Simulation method for distribution network regulation requirement and potential of electric vehicle cluster

Abstract

The application discloses a power distribution network regulation demand and electric vehicle cluster potential simulation evaluation method, which comprises the following steps: collecting multi-source data, constructing a behavior portrait reflecting the real charging habits of users, and generating a credible simulation scene with two-dimensional randomness of space and time; for a single electric vehicle, the power feasible region of the single electric vehicle is defined through a virtual battery model; for each group, the power feasible region of the single electric vehicle is aggregated through a relaxed Minkowski sum algorithm; the power distribution network time sequence power flow simulation is carried out in the scene without electric vehicle regulation participation, and the power distribution network in the corresponding multi-dimensional regulation demand matrix is output in combination with node voltage-power sensitivity analysis; a double-layer optimization model corresponding to the power distribution network and nodes is constructed, and the node margin interval of each node is output; spatial matching and economic comparison are carried out, and the corresponding charging operation strategy is generated. The application can realize accurate prediction of node regulation demand gap.

Description

Technical Field

[0001] This application relates to the field of power grid regulation, specifically to a simulation evaluation method for distribution network regulation demand and the potential of electric vehicle clusters. Background Technology

[0002] With technological advancements, distributed and flexible resources, exemplified by electric vehicles, are experiencing explosive growth. The number of charging infrastructures is also increasing accordingly. While the large-scale integration of electric vehicles promotes green and low-carbon development through electrification in the transportation sector, the highly random nature of their charging behavior in time and space poses a significant challenge to the safe and stable operation of the power distribution network. Unordered integration of electric vehicles on a large scale can lead to problems such as increased peak-valley load differences in the power distribution network, transformer overload, voltage exceeding limits, and degraded power quality.

[0003] To address this challenge, vehicle-to-grid (V2G) technology has emerged. This technology treats electric vehicles as distributed mobile energy storage units, participating in ancillary services such as peak shaving and frequency regulation of the power grid through bidirectional charging and discharging (V2G). However, in practical engineering applications, how to scientifically assess the regulation demands on the distribution network after large-scale electric vehicle integration, and how to quantify the adjustability potential of electric vehicle clusters as a flexibility resource, have become key bottlenecks restricting the coordinated development of V2G.

[0004] Traditional approaches to evaluating this problem have the following main drawbacks: 1. Fragmentation of assessment dimensions: Traditional distribution network planning and demand-side management are often relatively independent. Planning departments usually expand and upgrade the power grid based on peak load growth, ignoring the substitutability of demand-side resources; while demand-side management focuses on short-term orderly charging guidance, lacking linkage analysis with medium- and long-term power grid planning (such as substation capacity expansion and line upgrading).

[0005] 2. Ambiguity and Node Mismatch in Potential Quantification: Existing research on the regulation potential assessment of electric vehicle clusters often employs static, deterministic methods (such as simply accumulating battery capacity) and generally treats the cluster as a single-point resource connected to the power grid (i.e., a bus model), ignoring the spatial distribution differences of electric vehicles at different physical nodes in the distribution network. This "node mismatch" assessment leads to a situation where, although the dispatch center knows the overall regulation capacity of the cluster, it is difficult to accurately determine whether this capacity can be delivered to the "bottleneck nodes" that are actually overloaded or have voltage exceeding limits, resulting in overly optimistic assessment results.

[0006] 3. Limitations of simulation methods: There is a lack of a system capable of "credible simulation". Most simulation tools struggle to simultaneously simulate the complex interactive characteristics of the convergence of the "transportation network, information network, and power network", especially in accurately simulating the real-time power feasible domain of a massive number of electric vehicles as heterogeneous resources aggregated under power grid dispatch commands, and the feasible power allocation intervals at various nodes of the distribution network. This makes it impossible for the power grid dispatch center to accurately grasp the specific spatial margin of adjustable resources. Summary of the Invention

[0007] To address the aforementioned issues, this application proposes a simulation evaluation method for distribution network regulation demand and the potential of electric vehicle clusters, including: Collect multi-source data, construct a behavioral profile reflecting the user's real charging habits based on the multi-source data, and generate a reliable simulation scenario with spatiotemporal two-dimensional randomness; Based on the aforementioned reliable simulation scenario, for a single electric vehicle, a standardized virtual battery model is transformed according to the battery's physical characteristics and user travel constraints. This virtual battery model is used to define the power feasible region of the single electric vehicle; the power feasible region is the power regulation range and energy boundary that the single electric vehicle can provide when connected to the grid. For individual electric vehicles, the power feasible regions of the individual electric vehicles are grouped, and for each group, the power feasible regions of the individual electric vehicles are aggregated using the relaxed Minkowski sum algorithm to obtain the aggregated feasible region with power allocation information of each node. Based on the aforementioned reliable simulation scenario, a distribution network time-series power flow simulation is performed in a scenario without electric vehicle regulation. Combined with node voltage-power sensitivity analysis, the key nodes and key time periods that cause branch overruns or voltage overruns are located, and the corresponding multi-dimensional regulation demand matrix of the distribution network is output. By using the power flow constraints corresponding to the power flow simulation and the aggregated feasible region, a two-layer optimization model corresponding to the distribution network and nodes is constructed. Uncertainty is handled by the split-blown bar chance constraint, and the node margin interval of each node is output. By spatially matching and economically comparing the node margin range with the adjustment demand matrix, the substitution value of electric vehicles for physical capacity expansion at specific nodes is quantified, and corresponding charging operation strategies are generated.

[0008] In one example, multi-source data is collected, and a behavioral profile reflecting the user's actual charging habits is constructed based on the multi-source data, specifically including: The charging operation platform, vehicle network platform and traffic condition system collect corresponding multi-source data; the multi-source data includes: vehicle arrival time, departure time, state of charge at arrival, expected state of charge at departure, daily mileage, charging power type, and the distribution network node number where the charging pile connected to the vehicle is located. Outliers in the multi-source data are removed by using a local outlier factor, and missing values ​​in the multi-source data are filled by using the k-nearest neighbor algorithm. Based on the multi-source data, users are classified into corresponding user types using a spectral clustering algorithm; For each user type, a spatiotemporal transition model based on a semi-Markov chain is established to describe the probability distribution of the transition probability and state duration of the vehicle between different states. Based on the spatiotemporal transition model, a behavioral profile reflecting the user's actual charging habits is obtained.

[0009] In one example, a reliable simulation scenario with two-dimensional spatiotemporal randomness is generated, specifically including: Based on the spatiotemporal transition model, a joint probability distribution among user behavior variables is constructed using a Copula function; the user behavior variables include: arrival time, initial state of charge, and access node. Based on the joint probability distribution, an initial scenario set containing multiple scenarios is generated through sampling, serving as a reliable simulation scenario with spatiotemporal two-dimensional randomness; a single scenario includes: electric vehicle access status, parking duration, initial state of charge, travel demand electricity and its spatial distribution.

[0010] In one example, based on the aforementioned reliable simulation scenario, for a single electric vehicle, a standardized virtual battery model is generated according to battery physical characteristics and user travel constraints. This virtual battery model is then used to define the power feasible region of the single electric vehicle, specifically including: Based on the access time, departure time, initial battery charge, and user-defined target battery charge at the time of disconnection of a single electric vehicle in the aforementioned reliable simulation scenario, a dynamic equation for battery energy is established. Based on the battery energy dynamic equation and the energy boundary constraints of user travel demand, an energy storage dynamic model for user rigid travel demand is constructed. Based on the energy storage dynamic model, by extending the static energy boundary constraints to a dynamic power boundary, the minimum energy required and the maximum allowable energy of a single electric vehicle at any given time are derived, thereby obtaining the power feasible region for describing the power adjustment range at that time. The power feasible domain and the energy storage dynamic model are described by standardized parameters, and standardized virtual battery parameters of a single electric vehicle are output. The standardized parameters include: rated power, available capacity, maximum charge / discharge rate, connection / disconnection time, and the distribution network node number where the charging pile to which the vehicle is connected is located.

[0011] In one example, individual electric vehicles are grouped, and for each group, the power feasible regions of the individual electric vehicles are aggregated using the relaxed Minkowski sum algorithm to obtain an aggregated feasible region with power allocation information for each node, specifically including: The power feasible region of a single electric vehicle is expanded into a polyhedron in a high-dimensional space on the time axis, and the polyhedron is approximated internally or externally using axis-aligned rectangles to output a set of single feasible regions with regular geometric shapes. For the set of feasible domains for each individual vehicle, each individual electric vehicle is grouped according to the node number of the power distribution network where the charging pile to which the vehicle is connected is located; Treating each group as a single node, the power feasible region of each node is aggregated using the relaxed Minkowski algorithm to obtain the aggregated feasible region with node power allocation information.

[0012] In one example, the method further includes: By using the edge, local data and access node information of connected vehicles are collected in real time to determine the parameters of individual virtual batteries and report them to the edge aggregator. The feasible domain for aggregation is obtained through the edge aggregator and reported to the cloud scheduling center; The real-time adjustment capabilities of the electric vehicle clusters at each node are obtained through the cloud-based dispatch center.

[0013] In one example, based on the aforementioned reliable simulation scenario, a distribution network time-series power flow simulation is performed in a scenario without electric vehicle regulation. Combined with node voltage-power sensitivity analysis, the key nodes and critical time periods leading to branch overruns or voltage overruns are located, and the corresponding multi-dimensional regulation demand matrix of the distribution network is output, specifically including: A power distribution network simulation model is built in the simulation environment. The typical daily load of the credible simulation scenario, excluding electric vehicles, is used as the base load as input to perform power flow simulation and obtain the voltage and branch load rate of each node in each time period, which serves as the reference state of the power distribution network. The charging load of electric vehicles in the disordered charging mode in the trusted simulation scenario is superimposed with the basic load of the corresponding node, and power flow simulation is performed to determine the degree of change and spatial difference of voltage and branch load rate of each node in each time period. If the simulation results exceed the preset safe operation threshold, then the corresponding node, time period or branch is determined to have adjustment requirements; Based on the aforementioned regulation requirements, the node voltage-power sensitivity matrix is ​​obtained by calculating the inverse of the power flow Jacobian matrix, and the four-dimensional regulation requirement matrix of time, space, node, and power is output.

[0014] In one example, a two-layer optimization model corresponding to the distribution network and nodes is constructed using the power flow constraints corresponding to the power flow simulation and the aggregated feasible region. Uncertainty is handled by the split-blown bar chance constraint, and the node margin interval of each node is output, specifically including: Based on the credible simulation scenario, as well as the prediction errors of photovoltaic output and base load, the shape similarity between the corresponding historical error curves is determined by the dynamic time warping algorithm, and the random scenarios in the credible simulation scenario are reduced by the clustering algorithm, resulting in a finite number of discrete uncertainty scenarios and a scenario tree with their corresponding probabilities. The power flow constraints corresponding to the power flow simulation are used as the upper-level model, and the aggregated feasible region is used as the lower-level model to construct a two-level optimization model for the distribution network and nodes; wherein, the node injection power called by the upper-level model belongs to the aggregated feasible region corresponding to the lower-level model. The scenario tree and the two-layer optimization model are used as inputs. The security constraints corresponding to the power grid flow constraints are converted into opportunity constraints. The opportunity constraints are converted into a deterministic convex optimization problem and solved by calling the solver. The node margin interval of each node in each time period is output.

[0015] In one example, the node margin interval is spatially matched and economically compared with the adjustment demand matrix to quantify the substitution value of electric vehicles for physical capacity expansion at a specific node, generating a corresponding charging operation strategy, specifically including: Using the aforementioned adjustment demand matrix as input, the corresponding equivalent physical expansion cost is determined for specific bottleneck nodes. Using the node margin range and the equivalent physical expansion cost as inputs, determine the total node-differentiated cost required to call the node margin range of each node, and obtain the replacement value of the node based on the total node-differentiated cost. Based on the aforementioned alternative value, a corresponding node-differentiated charging operation strategy is generated.

[0016] The simulation evaluation method for distribution network regulation demand and electric vehicle cluster potential proposed in this application can bring the following benefits: 1. Accurately predict the demand gap for node regulation: By constructing a simulation system with multiple time scales of "day-to-day-real-time", the identification of regulation demand can be refined to each physical node and each key branch of the distribution network. It can accurately identify the spatial heterogeneity and flexibility regulation demand (upward / downward power gap) of the distribution network at different time sections after the large-scale access of electric vehicles, providing quantitative basis with spatial coordinates for power grid planning departments.

[0017] 2. Quantifying the Node-Accessible Substitution Value of Electric Vehicle Clusters: A three-layer coupling model of "node-branch-cluster" is proposed, deeply coupling the electric vehicle cluster with the node power balance equations and branch power flow constraints of the distribution network through the Minkowski equation and aggregation to obtain a "super virtual battery". This model not only assesses the total regulation potential of the cluster but also scientifically quantifies the extent to which this potential can be delivered to the "bottleneck nodes" of the grid, thereby replacing traditional physical grid expansion schemes. Through equivalent economic comparison, the model accurately calculates the investment savings in distribution network upgrades that can be achieved by tapping demand-side resources. The "substitution value" of the cluster is defined from the perspective of grid asset investment.

[0018] 3. Improve the economy and safety of vehicle-grid collaborative operation: Through refined evaluation results, guide charging operators to formulate differentiated charging operation strategies for nodes (such as price incentives based on the marginal electricity price of nodes and direct load control for heavily loaded nodes). Under the premise of meeting users' travel needs, maximize the use of the "spatial accessibility" of the cluster's flexibility, strengthen vehicle-grid collaboration, and smooth out the fluctuations of new energy. Attached Figure Description

[0019] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a flowchart illustrating the simulation evaluation method for power distribution network regulation demand and electric vehicle cluster potential in the embodiments of this application. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0021] The technical solutions provided by the various embodiments of this application are described in detail below with reference to the accompanying drawings.

[0022] To address the problems mentioned in the background, this paper provides a quantitative assessment method for the multi-faceted flexibility of electric vehicle (EV) clusters. This method establishes an operational boundary model for EV clusters and embeds it into the joint clearing optimization of the electricity market. Using flexibility cost reduction as a quantitative indicator, and combined with Monte Carlo stochastic scenario simulation, it achieves a probabilistic assessment of the potential of EV clusters to provide diverse ancillary services (frequency regulation, reserve, etc.) at different time periods. However, it still has the following shortcomings: 1. The cluster aggregation results are not coupled with the power flow constraints of specific nodes in the distribution network, and the potential evaluated is an "ideal potential" that is detached from the physical carrying capacity of the power grid.

[0023] 2. The spatial identification of the distribution network regulation demand gap (i.e., which node and which branch needs regulation) is not involved.

[0024] 3. The economic comparison of the cluster's regulation potential as an alternative to the physical power grid expansion was not conducted, and the "alternative value" was not quantified from the perspective of power grid planning.

[0025] Based on this, such as Figure 1 As shown in the embodiments of this application, a simulation evaluation method for distribution network regulation demand and electric vehicle cluster potential is provided, including: S101: Collect multi-source data, construct a behavioral profile reflecting the user's real charging habits based on the multi-source data, and generate a reliable simulation scenario with spatiotemporal two-dimensional randomness.

[0026] This step primarily addresses the issue of the randomness in electric vehicle charging behavior. By cleaning and mining massive amounts of historical data, a behavioral profile reflecting users' real habits is constructed, and a reliable simulation scenario with probabilistic characteristics is generated. This provides realistic input boundaries for subsequent flexible modeling, avoiding evaluation distortion caused by overly idealistic assumptions.

[0027] Specifically, multi-source data is collected through the charging operation platform, vehicle network platform, and traffic condition system. The multi-source data includes: vehicle arrival time, departure time, state of charge upon arrival, expected state of charge upon departure, daily mileage, charging power type (AC / DC), and the distribution network node number where the charging pile connected to the vehicle is located.

[0028] Outliers are removed from multi-source data by using local outlier factors, and missing values ​​in the multi-source data are filled by using the k-nearest neighbor algorithm.

[0029] The Local Outlier Factor (LOF) algorithm was used to identify and remove outliers caused by equipment malfunctions or recording errors. For missing values, an imputation method based on the k-Nearest Neighbors (k-NN) algorithm was used to ensure the integrity of the dataset.

[0030] At this point, a clean, well-organized dataset of bicycle charging event records with spatial location labels can be obtained, providing a high-quality data source for subsequent mining.

[0031] Based on multi-source data, users are categorized into corresponding user types using a spectral clustering algorithm. This algorithm categorizes users into different types, such as commuter users, business users, and private non-commuter users.

[0032] For each user type, a spatiotemporal transition model based on a semi-Markov chain is established to describe the probability distribution of vehicle transitions between different states and the duration of each state. For each user class, a spatiotemporal transition model based on a semi-Markov chain is established. This model can not only describe the transition probabilities of the vehicle between different states (driving, stationary) but also characterize the probability distribution of the duration of each state.

[0033] Based on the spatiotemporal transition model, a behavioral profile reflecting users' actual charging habits is obtained. By refining chaotic user behavior into a computable mathematical model, this behavioral profile reflects users' actual charging habits, enabling simulations to distinguish the differences in the contributions of different user groups to grid regulation.

[0034] Based on the spatiotemporal transition model, a joint probability distribution among user behavior variables is constructed using a Copula function. These user behavior variables include arrival time, initial state of charge, and access node. Based on the joint probability distribution, an initial scenario set containing multiple scenarios is generated through sampling, serving as a reliable simulation scenario with spatiotemporal two-dimensional randomness. A single scenario includes: electric vehicle access status, dwell time, initial state of charge, travel demand electricity, and its spatial distribution.

[0035] Copula functions (such as t-Copula or Clayton-Copula) are used to construct the joint probability distribution among user behavior variables (such as arrival time, initial SOC, and access node). An initial scenario set containing N scenarios is generated through sampling. Each scenario contains the electric vehicle access status, dwell time, initial SOC, travel demand electricity, and their spatial distribution at 96 points (one section every 15 minutes) in the future day.

[0036] Using Copula theory to generate arrival time considerations Initial SOC and access nodes Joint distribution of correlations. Marginal distribution. , and The joint distribution, obtained through nonparametric kernel density estimation, is shown in the formula: .

[0037] in, The moment when an electric vehicle is connected to the power grid; State of Charge (SOC) of the battery when an electric vehicle is connected to the grid. The physical node number of the power distribution network to which the electric vehicle is connected; Time variable; : Charged state variables; Node number variable; :random variable , and The joint cumulative distribution function; Arrival Time The marginal cumulative distribution function; Initial SOC The marginal cumulative distribution function; Access Node The marginal cumulative distribution function; : A three-dimensional Copula function used to connect three marginal distributions to construct a joint distribution, where The correlation coefficient parameter is the Copula function.

[0038] In this way, thousands of reliable electric vehicle access scenarios with spatiotemporal two-dimensional randomness were generated, providing a foundation for subsequent spatial robustness assessment.

[0039] S102: Based on the aforementioned credible simulation scenario, for a single electric vehicle, a standardized virtual battery model is generated according to the battery's physical characteristics and user travel constraints. This virtual battery model is used to define the power feasible region of the single electric vehicle. The power feasible region is the power regulation range and energy boundary that the single electric vehicle can provide when connected to the power grid.

[0040] This step targets individual electric vehicles, transforming their complex battery physics and user mobility constraints into a standardized, easy-to-understand "virtual battery" model. This model defines the power regulation range and energy boundary that an individual electric vehicle can provide when connected to the grid, i.e., its "power feasible region."

[0041] Specifically, based on a trusted simulation scenario, the battery energy dynamic equation is established using the connection time, departure time, initial battery level, and user-defined target battery level for a single electric vehicle upon disconnection. The battery energy dynamic equation is shown in the formula: .

[0042] in, The moment when an electric vehicle is connected to the power grid; The expected time when electric vehicles will be disconnected from the power grid; : The current moment; :exist Battery charge at any given time (unit: kWh); : The time interval for simulation or control (unit: hours or minutes); Charging efficiency ( ); :exist The charging power at any given moment (unit: kW) is a positive value; Discharge efficiency ( ); :exist The discharge power at any given time (unit: kW) is a positive value.

[0043] Based on the battery energy dynamic equation and the energy boundary constraints of user travel demand, a dynamic model of energy storage is constructed to meet the rigid travel demand of users. The energy boundary constraints (meeting user travel demand) are shown in the formula: .

[0044] in, Access time Initial battery capacity (unit: kWh); Departure time Minimum battery capacity required by the user (unit: kWh); The minimum charge capacity allowed by battery technology (usually corresponding to the lower limit of SOC, unit: kWh); The maximum capacity allowed by battery technology (usually corresponding to the upper limit of SOC, unit: kWh).

[0045] At this point, the battery energy dynamic equation is combined with energy boundary constraints to form an energy storage dynamic model, also known as the model based on the vehicle's connection to the grid. departure time Initial power and the user-defined target power level when offline. A generalized energy storage model for a single EV was established.

[0046] Based on the energy storage dynamic model, by extending the static energy boundary constraints to a dynamic power boundary, the minimum energy required and the maximum allowable energy of a single electric vehicle at any given time are derived, thereby obtaining the power feasible region for describing the power adjustment range at that time.

[0047] The energy constraint described above is transformed into a power feasible region at any time t. Introducing the concept of cumulative energy demand, the minimum energy that must be achieved at time t is derived. and the maximum allowed energy As shown in the formula: ; .

[0048] in, :exist At any given time, the minimum amount of energy a battery must achieve to meet future off-grid requirements (unit: kWh). :exist At any given moment, the maximum amount of charge (in kWh) that the battery can continuously discharge from its initial moment is defined. This constraint prevents over-discharge.

[0049] The definition has been described above and will not be repeated here; Maximum charging power of electric vehicles (unit: kW); The maximum discharge power of the electric vehicle (unit: kW). Therefore, the feasible power region at time t is the set of power values ​​that satisfy the energy constraints. The static energy boundary constraint (also known as the SOC constraint) is extended to a dynamic power boundary.

[0050] By using standardized parameters, the feasible power domain and energy storage dynamic model are described, and standardized virtual battery parameters for a single electric vehicle are output. The standardized parameters include: rated power, available capacity, maximum charge / discharge rate, connection / disconnection time, and the distribution network node number where the charging pile to which the vehicle is connected is located.

[0051] A set of standardized parameters is used to summarize the complex model described above (including the power feasible region and energy storage dynamics model), forming an equivalent description of the "virtual battery," which is called the standardized virtual battery parameters for a single electric vehicle. This set of standardized parameters includes: rated power... Available capacity Maximum charge / discharge rate, access / disconnection time, and access node number. .

[0052] The formula for calculating the usable capacity of a virtual battery is: (This formula is for charging scenarios.)

[0053] in, The available capacity of the virtual battery is the maximum amount of electricity that the electric vehicle can flexibly adjust during the current access period (unit: kWh). Departure time Minimum battery capacity required by the user (unit: kWh).

[0054] The three terms in the formula , and These represent: the charging capacity limited by the battery's physical limits, the charging capacity limited by the maximum charging power and time, and the minimum charging amount required to meet user travel needs. The minimum value of these three is the maximum amount of electricity that the vehicle can flexibly participate in grid interaction without affecting user travel.

[0055] In this way, a unified mathematical description of heterogeneous resources is achieved, and key spatial location information is preserved, laying the foundation for node aggregation in subsequent steps.

[0056] S103: Group individual electric vehicles, and for each group, aggregate the power feasible regions of the individual electric vehicles using the relaxed Minkowski sum algorithm to obtain an aggregated feasible region with power allocation information for each node.

[0057] This step primarily involves grouping thousands of individual resources, described by "virtual batteries," according to the physical nodes of the distribution network they are connected to. Then, Minkowski aggregation is performed within each node, ultimately forming a "node-decoupled super virtual battery (referred to as a virtual battery)" that reflects the overall regulation capability of the cluster and includes node power allocation information. This addresses the problem of traditional aggregation models neglecting the spatial distribution of resources.

[0058] Specifically, the power feasible region of a single electric vehicle is expanded into a polyhedron in a high-dimensional space along the time axis. An axis-aligned rectangle is used to approximate the polyhedron internally or externally, outputting a set of geometrically regular single-unit feasible regions. The obtained single-unit power feasible region is expanded along the time axis and represented by a polyhedron in a high-dimensional space. To reduce the computational complexity of subsequent aggregation, an axis-aligned rectangle (zonotope) can be used to approximate the polyhedron internally or externally. In this way, the complex time-varying power range of a single unit is simplified into a geometrically regular set.

[0059] For each individual feasible region set, the individual electric vehicles are grouped according to the distribution network node number where the charging pile to which the vehicle is connected is located. This is based on the access node number of each virtual battery (which is also the distribution network node number where the charging pile to which the vehicle is connected is located). Divide all electric vehicles into There are 1 group, and each group corresponds to a physical node (referred to as a node) in the distribution network.

[0060] Treating each group as a single node, the relaxed Minkowski sum algorithm is used to aggregate the power feasible regions of each node, resulting in an aggregated feasible region with node power allocation information. Within each node group... Internally, the relaxed Minkowski algorithm is used to aggregate the feasible regions of all connected electric vehicles on the node, resulting in the node's aggregated feasible region (also known as the cluster virtual battery model). The aggregation process is executed on the cloud scheduling platform or at the edge aggregator.

[0061] For accessing the same node The feasible domain of two independent resources and The exact feasible region after aggregation Satisfying the formula: .

[0062] in, Access Node The power feasible domain of the first individual resource; Access Node The power feasible domain of the second individual resource; :node The total power feasible region after aggregating all resources; : Represents the Minkowski sum operator; : Each belongs to the feasible region and Any power vector within.

[0063] Apply the relaxation approximation independently to each node group and aggregate the feasible region. This formula can be used for quick calculation: .

[0064] in, Nodes obtained through relaxation of Minkowski and approximation The aggregate feasible region; :node Total charging and discharging power of the electric vehicle cluster (unit: kW); :node The lower bound of the total power of the cluster is the algebraic sum of the lower bounds of the individual power of all nodes on that node; :node The upper limit of the total power of the cluster is the algebraic sum of the upper limits of the power of all individual nodes of this node; :node The lower limit of the cluster's total energy throughput capacity; :node The upper limit of the cluster's total energy throughput capacity.

[0065] In this way, the cluster aggregation results are decoupled from the physical nodes of the distribution network. The final result is a vector set. This is not a single lumped model, but rather a more comprehensive one. This provides a fundamental guarantee for subsequent steps to accurately assess "where the electricity can be adjusted".

[0066] Furthermore, it can also be based on real-time status updates and node mapping based on a cloud-edge collaborative architecture.

[0067] By collecting local data and access node information of connected vehicles in real time through the edge, the parameters of individual virtual batteries are determined and reported to the edge aggregator; through the edge aggregator, the feasible domain for aggregation is obtained and reported to the cloud dispatch center; through the cloud dispatch center, the real-time adjustment capability of the electric vehicle cluster on each node is obtained.

[0068] A cloud-edge collaborative architecture of "charging pile - edge aggregator - cloud" is adopted. The edge (charging pile or community aggregator) collects local data and access node information of connected vehicles in real time, calculates individual virtual battery parameters, and reports them to the edge aggregator. The edge aggregator executes the node grouping Minkowski algorithm and aggregation in step S103, reporting the aggregation feasible domains by node within its region to the cloud dispatch center. The cloud dispatch center thus gains real-time adjustment capabilities for the electric vehicle cluster on each physical node.

[0069] In this way, a near real-time, spatially refined perception and update of cluster flexibility is achieved, enabling the simulation system to dynamically respond to random access, exit, and spatial distribution changes of EVs, ensuring the spatiotemporal accuracy of the evaluation results.

[0070] S104: Based on the aforementioned reliable simulation scenario, perform distribution network time-series power flow simulation in a scenario without electric vehicle regulation participation, and combine node voltage-power sensitivity analysis to locate the key nodes and key time periods that cause branch over-limit or voltage over-limit, and output the corresponding multi-dimensional regulation demand matrix of the distribution network.

[0071] This step is used to assess the "native" operating state of the distribution network after the electric vehicle cluster is connected. By simulating a scenario where electric vehicles do not participate in regulation (i.e., conventional disordered charging), the load curve of the distribution network and the voltage levels of key nodes are obtained. Then, using node sensitivity analysis, the key nodes that cause branch over-limits or voltage over-limits are accurately located, thereby identifying exactly how much power the system is "lacking" and "what direction of regulation is missing" in space.

[0072] Specifically, a distribution network simulation model is built in the simulation environment. The typical daily load in the credible simulation scenario, excluding electric vehicles, is used as the base load as input to perform power flow simulation and obtain the voltage and branch load rate of each node in each time period, which serves as the baseline state of the distribution network.

[0073] Build a distribution network simulation model in a Matlab / OpenDSS or Python / Pandapower environment. Using the typical scenario (excluding EV loads) generated from the aforementioned reliable simulation scenario as the base load input, perform power flow calculations using the forward-backward substitution method or the Newton-Raphson method to obtain the voltage at each node in each time period. Branch load rate This is the baseline state of the "zero-scheme" for the distribution network.

[0074] In a reliable simulation scenario, the charging load of electric vehicles in disordered charging mode is superimposed with the base load of the corresponding node, and power flow simulation is performed to determine the degree of change and spatial difference of voltage and branch load rate of each node in each time period.

[0075] The electric vehicle charging load curves with node affiliation information from the aforementioned reliable simulation scenario (assuming an unordered mode) are superimposed onto the base load. Time-series power flow simulation is then performed again. This time, significant differences in the degree of variation in voltage at different nodes and load rates in different branches can be observed.

[0076] compute nodes The net load is shown in the formula: .

[0077] in, :exist Time, node The total net active power (unit: kW) includes the base load and the disordered charging load of electric vehicles at this node. :exist Time, node Active power of the basic load of the distribution network excluding electric vehicles; :exist Moment, belonging to the node The total active power of unordered charging of electric vehicles.

[0078] In this way, the hotspots caused by the disorderly access of electric vehicles in space can be visually displayed, and the negative impact of EV access at each node can be quantified.

[0079] If the simulation results exceed the preset safe operation threshold, it is determined that there is a regulation requirement for the corresponding node, time period, or branch. Based on the regulation requirement, the node voltage-power sensitivity matrix is ​​obtained by calculating the inverse of the power flow Jacobian matrix, and the four-dimensional regulation requirement matrix of time, space, node, and power is output.

[0080] Thresholds for safe operation of the distribution network are set (e.g., voltage deviation ≤ ±5%, line load rate ≤ 80%). When simulation results under disordered charging scenarios exceed these thresholds, it is determined that there is a need for regulation at that time period, at that node, or on that branch. To accurately calculate how much power needs to be injected at which node to eliminate over-limits, a node voltage-power sensitivity matrix is ​​introduced. By calculating the inverse of the power flow Jacobian matrix, the sensitivity of voltage changes at each node to changes in injected power at each node can be obtained.

[0081] For nodes The minimum active / reactive power regulation required to address the voltage exceeding the lower limit can be approximated by the following formula: .

[0082] in, To eliminate nodes If the voltage exceeds the limit, it needs to be checked at the node. The active power regulation demand injected (or its neighboring nodes) (unit: kW). A positive value indicates that the injection needs to be increased (demand reduction), and a negative value indicates that the injection needs to be reduced (demand increase). To eliminate nodes If the voltage exceeds the limit, it needs to be checked at the node. Injected reactive power regulation demand (unit: kVar); : Obtained in power flow calculation, and related to nodes The voltage-dependent Jacobian submatrix describes the linearized relationship between the node injected power variation and the node voltage variation; :node The difference between the current voltage and the upper or lower safety limit (unit: kV); : Represents the matrix inversion operation.

[0083] For branch roads The overload problem, which requires power reduction (demand reduction), can be broken down to its two endpoints. and As shown in the formula below: ; .

[0084] in, :exist At any time, to eliminate branch roads Overload requires attention at both ends of the node. and The respective reductions in load power (i.e., the upward adjustment of demand, unit: kW); :exist At that moment, the side road Load rate (%); Branch road Maximum permissible safe load rate (%) Branch road Rated capacity (unit: kVA); The node adjustment demand allocation coefficient is determined based on the sensitivity of branch power flow to the injected power at the node, and satisfies... For example, if the overload is caused by a node If the load on one side is dominant, then Larger; the max function ensures that non-zero demand is generated only when the load rate exceeds the limit.

[0085] Thus, the output distribution network's adjustment demand matrix is ​​calculated in four dimensions: time, space, node, and power direction. ,in This is the "bullseye" that guides the assessment of subsequent node coupling potential.

[0086] S105: Construct a two-layer optimization model for the distribution network and nodes by using the power flow constraints corresponding to the power flow simulation and the aggregated feasible region, and handle uncertainty by using the distributed bar chance constraint to output the node margin interval of each node.

[0087] This step is used to evaluate the actual supply capacity (corresponding to the aggregated feasible region) of the virtual batteries of each node cluster after the spatial demand gap has been identified. This step not only evaluates the total regulation capacity of each node cluster, but also evaluates the "achievable regulation margin" of the cluster as a whole under the constraints of node power allocation and grid power flow by constructing a two-layer "distribution network-cluster" model, considering the feasible region of cluster power allocation among each node.

[0088] Specifically, based on the credible simulation scenario, as well as the prediction errors of photovoltaic power output and base load, the shape similarity between the corresponding historical error curves is determined by the dynamic time warping algorithm, and the random scenarios in the credible simulation scenario are reduced by the clustering algorithm, resulting in a finite number of discrete uncertain scenarios and a scenario tree with their corresponding probabilities.

[0089] In addition to the uncertainty of EV behavior, the prediction errors of photovoltaic output and base load are further considered. The Dynamic Time Warping (DTW) algorithm is used to measure the shape similarity of historical error curves, and the K-means clustering algorithm is combined to reduce the massive random scenarios, extracting several representative typical scenarios and their corresponding probabilities, forming a "scenario tree" describing the fluctuations of uncertainty.

[0090] DTW distance is used to compare two time series. and Based on similarity, find the optimal curved path. As shown in the formula: .

[0091] in, Two time series to be compared (e.g., two sets of photovoltaic power output prediction error curves). In the DTW algorithm, a curved path is used for nonlinear matching of points from two time series. : curved path The total length; The first on the curved path Each point represents a sequence. A point and sequence in Pairing of a point in; : The first on the path Points The distance between the two corresponding sequence points (usually Euclidean distance); :sequence and The dynamic time-warped distance between them is determined by finding the optimal matching path. To minimize the cumulative distance.

[0092] In this way, complex uncertainties are transformed into a finite number of discrete scenarios.

[0093] The power flow constraints corresponding to the power flow simulation are used as the upper-level model, and the aggregated feasible region is used as the lower-level model to construct a two-level optimization model for the distribution network and nodes. The node injection power called by the upper-level model belongs to the aggregated feasible region corresponding to the lower-level model.

[0094] A two-level optimization model is constructed to accurately solve for the regulation margin considering spatial accessibility. The upper-level model represents the distribution network dispatch center, and its objective is to maximize the total upward / downward power margin provided by all electric vehicle clusters, or to minimize the regulation cost given a given regulation demand, while satisfying distribution network power flow constraints (especially node voltage and branch power flow constraints). The lower-level model consists of all clustered virtual batteries (i.e., aggregated feasible regions), and its constraints are the clustered power / energy feasible regions of each node. The key coupling constraint is the node injection power called by the upper-layer model. It must belong to the feasible region of the node in the lower-level model. .

[0095] The overall upward adjustment potential maximization target, as shown in the formula in the upper-level model (distribution network dispatch center): .

[0096] Constrained by the power balance equations of distribution network nodes and the power flow equations of branches, as shown in the formula: ; ; .

[0097] The lower-level model (electric vehicle clusters at each node) is shown in the formula: .

[0098] The coupling constraints are shown in the formula: .

[0099] in, : The set of time periods in which adjustment needs exist; The set of all nodes in the power distribution network; :exist Time, node The up-charge power (i.e., discharge / reduction charging power, positive value, unit: kW) is called upon by the electric vehicle cluster. :exist At any time, injected from the upstream power grid into the node Power (unit: kW); :exist At any time, access node The Output of distributed power sources (such as photovoltaics) (unit: kW); :exist Time, node Basic load (unit: kW); :exist At any given moment, the decision-making nodes of the upper-level model... The total charging and discharging power of the electric vehicle cluster (charging is positive, discharging is negative, unit: kW). This variable is key to the interaction between the upper and lower layers; : with nodes A set of connected branches; :exist Time, from the node Outflow branch Active power (unit: kW); :node exist Voltage at any given moment; Branch road exist Load rate at any given time; :node The aggregated feasible region of the electric vehicle cluster; the max function ensures Only during discharge ( Only then is it correct.

[0100] In this way, the model couples the regulation capacity of the power cluster with the node carrying capacity and power allocation feasibility of the power grid. It can not only assess "how much the cluster can regulate", but also "how much the power grid can accept and where it can be transmitted". The assessment results are no longer castles in the air detached from the physical network, but highly reliable and actionable spatial regulation margins.

[0101] Using the scenario tree and the two-layer optimization model as input, the security constraints corresponding to the power grid flow constraints are converted into opportunity constraints, and the opportunity constraints are converted into a deterministic convex optimization problem. The solver is then called to solve the problem and output the node margin interval of each node in each time period.

[0102] Given the difficulty in obtaining the precise probability distribution of uncertainty, a distributed bar chance constraint is introduced. In the two-level optimization model, safety constraints such as voltage and branch power flow are rewritten as chance constraints. The requirement is that the regulation scheme must satisfy the safety constraints with a probability no less than a given confidence level in all scenarios corresponding to the scenario tree. Using dual transformation or linearization techniques, this complex chance constraint is transformed into a deterministic convex optimization problem, which is then solved using Cplex or Gurobi solvers, as shown in the formula: .

[0103] in, The true probability distribution of uncertain parameters (such as photovoltaic output and EV behavior); A "fuzzy set" that contains all possible distributions, constructed based on historical data; In fuzzy sets In the middle, take the infimum of the most unfavorable distribution; Given the EV charging and discharging power at each node In the case of nodes obtained through power flow calculation exist Voltage at any given moment; :node Maximum permissible safe voltage; : Permissible voltage exceedance risk probability.

[0104] The meaning of this formula is that even under the worst probability distribution, the probability that the voltage at each node will not exceed the limit must not be less than [a certain value]. .

[0105] Thus, the final output is not a single potential value, but a potential range that takes into account spatial uncertainties (e.g., at a 90% confidence level, the power at one node can be increased by 0.5-0.8MW at a certain time period, while at another node it can be increased by 0.1-0.2MW). This provides a valuable reference for conservative and refined power grid planning.

[0106] S106: Spatial matching and economic comparison are performed between the node margin interval and the adjustment demand matrix to quantify the substitution value of electric vehicles replacing physical expansion at specific nodes and generate corresponding charging operation strategies.

[0107] This step, as the final value realization stage, spatially matches and economically compares the assessed node refinement potential with traditional power grid planning schemes, quantifies the "substitution value" of electric vehicle clusters at specific nodes, and provides guidance on how to realize this potential through node-differentiated operational strategies. This step is a fundamental extension and concretization of the "cost reduction items" concept in the comparative document, moving from abstract market cost reduction to concrete power grid asset investment substitution.

[0108] Specifically, the adjustment demand matrix is ​​used as input to determine the corresponding equivalent physical expansion cost for specific bottleneck nodes. This is done for identified adjustment demand gaps with spatial attributes (such as those at a certain node). Voltage exceeding limit, or a certain branch (Overload), calculate the cost of traditional physical upgrades required to eliminate this spatial gap, based on the typical cost standards of the local power grid company. This includes: nodes. Transformer capacity expansion costs, branch circuits The costs of line renovation, land acquisition, etc., are converted into annualized costs. or Thus, a benchmark is established for assessing the value of spatially heterogeneous demand-side resources.

[0109] Using node margin ranges (also known as adjustment potential) and equivalent physical expansion costs as inputs, the total differentiated cost required to call up the node margin range of each node is determined, and the substitution value of that node is obtained based on the total differentiated cost. The adjustment potential of each node cluster (e.g., at bottleneck nodes) is then assessed. Peak load that can be reduced MW (Mean Power Scale) is considered an alternative to physical expansion of the node. The node-differentiated cost required to utilize this potential is calculated. This includes: incentive fees paid to users in the node area, operating costs of aggregators, and additional battery wear costs caused by V2G charging and discharging.

[0110] Calculate the calling node The substitution value of electric vehicle clusters As shown in the formula: ; .

[0111] in, : Calling Node Total annualized cost required to realize the adjustment potential of electric vehicle clusters (unit: yuan / year); :exist During the time period, payments are made to the nodes. The incentive electricity price for regional users (unit: yuan / kWh) can be set differently based on the supply and demand tension at that node; :exist Time period, from node The actual regulating power called by the cluster (unit: kW); : Duration of this period (unit: hours); : Due to the calling node The annualized cost of additional battery wear caused by adjusting the cluster (unit: yuan / year). To solve the node Annualized investment cost for physical expansion due to capacity / voltage issues (unit: yuan / year); : Utilizing nodes The net benefit of replacing the physical expansion of the node with an electric vehicle cluster is the node replacement value (unit: yuan / year). If the value is positive, it means that using EV regulation is more economical than physical expansion at the node.

[0112] The total substitution value of the entire network is shown in the formula: .

[0113] in, This is the set of all nodes with significant adjustment demand gaps.

[0114] Thus, from the perspective of power grid asset investment, the economic viability and spatial distribution of "replacing construction with management" have been demonstrated. Power grid planning departments can intuitively see at which node it is most cost-effective to invest in vehicle-grid interaction, and which node still needs capacity expansion. This provides direct and quantifiable evidence for saving investment in upgrading and transforming distribution power grids.

[0115] Based on substitution value, corresponding node-differentiated charging operation strategies are generated. The evaluation results are then fed back to the charging operator's decision-making system. If a certain node... Substitute value If the substitution value is significantly positive, a high-compensation incentive strategy (such as a high discharge subsidy) is formulated for that node to guide users to discharge or charge during off-peak hours at that node. If the substitution value of a node is negative or its potential is insufficient, but the grid demand is urgent, a "direct load control" or "V2G forced dispatch" protocol is initiated at that node (which needs to be signed with the user in advance). Through model predictive control rolling optimization, under the premise of satisfying grid constraints, the differentiated allocation of operating costs among nodes is achieved, maximizing the total substitution value of the entire network. .

[0116] In this way, a spatially refined closed loop is formed. The evaluation results directly guide different nodes to adopt different operational strategies, realizing the strengthening of vehicle-network collaboration from "one-size-fits-all" to "one policy per node" through "evaluation-decision-control".

[0117] 1. Accurately predict the demand gap for node regulation: By constructing a simulation system with multiple time scales of "day-to-day-real-time", the identification of regulation demand can be refined to each physical node and each key branch of the distribution network. It can accurately identify the spatial heterogeneity and flexibility regulation demand (upward / downward power gap) of the distribution network at different time sections after the large-scale access of electric vehicles, providing quantitative basis with spatial coordinates for power grid planning departments.

[0118] 2. Quantifying the Node-Accessible Substitution Value of Electric Vehicle Clusters: A three-layer coupling model of "node-branch-cluster" is proposed, deeply coupling the electric vehicle cluster with the node power balance equations and branch power flow constraints of the distribution network through the Minkowski equation and aggregation to obtain a "super virtual battery". This model not only assesses the total regulation potential of the cluster but also scientifically quantifies the extent to which this potential can be delivered to the "bottleneck nodes" of the grid, thereby replacing traditional physical grid expansion schemes. Through equivalent economic comparison, the model accurately calculates the investment savings in distribution network upgrades that can be achieved by tapping demand-side resources. The "substitution value" of the cluster is defined from the perspective of grid asset investment.

[0119] 3. Improve the economy and safety of vehicle-grid collaborative operation: Through refined evaluation results, guide charging operators to formulate differentiated charging operation strategies for nodes (such as price incentives based on the marginal electricity price of nodes and direct load control for heavily loaded nodes). Under the premise of meeting users' travel needs, maximize the use of the "spatial accessibility" of the cluster's flexibility, strengthen vehicle-grid collaboration, and smooth out the fluctuations of new energy.

[0120] In one embodiment, taking a new urban district as an example, the district is expected to have 20,000 electric vehicles by 2026. After applying this application: 1. Through step S101, it was identified that 70% of the users in the area are regular commuters, with a nighttime home parking rate as high as 90%, and 60% of the vehicles are concentrated at the power distribution network nodes of the three residential areas A, B, and C.

[0121] 2. Simulation in step S104 revealed that without control, the load rate of the 10kV line connected to node A will reach 112% during the evening peak hours of 19:30-21:00; the voltage of node C will drop to 0.92pu, which is lower than the lower limit of 0.93pu.

[0122] 3. Based on the evaluation in step S105, at a 95% confidence level, the cluster of node A can provide an upward adjustment potential of 2MW at night (discharge / reduction of charging); the cluster of node C can provide an upward adjustment potential of 1.5MW; while the cluster of node B mainly has a downward adjustment potential (increase in charging).

[0123] 4. Through the comparison in step S106, solving the overload problem of node A requires a physical expansion investment of 5 million yuan; while utilizing the 2MW EV upgrade potential of node A only requires an annual incentive fee of approximately 300,000 yuan. Solving the voltage problem of node C requires the installation of a Static Synchronous Compensator (STATCOM) costing 1 million yuan, while utilizing its EV potential only requires 150,000 yuan annually. The replacement value of node A is calculated. (Net Present Value), the substitution value of node C (Net present value).

[0124] Final conclusion: It is recommended to postpone the physical expansion investment for node A and launch a pilot project for vehicle-to-grid interaction based on V2G; for node C, since the substitution value is negative, it is recommended to continue using the traditional reactive power compensation scheme. This node-differentiated decision is expected to save 2 million yuan in initial investment and verifies the scientific nature and practicality of this application in spatial fine-grained planning.

[0125] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A simulation evaluation method for distribution network regulation demand and electric vehicle cluster potential, characterized in that, include: Collect multi-source data, construct a behavioral profile reflecting the user's real charging habits based on the multi-source data, and generate a reliable simulation scenario with spatiotemporal two-dimensional randomness; Based on the aforementioned reliable simulation scenario, for a single electric vehicle, a standardized virtual battery model is transformed according to the battery's physical characteristics and user travel constraints. This virtual battery model is used to define the power feasible region of the single electric vehicle; the power feasible region is the power regulation range and energy boundary that the single electric vehicle can provide when connected to the grid. For individual electric vehicles, the power feasible regions of the individual electric vehicles are grouped, and for each group, the power feasible regions of the individual electric vehicles are aggregated using the relaxed Minkowski sum algorithm to obtain the aggregated feasible region with power allocation information of each node. Based on the aforementioned reliable simulation scenario, a distribution network time-series power flow simulation is performed in a scenario without electric vehicle regulation. Combined with node voltage-power sensitivity analysis, the key nodes and key time periods that cause branch overruns or voltage overruns are located, and the corresponding multi-dimensional regulation demand matrix of the distribution network is output. By using the power flow constraints corresponding to the power flow simulation and the aggregated feasible region, a two-layer optimization model corresponding to the distribution network and nodes is constructed. Uncertainty is handled by the split-blown bar chance constraint, and the node margin interval of each node is output. By spatially matching and economically comparing the node margin range with the adjustment demand matrix, the substitution value of electric vehicles for physical capacity expansion at specific nodes is quantified, and corresponding charging operation strategies are generated.

2. The method according to claim 1, characterized in that, Collect multi-source data and construct a behavioral profile reflecting the user's real charging habits based on the multi-source data, specifically including: The charging operation platform, vehicle network platform and traffic condition system collect corresponding multi-source data; the multi-source data includes: vehicle arrival time, departure time, state of charge at arrival, expected state of charge at departure, daily mileage, charging power type, and the distribution network node number where the charging pile connected to the vehicle is located. Outliers in the multi-source data are removed by using a local outlier factor, and missing values ​​in the multi-source data are filled by using the k-nearest neighbor algorithm. Based on the multi-source data, users are classified into corresponding user types using a spectral clustering algorithm; For each user type, a spatiotemporal transition model based on a semi-Markov chain is established to describe the probability distribution of the transition probability and state duration of the vehicle between different states. Based on the spatiotemporal transition model, a behavioral profile reflecting the user's actual charging habits is obtained.

3. The method according to claim 2, characterized in that, Generate a reliable simulation scenario with two-dimensional spatiotemporal randomness, specifically including: Based on the spatiotemporal transition model, a joint probability distribution among user behavior variables is constructed using a Copula function; the user behavior variables include: arrival time, initial state of charge, and access node. Based on the joint probability distribution, an initial scenario set containing multiple scenarios is generated through sampling, serving as a reliable simulation scenario with spatiotemporal two-dimensional randomness; a single scenario includes: electric vehicle access status, parking duration, initial state of charge, travel demand electricity and its spatial distribution.

4. The method according to claim 2, characterized in that, Based on the aforementioned reliable simulation scenario, for a single electric vehicle, a standardized virtual battery model is generated according to the battery's physical characteristics and user travel constraints. This virtual battery model is then used to define the power feasible region of the single electric vehicle, specifically including: Based on the access time, departure time, initial battery charge, and user-defined target battery charge at the time of disconnection of a single electric vehicle in the aforementioned reliable simulation scenario, a dynamic equation for battery energy is established. Based on the battery energy dynamic equation and the energy boundary constraints of user travel demand, an energy storage dynamic model for user rigid travel demand is constructed. Based on the energy storage dynamic model, by extending the static energy boundary constraints to a dynamic power boundary, the minimum energy required and the maximum allowable energy of a single electric vehicle at any given time are derived, thereby obtaining the power feasible region for describing the power adjustment range at that time. The power feasible domain and the energy storage dynamic model are described by standardized parameters, and standardized virtual battery parameters of a single electric vehicle are output. The standardized parameters include: rated power, available capacity, maximum charge / discharge rate, connection / disconnection time, and the distribution network node number where the charging pile to which the vehicle is connected is located.

5. The method according to claim 1, characterized in that, For individual electric vehicles, the power feasible regions of each individual electric vehicle are grouped. For each group, the power feasible region of the individual electric vehicle is aggregated using the relaxed Minkowski sum algorithm to obtain an aggregated feasible region with power allocation information for each node, specifically including: The power feasible region of a single electric vehicle is expanded into a polyhedron in a high-dimensional space on the time axis, and the polyhedron is approximated internally or externally using axis-aligned rectangles to output a set of single feasible regions with regular geometric shapes. For the set of feasible domains for each individual vehicle, each individual electric vehicle is grouped according to the node number of the power distribution network where the charging pile to which the vehicle is connected is located; Treating each group as a single node, the power feasible region of each node is aggregated using the relaxed Minkowski algorithm to obtain the aggregated feasible region with node power allocation information.

6. The method according to claim 5, characterized in that, The method further includes: By using the edge, local data and access node information of connected vehicles are collected in real time to determine the parameters of individual virtual batteries and report them to the edge aggregator. The feasible domain for aggregation is obtained through the edge aggregator and reported to the cloud scheduling center; The real-time adjustment capabilities of the electric vehicle clusters at each node are obtained through the cloud-based dispatch center.

7. The method according to claim 1, characterized in that, Based on the aforementioned reliable simulation scenario, a distribution network time-series power flow simulation is performed in a scenario without electric vehicle regulation. Combined with node voltage-power sensitivity analysis, the key nodes and critical time periods leading to branch overruns or voltage overruns are located, and a multi-dimensional regulation demand matrix for the distribution network is output, specifically including: A power distribution network simulation model is built in the simulation environment. The typical daily load of the credible simulation scenario, excluding electric vehicles, is used as the base load as input to perform power flow simulation and obtain the voltage and branch load rate of each node in each time period, which serves as the reference state of the power distribution network. The charging load of electric vehicles in the disordered charging mode in the trusted simulation scenario is superimposed with the basic load of the corresponding node, and power flow simulation is performed to determine the degree of change and spatial difference of voltage and branch load rate of each node in each time period. If the simulation results exceed the preset safe operation threshold, then the corresponding node, time period or branch is determined to have adjustment requirements; Based on the aforementioned regulation requirements, the node voltage-power sensitivity matrix is ​​obtained by calculating the inverse of the power flow Jacobian matrix, and the four-dimensional regulation requirement matrix of time, space, node, and power is output.

8. The method according to claim 1, characterized in that, Based on the power flow constraints corresponding to the power flow simulation and the aggregated feasible region, a two-level optimization model corresponding to the distribution network and nodes is constructed. Uncertainty is handled by the split-blown bar chance constraint, and the node margin interval of each node is output, specifically including: Based on the credible simulation scenario, as well as the prediction errors of photovoltaic output and base load, the shape similarity between the corresponding historical error curves is determined by the dynamic time warping algorithm, and the random scenarios in the credible simulation scenario are reduced by the clustering algorithm, resulting in a finite number of discrete uncertainty scenarios and a scenario tree with their corresponding probabilities. The power flow constraints corresponding to the power flow simulation are used as the upper-level model, and the aggregated feasible region is used as the lower-level model to construct a two-level optimization model for the distribution network and nodes; wherein, the node injection power called by the upper-level model belongs to the aggregated feasible region corresponding to the lower-level model. The scenario tree and the two-layer optimization model are used as inputs. The security constraints corresponding to the power grid flow constraints are converted into opportunity constraints. The opportunity constraints are converted into a deterministic convex optimization problem and solved by calling the solver. The node margin interval of each node in each time period is output.

9. The method according to claim 1, characterized in that, By spatially matching and economically comparing the node margin interval with the adjustment demand matrix, the substitution value of electric vehicles for physical capacity expansion at specific nodes is quantified, and corresponding charging operation strategies are generated, specifically including: Using the aforementioned adjustment demand matrix as input, the corresponding equivalent physical expansion cost is determined for specific bottleneck nodes. Using the node margin range and the equivalent physical expansion cost as inputs, determine the total node-differentiated cost required to call the node margin range of each node, and obtain the replacement value of the node based on the total node-differentiated cost. Based on the aforementioned alternative value, a corresponding node-differentiated charging operation strategy is generated.

Images