Virtual power plant optimization scheduling method and device

Abstract

The invention discloses a virtual power plant optimization scheduling method and device, and the method comprises the steps: obtaining virtual power plant data and power market data; according to the virtual power plant data and the power market data, an optimal scheduling model of the virtual power plant is obtained, and the optimal scheduling model comprises a target function and constraint conditions; whether malicious agent nodes exist in the agent nodes or not is judged, an adjustment strategy is executed when the malicious agent nodes exist, and the adjustment strategy at least comprises the step of adjusting the optimal scheduling model; and solving the optimal scheduling model or the adjusted optimal scheduling model, and distributing the solved scheduling scheme to each agent node for execution. According to the method, the robustness of the system can be improved, so that stable operation of the virtual power plant is ensured.

Description

Technical Field

[0001] The present application relates to the technical field of power system dispatching, and in particular to a virtual power plant optimization dispatching method and device. Background Art

[0002] With the continuous attention and vigorous promotion of renewable energy around the world, the rapid rise of clean energy such as solar energy and wind energy, and the gradual opening of the power market structure and the introduction of competition mechanisms, the strategic significance and technical value of virtual power plants (VPPs) as a key bridge connecting distributed energy resources and power market demand are becoming increasingly prominent. Virtual power plants highly integrate diversified energy resources including distributed power generation devices, energy storage systems, and demand response strategies to build a flexible, adjustable, and intelligently controlled energy network, aiming to improve the comprehensive utilization efficiency of energy and actively participate in the operation and competition of the power market.

[0003] Driven by smart grid technology and distributed energy technology, virtual power plants have become a hot area of ​​current research and application. Its core functions include effective resource aggregation, intelligent scheduling optimization, market trading strategy formulation, and comprehensive risk management. Relying on advanced communication and information technology, virtual power plants can achieve real-time status monitoring, efficient scheduling and remote control of various distributed energy resources, thereby showing significant advantages in reducing costs, improving operational efficiency, and enhancing the stability and security of power systems.

[0004] However, despite the significant progress in virtual power plant technology, existing technologies still face many challenges and limitations in terms of optimal scheduling, which mainly stem from the traditional centralized scheduling method. In other words, in existing technologies, the scheduling of virtual power plants mainly relies on a central control node, which is responsible for collecting information on all distributed energy resources (such as solar power generation, wind power generation, energy storage systems, etc.) and formulating a unified scheduling strategy based on this information. However, this centralized scheduling method has the following significant disadvantages:

[0005] Single target. In the centralized dispatch mode, the setting of dispatch targets is often limited by the original intention of system design and the need for simplification. In order to reduce system complexity and improve computing efficiency, the system usually only selects a single target such as cost minimization or profit maximization for optimization. This single-target optimization strategy ignores multiple targets such as system stability and environmental benefits, resulting in serious impact on the overall operating efficiency and comprehensive benefits of the virtual power plant.

[0006] Insufficient robustness. Since all information on distributed energy resources needs to be aggregated to the central control node for processing, the system is highly dependent on the central node. Once the node fails, suffers a malicious attack, or data is leaked, the scheduling of the entire system will fall into chaos, significantly reducing the robustness of the system, making it difficult to cope with complex and changing power market demands and potential malicious behaviors, seriously affecting the stable operation of the virtual power plant.

[0007] Low solution efficiency. The solution algorithms in the existing technology mostly use general optimization methods, and fail to make customized improvements based on the particularity and complexity of the virtual power plant optimization scheduling problem. When these algorithms deal with large-scale, multi-dimensional, and nonlinear virtual power plant optimization scheduling problems, the solution process may become lengthy and inefficient, making it difficult to meet the needs of real-time scheduling.

[0008] Lack of coordination in strategy execution. In the centralized scheduling mode, the formulation and execution of scheduling strategies are mainly completed by the central control node alone, lacking effective communication and coordination with other distributed energy resources or agent nodes. This lack of coordination in scheduling may lead to inconsistencies or conflicts when distributed energy resources execute scheduling strategies, which seriously affects the overall scheduling effect and operating efficiency of the system.

[0009] In response to the above challenges, the industry and academia are constantly exploring improvement solutions, such as building more sophisticated optimization models, introducing advanced solution algorithms, optimizing communication protocol design, etc. However, most of these efforts have only improved specific problems and have failed to fundamentally and comprehensively solve the multiple challenges faced by virtual power plant optimization scheduling. In particular, in recent years, the application of distributed consensus algorithms (such as Raft, Paxos, etc.) in the field of virtual power plant optimization scheduling has provided a new perspective for solving scheduling consistency problems. These algorithms effectively improve the synchronization and efficiency of scheduling by promoting information sharing and collaborative decision-making among agent nodes. However, existing research has mostly focused on the theoretical discussion and basic implementation of the algorithm. There is still a lack of research on how to deeply integrate it into the complex application scenarios of virtual power plants and how to perform adaptive optimization to overcome specific problems in actual operation.

[0010] In summary, there is an urgent need for an improved virtual power plant optimization scheduling method that can comprehensively consider multiple objectives such as economy and environment, ensure the consistency and efficient execution of scheduling strategies, and especially improve the robustness of the system. Summary of the invention

[0011] In order to overcome the shortcomings of the prior art, the purpose of this application is to provide a virtual power plant optimization scheduling method, device, equipment and computer storage medium, which can improve the robustness of the system to ensure the stable operation of the virtual power plant.

[0012] The purpose of this application is achieved by the following technical solutions:

[0013] In a first aspect, the present application provides a virtual power plant optimization scheduling method, the virtual power plant optimization scheduling method is implemented based on a multi-agent system, the multi-agent system includes multiple agent nodes, each agent node corresponds to a distributed energy resource of the virtual power plant, the method includes: obtaining virtual power plant data and electricity market data, the virtual power plant data includes at least one of the power generation, energy storage status, and equipment failure rate of each agent node, and the electricity market data includes at least one of price information and load forecast data; according to the virtual power plant data and the electricity market data, an optimization scheduling model of the virtual power plant is obtained, the optimization scheduling model includes an objective function and constraints, wherein the objective function is expressed as a form of optimization of multiple objectives The multiple objectives include at least one of cost minimization, benefit maximization, environmental benefit and decision variables, the decision variables include at least one of output power, charge and discharge power, power trading volume, load adjustment volume and auxiliary service provision, and the constraint conditions include at least one of output power constraint, charge and discharge power constraint, load adjustment volume constraint, market trading volume limit constraint, ramp constraint and cycle life constraint; it is determined whether there are malicious proxy nodes in each proxy node, and an adjustment strategy is executed when there are malicious proxy nodes, and the adjustment strategy at least includes: adjusting the optimization scheduling model; solving the optimization scheduling model or the adjusted optimization scheduling model, and distributing the solved scheduling scheme to each proxy node for execution.

[0014] In one possible implementation, the determining whether there is a malicious proxy node among the proxy nodes and executing an adjustment strategy when a malicious proxy node exists includes: determining whether each proxy node is abnormal based on whether a historical behavior of each proxy node is abnormal, and identifying the proxy node determined to be abnormal as a malicious proxy node, the historical behavior including at least one of transaction time, transaction volume, and transaction price; obtaining a threat level of the malicious proxy node based on characteristics of the abnormal historical behavior, and executing an adjustment strategy based on the threat level of the malicious proxy node, the characteristics including at least one of a behavior type, a behavior frequency, and a behavior duration.

[0015] In one possible implementation, the adjusting the optimization scheduling model includes: adjusting constraints, wherein the adjusting constraints include one of increasing the power generation of normal proxy nodes and reducing the power generation of normal proxy nodes.

[0016] In one possible implementation manner, the adjusting the optimization scheduling model further includes: adjusting an objective function, wherein the adjusting the objective function includes adding a risk function to the objective function.

[0017] In one possible implementation manner, the adjustment strategy further includes replacing the proxy node, and the replacing the proxy node includes replacing the malicious proxy node with a backup proxy node.

[0018] In one possible implementation manner, the adjustment strategy further includes regularly backing up data, and the regularly backing up data includes backing up data of the multi-agent system according to a preset period.

[0019] In one possible implementation, the multi-agent system includes at least one of the following agent layers: a virtual power plant agent layer, which is at least used to receive electricity price information from the electricity market and bidding information from virtual power plants at the same layer to adjust the operating strategies of the distributed energy agent layer, the energy storage system agent layer and the load demand agent layer; an electricity market agent layer, which is at least used to collect the electricity purchase and sale willingness and electricity quantity information of the electricity market, and exchange data with the virtual power plant agent layer; a distributed energy agent layer, which is at least used to monitor the output of distributed energy resources and adjust them according to the scheduling strategy of the virtual power plant, wherein the distributed energy resources include at least one of solar power generation and wind power generation; an energy storage system agent layer, which is at least used to monitor the status of the energy storage equipment, wherein the status includes at least one of the charging and discharging strategy and the remaining capacity, and exchange real-time data with the virtual power plant agent layer; a load demand agent layer, which is at least used to receive data related to load demand to predict the load change trend and adjust the load demand according to the scheduling strategy of the virtual power plant.

[0020] In a second aspect, the present application further provides a virtual power plant optimization scheduling device, which is implemented based on a multi-agent system, wherein the multi-agent system includes a plurality of agent nodes, each agent node corresponds to a distributed energy resource of the virtual power plant, and the device includes:

[0021] A data collection module, used to obtain virtual power plant data and power market data, wherein the virtual power plant data includes at least one of power generation, energy storage status, and equipment failure rate of each proxy node, and the power market data includes at least one of price information and load forecast data;

[0022] A model building module is used to obtain an optimal dispatching model of a virtual power plant according to the virtual power plant data and the power market data, wherein the optimal dispatching model includes an objective function and constraints, wherein the objective function is expressed in the form of optimizing multiple objectives, wherein the multiple objectives include at least one of cost minimization, benefit maximization, environmental benefit and decision variables, wherein the decision variables include at least one of output power, charge and discharge power, power trading volume, load adjustment volume and auxiliary service provision volume, and wherein the constraints include at least one of output power constraint, charge and discharge power constraint, load adjustment volume constraint, market transaction volume limit constraint, ramp constraint and cycle life constraint;

[0023] The risk assessment module is used to determine whether there are malicious proxy nodes among the proxy nodes, and to execute an adjustment strategy when there are malicious proxy nodes, wherein the adjustment strategy at least includes: adjusting the optimization scheduling model;

[0024] The model solving module is used to solve the optimization scheduling model or the adjusted optimization scheduling model, and distribute the solved scheduling solution to each proxy node for execution.

[0025] In a third aspect, the present application further provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method described in any one of the first aspects above is implemented.

[0026] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements any of the methods described in the first aspect above.

[0027] Compared with the prior art, the beneficial effects of the present application include at least:

[0028] The virtual power plant optimization scheduling method of the present application is implemented based on a multi-agent system, and the optimization scheduling model comprehensively considers multiple goals such as cost minimization and profit maximization. It executes adjustment strategies by judging the mechanism of malicious agent nodes, such as adjusting the optimization scheduling model, which can effectively enhance the robustness of the system and ensure the stable operation of the virtual power plant.

[0029] In the virtual power plant optimization scheduling method of the present application, by monitoring and analyzing the historical behavior of each agent node, the agent nodes with abnormal behavior are identified, and their threat level is evaluated based on their behavioral characteristics. Adjustment strategies can be implemented in a targeted manner, and malicious behavior can be responded to quickly to reduce its impact on system stability.

[0030] In the virtual power plant optimization scheduling method of the present application, by controlling the power generation of normal proxy nodes as a constraint condition, it is possible to reduce system fluctuations and instability caused by supply and demand imbalance, better maintain the supply and demand balance of the system, and thus ensure the stable operation of the virtual power plant. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The following is a brief introduction to the drawings involved in the embodiments of the present application.

[0032] Figure 1 It is a structural diagram of a multi-agent system for implementing a virtual power plant optimization scheduling method provided by an embodiment of the present application;

[0033] Figure 2is a flow chart of a virtual power plant optimization scheduling method provided by an embodiment of the present application;

[0034] Figure 3 is a flowchart of step S300 of a virtual power plant optimization scheduling method provided by an embodiment of the present application;

[0035] Figure 4 It is a schematic diagram of historical behavior data and key features of an agent node for a virtual power plant optimization scheduling method provided by an embodiment of the present application;

[0036] Figure 5 It is a structural framework diagram of a virtual power plant optimization scheduling device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0037] In order to facilitate the understanding of the present application, the present application will be described more fully below with reference to the relevant drawings. The preferred embodiments of the present application are provided in the drawings. However, the present application can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive.

[0038] Example 1

[0039] An embodiment of the present application provides a virtual power plant optimization scheduling method. The virtual power plant optimization scheduling method can be implemented based on a multi-agent system, and the multi-agent system can include multiple agent nodes, each agent node corresponding to a distributed energy resource of the virtual power plant.

[0040] Among them, Virtual Power Plant (VPP), as a system that integrates and manages Distributed Energy Resources (DERs), can achieve the optimal scheduling and coordinated operation of these resources. The optimal scheduling of virtual power plants based on Multi-Agent System (MAS) has the advantages of high efficiency and flexibility.

[0041] A multi-agent system can be a collection of multiple agent nodes with autonomous decision-making capabilities. Each agent node can sense the environment, make decisions, execute actions, and communicate and collaborate with other agent nodes to jointly complete complex tasks. In the application scenario of a virtual power plant, each agent node can represent a distributed energy resource, such as a solar power generation system, a wind power generation system, an energy storage device, an electric vehicle charging and discharging device, etc.

[0042] The proxy nodes can have the following functions: each proxy node can make decisions independently according to the characteristics and current status of the distributed energy resources it represents; the proxy nodes can respond quickly to external changes (such as electricity price fluctuations, weather changes, changes in user demand, etc.); the proxy nodes can achieve optimal resource allocation and coordinated operation through communication and negotiation; the proxy nodes can learn and adapt to the environment and continuously optimize their decision-making strategies.

[0043] The multi-agent system may include one or more of the following agent layers:

[0044] The virtual power plant agent layer is at least used to receive electricity price information from the power market and bidding information from virtual power plants at the same layer to adjust the operating strategies of the distributed energy agent layer, energy storage system agent layer and load demand agent layer. Specifically, as the "brain" of the entire multi-agent system, the virtual power plant agent layer is responsible for integrating, analyzing and processing information from other agent layers and formulating the global optimal scheduling strategy. By receiving electricity price information from the power market and bidding information from virtual power plants at the same layer, the virtual power plant agent layer can understand market dynamics and adjust the operating strategies of each agent layer to maximize economic benefits or meet specific operating goals. Based on the collected information, the virtual power plant agent layer can dynamically adjust the operating strategies of the distributed energy agent layer, energy storage system agent layer and load demand agent layer to ensure efficient and stable operation of the entire virtual power plant.

[0045] The power market agent layer is at least used to collect the willingness to buy and sell electricity and electricity information in the power market, and to exchange data with the virtual power plant agent layer. Among them, the willingness to buy and sell electricity refers to the willingness to buy or sell electricity. Specifically, the power market agent layer is responsible for collecting the willingness to buy or sell electricity, electricity information, and electricity price trends in the power market, and providing decision support for the virtual power plant agent layer. The power market agent layer also maintains real-time data exchange with the virtual power plant agent layer to ensure that the virtual power plant can adjust its strategy in a timely manner according to market changes.

[0046] The distributed energy proxy layer is at least used to monitor the output of distributed energy resources and adjust them according to the scheduling strategy of the virtual power plant. The distributed energy resources include at least one of solar power generation and wind power generation. Specifically, the distributed energy proxy layer monitors the output of distributed energy resources such as solar power generation and wind power generation in real time, and adjusts the output according to the scheduling strategy of the virtual power plant to ensure efficient use of energy and balance between supply and demand, and, through collaboration with other proxy layers, achieves optimal configuration and maximum utilization of distributed energy resources.

[0047] The energy storage system proxy layer is at least used to monitor the status of the energy storage device, which includes at least one of the charging and discharging strategy and the remaining capacity, and to exchange data with the virtual power plant proxy layer in real time. Specifically, the energy storage system proxy layer monitors the key status information such as the charging and discharging strategy and the remaining capacity of the energy storage device in real time, and provides accurate energy storage resource information to the virtual power plant proxy layer. It also exchanges data with the virtual power plant proxy layer in real time to ensure that the energy storage device can perform charging and discharging operations according to the global scheduling strategy to support the stable operation of the power system and the efficient use of energy.

[0048] The load demand proxy layer is at least used to receive data related to load demand to predict the changing trend of load and adjust the load demand according to the scheduling strategy of the virtual power plant. Specifically, the load demand proxy layer receives data related to load demand (such as historical load data, weather forecasts, etc.) and uses these data to predict the changing trend of load. And according to the scheduling strategy of the virtual power plant and the load forecast results, the load demand is adjusted to achieve supply and demand balance and efficient use of energy. When necessary, by implementing demand response measures (such as peak and valley time-of-use electricity prices, demand-side management, etc.), guide users to reasonably adjust their electricity consumption behavior and further optimize the allocation of power resources.

[0049] The virtual power plant optimization scheduling method based on multi-agent system realizes the comprehensive management and optimization scheduling of distributed energy resources, energy storage systems, load demand and power market through hierarchical design and cooperation and coordination between agent layers. This method can not only improve the flexibility and reliability of the power system, but also promote the efficient use and sustainable development of energy.

[0050] Figure 1 The structure diagram of a multi-agent system for implementing a virtual power plant optimization scheduling method provided by an embodiment of the present application is shown. Figure 1 As shown, the multi-agent system includes:

[0051] The VPP agent layer, namely the virtual power plant agent layer, is responsible for receiving the electricity price information issued by the power market agent layer and the bidding information of other VPPs in the same layer. At the same time, it formulates and feeds back corresponding strategies to the upper agent layer based on its own power generation capacity and load forecast data. Its core component is the VPP Central Control System (VPP Central Controller, VPPCC), which is responsible for summarizing and adjusting the operation strategies of the lower-level agents (including DER, ESS), as well as the load demand response (Load Demand Response, LDR) strategy.

[0052] The power market agent layer mainly collects the willingness to purchase and sell power in the power market, as well as related power information. By exchanging data and information with the VPP agent layer, it provides market basis for VPP scheduling decisions.

[0053] DER agent layer, that is, distributed energy agent layer, this layer represents distributed energy resources (Distributed Energy Resources, DER) and exchanges information with the VPP agent layer. During the VPP scheduling process, it is responsible for regulating the output of each DER to respond to the VPP scheduling instructions.

[0054] The ESS agent layer, or energy storage system agent layer, focuses on monitoring and managing the status of the energy storage system (ESS), including key parameters such as the energy storage charging and discharging strategy and remaining capacity. It maintains real-time data exchange with the VPP agent layer to ensure that the contribution and constraints of the energy storage system are fully considered during the scheduling process.

[0055] The LDR agent layer, or load demand agent layer, is responsible for receiving and analyzing load demand response (LDR) signals. This layer has the ability to predict load change trends and flexibly adjust load demand according to the VPP's dispatch strategy to achieve a balance between power supply and demand.

[0056] In the multi-agent system of the present application, a distributed decision-making mechanism and a dynamic adjustment mechanism may be adopted.

[0057] in,

[0058] The distributed decision-making mechanism can rely on each agent node in the multi-agent system (MAS) to make independent decisions and achieve global optimization through efficient collaborative control algorithms. Every distributed resource, whether it is photovoltaic power generation, wind power generation or energy storage system, and even every load node, is given the identity of an agent node in this system.

[0059] These proxy nodes have excellent information processing and decision-making capabilities. They can quickly make the most intelligent decisions based on the local information and real-time data they have, such as the current status of energy resources, specific load requirements, and market price dynamics. In order to promote the comprehensive circulation and sharing of information, the proxy nodes are closely connected through advanced communication networks to ensure that each node can accurately grasp the overall situation and make more reasonable and forward-looking decisions.

[0060] Under the framework of MAS, each agent node relies on collaborative control algorithms to work together towards the goal of global optimization. These algorithms can flexibly and dynamically adjust the scheduling strategy based on the real-time data and feedback information provided by the agent nodes, striving to achieve the optimal state in energy distribution and load balance, so as to enable the efficient and stable operation of the virtual power plant.

[0061] The dynamic adjustment mechanism enables the agent nodes to be given real-time monitoring capabilities, which can instantly capture critical information such as the current status of energy resources and subtle changes in load demand. This information is then quickly transmitted to other agent nodes and the central controller through the communication network, providing data support for the decision-making of the entire system.

[0062] Based on the data obtained from these real-time monitoring and the feedback information from other nodes, the proxy node can flexibly and dynamically adjust the scheduling strategy. This adjustment can be fine-tuning for local conditions to ensure the optimal operation of a single resource or load node, or it can be a global strategy reconstruction to achieve the optimization of energy distribution and load balance of the entire virtual power plant.

[0063] It should be pointed out that under the framework of MAS (multi-agent system), agent nodes also have the ability of adaptive learning. They can deeply analyze historical data and real-time data to dig out potential optimization space and opportunities. Through continuous learning, agent nodes can gradually optimize their scheduling strategies and further improve scheduling efficiency.

[0064] Figure 2 FIG. 1 is a flow chart of a virtual power plant optimization scheduling method provided by an embodiment of the present application. Figure 2 As shown, the virtual power plant optimization scheduling method provided by the embodiment of the present application may include:

[0065] Step S100, obtain virtual power plant data and electricity market data, the virtual power plant data includes at least one of the power generation, energy storage status, and equipment failure rate of each proxy node, and the electricity market data includes at least one of price information and load forecast data. Specifically, the virtual power plant data covers the key information of each proxy node, such as power generation, energy storage status, equipment failure rate, etc. The power generation reflects the current and future power generation potential, the energy storage status may include the remaining capacity and the charge and discharge status, and the equipment failure rate is used to evaluate the reliability and availability of the equipment. These data are the basis for formulating scheduling strategies. The electricity market data includes price information, load forecast data, etc. The price information specifically includes real-time electricity prices, peak and valley electricity prices, etc., which have a direct impact on the economy of the scheduling strategy. The load forecast data is used to predict future electricity demand, thereby adjusting the power generation and energy storage strategies. These data can be used for virtual power plants to better adapt to market changes.

[0066] Step S200, based on the virtual power plant data and the electricity market data, obtain an optimization scheduling model of the virtual power plant, the optimization scheduling model includes an objective function and constraints, wherein the objective function is expressed in the form of optimization of multiple objectives, the multiple objectives include cost minimization, benefit maximization, environmental benefits and at least one of decision variables, the decision variables include output power, charging and discharging power, power trading volume, load adjustment, and auxiliary service provision. At least one of the constraints includes output power constraints, charging and discharging power constraints, load adjustment constraints, market trading volume limit constraints, ramp constraints, and cycle life constraints.

[0067] Among them, the objective function is expressed in the form of optimizing multiple objectives, such as cost minimization, benefit maximization, environmental benefits and decision variables. Cost minimization specifically considers power generation costs and energy storage costs; benefit maximization is achieved by obtaining benefits through power trading, ancillary services and other means; environmental benefits include reducing carbon emissions, etc.; decision variables can include output power, charging and discharging power, power trading volume, load adjustment, and ancillary service provision. Among them, charging and discharging power determines the charging and discharging behavior of energy storage equipment, power trading volume is the amount of electricity bought and sold with the market, load adjustment is to adjust the load through demand response and other means, and ancillary service provision includes frequency regulation, voltage regulation, etc.

[0068] These objectives may conflict with each other and can be weighed using multi-objective optimization algorithms.

[0069] In a specific implementation, it can be assumed that the dispatch interval consists of T consecutive time periods, and t∈{1,2,…,T} can represent a certain moment in the dispatch interval. The optimal dispatch of a virtual power plant involves multiple objectives, including cost minimization and environmental benefits, which can be expressed as:

[0070] min F(x)=[C(x)-R(x),-EnvBenefit(x)]

[0071] Where: C(x) represents the total cost function, R(x) represents the total benefit function, EnvBenefit(x) represents the environmental benefit function, and x is the decision variable vector.

[0072] Cost minimization is one of the primary goals of virtual power plant operation. The total cost function C(x) can be expressed as:

[0073] C(x)=C DER (x)+C buy (x)+C ESS (x)+C LDR (x)

[0074]

[0075] in:

[0076] C DER (x), C buy (x), C ESS (x), C LDR (x) are the generation cost of distributed energy resources (DER), the electricity purchase cost, the charging and discharging depreciation cost of the energy storage system (ESS), and the incentive cost of load demand response (LDR);

[0077] N DER is the number of DERs, c DER,i is the unit electricity generation cost of the i-th DER, P DER,i (t) is the output power of the i-th DER at time t;

[0078] λ buy (t) is the electricity purchase price at time t, P buy (t) is the electricity purchase amount at time t;

[0079] are the unit charging depreciation cost and unit discharging depreciation cost of ESS respectively, is the charging power and discharging power of ESS at time t;

[0080] λ LDR (t) is the incentive price of LDR at time t, ΔP LDR (t) is the load adjustment at time t.

[0081] Revenue maximization is achieved by virtual power plants through participating in electricity market transactions and providing ancillary services. The total revenue function R(x) can be expressed as:

[0082] R(x)=R sell (x)+R ancillary (x)

[0083]

[0084]

[0085] in:

[0086] R sell (x), R ancillary (x) are the revenue from the sale of electricity and the revenue from the provision of ancillary services;

[0087] λ sell (t) is the electricity price at time t, Psell (t) is the electricity sales volume at time t;

[0088] S(t) is the subsidy income for providing ancillary services at time t, N ancilalry is the number of auxiliary services provided, π j (t) is the unit price for providing the jth ancillary service, Q j (t) is the provision of the jth ancillary service.

[0089] Environmental benefits can be the impact of virtual power plants on the environment by reducing greenhouse gas emissions (mainly carbon dioxide here), which can be expressed as:

[0090]

[0091] Among them, e DER,i is the carbon emission factor of the i-th DER, e grid is the carbon emission factor of the power grid. It is assumed here that the electricity purchased by the virtual power plant from the power grid will also indirectly generate carbon emissions.

[0092] The decision variable vector x contains the parameters or strategies that can be adjusted in the optimal scheduling of the virtual power plant, including the following variables:

[0093] DER output power:

[0094] Charging and discharging power of ESS equipment:

[0095] Electricity trading volume:

[0096] LDR load adjustment:

[0097] Amount of auxiliary services provided:

[0098] Constraints can be power limits, power market rules, grid security constraints, etc., and can specifically include output power constraints, charge and discharge power constraints, load adjustment constraints, market transaction volume limit constraints, ramp constraints, cycle life constraints, etc. Among them, output power constraints include the output range of generator sets, charge and discharge power constraints include the charge and discharge rate and capacity limits of energy storage devices, load adjustment constraints include the load adjustment range acceptable to users, market transaction volume limit constraints include the transaction rules of the power market, ramp constraints include the ramp rate limits of generator sets, and cycle life constraints include the cycle life limits of energy storage devices. These constraints ensure the feasibility and safety of the scheduling strategy.

[0099] In a specific implementation, to ensure the feasibility of optimal scheduling, the following constraints may be introduced:

[0100] The output power constraint can be expressed as:

[0101]

[0102] in, are the minimum and maximum output powers of the i-th DER.

[0103] The charge and discharge power constraints can be expressed as:

[0104]

[0105] SoC min ≤SoC(t)≤SoC max

[0106]

[0107] in:

[0108] are the minimum and maximum charging power of ESS respectively;

[0109] are the minimum and maximum discharge powers of ESS respectively;

[0110] SoC(t) is the charge state of ESS at time t. min 、SoC max are the minimum and maximum permissible state of charge of the ESS;

[0111] Q ESS is the capacity of ESS, η ch , η dch are the charging and discharging efficiencies of ESS, respectively, and δ(t) is the power loss of ESS at time t.

[0112] The load adjustment constraint can be expressed as:

[0113]

[0114] in, are the minimum and maximum load adjustments.

[0115] The power balance constraint can be expressed as:

[0116]

[0117] Where D(t) is the load demand at time t.

[0118] The market transaction volume limit constraint can be expressed as:

[0119]

[0120] in, They are the lower and upper limits of the purchase transaction volume respectively; They are the lower and upper limits of the sold transaction volume respectively.

[0121] The climbing constraint can be expressed as:

[0122]

[0123] in, are the downward and upward ramp rate limits of the i-th DER.

[0124] The cycle life constraint can be expressed as:

[0125]

[0126] in, It is the maximum number of charge and discharge cycles allowed within the scheduling interval.

[0127] Step S300, determining whether there is a malicious proxy node among the proxy nodes, and executing an adjustment strategy when a malicious proxy node exists, the adjustment strategy at least comprising: adjusting the optimization scheduling model.

[0128] By monitoring the behavior and data of the proxy nodes, it is possible to determine whether there is malicious behavior or abnormal data, which helps maintain the stability and security of the virtual power plant.

[0129] Specifically, if a malicious proxy node is detected, a variety of adjustment strategies can be adopted.

[0130] For example, the adjustment strategy may include adjusting the optimization scheduling model, that is, considering the influence of malicious nodes and re-formulating the scheduling strategy.

[0131] For another example, the adjustment strategy may include replacing proxy nodes, and the replacement of proxy nodes may specifically include: replacing malicious proxy nodes with standby proxy nodes. In other words, replacing malicious proxy nodes means using standby proxy nodes to replace malicious proxy nodes to ensure the normal operation of the system. Specifically, a group of safety-verified standby proxy nodes may be prepared in advance to ensure that they meet system requirements in terms of performance, stability, and security. When a malicious proxy node is discovered, the traffic is immediately migrated from the node to the standby node, and the network configuration is updated to point to the new proxy node. Ensure that the standby proxy node can seamlessly take over the tasks of the malicious node to maintain the continuity and stability of the system.

[0132] In other embodiments, the adjustment strategy may also include regular data backup, and the regular data backup includes backing up the data of the multi-agent system according to a preset period. The preset period can be set according to the actual situation and needs of the system. This period can be defined daily, weekly, monthly, or according to specific business logic. For example, for critical business data, it may be necessary to back up every day or even every hour; for less important data, a longer backup period can be selected. As an additional safety measure, regular backing up of the data of the multi-agent system can restore the system status in the event of an unexpected situation.

[0133] Step S400, solving the optimization scheduling model or the adjusted optimization scheduling model, and distributing the solved scheduling solution to each proxy node for execution.

[0134] Optimization algorithms, such as the NSGA-II algorithm, can be used to solve the optimization scheduling model and obtain the optimal or suboptimal scheduling solution. The NSGA-II algorithm, namely the Non-dominated Sorting Genetic Algorithm II, is a multi-objective genetic algorithm that can be used to solve complex optimization problems involving multiple conflicting objectives (such as cost minimization and revenue maximization).

[0135] It should be noted that the NSGA-II algorithm can be used to solve the multi-objective optimization problem of virtual power plant optimal scheduling constructed in this application. As a widely used evolutionary algorithm, NSGA-II performs well in dealing with complex multi-objective optimization problems. It can effectively find the Pareto frontier in the multi-objective space and provide a set of compromise solutions that balance the various objectives. Since the NSGA-II algorithm is relatively mature, it will not be described in detail.

[0136] The solved dispatching scheme is distributed to each agent node for execution, guiding them to perform power generation, energy storage, trading and other behaviors to ensure the coordinated operation and overall optimization of the various components of the virtual power plant. Specifically, a stable and efficient communication mechanism can be established to ensure that the dispatching center can accurately and timely send the dispatching scheme to each agent node. A unified data format and transmission protocol can be formulated so that the agent node can correctly parse and receive the dispatching scheme. A confirmation mechanism can be implemented to ensure that the agent node has successfully received and understood the dispatching scheme.

[0137] Preferably, a distributed consensus algorithm can be used to synchronize the solved scheduling plan to each proxy node to ensure that the scheduling plan can be generated and executed in the presence of malicious proxy nodes.

[0138] The distributed consensus algorithm enables multiple nodes in a distributed system to reach a consistent state, ensuring that the multi-agent system of the virtual power plant can maintain consistency and collaboratively execute adjustment strategies under potential threats. The working principle of the algorithm is that each agent node regularly exchanges information with other agent nodes to jointly grasp the system status, so as to timely detect potential threats such as agent node failures or malicious behavior, and take countermeasures quickly.

[0139] With the help of distributed consensus algorithms, all proxy nodes can reach a consensus on the adjustment strategy when potential threats exist, and schedule according to the same strategy, effectively ensuring the overall stability and efficiency of the system. In addition, distributed consensus algorithms usually have built-in fault tolerance mechanisms that can resist a certain number of faulty or malicious proxy nodes, ensuring that even if some proxies are attacked or fail, the system can continue to operate and reach a consensus on the adjustment strategy.

[0140] The coordinated execution process of the adjustment strategy based on the distributed consensus algorithm is as follows:

[0141] Information broadcast: Each proxy node regularly sends its own status information to other nodes, covering key data such as power generation, load demand, and energy storage status.

[0142] Information reception and verification: After receiving the broadcast information, the proxy node immediately verifies it to ensure the authenticity and integrity of the information.

[0143] Status update: Based on the verified information, each proxy node updates its own status.

[0144] Strategy calculation: The agent node combines the current state information with the preset optimization goals, such as minimizing costs and maximizing environmental benefits, to calculate its own adjustment strategy.

[0145] Policy sharing: Using distributed consensus algorithms, such as the Raft algorithm, has the advantages of being easy to understand, fast to implement, fast to recover from failures, and highly scalable. The Raft algorithm can be used to share the calculated adjustment strategy to ensure that all proxy nodes have a unified understanding of the adjustment strategy.

[0146] Collaborative execution of strategies: After reaching a consensus, all agent nodes act in unison based on the coordinated adjustment strategy, such as adjusting power generation, distributing loads, etc.

[0147] Through the above process, the multi-agent system can achieve the coordinated execution of adjustment strategies, significantly improving the stability, efficiency and security of the system.

[0148] In addition, optionally, the operating status of the virtual power plant can be monitored in real time, especially focusing on the behavior of malicious proxy nodes, and the execution effect of the adjustment strategy can be evaluated by comparing and analyzing the actual operating data with the expected goals.

[0149] In the virtual power plant optimization scheduling method of the present application, by abstracting each distributed energy resource in the virtual power plant as a proxy node, this distributed management method improves the flexibility and response speed of scheduling, so that the virtual power plant can adapt to changes in the electricity market and fluctuations in the internal resource status more quickly. The optimization scheduling model in the method comprehensively considers multiple objectives such as cost minimization and profit maximization, as well as multiple decision variables including output power, charging and discharging power, and power trading volume, which helps to allocate resources more reasonably. At the same time, by considering market transaction prices and load forecast data, the power trading strategy can be further optimized and the economic benefits can be improved. In addition, the method includes a mechanism for judging malicious proxy nodes, and accordingly executes adjustment strategies, such as adjusting the optimization scheduling model, to effectively enhance the robustness of the system and ensure the stable operation of the virtual power plant.

[0150] Figure 3 FIG. 1 is a flow chart of step S300 of a virtual power plant optimization scheduling method provided by an embodiment of the present application. Figure 3 As shown, the determining whether there is a malicious proxy node among the proxy nodes and executing the adjustment strategy when there is a malicious proxy node may include:

[0151] Step S310, judging whether each proxy node is abnormal according to whether the historical behavior of each proxy node is abnormal, and identifying the proxy node judged to be abnormal as a malicious proxy node, wherein the historical behavior includes at least one of transaction time, transaction volume, and transaction price.

[0152] For transaction time, the timestamps of all transactions conducted by each proxy node in the past period of time can be recorded. Check whether the transaction time of the proxy node conforms to the normal transaction pattern. For example, if a proxy node often conducts a large number of transactions during non-normal trading hours (such as late at night or on holidays), this may indicate abnormal behavior.

[0153] For transaction volume, you can count the transaction volume of each proxy node in a specific time period. Compare the transaction volume of each proxy node with the historical average or industry average. If the transaction volume of a proxy node suddenly surges or drops, this may indicate abnormal behavior.

[0154] For transaction prices, the prices of all transactions conducted by each proxy node can be recorded, and their average price, price fluctuation range, etc. can be calculated. Analyze whether the transaction price of the proxy node deviates from the market average or has abnormal fluctuations. For example, if a proxy node frequently trades at extremely high or low prices, this may mean that its behavior is abnormal.

[0155] Based on historical data and analysis results, reasonable abnormal thresholds can be set for each key indicator. These thresholds should be able to capture most abnormal behaviors while avoiding false positives. Monitor the transaction behavior of each proxy node in real time and compare it with historical behavior patterns. Once the behavior of a proxy node is found to exceed the set threshold range, it will be marked as abnormal.

[0156] Step S320, obtaining the threat level of the malicious proxy node according to the characteristics of the abnormal historical behavior, and executing an adjustment strategy according to the threat level of the malicious proxy node, wherein the characteristics include at least one of the behavior type, behavior frequency, and behavior duration.

[0157] Among them, for behavior types, different abnormal behavior types may represent different threat levels. For example, maliciously raising or lowering transaction prices may be regarded as market manipulation, and its threat level is usually high; while occasional abnormal trading time may be caused by individual circumstances, and the threat level is relatively low. According to the behavior type, each abnormal behavior can be assigned a threat level, such as high, medium, and low.

[0158] For behavior frequency, frequently occurring abnormal behaviors may indicate that malicious proxy nodes are continuously trying to disrupt the system or manipulate the market. The frequency of each abnormal behavior can be recorded, and the threat level can be adjusted based on the frequency.

[0159] For behavior duration, abnormal behaviors that persist for a long time may mean that the malicious agent node has a stronger intention to destroy or a more complex attack strategy. The duration of each abnormal behavior can be measured and the threat level can be adjusted accordingly.

[0160] A threat level assessment model can be established based on the behavior type, behavior frequency, and behavior duration. The model can be rule-based, statistical, or machine learning, depending on the complexity of the system and the available data. The assessment model is used to convert feature data into quantitative threat levels, such as low threat, medium threat, and high threat. This helps to more intuitively understand the potential harm of malicious proxy nodes and provide a decision-making basis for subsequent adjustment strategies.

[0161] In a specific implementation, step 300, i.e., the step of "determining whether there are malicious proxy nodes among the proxy nodes, and executing the adjustment strategy when there are malicious proxy nodes", can also be implemented in the following manner, i.e., step 300 can specifically include: risk assessment, risk response mechanism, fault tolerance strategy and recovery strategy.

[0162] In step 300, the purpose of risk assessment is to predict and quantify potential malicious behavior so that timely measures can be taken to prevent it, which can be achieved through behavior pattern monitoring, threat level assessment and early warning mechanism.

[0163] In the risk assessment step, behavior pattern monitoring is the first step of risk assessment. It aims to learn the historical behavior data of agent nodes through machine learning algorithms, establish a normal behavior pattern library, and monitor the behavior of agent nodes in real time and compare it with normal behavior patterns to detect abnormal behavior.

[0164] See also Figure 4 As shown, we collect historical behavior data of agent nodes and extract key features from them.

[0165] The extracted features are trained using machine learning algorithms such as cluster analysis and anomaly detection to establish a normal behavior pattern library. At the same time, the corresponding behavior data of the proxy node is monitored in real time, and key features are extracted. The features monitored in real time are compared with the normal behavior pattern library to detect abnormal behavior.

[0166] In the risk assessment step, the threat level assessment aims to design a threat assessment index system based on factors such as the type, frequency, and duration of abnormal behaviors, and use the fuzzy comprehensive evaluation method to quantitatively assess the threat level of malicious nodes.

[0167] Threat assessment indicators may include abnormal behavior.

[0168] Abnormal behavior refers to the deviation of agent nodes during scheduling.

[0169] Abnormal behavior involves abnormal behavior types. Abnormal behavior types, that is, classification based on the nature of abnormal behavior, can include power anomalies, communication failures, non-execution of scheduling instructions, abnormal resource utilization, violation of security policies, data tampering or falsification, network attack behavior, abnormal system resource usage, equipment failures or anomalies, and non-compliance with protocols or specifications.

[0170] Power anomalies include: unstable power output, that is, the power output of proxy nodes (especially DER) fluctuates frequently and exceeds the normal range; insufficient or excessive power output, that is, the actual power output of proxy nodes is lower or higher than expected, resulting in an imbalance between supply and demand in the system; sudden change in power output, that is, the power output of proxy nodes suddenly changes drastically in a short period of time, exceeding the normal fluctuation range; abnormal power output cycle, that is, the periodic change pattern of power output of proxy nodes does not match expectations; ineffective power output, that is, the power output generated by proxy nodes has no actual contribution to the system; power factor abnormality, that is, the power factor (ratio of active power to apparent power) of proxy nodes deviates from the normal range, affecting system efficiency.

[0171] Communication failures include: communication interruption, that is, the communication link between the proxy node and other nodes is suddenly interrupted, resulting in the inability to transmit information; communication delay, that is, the communication response time of the proxy node is too long, exceeding the preset threshold. Data loss, that is, during the communication process, part or all of the data sent or received by the proxy node is lost; communication protocol mismatch, that is, the communication protocol used by the proxy node is incompatible with the system or other nodes, resulting in communication failure; communication data format error, that is, the data format sent or received by the proxy node does not comply with the system regulations, resulting in the inability to correctly parse the data; unstable communication link, that is, the communication link between the proxy node and other nodes is frequently disconnected or reconnected; communication encryption failure, that is, during the communication process, the encryption or decryption operation of the proxy node fails, resulting in data leakage or tampering.

[0172] The scheduling instructions are not executed, including: refusing to execute instructions, that is, the agent node refuses to execute or delays execution after receiving the scheduling instructions; execution error, that is, the agent node encounters an operation error or fails to execute the scheduling instructions in place; partial execution of instructions, that is, the agent node only executes part of the content of the scheduling instructions and does not operate completely according to the instructions; delayed execution of instructions, that is, the agent node does not execute the scheduling instructions within the specified time after receiving them; misunderstanding of instructions, that is, the agent node has a wrong understanding of the scheduling instructions, resulting in the execution result not meeting the expectations; instruction conflict, that is, the agent node receives multiple conflicting scheduling instructions at the same time, resulting in the inability to determine which instruction to execute.

[0173] Abnormal resource utilization includes: excessive use of resources, that is, the proxy node exceeds the preset limit or restriction when using system resources; inefficient use of resources, that is, the proxy node's resource utilization efficiency is low, resulting in resource waste; resource occupation conflict, that is, the proxy node conflicts with other nodes in resource use, resulting in the ineffective use of resources; uneven resource allocation, that is, the system resources are unevenly distributed among proxy nodes, resulting in excess resources in some nodes and insufficient resources in some nodes; resource exhaustion, that is, the proxy node exhausts resources due to excessive use of resources, affecting the normal operation of the system; resource leakage, that is, the proxy node fails to release resources in time after using them, resulting in resource waste and system performance degradation.

[0174] Violations of security policies include: failure to comply with security policies, i.e., the proxy node fails to operate in accordance with the security policies specified by the system; exploitation of security vulnerabilities, i.e., the proxy node is maliciously exploited to become a springboard for attacking the system or the source of an attack; unauthorized access, i.e., the proxy node accesses the resources of the system or other nodes without authorization; malicious code injection, i.e., the proxy node injects malicious code into the system or other nodes; privilege escalation, i.e., the proxy node attempts to elevate its own privileges to access or modify data or resources that should not be accessed; and sensitive information leakage, i.e., the proxy node leaks sensitive information of the system or other nodes.

[0175] Data tampering or forgery, including: data integrity destruction, that is, the proxy node illegally modifies the data when transmitting or storing data, thereby destroying the integrity of the data; data injection attack, that is, the proxy node injects false data into the system or other nodes to interfere with the normal operation of the system or mislead other nodes; data replay attack, that is, the proxy node resends the previously captured data to deceive the system or other nodes; data forged signature, that is, the proxy node forges the signature of the data to make the data look like it comes from other legitimate nodes.

[0176] Network attack behaviors include: DDoS attack, that is, the proxy node becomes the initiator of DDoS attack and sends a large amount of malicious traffic to the system or other nodes; man-in-the-middle attack, that is, the proxy node pretends to be other nodes to communicate during network communication, steals or tampers with information; ARP spoofing, that is, the proxy node deceives other nodes into treating it as the default gateway by sending forged ARP response packets; DNS hijacking, that is, the proxy node tampers with DNS resolution results and redirects users to malicious websites or servers; session hijacking, that is, the proxy node steals the session information of other nodes to impersonate the node to communicate; port scanning, that is, the proxy node scans the open ports of the system or other nodes to find potential security vulnerabilities.

[0177] Abnormal system resource usage, including: CPU overload, that is, the CPU usage of the proxy node is continuously too high, resulting in slow system response or crash; memory leak, that is, the memory usage of the proxy node continues to increase during operation, resulting in insufficient system memory; disk space exhaustion, that is, the proxy node occupies a large amount of disk space, resulting in the system being unable to store new data or log files; network bandwidth usage, that is, the proxy node occupies a large amount of network bandwidth, resulting in other nodes being unable to communicate normally.

[0178] Equipment failure or abnormality, including: hardware failure, the hardware equipment of the proxy node fails; software abnormality, that is, the running software of the proxy node has an abnormality; equipment overheating, that is, the performance of the hardware equipment of the proxy node is degraded or damaged due to overheating; power failure, that is, there is a problem with the power supply of the proxy node, causing the equipment to fail to work normally.

[0179] Failure to comply with protocols or specifications includes: protocol version mismatch, i.e., the protocol version used by the proxy node is incompatible with the system or other nodes; message format error, i.e., the message format sent or received by the proxy node does not conform to the format specified by the system; timeout failure to respond, i.e., the proxy node fails to respond within the specified time after receiving the request; illegal request, i.e., the proxy node sends an illegal request to the system or other nodes, such as requesting a non-existent resource or performing an illegal operation.

[0180] Abnormal behavior also involves abnormal behavior frequency, which is the number of times abnormal behavior occurs in a unit of time.

[0181] Abnormal behavior also involves abnormal behavior duration, which records the length of time from the start to the end of the abnormal behavior.

[0182] Abnormal behavior also involves the degree of deviation of behavior patterns. The degree of deviation of behavior patterns refers to the degree of difference between the current behavior of the proxy node and the normal behavior pattern. The historical behavior data of the proxy node can be learned through machine learning algorithms (such as cluster analysis, anomaly detection, etc.) to establish a normal behavior pattern library. Then, the behavior of the proxy node is monitored in real time and compared with the normal behavior pattern to calculate the degree of deviation. The greater the degree of deviation, the more abnormal the behavior of the node is and the higher the threat level is.

[0183] Threat assessment indicators can also include economic impact.

[0184] Economic impact, that is, the impact of the malicious node's behavior on the overall economic benefits of the virtual power plant.

[0185] Economic impact involves economic losses, that is, assessing the economic losses caused by the behavior of malicious nodes. Economic losses can include direct losses and indirect losses.

[0186] Among them, direct losses may include: repair costs, that is, the costs required to repair systems, equipment or data caused by malicious node behavior; loss of revenue, that is, the reduction in revenue due to service interruption or quality degradation; fines and compensation, that is, fines or compensation paid for violation of contracts or regulations.

[0187] Indirect losses may include: reputation loss valuation, that is, reputation loss estimated through market research, brand evaluation and other methods; business losses caused by decreased customer satisfaction, which can be measured through indicators such as customer churn rate and order reduction.

[0188] Economic impact also involves market impact, analyzing the impact of malicious node behavior on electricity market transactions. Market impact can include price fluctuations, changes in trading volume, and analysis of market participant behavior.

[0189] Among them, price fluctuations may include: the number of abnormal price fluctuations, that is, the number of abnormal fluctuations in electricity market transaction prices caused by malicious node behavior; the price fluctuation amplitude, that is, the specific amplitude of fluctuations in electricity market transaction prices before and after malicious node behavior; price volatility, that is, a quantitative indicator of price fluctuation amplitude.

[0190] Changes in trading volume may include: the number of abnormal reductions in trading volume, i.e. the number of abnormal reductions in power market trading volume caused by malicious node behavior; the proportion of trading volume reduction, i.e. the specific proportion of power market trading volume reduction after malicious node behavior.

[0191] The behavior analysis of market participants can include: the proportion of market participants' trust decline, that is, the specific proportion of market participants' trust in virtual power plants decline obtained through questionnaires, interviews, etc.; the trend of market share changes, that is, the specific impact trend of malicious node behavior on the market share of virtual power plants.

[0192] The economic impact also involves operating costs, that is, the increase in the operating costs of the virtual power plant due to the behavior of malicious nodes. Operating costs can include system maintenance, security reinforcement, and additional monitoring costs.

[0193] Among them, system maintenance may include: system maintenance costs, that is, the increased maintenance costs to prevent similar incidents from happening again; system repair cycle, that is, the specific time period required for system repair after malicious node behavior; increased system maintenance difficulty, that is, a specific assessment of the increased difficulty of system maintenance caused by malicious node behavior.

[0194] Security reinforcement may include: the purchase cost of security equipment, such as firewalls, intrusion detection systems and other equipment; security service costs, such as security consulting, vulnerability scanning and other service costs; security reinforcement effect, that is, the security reinforcement effect evaluated through security testing, vulnerability scanning and other methods.

[0195] Additional monitoring costs may include: monitoring equipment purchase costs, that is, the cost of additional monitoring equipment; monitoring personnel costs, that is, the cost of additional personnel for monitoring malicious node behavior; and monitoring efficiency improvement, that is, the monitoring efficiency improved by adding monitoring equipment or personnel.

[0196] Economic impacts also involve energy efficiency impacts, which can specifically include reduced energy efficiency, energy scheduling impacts, and system stability impacts.

[0197] Among them, energy efficiency reduction can include: the amount of energy waste, that is, the amount of energy waste caused by malicious node behavior; the energy waste ratio, that is, the ratio of energy waste caused by malicious node behavior to total energy; the extent of energy efficiency reduction, that is, the specific extent of energy efficiency reduction before and after malicious node behavior; energy efficiency reduction rate, that is, the comparison of energy efficiency before and after malicious node behavior.

[0198] Energy scheduling impacts may include: energy scheduling error rate, that is, the energy scheduling error rate caused by malicious node behavior; energy scheduling response time, that is, the specific time required for energy scheduling response after malicious node behavior.

[0199] The impact on system stability may include: the number of system failures, that is, the number of system failures caused by malicious node behavior; the system failure rate, that is, the frequency of system failures caused by malicious node behavior; the system recovery time, that is, the time required for the system to recover from a failure; the system recovery capability, that is, the assessment of the system's ability to recover from a failure after malicious node behavior; the system reliability, that is, the system reliability assessed through simulation tests and other methods; the proportion of system reliability reduction, that is, the specific proportion of system reliability reduction caused by malicious node behavior.

[0200] The threat assessment indicator system can include the correlation between nodes.

[0201] The inter-node correlation refers to the correlation degree and mutual influence between the proxy node and other nodes. Specifically, the inter-node correlation may include connectivity, influence range, and dependency.

[0202] Among them, connectivity, which is to evaluate the connection status of the proxy node in the network, can include: node degree, which is the number of nodes directly connected to the proxy node; degree centrality, which is the node degree of the proxy node divided by the maximum number of connections that can be formed by all nodes in the network.

[0203] The scope of impact, that is, analyzing the scope of impact of the malicious node's behavior on other nodes or the entire virtual power plant, may include: the number of affected nodes, that is, the number of nodes directly or indirectly affected after the malicious node behavior occurs; the speed of impact spread, that is, the time required for the malicious behavior to affect the maximum scope; the degree of impact, that is, evaluating the impact of malicious behavior on the overall performance of the virtual power plant through quantitative indicators (such as power load changes, voltage fluctuations, etc.).

[0204] Dependency, that is, evaluating the degree of dependence of the proxy node on other nodes, can include: the number of dependent nodes, that is, the number of other nodes that the proxy node depends on to complete a task or function; the length of the dependent path, that is, the shortest path length from the proxy node to its dependent nodes; the dependence intensity, that is, evaluating the degree of dependence of the proxy node on its dependent nodes through quantitative indicators (such as data transmission volume, energy exchange volume, etc.).

[0205] According to the threat assessment index system, a fuzzy membership function can be set for each indicator to convert the qualitative evaluation into a quantitative evaluation. Then, the comprehensive score is calculated using methods such as weighted average or geometric average to quantitatively assess the threat level of malicious nodes:

[0206] The overall assessment value of the node threat level = ∑ single indicator assessment value × indicator weight. After receiving the alarm information, the system administrator should immediately initiate the response mechanism, including adjusting the constraints and objective function of the optimization model, enabling spare nodes, performing data recovery, etc.

[0207] In the risk assessment step, the early warning mechanism, that is, when the threat level exceeds the preset threshold, triggers the early warning mechanism, sends an alarm to the system administrator, and initiates the response mechanism. A reasonable threat level threshold θ can be set according to the operation needs and safety requirements of the virtual power plant. When the threat level exceeds the threshold θ, the early warning mechanism is automatically triggered to send an alarm message to the system administrator, including detailed information such as the identity of the malicious node, the threat level, and the type of abnormal behavior.

[0208] In step 300, the risk response mechanism aims to reduce the impact of malicious nodes on the system by adjusting the constraints and objective functions of the optimization model, which may include dynamically adjusting the constraints and optimizing the objective.

[0209] Among them, dynamically adjust the constraints, that is, dynamically adjust the constraints of the optimization model according to the threat level of the malicious nodes to ensure that the system can still operate stably in the presence of malicious nodes. You can consider adjusting power balance constraints, voltage stability constraints, etc.

[0210] Power balance constraint adjustment, that is, when malicious nodes exist, it may disturb the balance between power generation and power consumption, causing system instability. Therefore, the system needs to dynamically adjust the various components in the power balance constraint according to the threat level of the malicious node. For example, when a malicious node is detected trying to disrupt the power balance, the system can automatically increase or decrease the power generation of other normal nodes to maintain the equality.

[0211] The voltage stability constraint adjustment can be expressed as:

[0212] V j,min ≤V j (t)≤V j,max

[0213] Among them, V j (t) is the voltage of system node j at time t, V j,min 、V j,max are the minimum and maximum voltages allowed at node j, respectively. Malicious nodes may disrupt voltage stability by manipulating certain parameters or devices. To this end, the system can dynamically adjust voltage stability constraints. When a malicious node is detected, the system can dynamically adjust these limits, such as by adjusting the tap position of the transformer, the switching of reactive compensation equipment, etc., to ensure that the voltage remains within the stable range.

[0214] a) Optimization objective adjustment: In order to reduce the impact of malicious nodes on system performance, the objective function and weight related to the malicious node behavior can be introduced into the objective function. For example, the optimization scheduling objective function (1) can be modified as follows:

[0215] min F′(x)=[C(x)-R(x),-EnvBenefit(x),w·risk(x)] (24)

[0216] Where risk(x) is the risk function and w is the risk weight. The risk function can be calculated using the following method:

[0217]

[0218] The threat level of the proxy node is the comprehensive evaluation value of the threat level of the proxy node in the following expression.

[0219] Overall evaluation value of node threat level = ∑ single indicator evaluation value × indicator weight

[0220] When the system detects a malicious node and confirms its existence, the risk weight w can be appropriately increased to increase the system's sensitivity to risk. Under normal operating conditions, the risk weight w can be appropriately reduced or directly set to 0.

[0221] In step 300, the fault-tolerant strategy is that when a malicious node is found, the system should immediately enable a backup node to replace the malicious node. The performance of the backup node should match the replaced malicious node to ensure that the system performance is not affected. And the backup node should undergo strict security testing to ensure that it does not have malicious behavior.

[0222] In step 300, a recovery strategy, i.e., a data recovery mechanism is designed to recover lost or damaged data to avoid the possibility that malicious nodes may cause system data loss or damage. The recovery strategy may include data backup, data verification, and data recovery.

[0223] Among them, data backup means regularly backing up system data to ensure that data can be restored when it is lost or damaged.

[0224] Data verification is to verify the backup data using methods such as checksums or hash values ​​to ensure the integrity and accuracy of the backup data.

[0225] Data recovery means using backup data to recover data when data is lost or damaged. First, stop the system to avoid data conflicts during the data recovery process. Then recover the lost or damaged data from the backup data, verify the recovered data to ensure the integrity and accuracy of the data, and finally restart the system.

[0226] In a virtual power plant, the presence of malicious proxy nodes may cause the adjustment strategy to fail or the system to crash, thus affecting the stability and safety of the entire virtual power plant. The present invention introduces risk assessment and response mechanisms into the optimization and control model to predict and prevent potential malicious behaviors. By monitoring the behavior patterns and data changes of proxy nodes, the threat level of malicious nodes can be discovered and evaluated in a timely manner. After discovering malicious nodes, the system adjusts the constraints and objective functions of the optimization model to ensure that the optimization goals can still be achieved in the presence of malicious nodes. At the same time, fault-tolerant and recovery strategies are designed, such as the activation of standby nodes and data recovery mechanisms, to deal with problems such as system failures or data loss caused by malicious nodes.

[0227] In a specific implementation, one or more of the adjustment strategies of adjusting the optimization scheduling model, replacing the proxy node, and regularly backing up data may be executed according to the threat level of the malicious proxy node.

[0228] Wherein, adjusting the optimization scheduling model may specifically include:

[0229] The constraint condition is adjusted, and the adjustment constraint condition includes one of increasing the power generation of the normal proxy node and reducing the power generation of the normal proxy node. By accurately controlling the power generation of the normal proxy node, the supply and demand balance of the system can be better maintained, the system fluctuation and instability caused by the imbalance of supply and demand can be reduced, and the stable operation of the virtual power plant can be guaranteed.

[0230] Adjust the objective function, wherein the adjusting objective function includes adding a risk function to the objective function. Specifically, the objective function can be denoted as f(x). This function may be a certain performance indicator to be maximized or minimized, such as profit, cost, time, etc. Next, define a risk function, denoted as r(x). This function is used to quantify the risk associated with decision x. The specific form of the risk function depends on the context of the problem, and it may be variance, maximum possible loss, uncertainty measure, etc. In order to integrate the risk function into the objective function, a weight λ can be set, which is usually a number between 0 and 1, or a positive real number, depending on the specific application and preference. The weight λ reflects the degree of importance we attach to risk. A higher λ value means that we pay more attention to risk, while a lower λ value means that we pay more attention to the original objective function. Combine the objective function and the risk function to construct an adjusted objective function. This new objective function can be expressed as:

[0231] F(x)=f(x)-λr(x)

[0232] Alternatively, if one needs to maximize the original objective function and minimize the risk at the same time, one can write it as:

[0233] F(x)=f(x)+λ(-r(x))

[0234] Here, λ(-r(x)) represents the negative value of the risk function multiplied by the weight, so that both the original objective function and the risk can be considered by maximizing F(x).

[0235] An appropriate optimization algorithm can be used to maximize (or minimize) the adjusted objective function F(x). This process involves taking the derivative of F(x), gradient descent, simulated annealing, genetic algorithms, and other optimization techniques, depending on the complexity of the problem and the nature of the objective function.

[0236] Just for example, assume that the original objective function is f(x)=10x-x2, the risk function is the variance r(x)=x2, and the weight λ=0.5. Then the adjusted objective function is:

[0237] F(x)=10x-x2-0.5x2=10x-1.5x2

[0238] The goal is to find the value of x that maximizes F(x).

[0239] Through the above steps, the risk function can be integrated into the original objective function, and the importance of risk in the decision-making process can be adjusted by setting weights. This method can better manage and control risks while pursuing goals.

[0240] It should be noted that by monitoring and analyzing the historical behavior of each agent node, it is possible to timely discover and identify agent nodes with abnormal behavior, evaluate their threat level based on their behavioral characteristics, implement adjustment strategies more specifically, and respond quickly to malicious behavior, reducing its impact on system stability and ensuring the stable operation of the virtual power plant.

[0241] Example 2

[0242] Figure 5 : is a structural framework diagram of a virtual power plant optimization scheduling device provided by an embodiment of the present application. The virtual power plant optimization scheduling device is implemented based on a multi-agent system, and the multi-agent system includes multiple agent nodes, each agent node corresponds to a distributed energy resource of the virtual power plant. Figure 5 As shown, the device may include:

[0243] The data collection module 100 is used to obtain virtual power plant data and electricity market data. The virtual power plant data includes at least one of the power generation, energy storage status, and equipment failure rate of each proxy node, and the electricity market data includes at least one of price information and load forecast data.

[0244] A model building module 200 is used to obtain an optimal scheduling model of a virtual power plant based on the virtual power plant data and the electricity market data, wherein the optimal scheduling model includes an objective function and constraints, wherein the objective function is expressed in the form of optimization of multiple objectives, and the multiple objectives include cost minimization, benefit maximization, environmental benefits and at least one of decision variables, the decision variables include at least one of output power, charge and discharge power, power trading volume, load adjustment amount, and auxiliary service provision amount, and the constraints include at least one of output power constraints, charge and discharge power constraints, load adjustment amount constraints, market trading volume limit constraints, ramp constraints, and cycle life constraints.

[0245] The risk assessment module 300 is used to determine whether there are malicious proxy nodes among the proxy nodes, and to execute an adjustment strategy when there are malicious proxy nodes. The adjustment strategy at least includes: adjusting the optimization scheduling model.

[0246] The model solving module 400 is used to solve the optimization scheduling model or the adjusted optimization scheduling model, and distribute the solved scheduling solution to each proxy node for execution.

[0247] It should be noted that the examples and application scenarios implemented by the above modules and corresponding steps are the same, but are not limited to the contents disclosed in the method of the above embodiment. It should be noted that the above modules as part of the device can be run in Figure 2 In the environment shown, it can be implemented by software or by hardware.

[0248] Example 3

[0249] One embodiment of the present application further provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the virtual power plant optimization scheduling method in the above-mentioned embodiment 1 is implemented. The specific examples in this embodiment can refer to the examples described in the above-mentioned embodiment 1, and this embodiment will not be repeated here.

[0250] Example 4

[0251] One embodiment of the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the virtual power plant optimization scheduling method in the above-mentioned embodiment 1 is implemented. Similarly, the specific examples in this embodiment can refer to the examples described in the above-mentioned embodiment 1, and this embodiment will not be repeated here.

[0252] In addition, those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0253] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A virtual power plant optimization scheduling method, characterized in that: The virtual power plant optimization scheduling method is implemented based on a multi-agent system, wherein the multi-agent system includes a plurality of agent nodes, each agent node corresponds to a distributed energy resource of the virtual power plant, and the method includes: Acquire virtual power plant data and power market data, wherein the virtual power plant data includes at least one of power generation, energy storage status, and equipment failure rate of each proxy node, and the power market data includes at least one of price information and load forecast data; According to the virtual power plant data and the power market data, an optimization scheduling model of the virtual power plant is obtained, wherein the optimization scheduling model includes an objective function and constraints, wherein the objective function is expressed in the form of optimizing multiple objectives, wherein the multiple objectives include at least one of cost minimization, benefit maximization, environmental benefit and decision variables, and the constraints include at least one of output power constraints, charge and discharge power constraints, load adjustment constraints, market transaction volume limit constraints, ramp constraints and cycle life constraints; Determine whether there is a malicious proxy node among the proxy nodes, and execute an adjustment strategy when there is a malicious proxy node, the adjustment strategy at least comprising: adjusting the optimization scheduling model; Solve the optimization scheduling model or the adjusted optimization scheduling model, and distribute the solved scheduling solution to each proxy node for execution.

2. The method according to claim 1 is characterized in that The determining whether there is a malicious proxy node among the proxy nodes and executing the adjustment strategy when there is a malicious proxy node includes: Whether each proxy node is abnormal is judged according to whether its historical behavior is abnormal, and the proxy node judged to be abnormal is identified as a malicious proxy node, wherein the historical behavior includes at least one of transaction time, transaction volume, and transaction price; according to the characteristics of the abnormal historical behavior, the threat level of the malicious proxy node is obtained, and according to the threat level of the malicious proxy node, an adjustment strategy is executed, wherein the characteristics include at least one of the behavior type, behavior frequency, and behavior duration.

3. The method according to claim 2, characterized in that The adjusting the optimization scheduling model includes: The constraint condition is adjusted, and the constraint condition is adjusted to include one of increasing the power generation of the normal proxy node and reducing the power generation of the normal proxy node.

4. The method according to claim 2, characterized in that: The adjusting the optimization scheduling model further includes: The objective function is adjusted, wherein the adjusting the objective function includes adding a risk function to the objective function.

5. The method according to claim 1, characterized in that The adjustment strategy also includes replacing the proxy node, and the replacing the proxy node includes replacing the malicious proxy node with a backup proxy node.

6. The method according to claim 1, characterized in that The adjustment strategy also includes regularly backing up data, and the regularly backing up data includes backing up data of the multi-agent system according to a preset period.

7. The method according to any one of claims 1 to 6, characterized in that: The multi-agent system comprises at least one of the following agent layers: The virtual power plant agent layer is at least used to receive electricity price information from the power market and bidding information from virtual power plants at the same layer, so as to adjust the operation strategies of the distributed energy agent layer, the energy storage system agent layer and the load demand agent layer; The power market agent layer is used to collect information on the power market's willingness to buy and sell power and the amount of power consumed, and to exchange data with the virtual power plant agent layer; A distributed energy proxy layer, at least for monitoring the output of distributed energy resources and adjusting them according to the scheduling strategy of the virtual power plant, wherein the distributed energy resources include at least one of solar power generation and wind power generation; The energy storage system agent layer is at least used to monitor the status of the energy storage device, the status including at least one of the charging and discharging strategy and the remaining capacity, and to exchange real-time data with the virtual power plant agent layer; The load demand agent layer is at least used to receive data related to load demand to predict the change trend of load and adjust the load demand according to the scheduling strategy of the virtual power plant.

8. A virtual power plant optimization scheduling device, characterized in that: The virtual power plant optimization scheduling device is implemented based on a multi-agent system, the multi-agent system includes a plurality of agent nodes, each agent node corresponds to a distributed energy resource of the virtual power plant, and the device includes: A data collection module, used to obtain virtual power plant data and power market data, wherein the virtual power plant data includes at least one of power generation, energy storage status, and equipment failure rate of each proxy node, and the power market data includes at least one of price information and load forecast data; A model building module is used to obtain an optimal dispatching model of a virtual power plant according to the virtual power plant data and the power market data, wherein the optimal dispatching model includes an objective function and constraints, wherein the objective function is expressed in the form of optimizing multiple objectives, wherein the multiple objectives include at least one of cost minimization, benefit maximization, environmental benefit and decision variables, wherein the decision variables include at least one of output power, charge and discharge power, power transaction volume, load adjustment volume and auxiliary service provision volume, and wherein the constraints include at least one of output power constraint, charge and discharge power constraint, load adjustment volume constraint, market transaction volume limit constraint, ramp constraint and cycle life constraint; The risk assessment module is used to determine whether there are malicious proxy nodes among the proxy nodes, and to execute an adjustment strategy when there are malicious proxy nodes, wherein the adjustment strategy at least includes: adjusting the optimization scheduling model; The model solving module is used to solve the optimization scheduling model or the adjusted optimization scheduling model, and distribute the solved scheduling solution to each proxy node for execution.

9. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images