighboring nodes at the last iteration; updating the real-time node information and real-time power of the distributed generation nodes based on the preparation node information of the review result, and re-executing the above steps.

[0012] Preferably, the preparation node information of the current distributed generation node is calculated and reviewed based on the real-time node information of all neighboring nodes at the last iteration, including: the preparation node information of the current distributed generation node for:

[0013]

[0014] In the formula, is the real-time node information sent by the jth node after communication review in the previous iteration of the kth time, and node j belongs to the set are all honest nodes judged by node i in the previous iteration, is the real-time power change of distributed generation node i, is the set of nodes in the network topology, for , b is the estimated upper bound of the number of Byzantine neighbors of each honest node.

[0015] Preferably, calculating and reviewing the preliminary node information of a current distributed generation node based on the real-time node information of all neighboring nodes at the last iteration includes: reviewing whether the preliminary node information of the current distributed generation node meets the review conditions and generating a review result; the review conditions are:

[0016]

[0017] In the formula, is the real-time node information sent by the i-th node in the previous k-th iteration, τ k-1 Parameters for review conditions.

[0018] The review results are:

[0019]

[0020] Preferably, based on the review result, the real-time node information and real-time power of the distributed power generation node are updated using the preparation node information, including: if the review result is that the review conditions are met, the preparation node information is assigned to the real-time node information of the distributed power generation node to update the real-time node information of the distributed power generation node; if the review result is that the review conditions are not met, the real-time node information obtained during the last iteration is retained.

[0021] Preferably, based on the review result, the real-time node information and real-time power of the distributed generation node are updated using the preliminary node information, including: the real-time power of the distributed generation node is:

[0022]

[0023] In the formula, is the real-time power of the ith node in the kth iteration,

[0024] is the real-time power of the ith node in the k-1th iteration,

[0025] γ k-1 >0 is the iteration step size in the k-1th iteration process, which decays with the number of iterations k.

[0026] N node is a constant, is the number of distributed generation nodes,

[0027] f i (·) is the local cost function of the ith node, which constitutes part of the objective function.

[0028] is the dual variable of the ith node updated by the dual ascent method in the k-1th iteration,

[0029] v 1 >0 is the regularization parameter,

[0030] is the real-time power lower limit of the i-th node,

[0031] is the real-time power upper limit of the ith node.

[0032] Preferably, the dual variables are updated as follows:

[0033]

[0034] In the formula, is the dual variable of the ith node in the k+1th iteration,

[0035] It is related to the constraints of the objective function.

[0036] v 2 >0 is the regularization parameter.

[0037] The second aspect of the present invention relates to a virtual power plant dispatching system with a communication review mechanism using the method in the first aspect of the present invention, the system comprising a collection module and a control module; wherein the collection module is used to collect the real-time power of each distributed power generation node in the virtual power plant, as well as the connection topology information between the distributed power generation nodes, and obtain the adjacency matrix of the distributed power generation nodes based on the connection topology information; the control module is used to minimize the power generation cost of all distributed power generation nodes, use Byzantine nodes to simulate abnormal distributed power generation nodes, and use dual variables to update the real-time power of honest nodes to perform dispatching control on the virtual power plant.

[0038] The beneficial effect of the present invention is that, compared with the prior art, the present invention provides a virtual power plant dispatching method and system with a communication review mechanism, which uses Byzantine attacks to simulate the risk of failure or malicious nodes in a distributed dispatching network, takes the minimum power generation cost as the goal, and allows the dual variables and local cost gradients to participate in the update of the real-time power of the nodes, providing a basis for virtual power plant dispatching. The present invention is effective and reliable, and judges Byzantine attacks through a distributed algorithm, provides power dispatching involving all power generation nodes under the premise of overcoming risks, and provides the best potential dispatching method for virtual power plants.

[0039] The beneficial effects of the present invention also include:

[0040] 1. After each honest node has reviewed the node information, the method of the present invention sends the node information that meets the review conditions to the neighboring node, which reduces the possibility of the honest node itself being deceived by the Byzantine node and improves the security of the overall decision-making of the network. By adjusting the review method, the limited communication resources occupied in the scheduling process can be effectively reduced under the premise of minimizing the scheduling error.

[0041] 2. In the process of communicating with neighboring nodes, the method of the present invention improves the accuracy of neighboring node information by eliminating several maximum messages and minimum messages. The most accurate neighbor status is obtained through iterative communication and improvement, which improves the consistency of decision-making methods among all honest nodes in the system and provides basic support for the operation of optimal scheduling. In addition, this filtering method significantly reduces the number of communications between nodes and improves communication efficiency on the one hand, and on the other hand, it achieves the advantage of being able to achieve correct economic scheduling under Byzantine attacks, improves the robustness of virtual power plants against Byzantine attacks, and has strong practical significance.

[0042] 3. The present invention uses the Lagrangian dual method to iteratively calculate the real-time power of each distributed generation node. In the dual process and the target solution process, the constraint set of the original power variable and the position of each element in the projection set of the dual variable are considered respectively, and a more accurate optimal iterative path of the dual variable and the original power variable is fitted by multiplication. This process eliminates the contradiction between the complexity and solvability of the economic dispatch model, and ensures that the calculation process follows the optimal solution. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a flow chart of a virtual power plant dispatching method with a communication review mechanism in the present invention;

[0044] Figure 2 A schematic diagram of a network topology structure of distributed power generation nodes of a virtual power plant in a virtual power plant scheduling method with a communication review mechanism of the present invention;

[0045] Figure 3 It is a schematic diagram of an iterative process curve of output power under different review conditions when simulating a maximum value attack in a virtual power plant scheduling method with a communication review mechanism of the present invention;

[0046] Figure 4 It is a schematic diagram of an iterative process curve of the number of node communications under different review conditions when simulating a maximum value attack in a virtual power plant scheduling method with a communication review mechanism of the present invention;

[0047] Figure 5 It is a schematic diagram of the iterative process curve of the error under different review conditions when simulating the maximum attack in a virtual power plant scheduling method with a communication review mechanism;

[0048] Figure 6 It is a schematic diagram of the iterative process curve of output power under different review conditions when simulating the clipping of Gaussian attacks in a virtual power plant scheduling method with a communication review mechanism;

[0049] Figure 7It is a schematic diagram of the iterative process curve of the number of node communications under different review conditions when simulating the clipped Gaussian attack in a virtual power plant scheduling method with a communication review mechanism;

[0050] Figure 8 It is a schematic diagram of the iterative process curve of the error under different review conditions when simulating the clipping of Gaussian attacks in a virtual power plant scheduling method with a communication review mechanism;

[0051] Fig. 9 A schematic diagram of the module structure of a virtual power plant dispatching system with a communication review mechanism in the present invention. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical scheme and advantages of the present invention clearer, the technical scheme of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. The embodiments described in the present invention are only embodiments of a part of the present invention, not all embodiments. Based on the spirit of the present invention, all other embodiments not recorded in the present invention obtained by ordinary technicians in this field according to the embodiments recorded in the present invention without making creative work should belong to the protection scope of the present invention.

[0053] Figure 1 FIG. 1 is a flow chart of a virtual power plant dispatching method with a communication review mechanism in the present invention. Figure 1 As shown, the first aspect of the present invention relates to a virtual power plant scheduling method with a communication review mechanism, and the method includes step 1 and step 2.

[0054] Step 1: collect the real-time power of each distributed generation node in the virtual power plant and the connection topology information between the distributed generation nodes, and obtain the adjacency matrix of the distributed generation nodes according to the connection topology information.

[0055] In order to simulate the decentralized economic dispatch process of virtual power plants based on event triggering under Byzantine attacks, it is necessary to obtain the information of neighboring nodes in a decentralized way. First, the distributed power generation network environment in the virtual power plant is modeled, and a node number i is assigned to each distributed power generation node. At the same time, the node communication method is used to obtain the neighbor nodes of each distributed power generation node, so as to obtain the communication connection relationship between the node and the neighbor node.

[0056] On this basis, the real-time power of each distributed power generation node in the virtual power plant is collected. Before the Byzantine attack and calculation are implemented, the real-time power collected is recorded as the initialization node power. According to the actual situation of each distributed generation node, the power constraint condition of the distributed generation node is obtained as follows:

[0057]

[0058] In the above formula, the distributed generation node numbering meets i=1,…,N node . N node is the total number of distributed generation nodes in the virtual power plant. and They are respectively the lower and upper limits of the power of a current distributed generation node.

[0059] The adjacency matrix between nodes can be obtained through the communication process between nodes. The dimension of this adjacency matrix is ​​related to the number of nodes. The row number and column number represent two different nodes respectively. The elements in the matrix are 0 or 1, which are used to represent whether there is an adjacency relationship between two nodes.

[0060] Figure 2 Schematic diagram of the network topology structure of distributed power generation nodes in a virtual power plant scheduling method with a communication review mechanism of the present invention. Figure 2 As shown, through the adjacency relationship and the pre-assigned numbers of distributed generation nodes, the network graph structure of the virtual electric field distributed generation nodes can be obtained. In the above embodiment, the 54 distributed generation nodes are associated with each other through regional adjacency and multiple adjacency between regions.

[0061] The above content provides the computational basis for the Byzantine attack simulation process in step 2 and the scheduling and control operation of the virtual electric field.

[0062] Step 2, with the goal of minimizing the power generation cost of all distributed generation nodes, Byzantine nodes are used to simulate abnormal distributed generation nodes, and the Lagrangian duality method is used to update the real-time power of distributed generation nodes under the premise of the positions of all elements in the reference constraint set to dispatch and control the virtual power plant.

[0063] After obtaining the basic data of distributed generation nodes, an economic model is constructed through objective functions and constraints to solve the optimal scheduling of multiple nodes. The scheduling goal is the optimal power of each node, so the objective function and constraints are based on the real-time power settings of each node.

[0064] In the present invention, the objective function and constraints of the distributed generation economic dispatch model in the virtual power plant are:

[0065]

[0066] The objective function is the local cost f of all nodes i (·) function, where the node is the set of virtual nodes in the virtual power plant after excluding Byzantine nodes. P is Pi A multidimensional vector. It is the average value of the sum of resources of all nodes, constraining the average upper limit of the output power of multiple distributed generation nodes. is the power constraint of the node.

[0067] During the calculation process, according to the information of the adjacency matrix, N is obtained. node In an undirected graph, the set of edges is represented by If there is a Then it is judged that nodes i and j are neighbors and support mutual communication. Therefore, for any node i, the set of its neighbor nodes is The degree of node i is the number of neighboring nodes An undirected graph consisting of nodes and edges is connected.

[0068] Byzantine attacks are used to simulate the failures, risks, and malicious attacks in virtual power plants. These situations are different from the pre-scheduling method of virtual power plants and the prediction of power consumption of power users in power plants.

[0069] Preferably, Byzantine nodes are used to simulate abnormal distributed power generation nodes, including: using maximum attack or clipped Gaussian attack to attack some of the distributed power generation nodes to modify the power of some nodes and modify some nodes to the Byzantine nodes; the attack method is obtained by analyzing the abnormal distribution of distributed power generation nodes in the virtual power plant.

[0070] In the present invention, it is supported to implement the simulation of such an attack by collecting the historical power consumption deviation in the virtual power plant. The historical power consumption deviation can be obtained by the statistics of the historical power supply and power consumption deviation of the virtual power plant. Such insufficient or excessive power supply will not only cause economic losses to the virtual power plant, but also may cause potential operational risks of the virtual power plant. It is easy to think that by collecting the characteristics of the deviation, it is possible to obtain mutations of certain power users and mutations of certain power generation nodes. On this basis, the simulation of the attack can be obtained according to the frequency of mutations, the number of nodes, and the power scale.

[0071] The maximum attack and the clipped Gaussian attack in the present invention are both obtained based on the above historical analysis process. In practical applications, other types of attacks can also be simulated according to the attack conditions of malicious nodes, the failures and risks of distributed power generation equipment.

[0072] In one embodiment, the maximum attack is implemented by randomly selecting a node in the virtual power plant as a Byzantine node. The real-time power information sent by the node to other adjacent nodes is a maximum value predetermined by the maximum attack method. In another embodiment, the clipped Gaussian attack is also implemented by randomly selecting a node. During the attack, the message sent by the Byzantine node conforms to a preset Gaussian distribution. The message is clipped before being sent, and only data within the preset power range is sent to the neighboring node.

[0073] Preferably, the Lagrangian dual method is used to update the real-time power of the distributed generation nodes under the premise of referring to the positions of all elements in the constraint set, including: instructing each distributed generation node in the virtual power plant to send real-time node information to its neighboring nodes; calculating and reviewing the preparation node information of the current distributed generation node based on the real-time node information of all neighboring nodes at the last iteration; based on the review results, using the preparation node information to update the real-time node information and real-time power of the distributed generation node, and re-executing the above steps.

[0074] Each node sends real-time node information to its neighbor nodes to inform them of its own status. The neighbor nodes then determine whether the node is trustworthy by summarizing all the real-time node information. The communication process and the adjustment process of the real-time node information after each communication are iterated multiple times to gradually update the node information and the power of the honest nodes.

[0075] After each node sends out the real-time node information, it will also collect the real-time node information from all neighboring nodes. Therefore, each node will update its own real-time node information according to the collected real-time node information of all neighboring nodes in each iteration to obtain a preliminary node information. This preliminary node information may not be accurate, so it will be sent to other neighboring nodes as the real-time node information of the next iteration after review.

[0076] In one embodiment, the preparation node information of the current distributed generation node is calculated and reviewed based on the real-time node information of all neighboring nodes at the last iteration, including: the preparation node information of the current distributed generation node for:

[0077]

[0078] In the formula, is the real-time node information sent by the jth node after communication review in the previous iteration of the kth time, and node j belongs to the set are all honest nodes judged by node i in the previous iteration,

[0079] is the real-time power change of distributed generation node i,

[0080] is the set of nodes in the network topology, for , b is the estimated upper bound of the number of Byzantine neighbors of each honest node.

[0081] In the above formula, the current node will first exclude the b largest message nodes and b smallest message nodes among the d neighbor nodes. In this way, the real-time node information of the remaining nodes including the current node itself is obtained, and the sum of these node information is added to the power change degree of the node in the last iteration. Thus, the preparation node information of the current node is obtained. The real-time node information updated here is actually related to the power that the node can assume to send to the neighboring node in the next iteration process, which represents the constraint of the transfer power.

[0082] Preferably, calculating and reviewing the preliminary node information of a current distributed generation node based on the real-time node information of all neighboring nodes at the last iteration includes: reviewing whether the preliminary node information of the current distributed generation node meets the review conditions and generating a review result; the review conditions are:

[0083]

[0084] In the formula, is the real-time node information sent by the i-th node in the previous k-th iteration.

[0085] τ k-1 Parameters for review conditions.

[0086] The review results are:

[0087]

[0088] In the above formula, the parameter τ k It can be a quantity that changes with the number of iterations, such as a non-negative non-increasing sequence. In one embodiment, τ k-1 =α×ρ k-1 , α and ρ are constants, and α>0, 0<ρ<1, then 0≤τ k ≤…≤τ 0 .

[0089] After obtaining the preliminary node information, it is not directly converted into the real-time node information of the next iteration process. It is necessary to review and determine whether the preliminary node information meets the power transfer constraints between nodes in the iterative change process. In the above conditions, It represents the degree of change of node information between two consecutive iterations. If the degree of change is higher than the constraint condition, communication between nodes is implemented. Otherwise, in order to save node communication resources, such change is ignored.

[0090] This algorithm will eventually converge to the neighborhood of the optimal value, and the error from the optimal value is related to the parameters α and ρ of the review function. If α is too large, the algorithm will not converge directly, and the value of ρ directly affects the decay rate of the review threshold function. The closer ρ is to 1, the stricter the communication review is, and the fewer the number of communications between nodes is, resulting in a greater deviation from the saddle point. On the contrary, the closer ρ is to 0, the looser the communication review is, the more frequent the communication between nodes is, and the smaller the deviation from the optimal value is. Therefore, the reduction of communication cost means the loss of some accuracy, which requires a compromise.

[0091] Preferably, based on the review result, the real-time node information and real-time power of the distributed generation node are updated using the reserve node information, including: if the review result is in compliance with the review condition, the reserve node information is updated. Provide real-time node information to distributed generation nodes To update the real-time node information of the distributed generation node; if the review result is not in compliance with the review conditions, the real-time node information obtained in the last iteration is retained.

[0092] Preferably, based on the review result, the real-time node information and real-time power of the distributed generation node are updated using the preliminary node information, including: the real-time power of the distributed generation node is:

[0093]

[0094] In the formula, is the real-time power of the ith node in the kth iteration,

[0095] is the real-time power of the ith node in the k-1th iteration,

[0096] γ k-1 >0 is the iteration step size in the k-1th iteration process, which decays with the number of iterations k.

[0097] N node is a constant, is the number of distributed generation nodes,

[0098] f i (·) is the local cost function of the ith node, which constitutes part of the objective function.

[0099] is the dual variable of the ith node updated by the dual ascent method in the k-1th iteration,

[0100] v 1 >0 is the regularization parameter,

[0101] is the real-time power lower limit of the i-th node,

[0102] is the real-time power upper limit of the ith node.

[0103] In the above formula, the power of each iteration change is obtained based on the gradient descent of the local cost function and the gradient ascent of the dual variable.

[0104] Preferably, the dual variables are updated as follows:

[0105]

[0106] In the formula, is the dual variable of the ith node in the k+1th iteration,

[0107] It is related to the constraints of the objective function.

[0108] v 2 >0 is the regularization parameter.

[0109] Similarly, the dual variable is fitted with the update process of the dual variable in the dual variable based on the gradient ascent. For multiple nodes, the feasible domain of the economic dispatch problem is not empty, and the local cost function of each node is a convex function, and The gradient satisfies the Lipschitz continuity with a constant L, and the learning rate is designed accordingly.

[0110] In this way, in each iteration, only the real-time node information is updated based on the honest nodes, and the dual variables are solved based on the real-time node information, and finally the power scheduling method is obtained. The above process is continuously iterated, and the nodes in the network can be divided into real nodes and Byzantine nodes. Here, it is considered that the Byzantine nodes may not comply with the protocol rules and send wrong or contradictory information to other nodes.

[0111] The implementation process of the above method will be further described below with an example. First, the node information of the distributed power generation nodes in the virtual power plant is collected to build a network topology diagram of 54 nodes. The maximum attack and the clipped Gaussian attack are added to the Byzantine nodes respectively. For the parameter γ k , it is agreed to be set to 0.5 in each iteration, or set to the change amount. To simplify the calculation, v 1 and v 2 The value is 0. Set three different review functions to judge the algorithm effect based on whether the review is strict or not. The parameters of the three review functions are and

[0112] Compare the performance of the algorithms under different review functions, and define the error performance index as:

[0113]

[0114] In the formula, P(k) and P(0) are the kth iteration and initial power respectively. By calculating the multiple powers under the current iteration and the actual load or optimal power P * The difference between the two is used to determine the percentage of the error of the current iteration relative to the initial error.

[0115] In this embodiment, the local cost function of the power generation of each node is characterized as:

[0116]

[0117] According to the local cost constraint, the constraint range of each coefficient in the above formula is c i ∈

[0118] [6.78,74.33],b i ∈[8.3391,37.6968],a i ∈[0.0024,0.0697]. Parameter P i Satisfy constraints And there is The total load that the virtual power plant can carry is 6000kW. In the embodiment of the present invention, only one Byzantine node is selected, so the number of honest nodes is 53, and the number of Byzantine nodes is 1. Substitute the above parameters into the objective function to solve the objective function.

[0119] Using the above method, the power constraints and initial power of each distributed generation node are collected, and the local cost function of each node is calculated based on historical data, as shown in Table 1.

[0120]

[0121]

[0122] Table 1 Distributed generation node parameters

[0123] According to different attack forms and different review condition parameters, the optimization results of the economic dispatch of the virtual power plant, that is, the optimal output power of each distributed generation, are shown in Table 2.

[0124]

[0125]

[0126] Table 2 Optimal output power of distributed generation nodes

[0127] In the above attack, the maximum value of the maximum attack is 420kW.

[0128] Figure 3 It is a schematic diagram of the iterative process curve of output power under different review conditions when simulating a maximum value attack in a virtual power plant dispatching method with a communication review mechanism of the present invention. Figure 3 As shown, as the number of iterations increases, the output power region of the Byzantine node becomes stable, and when the review method becomes stricter, the output power deviation becomes smaller. The method of the present invention is generally superior to other scheduling methods.

[0129] Figure 4 It is a schematic diagram of the iterative process curve of the number of node communications under different review conditions when simulating a maximum attack in a virtual power plant scheduling method with a communication review mechanism of the present invention. Figure 4 As shown, with the increase of iteration times, if the review conditions are stricter, the increase of communication times curve will slow down.

[0130] Figure 5 The following is a schematic diagram of the iterative process curve of the error under different review conditions when simulating the maximum attack in a virtual power plant scheduling method with a communication review mechanism. Figure 5 As shown in the figure, as the value of ρ increases, the communication review becomes more and more stringent, the number of communications is greatly reduced, and the error ACC(k) gradually increases with the increase of ρ. Therefore, it is necessary to comprehensively consider the reduction of communication cost and the increase of error. Selecting appropriate α,ρ can achieve performance similar to that when the communication review mechanism is not adopted.

[0131] In the clipped Gaussian attack, each Byzantine 1 node is randomly selected as a Byzantine node, and the messages sent by this node conform to the Gaussian distribution with a mean of 200 and a variance of 1100. Then, it is clipped to the range of the power constraint [5,420].

[0132] Figure 6 Schematic diagram of the iterative process curve of output power under different censorship conditions when simulating clipping Gaussian attacks in a virtual power plant scheduling method with a communication censorship mechanism. Figure 7 Schematic diagram of the iterative process curve of the number of node communications under different censorship conditions when simulating a clipped Gaussian attack in a virtual power plant scheduling method with a communication censorship mechanism. Figure 8 The following is a schematic diagram of the iterative process curve of the error under different review conditions when simulating the clipping of Gaussian attacks in a virtual power plant scheduling method with a communication review mechanism. Figure 6-8 As shown in the figure, the impact of review conditions, errors, and iteration times on algorithm performance is similar when using different attack methods.

[0133] Through the simulation of the above two scenarios, the effectiveness of the algorithm proposed in the present invention in the economic dispatch of virtual power plants and the high efficiency of communication between nodes are verified. At the same time, the robust performance under Byzantine attacks is verified, which shows that it has strong practical significance when communication resources are limited and security threats exist.

[0134] Fig. 9 FIG. 1 is a schematic diagram of the module structure of a virtual power plant dispatching system with a communication review mechanism in the present invention. Fig. 9 As shown, the second aspect of the present invention relates to a virtual power plant dispatching system with a communication review mechanism using the method in the first aspect of the present invention, the system includes a collection module and a control module; wherein the collection module is used to collect the real-time power of each distributed power generation node in the virtual power plant, and the connection topology information between the distributed power generation nodes, and obtain the adjacency matrix of the distributed power generation nodes according to the connection topology information; the control module is used to minimize the power generation cost of all distributed power generation nodes, use Byzantine nodes to simulate abnormal distributed power generation nodes, and use dual variables to update the real-time power of honest nodes to perform dispatching control on the virtual power plant.

[0135] A third aspect of the present invention relates to a terminal, comprising a processor and a storage medium; the storage medium is used to store instructions; the processor is used to operate according to the instructions to execute the steps of the method in the first aspect of the present invention.

[0136] A fourth aspect of the present invention relates to a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method in the first aspect of the present invention.

[0137] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents, and any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A virtual power plant dispatching method with a communication review mechanism, characterized in that: The method comprises the following steps: Collecting the real-time power of each distributed power generation node in the virtual power plant and the connection topology information between the distributed power generation nodes, and obtaining the adjacency matrix of the distributed power generation nodes according to the connection topology information; With the goal of minimizing the power generation cost of all the distributed power generation nodes, Byzantine nodes are used to simulate the abnormal distributed power generation nodes. The Lagrangian dual method is used to update the real-time power of the distributed power generation nodes under the premise of the positions of all elements in the reference constraint set, so as to dispatch and control the virtual power plant.

2. A virtual power plant dispatching method with a communication review mechanism according to claim 1, characterized in that: The method of simulating the abnormal distributed generation node with a Byzantine node includes: Attacking some of the distributed generation nodes using a maximum attack or a clipped Gaussian attack to modify the power of some of the nodes and modify the some of the nodes into the Byzantine nodes; The attack method is obtained by analyzing the abnormal distribution of the distributed power generation nodes in the virtual power plant.

3. A virtual power plant dispatching method with a communication review mechanism according to claim 2, characterized in that: The method of updating the real-time power of the distributed generation node by using the Lagrangian dual method under the premise of the positions of all elements in the reference constraint set includes: Instruct each of the distributed power generation nodes in the virtual power plant to send real-time node information to its neighboring nodes; Calculate and review the preliminary node information of the current distributed generation node based on the real-time node information of all neighboring nodes at the last iteration; Based on the review result, the real-time node information and real-time power of the distributed power generation node are updated using the preliminary node information, and the above steps are re-executed.

4. A virtual power plant dispatching method with a communication review mechanism according to claim 3, characterized in that: The calculating and reviewing of the preliminary node information of the current distributed power generation node based on the real-time node information of all neighboring nodes at the last iteration includes: The backup node information of the current distributed generation node for: In the formula, is the real-time node information sent by the jth node after communication review in the previous iteration of the kth time, and node j belongs to the set are all honest nodes judged by node i in the previous iteration, is the real-time power change of the distributed generation node i, is the set of nodes in the network topology, for , b is the estimated upper bound of the number of Byzantine neighbors of each honest node.

5. A virtual power plant dispatching method with a communication review mechanism according to claim 4, characterized in that: The calculating and reviewing of the preliminary node information of the current distributed power generation node based on the real-time node information of all neighboring nodes at the last iteration includes: Reviewing whether the preliminary node information of the current distributed power generation node meets the review conditions and generating a review result; The review conditions are: In the formula, is the real-time node information sent by the i-th node in the previous k-th iteration. τ k-1 Parameters for review conditions.

6. A virtual power plant dispatching method with a communication review mechanism according to claim 5, characterized in that: The updating of the real-time node information and the real-time power of the distributed power generation node using the preliminary node information based on the review result includes: If the review result is that the review condition is met, the provisioning node information is assigned to the real-time node information of the distributed power generation node to update the real-time node information of the distributed power generation node; If the review result is that the review condition is not met, the real-time node information obtained during the last iteration is retained.

7. A virtual power plant dispatching method with a communication review mechanism according to any one of claims 4 to 6, characterized in that: The updating of the real-time node information and the real-time power of the distributed power generation node using the preliminary node information based on the review result includes: The real-time power of the distributed generation node is: In the formula, is the real-time power of the ith node in the kth iteration, is the real-time power of the ith node in the k-1th iteration, γ k-1 >0 is the iteration step size in the k-1th iteration process, which decays with the number of iterations k. N node is a constant, is the number of distributed generation nodes, f i (·) is the local cost function of the ith node, which constitutes part of the objective function. is the dual variable of the ith node updated by the dual ascent method in the k-1th iteration, v1>0 is the regularization parameter, is the real-time power lower limit of the i-th node, is the real-time power upper limit of the ith node.

8. A virtual power plant dispatching method with a communication review mechanism according to claim 7, characterized in that: The dual variables are updated as follows: In the formula, is the dual variable of the ith node in the k+1th iteration, It is related to the constraints of the objective function. v2>0 is the regularization parameter.

9. A virtual power plant dispatching system with a communication review mechanism using the method according to any one of claims 1 to 8, characterized in that: The system includes a collection module and a control module; wherein, The acquisition module is used to collect the real-time power of each distributed power generation node in the virtual power plant and the connection topology information between the distributed power generation nodes, and obtain the adjacency matrix of the distributed power generation nodes according to the connection topology information; The control module is used to minimize the power generation cost of all the distributed power generation nodes, use Byzantine nodes to simulate the abnormal distributed power generation nodes, and use dual variables to update the real-time power of honest nodes to schedule and control the virtual power plant.

Images