A two-stage optimization scheduling method for virtual power plants considering distribution networks
By employing a two-stage optimization scheduling method, combined with the spatiotemporal distribution model of electric vehicles and the flexible load theory adjustment capacity, the problems of physical constraints and multi-source uncertainties in the distribution network of the virtual power plant are solved, realizing the coordinated optimization operation of the virtual power plant and the distribution network, and improving the robustness and economy of scheduling.
Patent Information
- Application Number
- CN202611302945.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-26
- Publication Date
- 2026-11-17
Smart Images

Figure CN123121632A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of power system operation control and virtual power plant energy management technology, and in particular to a two-stage optimized scheduling method for virtual power plants that takes into account the distribution network. Background Technology
[0002] With the steady progress of the "dual carbon" goal, the penetration rate of distributed energy resources (DERs) in the power grid is increasing. Virtual power plants (VPPs), as a management model that can efficiently aggregate distributed generators (DGs), battery energy storage systems (BESSs), electric vehicles (EVs), and flexible loads, are gradually becoming a research hotspot. However, current virtual power plant dispatching technology still faces many severe challenges: First, it emphasizes logical aggregation while neglecting physical constraints. Existing virtual power plant dispatching often only focuses on the economic transaction logic at the upper level, ignoring the node voltage overruns, line overloads, and network losses of the underlying distribution network. This can easily lead to safety hazards in the physical power grid when implemented in actual projects. Second, there is insufficient coordination between the two stages. If the initial planning only uses a simple static queuing model and ignores the actual bidirectional responsiveness of electric vehicles, it will seriously reduce the operating efficiency of aggregators. Finally, the computational bottleneck in dealing with the dual uncertainties of source and load becomes prominent. Especially in a pure spot market environment that strictly does not involve medium- and long-term contracts, the base load is superimposed with high-frequency fluctuating spot time-of-use electricity prices. Commonly used point estimation methods or Monte Carlo simulations are difficult to defend against extreme and severe scenarios, and more timely and robust dispatching algorithms are urgently needed. Summary of the Invention
[0003] The purpose of this invention is to provide a two-stage robust optimization scheduling method for virtual power plants that takes into account the physical constraints of the distribution network, in order to solve the technical problems existing in the current virtual power plant scheduling process, such as insufficient consideration of the physical constraints of the distribution network, limited ability to handle multi-source uncertainties, and inaccurate modeling of electric vehicle response characteristics.
[0004] To achieve the above objectives, this invention provides a two-stage optimized scheduling method for virtual power plants considering distribution networks, comprising the following steps: S1. Obtain historical operation data, real-time status, distribution network topology parameters, source-load forecast data, and day-ahead spot market time-of-use electricity price data of the virtual power plant and its aggregated distributed energy, and establish a spatiotemporal distribution model of electric vehicle virtual energy storage based on the historical travel trajectory of electric vehicles. S2. Perform two-stage collaborative optimization of virtual power plants. The first stage determines the distributed power access capacity, battery energy storage capacity, number of electric vehicle charging station plugs, and dispatchable power boundary of electric vehicle virtual energy storage at the planning level. The second stage establishes a collaborative scheduling model at the day-ahead operation level that takes into account the cost of flexible load adjustment, electric vehicle battery life loss, spot price incentives, and distribution network physical constraints. S3. Construct a data-driven probability distribution fuzzy set based on the combined constraints of 1-norm and ∞-norm, generate typical scenarios using Latin hypercube sampling and K-means clustering, and correct the confidence interval boundary of the fuzzy set through a rolling update strategy. S4. Establish a two-stage sub-Blu-ray integrated objective function with the goal of minimizing the total system cost. The total cost of the scenario consists of the main grid power purchase cost, the operating cost of dispatchable generators, the charging and discharging scheduling cost of electric vehicles, the flexible load adjustment cost, the aging cost of battery energy storage, the network loss cost, and the distribution network reconfiguration cost. Under the constraints of node voltage, line capacity, power balance, and radial topology, the solution is obtained using a column and constraint generation algorithm, and the output is the resource allocation, day-ahead scheduling, and distribution network reconfiguration strategy.
[0005] Preferably, the distributed energy sources in S1 include solar photovoltaic power plants, wind power plants, dispatchable generators, electric vehicle charging stations, battery energy storage systems, and flexible loads; the strong random variables include at least wind speed, solar irradiance, load demand, and day-ahead spot market time-of-use electricity prices.
[0006] Preferably, the first stage of S2 is executed at the system planning level, using an optimization method based on the fuzzy maximum-minimum principle to determine the optimal access capacity of wind and solar generators, and determining the extreme boundary of dispatchable charging and discharging power and the minimum number of charging plugs for each charging station based on the spatiotemporal distribution model of electric vehicles and objective and subjective responsiveness indicators. At the same time, the optimal capacity of the battery energy storage system is determined by combining the load curve and the charging and discharging tank selection method. The second stage is executed at the system day-ahead operation level, including establishing a demand response model for the flexible load of different typical industries that is constrained by the heterogeneity theory adjustment capacity, and establishing a joint response model for the virtual energy storage of electric vehicles that takes into account the battery cycle aging cost and spot price incentives.
[0007] Preferably, in the first stage, a fuzzy maximum-minimum comprehensive evaluation index composed of the actual power loss index, node voltage, and distributed power penetration index is embedded into the capacity configuration model of the planning layer, and the optimal access capacity of wind and solar generators is determined by the two-stage sub-Bruker model within the feasible domain of the first stage. In the first phase of evaluating the virtual energy storage dispatchable boundary of electric vehicles, the state of charge when the vehicle is parked and the expected parking time are extracted. An objective responsiveness is calculated to assess the physical charging and discharging capacity, and a subjective responsiveness is calculated to assess user willingness. The subjective responsiveness is expressed using the following formula: ; ; ; in, For the final subjective response, This is a correction factor for berthing duration. The fundamental response coefficient is determined by the state of charge. For the first i electric vehicles t State of charge during a period of time To estimate the berthing duration, The minimum parking time required to participate in scheduled charging and discharging; based on this, the expression for the upper limit of the aggregated scheduled charging power of the electric vehicle virtual energy storage is: ; in, for t The maximum available charging power for electric vehicles during specific time periods. For the first n The rated charging power of an electric vehicle N EV This represents the total number of electric vehicles aggregated. n For electric vehicles, For the first n electric vehicles t Objective responsiveness over a given time period For the first n electric vehicles t Subjective charging response coefficients related to time period and parking duration.
[0008] Preferably, in the second-stage demand response model, the scheduling cost function for flexible loads participating in peak shaving and valley filling in the spot market... Based on the theoretical adjustment capacity boundary of typical industries, its expression is as follows: ; in, and These are the unit compensation cost coefficients for loads that can be reduced and loads that can be shifted, respectively. and They are respectively tThe total power reduction and shifting during the actual time period; for the textile industry load, extract the theoretical upper limit of its adjustment capacity based on production shift shifting and assembly line characteristics to constrain [the load]. For the load of the education park, the theoretical adjustment capacity boundary is extracted based on its strong temporal regularity and scheduling characteristics to constrain the load. ; In the joint response model, the day-ahead scheduling operation cost of the electric vehicle station aggregator Based on the incentive of charging and discharging electricity prices in the spot market and the battery life loss, its expression is: ; in, This refers to the current day spot electricity price. Service fee per unit of electricity. This is the battery cycle aging cost coefficient. and These represent the charge and discharge power after aggregation within the responsivity boundary. This refers to the power of a rigid charging load that cannot be scheduled.
[0009] Preferably, in S3, the handling of uncertainty propagation of strong random variables abandons the assumption of an exact probability distribution. It generates an initial scene based on Latin hypercube sampling and uses the K-means clustering algorithm to obtain typical discrete scenes. On this basis, it constructs a data-driven probability distribution fuzzy set Ψ containing combined constraints of the 1-norm and ∞-norm, with the constraint expression as follows: ; ; in, S This represents the total number of typical scenarios. For the first s The probability to be determined for each typical scenario. For the first s The empirical probability of a typical scenario The radius is constrained by the 1-norm. The radius is constrained by the ∞-norm. N For historical sample size, and Each represents a given confidence level.
[0010] Preferably, the comprehensive objective function in S4 is established based on minimizing the total expected operating cost of the system. The two-stage sub-Bruker model is decomposed into a main problem and sub-problems using a column and constraint generation algorithm, and solved iteratively. Simultaneously, the physical operating constraints of the underlying distribution network are used as rigid boundaries, and distribution network reconfiguration technology is introduced to find the radial topology with the minimum active power loss. The comprehensive objective function is transformed into the expression for minimizing the expected operating cost model under the worst-case probability distribution scenario within the sub-Bruker optimization framework: ; in, x Let p be the probability vector for the planning and allocation decision variables in the first stage. X For the first stage of planning and allocation decision variables x feasible domain, For the scene s The second phase of day-ahead operation and scheduling decision variables, Given the first-stage decision x Second stage variables feasible domain, For the scene s The total operating cost is expressed as follows: ; in, , , , , , and Scenes s The costs include: main grid power purchase cost, dispatchable generator operation cost, electric vehicle charging and discharging dispatch cost, flexible load adjustment cost, battery energy storage aging cost, grid loss cost, and distribution network reconfiguration cost. In addition, the physical operating constraints in S4 include: Node voltage The safety constraints are: ; in, and These are the lower and upper limits of the node voltage, respectively. Line transmission capacity The safety constraints are: ; in, This represents the maximum transmission capacity of the line. j For nodes i Adjacent node numbers; The active power balance constraint at the node is: ; in, For nodes i exist t Active power generation during a given time period For nodes i exist t Active power of load during a given period For nodes i The set of adjacent nodes, for t Time period is determined by nodes i Flow to Node j The active power of the branch; The reactive power balance constraint at the node is: ; in, For nodes i exist t Reactive power generation during a given period For nodes i exist t Reactive power of load during a given period for t Time period is determined by nodes i Flow to Node j The reactive power of the branch circuit.
[0011] Preferably, the radial topology is generated by introducing an improved Newton-Raphson power flow algorithm that includes balancing nodes, PQ nodes, and PV nodes. When optimizing the power flow calculation for each time period, all loops formed by closed tie switches are traversed, and power flow calculation is performed by disconnecting one branch at a time. The switch combination that satisfies all physical operating constraints and minimizes active power loss is selected, and the optimal radial distribution network topology under this switch combination is generated during the iteration of the main problem.
[0012] Therefore, the present invention employs the above-mentioned two-stage optimized scheduling method for virtual power plants considering the distribution network, which has the following beneficial effects: 1. This invention integrates the economic dispatch of virtual power plants with the voltage of distribution network nodes, line capacity, power balance and radial topology constraints in a unified model, avoiding the problem of only considering upper-level aggregation transactions while ignoring the underlying physical security, and improving the engineering feasibility of the dispatch scheme.
[0013] 2. This invention determines the dispatchable power boundary by combining the spatiotemporal distribution model of electric vehicles, objective responsiveness, and subjective responsiveness, and adjusts the capacity by combining flexible load theory, thereby improving the accuracy of multi-type aggregated resources participating in the day-ahead spot market dispatch.
[0014] 3. This invention employs Latin hypercube sampling, K-means clustering, and a data-driven probability distribution fuzzy set with combined constraints of 1-norm and ∞-norm to handle source load and electricity price uncertainties, thereby improving the robustness and economy of scheduling results under complex fluctuation scenarios. Attached Figure Description
[0015] Figure 1 This is a diagram of the overall operating architecture of the virtual power plant considering source-load uncertainty and active distribution network reconfiguration in this invention. Figure 2This is a schematic diagram illustrating the principle of fuzzy evaluation of the boundary of virtual energy storage aggregation and schedulable power for electric vehicles provided by the present invention. Figure 3 This is a diagram showing the upper limit of the schedulable charging power for electric vehicles using virtual energy storage aggregation, as presented in this invention. Figure 4 This is a comprehensive evaluation diagram of distributed power supply capacity configuration under the fuzzy maximum and minimum principle of the present invention; Figure 5 This is a diagram illustrating the optimal scheduling strategy for the day-ahead optimization of the distributed bar in this invention. Detailed Implementation
[0016] To make the objectives, technical solutions, and advantages disclosed in the embodiments of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the embodiments of the present invention and are not intended to limit the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of this application. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout.
[0017] It should be noted that the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion, such as a process, method, system, product, or server that includes a series of steps or units, not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such process, method, product, or device.
[0018] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.
[0019] like Figures 1-5As shown, this invention provides a two-stage optimization scheduling method for virtual power plants considering distribution networks, including the following steps: S1, acquiring historical operating data, real-time status, distribution network topology parameters, source-load forecast data, and day-ahead spot market time-of-use electricity price data of the virtual power plant and its aggregated distributed energy, and establishing a spatiotemporal distribution model of electric vehicle virtual energy storage based on the historical travel trajectories of electric vehicles; S2, performing two-stage collaborative optimization of the virtual power plant. In the first stage, at the planning level, the capacity of distributed power source access, battery energy storage capacity, number of electric vehicle charging station plugs, and dispatchable power boundary of electric vehicle virtual energy storage are determined. In the second stage, at the day-ahead operation level, a collaborative scheduling model considering flexible load adjustment costs, electric vehicle battery life loss, spot price incentives, and distribution network physical constraints is established; S3, constructing a data-driven probability distribution fuzzy set based on the comprehensive constraints of 1-norm and ∞-norm, generating typical scenarios using Latin hypercube sampling and K-means clustering, and correcting the confidence interval boundary of the fuzzy set through a rolling update strategy; S4, establishing and solving the two-stage sub-Bruker comprehensive objective function, and outputting resource allocation, day-ahead scheduling, and distribution network reconfiguration strategies.
[0020] This embodiment takes a distribution network with 33 nodes and its internally aggregated virtual power plant as an example. The virtual power plant aggregates photovoltaic (node 5), wind power (node 14), dispatchable diesel generators (node 6), 5 electric vehicle charging stations (nodes 8, 15, 21, 23, and 30), and some typical industry controllable loads. The specific implementation steps are as follows: S1. Acquire basic data and uncertainty characteristic parameters, including branch impedance data and node baseline load data of the active distribution network; acquire technical parameters of virtual power plant aggregated resources, such as the maximum grid-connected capacity limits of photovoltaic and wind power, and the charging and discharging efficiency of battery energy storage systems; acquire statistical data of electric vehicles in the road network, establish a spatiotemporal distribution model of electric vehicle virtual energy storage based on historical travel trajectories, and extract vehicle arrival and departure times and real-time state of charge; finally, acquire uncertainty variable parameters of source load and electricity price in the system, focusing on the high-frequency fluctuating day-ahead electricity market environment, and extract high-frequency historical discrete scenario data of wind speed, solar irradiance, load demand and time-of-use node marginal electricity price.
[0021] S2. The first phase is executed at the system planning level, determining multi-resource allocation; the second phase is executed at the system's day-to-day operation level, establishing specific demand response models, specifically: In the first phase of distributed generator capacity configuration, an optimization method based on the fuzzy maximum-minimum principle is introduced. For each group of candidate solar and wind power generator capacities, the system power flow is calculated, and three fuzzy membership functions are constructed, including the actual power loss index membership function, the node voltage membership function, and the distributed power penetration rate index membership function. For each group of candidate capacity solutions, its value under the three fuzzy membership functions is calculated, and the minimum value of the above three membership functions is determined as the comprehensive evaluation index. The optimal candidate capacity solution is obtained by maximizing this minimum and worst value through the optimization algorithm.
[0022] The charging station plug configuration abandons the static queuing theory and adopts a virtual energy storage spatiotemporal response model for electric vehicles. This model extracts the vehicle's state of charge and estimated parking time to calculate objective and subjective response. To avoid piecewise functions, a maximum-minimum function is used to achieve linear decrease. The relevant formula for subjective charging response is as follows: First, calculate the fundamental response coefficient determined by the state of charge. : ; in, For the first i electric vehicles t State of charge over a period of time; Multiply by the berthing duration correction factor : ; in, To estimate the berthing duration, In this embodiment, the minimum docking time required to participate in schedulable charging and discharging is specified. T min =2 hours; Obtain the final subjective response level : ; and with The minimum number of charging plugs and the optimal energy storage capacity required for dynamic extreme value boundary iteration; based on this, the expression for the upper limit of the aggregate schedulable charging power of the virtual energy storage of electric vehicles is: ; in, The maximum schedulable charging power for electric vehicles during time period t. For the first n The rated charging power of each electric vehicle, where NEV is the total number of electric vehicles aggregated. n For electric vehicles, For the first n electric vehicles tObjective responsiveness over a given time period For the first n electric vehicles t Subjective charging response coefficients related to time period and parking duration.
[0023] In the second-stage demand response model, the flexible load cost function is adopted for specific industries: ; in, and These are the unit compensation cost coefficients for loads that can be reduced and loads that can be shifted, respectively. and They are respectively The total power reduction and shifting during the actual time period; for the textile industry load, extract the theoretical upper limit of its adjustment capacity based on production shift shifting and assembly line characteristics to constrain [the load]. For the load of the education park, the theoretical adjustment capacity boundary is extracted based on its strong temporal regularity and scheduling characteristics to constrain the load. ; For the textile industry, the theoretical upper limit of the adjustment capacity of production shift shifting and assembly line characteristics is extracted to strictly constrain the amount of shifting. For educational parks, the theoretical adjustment capacity boundary is extracted based on their strong temporal regularity and scheduling characteristics to constrain the capacity that can be reduced. .
[0024] The second phase addresses the response to the participation of virtual energy storage for electric vehicles in day-ahead dispatch, and its dispatching and operating costs. The current dispatching operation cost has been fully accounted for, taking into full account dynamic electricity price incentives and battery aging losses. Including the battery cycle aging cost factor caused by charge and discharge cycles and unit service fee coefficient By optimizing the bidirectional charging and discharging power of electric vehicles through the recent electricity price directive, the dispatching and operation costs are reduced. The expression is: ; in, This refers to the current day spot electricity price. and These represent the charge and discharge power after aggregation within the responsivity boundary. This refers to the power of a rigid charging load that cannot be scheduled.
[0025] S3. For systems containing multiple strongly random input variables (wind, solar, load, and day-ahead dynamic electricity price), initial scenarios are generated using Latin hypercube sampling, and typical scenarios are obtained using K-means. A fuzzy set Ψ of probability distribution with combined constraints of 1-norm and ∞-norm is constructed: ; ; in, S This represents the total number of typical scenarios. For the first s The probability to be determined for each typical scenario. Let be the empirical probability of the s-th typical scenario. The radius is constrained by the 1-norm. The radius is constrained by the ∞-norm. N For historical sample size, and Given confidence levels, a rolling update strategy is adopted: after each scheduling cycle (24 hours), the newly obtained day-ahead spot electricity price, actual wind and solar power output, and actual load value are added to the historical database, and the calculation is recalculated. , To update Ψ. Preserving the characteristics of the sub-Bruker bar, the worst-case probability distribution of the comprehensive objective function is sought on the fuzzy set. We can then derive the worst-case expected boundary of the objective function's operating cost.
[0026] S4. The comprehensive objective function is transformed into a model that minimizes the expected operating cost under the worst-case probability distribution scenario within the sub-Bruker optimization framework. During the optimization process, the physical operating constraints of the underlying distribution network are treated as rigid boundaries, and distribution network reconfiguration technology is introduced to find the radial topology with the minimum active power loss. Finally, the column and constraint generation algorithm is used to decompose and solve the comprehensive objective function. The expected expression of the comprehensive objective function in the worst-case scenario is: ; in, x Let p be the probability vector for the planning and allocation decision variables in the first stage. X For the first stage of planning and allocation decision variables x feasible domain, For the scene s The second phase of day-ahead operation and scheduling decision variables, Given the first-stage decision x Second stage variables feasible domain, For the scene s The total operating cost is expressed as follows: ; in, , , , , , and Scenes sThe costs include the main grid power purchase cost, dispatchable generator operation cost, electric vehicle charging and discharging dispatch cost, flexible load adjustment cost, battery energy storage aging cost, grid loss cost, and distribution network reconfiguration cost.
[0027] The physical operating constraints set in S4 include node voltage safety constraints (0.94 pu~1.06 pu), line transmission capacity safety constraints, node active power balance constraints, and node reactive power balance constraints, as detailed below: Node voltage The safety constraints are: ; In the formula, and These are the lower and upper limits of the node voltage, respectively. Line transmission capacity The safety constraints are: ; In the formula, This represents the maximum transmission capacity of the line. j Number the node adjacent to node i; The active power balance constraint at the node is: ; In the formula, For nodes i exist t Active power generation during a given time period For nodes i exist t Active power of load during a given period For nodes i The set of adjacent nodes, for t The time period flows from node i to node i j The active power of the branch; The reactive power balance constraint at the node is: ; In the formula, For nodes i exist t Reactive power generation during a given period For nodes i exist t Reactive power of load during a given period for t Time period is determined by nodes i Flow to Node j The reactive power of the branch circuit.
[0028] Furthermore, the radial topology is generated by introducing an improved Newton-Raphson power flow algorithm that includes zero nodes (corresponding to dispatchable generators) and PQV nodes (corresponding to the main grid interface). During the iterative optimization of the power flow in each time period of the main problem, a loop consisting of all closed tie switches is sought. Branches are individually disconnected to maintain the radial topology of the distribution network, and the switch state with the minimum network loss is selected as the optimal physical topology for that iteration step. After the algorithm converges, it outputs the optimal scheduling strategy under the day-ahead electricity price environment, including resource charging and discharging plans and the optimal switch state sequence of the reconstructed distribution network.
[0029] In summary, this invention constructs a pre-planning mechanism based on the spatiotemporal responsiveness of electric vehicles, utilizes data-driven sub-Browser optimization to handle multiple random variables under specific short-term power trading environments, and combines active distribution network reconfiguration technology with industry-specific theoretical capacity adjustment mechanisms to achieve the economical, efficient, and physically secure coordinated operation of virtual power plants with the underlying distribution network.
[0030] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.
Claims
1. A two-stage optimal scheduling method for virtual power plants considering distribution networks, characterized in that, Includes the following steps: S1. Obtain historical operation data, real-time status, distribution network topology parameters, source-load forecast data, and day-ahead spot market time-of-use electricity price data of the virtual power plant and its aggregated distributed energy, and establish a spatiotemporal distribution model of electric vehicle virtual energy storage based on the historical travel trajectory of electric vehicles. S2. Perform two-stage collaborative optimization of virtual power plants. The first stage determines the distributed power access capacity, battery energy storage capacity, number of electric vehicle charging station plugs, and dispatchable power boundary of electric vehicle virtual energy storage at the planning level. The second stage establishes a collaborative scheduling model at the day-ahead operation level that takes into account the cost of flexible load adjustment, electric vehicle battery life loss, spot price incentives, and distribution network physical constraints. S3. Construct a data-driven probability distribution fuzzy set based on the combined constraints of 1-norm and ∞-norm, generate typical scenarios using Latin hypercube sampling and K-means clustering, and correct the confidence interval boundary of the fuzzy set through a rolling update strategy. S4. Establish a two-stage sub-Blu-ray integrated objective function with the goal of minimizing the total system cost. The total cost of the scenario consists of the main grid power purchase cost, the operating cost of dispatchable generators, the charging and discharging scheduling cost of electric vehicles, the flexible load adjustment cost, the aging cost of battery energy storage, the network loss cost, and the distribution network reconfiguration cost. Under the constraints of node voltage, line capacity, power balance, and radial topology, the solution is obtained using a column and constraint generation algorithm, and the output is the resource allocation, day-ahead scheduling, and distribution network reconfiguration strategy.
2. The two-stage optimized scheduling method for virtual power plants considering distribution networks according to claim 1, characterized in that: Distributed energy sources in S1 include solar photovoltaic power plants, wind power plants, dispatchable generators, electric vehicle charging stations, battery energy storage systems, and flexible loads; strong random variables include at least wind speed, solar irradiance, load demand, and day-ahead spot market time-of-use electricity prices.
3. The two-stage optimization scheduling method for virtual power plants considering distribution networks according to claim 1, characterized in that: The first phase of S2 is executed at the system planning level. It uses an optimization method based on the fuzzy maximum-min principle to determine the optimal access capacity of wind and solar generators. Based on the spatiotemporal distribution model of electric vehicles and objective and subjective responsiveness indicators, it determines the extreme boundary of dispatchable charging and discharging power and the minimum number of charging plugs for each charging station. At the same time, it combines load curves and charging and discharging tank selection methods to determine the optimal capacity of the battery energy storage system. The second phase is executed at the day-ahead operation level of the system. It includes establishing a demand response model for flexible loads of different typical industries that is constrained by the heterogeneity theory of adjustment capacity, and establishing a joint response model for virtual energy storage of electric vehicles that takes into account the battery cycle aging cost and spot price incentives.
4. The two-stage optimization scheduling method for virtual power plants considering distribution networks according to claim 3, characterized in that, In the first stage, a fuzzy maximum-minimum comprehensive evaluation index composed of actual power loss index, node voltage and distributed power penetration index is embedded into the capacity configuration model of the planning layer, and the optimal access capacity of wind and solar generators is determined by the two-stage sub-Bruker model in the feasible domain of the first stage. In the first phase of evaluating the virtual energy storage dispatchable boundary of electric vehicles, the state of charge when the vehicle is parked and the expected parking time are extracted. An objective responsiveness is calculated to assess the physical charging and discharging capacity, and a subjective responsiveness is calculated to assess user willingness. The subjective responsiveness is expressed using the following formula: ; ; ; in, For the final subjective response, This is a correction factor for berthing duration. The fundamental response coefficient is determined by the state of charge. For the first i electric vehicles t State of charge during a period of time To estimate the berthing duration, The minimum parking time required to participate in scheduled charging and discharging; based on this, the expression for the upper limit of the aggregated scheduled charging power of the electric vehicle virtual energy storage is: ; in, for t The maximum available charging power for electric vehicles during specific time periods. For the first n The rated charging power of an electric vehicle N EV This represents the total number of electric vehicles aggregated. n For electric vehicles, For the first n electric vehicles t Objective responsiveness over a given time period For the first n electric vehicles t Subjective charging response coefficients related to time period and parking duration.
5. The two-stage optimized scheduling method for virtual power plants considering distribution networks according to claim 3, characterized in that: In the second-stage demand response model, the scheduling cost function for flexible loads participating in peak shaving and valley filling in the spot market is... Based on the theoretical adjustment capacity boundary of typical industries, its expression is as follows: ; in, and These are the unit compensation cost coefficients for loads that can be reduced and loads that can be shifted, respectively. and They are respectively t The total power reduction and shifting during the actual time period; for the textile industry load, extract the theoretical upper limit of its adjustment capacity based on production shift shifting and assembly line characteristics to constrain [the load]. For the load of the education park, the theoretical adjustment capacity boundary is extracted based on its strong temporal regularity and scheduling characteristics to constrain the load. ; In the joint response model, the day-ahead scheduling operation cost of the electric vehicle station aggregator Based on the incentive of charging and discharging electricity prices in the spot market and the battery life loss, its expression is: ; in, This refers to the current day spot electricity price. Service fee per unit of electricity. This is the battery cycle aging cost coefficient. and These represent the charge and discharge power after aggregation within the responsivity boundary. This refers to the power of a rigid charging load that cannot be scheduled.
6. The two-stage optimization scheduling method for virtual power plants considering distribution networks according to claim 1, characterized in that: In S3, the handling of uncertainty propagation in strong random variables abandons the assumption of an exact probability distribution. It generates initial scenarios based on Latin hypercube sampling and uses the K-means clustering algorithm to obtain typical discrete scenarios. On this basis, it constructs a data-driven probability distribution fuzzy set Ψ containing combined constraints of the 1-norm and ∞-norm, with the constraint expression as follows: ; ; in, S This represents the total number of typical scenarios. For the first s The probability to be determined for each typical scenario. For the first s The empirical probability of a typical scenario The radius is constrained by the 1-norm. The radius is constrained by the ∞-norm. N For historical sample size, and Each represents a given confidence level.
7. The two-stage optimization scheduling method for virtual power plants considering distribution networks according to claim 1, characterized in that: The comprehensive objective function in S4 is based on minimizing the total expected operating cost of the system. A column and constraint generation algorithm is used to decompose the two-stage sub-Bruker model into a main problem and sub-problems for iterative solving. Simultaneously, the physical operating constraints of the underlying distribution network are used as rigid boundaries, and distribution network reconfiguration technology is introduced to find the radial topology with the minimum active power loss. The comprehensive objective function is transformed into the expression for minimizing the expected operating cost model under the worst-case probability distribution scenario within the sub-Bruker optimization framework: ; in, x Let p be the probability vector for the planning and allocation decision variables in the first stage. X For the first stage of planning and allocation decision variables x feasible domain, For the scene s The second phase of day-ahead operation and scheduling decision variables, Given the first-stage decision x Second stage variables feasible domain, For the scene s The total operating cost is expressed as follows: ; in, , , , , , and Scenes s The costs include: main grid power purchase cost, dispatchable generator operation cost, electric vehicle charging and discharging dispatch cost, flexible load adjustment cost, battery energy storage aging cost, grid loss cost, and distribution network reconfiguration cost. In addition, the physical operating constraints in S4 include: Node voltage The safety constraints are: ; in, and These are the lower and upper limits of the node voltage, respectively. Line transmission capacity The safety constraints are: ; in, This represents the maximum transmission capacity of the line. j For nodes i Adjacent node numbers; The active power balance constraint at the node is: ; in, For nodes i exist t Active power generation during a given time period For nodes i exist t Active power of load during a given period For nodes i The set of adjacent nodes, for t Time period is determined by nodes i Flow to Node j The active power of the branch; The reactive power balance constraint at the node is: ; in, For nodes i exist t Reactive power generation during a given period For nodes i exist t Reactive power of load during a given period for t Time period is determined by nodes i Flow to Node j The reactive power of the branch circuit.
8. The two-stage optimized scheduling method for virtual power plants considering distribution networks according to claim 7, characterized in that: The radial topology is generated by introducing an improved Newton-Raphson power flow algorithm that includes slack nodes, PQ nodes, and PV nodes. When optimizing the power flow calculation for each time period, all loops formed by closed tie switches are traversed, and power flow calculation is performed by disconnecting one branch at a time. The switch combination that satisfies all physical operating constraints and minimizes active power loss is selected, and the optimal radial distribution network topology under this switch combination is generated during the iteration of the main problem.
