A method, system, device and medium for dynamic cluster modeling of electric vehicles in multiple scenarios
By using Markov state transition models and probability sampling methods, and leveraging the macroscopic proportion parameters of electric vehicles, the state transition of the cluster is dynamically characterized, solving the problem of data loss in existing technologies. This enables high-precision multi-scenario electric vehicle cluster modeling, which is suitable for accurate assessment and planning of power distribution networks.
Patent Information
- Application Number
- CN202611291295.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-25
- Publication Date
- 2026-11-17
AI Technical Summary
Existing electric vehicle cluster modeling methods heavily rely on real-time data of all individual vehicles, which cannot be obtained in the actual environment of the power distribution network. This results in data gaps, an inability to accurately characterize the load impact characteristics and cluster state transitions in multiple scenarios, and a lack of dynamic description capabilities.
A two-stage cluster partitioning method based on Markov state transition model is adopted. By using macro-proportion parameters of electric vehicle traffic use and vehicle type, the cluster state transition is dynamically characterized through probability sampling and Markov state transition model, and a multi-scenario electric vehicle dynamic cluster modeling system is constructed.
This technology enables high-precision, low-threshold multi-scenario electric vehicle cluster modeling in environments with limited information, accurately quantifies cluster state evolution, improves the objectivity and robustness of cluster distributed computing, and is suitable for power grid carrying capacity assessment and charging facility planning.
Smart Images

Figure CN123115662A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of dynamic cluster modeling technology for electric vehicles, and in particular to a dynamic cluster modeling method, system, device and medium for electric vehicles in multiple scenarios. Background Technology
[0002] Electric vehicles, as the core carrier of low-carbon transportation, are experiencing explosive growth in their numbers. However, with large-scale integration of electric vehicles into the power grid, their charging behavior exhibits significant spatial randomness and temporal uncertainty, posing a severe challenge to the safe and stable operation of the power distribution network. Establishing an accurate electric vehicle cluster aggregation model is a key technology for achieving orderly charging control and grid dispatch support. However, existing electric vehicle cluster modeling and grouping methods still have significant shortcomings in practical applications: First, existing models heavily rely on full real-time data, resulting in massive computational demands. Traditional cluster modeling methods typically require high-precision real-time GPS trajectories for each electric vehicle, complete historical charging preference data, and constant battery state of charge. However, in actual power distribution network operation environments, limitations such as platform barriers and user privacy protection make it impossible to obtain such massive and complete micro-level data for individual vehicles. This severe data gap leads to application difficulties for traditional modeling methods in practical engineering. Second, there is a lack of multi-scenario characterization capabilities under data-constrained conditions. Existing solutions often treat all connected vehicles as homogeneous loads. This approach ignores the fundamental differences in parking time, average daily mileage, and application scenarios among vehicles with different transportation uses, such as logistics delivery vehicles, private commuter vehicles, and ride-hailing vehicles, resulting in distorted cluster load curves that fail to accurately reflect the load impact characteristics of different functional areas. Finally, existing clustering methods are mostly static, lacking dynamic description mechanisms for physical state transitions. Existing cluster partitioning methods typically rely on static classification based on a specific SOC threshold at a given moment, failing to establish a mathematical evolution model that describes the state transitions of cluster members over time and with varying power levels. This inability to track the evolution of internal cluster states leads to delays in long-term load forecasting and cluster characteristic characterization.
[0003] Existing technologies focus on the characteristics of electric vehicle user travel chains, analyze the Markov characteristics of charging and discharging behavior, and construct a cluster load evolution model and evaluation system for grid dispatch. This method discretizes the charge state through multi-state interval partitioning units, calculates the transition matrix using a transition probability derivation unit combined with historical data, establishes the load evolution equation through a cluster aggregation modeling unit, and finally quantifies the controllable range of the cluster's upward and downward adjustment capacity using an adjustable capacity assessment unit. This method provides a theoretical basis for cluster load prediction and regulation with detailed travel data. However, it has certain limitations: the model highly depends on individual vehicle travel chains and underlying real-time operational data, making data acquisition extremely difficult in the limited information environment of the actual distribution network. It is difficult to achieve rapid cluster modeling for multiple scenarios based solely on macro-level parameters such as vehicle type and traffic use. Other existing technologies analyze the response characteristics of individual electric vehicles, proposing three access states and four response modes; using finite state sub-intervals to describe the state distribution of large-scale electric vehicle clusters, thereby evaluating the cluster's output power and adjustable capacity; and finally designing a control signal composed of two probability values, controlling all vehicles to receive the same probability signal to reduce the complexity of large-scale cluster control and communication requirements. This method focuses on simplification of the control signal transmission process, treating the massive cluster as a homogeneous response whole. It fails to isolate and characterize the energy consumption characteristics and task constraints of various vehicle types, such as logistics vehicles, ride-hailing vehicles, and private cars, in different application scenarios. Furthermore, the model lacks mathematical tools for cluster internal state evolution driven by macroscopic parameters, making it unable to dynamically deduce state transitions over time under data-constrained conditions. This makes it difficult to directly apply to refined assessment of the carrying capacity of distribution networks in multiple scenarios lacking underlying interaction. Other existing technologies provide a theoretical basis for load forecasting with detailed travel data. However, they heavily rely on individual vehicle travel chains and underlying real-time operational data, making data acquisition extremely difficult in the limited information environment of actual distribution networks. This makes it difficult to achieve rapid cluster modeling in multiple scenarios based solely on macroscopic proportion parameters. Other methods utilize finite state sub-intervals to evaluate cluster distribution and simplify control signals, effectively reducing the complexity and communication requirements of large-scale control. However, they treat the cluster as a homogeneous whole, failing to characterize the energy consumption characteristics of various vehicle types and lacking mathematical tools for internal evolution driven by macroscopic parameters, making it impossible to deduce state transitions under data constraints.
[0004] Therefore, there is an urgent need to develop a multi-scenario dynamic cluster modeling method for electric vehicles that is adapted to environments with limited information. This method should rely solely on macro-level parameters such as vehicle type and traffic use, combined with Markov state transition tools, to fundamentally overcome the bottleneck of micro-level data acquisition and achieve refined characterization and dynamic deduction of state transitions in heterogeneous clusters across multiple scenarios. Summary of the Invention
[0005] In view of the above-mentioned existing problems, the present invention provides a dynamic cluster modeling method, system, device and medium for electric vehicles in multiple scenarios, which solves the problems that existing electric vehicle cluster modeling methods rely on the full range of operating parameters of a single vehicle, do not fully consider the differentiated energy consumption and travel characteristics under multiple application scenarios, and generally suffer from high data acquisition costs, large computational complexity and insufficient characterization of state transition uncertainties.
[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a dynamic cluster modeling method for electric vehicles in multiple scenarios, including: Obtain macro-level percentage parameters of electric vehicle categories within the target area; Based on the macro-proportion parameters of electric vehicles in the target area, the category of each electric vehicle is determined by probability sampling method, and the initial state of charge of each electric vehicle after completing its trip is calculated based on the statistical distribution of the travel patterns and energy consumption characteristics of various types of electric vehicles. Based on the initial state of charge, all electric vehicles connected to the grid are divided and the initial static cluster distribution is obtained. Based on the initial static cluster distribution, a Markov state transition model is constructed. Using the initial static cluster distribution as the initial state vector, the Markov state transition model is input for a first state transition, resulting in a second dynamically corrected cluster state distribution, thus realizing dynamic cluster modeling of electric vehicles in multiple scenarios.
[0007] As a preferred embodiment of the dynamic cluster modeling method for electric vehicles in multiple scenarios described in this invention, wherein: Determining the category of each electric vehicle using probability sampling methods includes: The macro-level percentage parameters of electric vehicles within the target area include the macro-level percentage parameters of electric vehicles used for transportation within the target area and the macro-level percentage parameters of electric vehicle types within the target area; Based on the macro-level proportion of electric vehicles in the target area, the traffic use categories of each electric vehicle are obtained through probability sampling. Based on the macro-proportion parameters of electric vehicle types within the target area, the physical type of each electric vehicle is obtained through probability sampling. The category of each electric vehicle is determined by combining its traffic use category and vehicle physical type.
[0008] The beneficial effect of this preferred technical solution is that it allocates vehicle categories based on macro-proportion sampling, without requiring individual vehicle data, and takes into account multi-dimensional vehicle differences.
[0009] As a preferred embodiment of the dynamic cluster modeling method for multi-scenario electric vehicles described in this invention, the calculation of the initial state of charge of each electric vehicle after it completes its trip and connects to the power grid includes: Based on the vehicle type of each electric vehicle, the battery capacity of the corresponding vehicle is determined by sampling from the probability distribution of battery capacity of the corresponding vehicle category. Based on the traffic use type of each electric vehicle, the daily travel distance of the corresponding vehicle is determined by sampling from the probability distribution of daily travel distance of the corresponding vehicle category. Based on the vehicle type of each electric vehicle, the energy consumption per unit mile is determined by sampling from the probability distribution of energy consumption per unit mile of the corresponding vehicle category. Obtain the initial state of charge of the corresponding electric vehicle before travel; Based on battery capacity, daily travel distance, energy consumption per unit mileage, and initial state of charge before travel, the initial state of charge of the corresponding electric vehicle after completing its trip is calculated according to the linear relationship between state of charge consumption and travel distance.
[0010] The beneficial effect of this preferred technical solution is that it calculates the network access SOC based on multi-dimensional probability sampling, which is in line with the actual energy consumption and travel characteristics of vehicles.
[0011] As a preferred embodiment of the dynamic cluster modeling method for multi-scenario electric vehicles described in this invention, the method involves dividing all connected electric vehicles into groups and statistically obtaining the initial static cluster distribution, including: Set power boundary thresholds, including a high power availability threshold and a low power safety threshold; In response to the initial state of charge of electric vehicles being less than the low-charge safety threshold, the corresponding electric vehicles are classified into a power shortage cluster. In response to the fact that the initial state of charge of an electric vehicle is greater than or equal to the low-charge safety threshold and less than or equal to the high-charge availability threshold, the corresponding electric vehicle is classified into a leisure cluster. In response to an electric vehicle's initial state of charge being greater than a high battery availability threshold, the corresponding electric vehicle is classified into a fully charged cluster. The number of electric vehicles in each cluster is counted to obtain the initial static cluster distribution.
[0012] As a preferred embodiment of the dynamic cluster modeling method for multi-scenario electric vehicles described in this invention, the construction of the Markov state transition model includes: The state spaces of the Markov state transition model are defined as the power shortage cluster, the leisure cluster, and the fully charged cluster. Define a set of state transition paths, which include at least the following: transition from a power-deficient cluster to a leisure cluster, a leisure cluster to its own self-sustaining transition, a leisure cluster to a fully charged cluster, a fully charged cluster to its own self-sustaining transition, and a fully charged cluster to a leisure cluster.
[0013] The beneficial effect of this preferred technical solution is that it characterizes the dynamic evolution behavior of EV clusters by defining multiple cluster states and transition paths.
[0014] As a preferred embodiment of the dynamic cluster modeling method for multi-scenario electric vehicles described in this invention, it further includes: Based on the current state of charge of each electric vehicle, the transition probability of each state transition path is determined by fitting a regression model. Construct a Markov state transition matrix based on each state transition path and its corresponding transition probability.
[0015] As a preferred embodiment of the dynamic cluster modeling method for multi-scenario electric vehicles described in this invention, the following steps are taken: Inputting the Markov state transition model and performing a single state transition to obtain the cluster state distribution after secondary dynamic correction includes: The initial state vector is constructed using the ratio of the number of vehicles in each type of cluster to the total number of vehicles entering the network in the initial static cluster distribution. Multiply the initial state vector by the Markov state transition matrix and perform a state transition calculation to obtain the state vector after secondary correction. Based on the state vector after the second correction and the total number of vehicles entering the network, the number of vehicles in each type of cluster after the second dynamic correction is calculated, and the cluster state distribution after the second dynamic correction is obtained.
[0016] Secondly, the present invention provides a dynamic cluster modeling system for electric vehicles in multiple scenarios, comprising: The data acquisition module is used to obtain macro-level percentage parameters of electric vehicles by category within the target area; The initial state of charge calculation module is used to determine the category of each electric vehicle based on the macro-proportion parameters of electric vehicle categories in the target area using a probability sampling method, and to calculate the initial state of charge of each electric vehicle after it completes its trip and is connected to the power grid based on the statistical distribution of the travel patterns and energy consumption characteristics of various types of electric vehicles. The initial static cluster partitioning module is used to partition all electric vehicles connected to the grid according to the initial state of charge and to obtain the initial static cluster distribution. The model building module is used to build a Markov state transition model based on the initial static cluster distribution. The dynamic cluster correction module is used to take the initial static cluster distribution as the initial state vector, input the Markov state transition model to perform a state transition, and obtain the cluster state distribution after secondary dynamic correction, so as to realize dynamic cluster modeling of electric vehicles in multiple scenarios.
[0017] Thirdly, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, characterized in that the processor executes the computer program to implement the steps of the dynamic cluster modeling method for multi-scenario electric vehicles.
[0018] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that the computer program, when executed by a processor, implements the steps of the dynamic cluster modeling method for multi-scenario electric vehicles.
[0019] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention constructs a multi-scenario electric vehicle cluster modeling system adapted to environments with limited information. For the first time, it uses only two macro-level proportion parameters—electric vehicle traffic use and vehicle type—as the sole input to complete the characterization of cluster travel patterns, energy consumption characteristics, and distribution features. This fills the gap in traditional electric vehicle cluster modeling, which heavily relies on real-time travel trajectories of individual vehicles and full charging / discharging data, making it unsuitable for engineering scenarios with insufficient data. This significantly lowers the data threshold for cluster modeling and better meets the practical engineering application needs such as power grid carrying capacity assessment and charging facility planning. This invention adopts a two-stage cluster partitioning method based on SOC threshold static cluster partitioning combined with Markov quadratic dynamic correction. It quantifies state transition probabilities based on the Logistic model, eliminating the need for users to manually input behavioral parameters, charging / discharging periods, or other additional information. It accurately quantifies the dynamic evolution of cluster states caused by user charging / discharging behavior. Compared with traditional static cluster partitioning and fixed behavioral assumptions, this invention effectively solves the problem of large deviations between cluster partitioning results and actual operating characteristics, significantly improving the objectivity, accuracy, and robustness of cluster distribution calculation results. This invention requires only two macro-level input parameters: the proportion of electric vehicles used for transportation in the target area and the proportion of different vehicle types. It models travel and energy consumption by purpose and vehicle type, and uses Monte Carlo batch sampling to calculate the initial State of Charge (SOC) for each vehicle entering the grid. A two-stage clustering method is employed to accurately characterize electric vehicle clusters across multiple scenarios. The final output shows the number and distribution characteristics of vehicles in three clusters: low-power, normal, and fully charged. This invention achieves low-threshold, high-precision, and highly adaptable modeling of electric vehicle clusters in multi-scenario environments with limited information. It effectively overcomes the core shortcomings of traditional modeling methods, such as high data dependence, insufficient dynamic characterization capabilities, and poor scenario adaptability. It can provide accurate input data for assessing the carrying capacity of electric vehicles in power distribution networks, optimizing the layout of charging networks, and scheduling power grid demand response. Attached Figure Description
[0020] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0021] Figure 1 A schematic diagram of the overall process logic of a dynamic cluster modeling method for electric vehicles in multiple scenarios provided in one embodiment of the present invention; Figure 2 A flowchart illustrating the dynamic cluster modeling process for electric vehicles in multiple scenarios, provided as an embodiment of the present invention. Figure 3 This is the initial static cluster partitioning of a dynamic cluster modeling method for electric vehicles in multiple scenarios, provided in one embodiment of the present invention. Detailed Implementation
[0022] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.
[0023] Example 1, referring to Figures 1-3 Tables 1 and 2 illustrate an embodiment of the present invention, providing a dynamic cluster modeling method for electric vehicles in multiple scenarios. Figure 2 As shown, it includes: S100: Obtain the macro-level percentage parameters of electric vehicles by category within the target area; In this embodiment of the invention, the macro-proportion parameter of electric vehicle category in the target area includes the macro-proportion parameter of electric vehicle traffic use in the target area and the macro-proportion parameter of electric vehicle type in the target area; The travel patterns of electric vehicles (EVs) are influenced by their intended use. According to market statistics, vehicles can be categorized into three types based on their transportation purpose, including private cars primarily used for commuting between home and work (HBW type), which account for a significant proportion. Private cars (HBO type) mainly used by retirees and the unemployed account for the following proportion: Non-private vehicles, such as company vehicles and public transportation vehicles (NHB type), account for the following proportion: The proportion of vehicles in the target area for each of the three transportation uses can be obtained through market statistics research. This classification of transportation uses is mainly used to obtain the distribution patterns of electric vehicle travel time and distance.
[0024] The energy consumption characteristics of electric vehicles are influenced by their vehicle type. According to market statistics, vehicles can be categorized into four types, including T1 type, which refers to four-wheeled small cargo vehicles with a maximum load capacity of 400kg or 500kg, accounting for [percentage missing]. Type P1 refers to four-wheeled passenger vehicles with a maximum of 8 seats, accounting for a certain percentage. Type T2 refers to four-wheeled cargo vehicles with a maximum cargo capacity of 3500kg, accounting for [percentage missing]. P2 type refers to four-wheeled cargo vehicles with a maximum cargo capacity between 3,500 and 12,000 kg, accounting for a certain percentage. The proportion of the four vehicle types in the target area can be obtained based on market statistics. This classification of four vehicle types is mainly used to obtain the probability distribution patterns of energy consumption parameters such as battery type, battery capacity, and travel energy consumption.
[0025] S200: Based on the macro-proportion parameters of electric vehicles in the target area, the category of each electric vehicle is determined by probability sampling method, and the initial state of charge of each electric vehicle after completing its trip is calculated based on the statistical distribution of the travel patterns and energy consumption characteristics of various types of electric vehicles. S300: Based on the initial state of charge, all electric vehicles connected to the grid are divided and the initial static cluster distribution is obtained. In one optional embodiment, all electric vehicles connected to the grid can be divided into fuzzy C-means clustering. Feature vectors of all connected electric vehicles are extracted, and the number of clusters is set to 3, corresponding to charging clusters, idle clusters, and discharging clusters, respectively. Fuzzy weighting coefficients and iteration termination conditions are set. Cluster centers are initialized, and the membership matrix of each EV to the 3 clusters is iteratively calculated and the cluster centers are updated until the convergence condition is met. Based on the principle of maximum membership, each EV is assigned to the cluster with the highest membership. The number of vehicles in each cluster is counted, and the initial static cluster distribution is output. In another optional embodiment, all electric vehicles connected to the network can be divided into operation condition event-driven divisions. For each connected EV, a network access trigger event is identified. If a charging demand event is triggered, the EV is directly assigned to a charging cluster; if a V2G discharge response event is triggered, the EV is assigned to a discharge cluster. Other parked vehicles without active charging or discharging needs are assigned to an idle cluster. The initial SOC is used as a constraint to filter out unreasonable divisions. The number of vehicles in the charging, idle, and discharging clusters is summarized and counted to obtain the initial static cluster distribution. In this embodiment of the invention, the classification of all electric vehicles connected to the network includes using a preset fixed dual threshold. Specifically, a high battery availability threshold and a low battery safety threshold are pre-set, with the low battery safety threshold being lower than the high battery availability threshold; the initial state of charge (SOC) of each connected electric vehicle is obtained; vehicles with an initial SOC lower than the low battery safety threshold are assigned to the charging cluster; vehicles with an initial SOC higher than the high battery availability threshold are assigned to the discharging cluster; and vehicles with an SOC between the two thresholds are assigned to the idle cluster; the number of vehicles in the charging, idle, and discharging clusters is counted to obtain the initial static cluster distribution.
[0026] S400: Based on the initial static cluster distribution, a Markov state transition model is constructed; S500: Using the initial static cluster distribution as the initial state vector, the Markov state transition model is input for a first state transition to obtain the cluster state distribution after a second dynamic correction, thus realizing dynamic cluster modeling of electric vehicles in multiple scenarios.
[0027] It should be noted that this invention relies on limited macroscopic proportion parameters, does not require actual measurement data of a single vehicle, and combines probability sampling with Markov models to take into account the differences of vehicles in multiple scenarios, thereby efficiently realizing dynamic extrapolation of electric vehicle clusters and improving the practicality of modeling.
[0028] In this embodiment of the invention, step S200 includes the following sub-steps A1-A4; In A1: The macro-level percentage parameters of electric vehicles in the target area include the macro-level percentage parameters of electric vehicles used for transportation and the macro-level percentage parameters of electric vehicle types in the target area; In A2: Based on the macro-proportion parameter of electric vehicle traffic use in the target area, the traffic use category of each electric vehicle is obtained through probability sampling. In A3: Based on the macro-proportion parameters of electric vehicle types within the target area, the physical type of each electric vehicle is obtained through probability sampling. In A4: Combine traffic use category and vehicle physical type to determine the category of each electric vehicle.
[0029] In this embodiment of the invention, based on the three classification methods of vehicle traffic use, the daily travel distance of each type of electric vehicle (EV) follows a certain probability distribution law, and the probability density function of the daily travel distance distribution is expressed as: in, For EV index, For the daily travel distance of EV-i, This represents the average daily travel distance for EVs. The standard deviation of daily travel distance for EVs. The log-normal distribution function is expressed as: in, For random variables, The mean of a log-normal distribution, The standard deviation of the log-normal distribution represents the start time of each type of vehicle. and end time of trip The distribution pattern can be obtained from traffic statistics.
[0030] The distribution pattern of battery capacity according to the four types of EV vehicles is shown in Table 1. Table 1. Distribution of EV Battery Capacity
[0031] As shown in Table 1, the gamma distribution function is expressed as: in, For the battery capacity of EV, This represents the minimum capacity of the EV battery. This represents the maximum capacity of the EV battery. The gamma distribution function is... For shape parameters, This is the scale parameter.
[0032] The gamma distribution function is expressed as: in, For random variables, For shape parameters, For scale parameters, It is the factorial function.
[0033] The normal distribution function is expressed as: in, This represents the average battery capacity. The standard deviation of battery capacity, The normal distribution function is expressed as: in, For stacked random variables, The mean of a normal distribution. is the standard deviation of the normal distribution.
[0034] According to statistics, the energy consumption per kilometer during EV travel is... The distribution pattern is shown in Table 2; Table 2 Distribution of energy consumption per kilometer
[0035] According to Table 2, the state of charge (SOC) of a battery is defined as follows: in, Let t be the SOC value of the battery at time t. This represents the remaining battery power.
[0036] Due to differences in individual travel habits and preferences, users have different SOC (State of Occupation) needs before traveling. The distribution pattern of users' SOC needs can be represented as follows: in, for SOC status of user travel demand This represents the average value of the user demand SOC status. The standard deviation of the user's required SOC status.
[0037] In one alternative embodiment, the probability sampling method can be stratified sampling, based on proportion. , Calculate the number of vehicles to be extracted for each layer (HBW, HBO, NHB) separately, and within each traffic use layer, further calculate the proportion of vehicles by type. , The system performs a second-level hierarchical calculation to determine the number of vehicles that should be allocated to each physical type within that level. Within each level, category labels are randomly assigned to vehicles to complete the category assignment for all electric vehicles in terms of traffic use and vehicle physical type. The system then outputs the traffic use category and vehicle physical type corresponding to each electric vehicle. In another alternative embodiment, the probability sampling method can also be multinomial distribution sampling, based on the proportion of traffic use. , As a probability parameter of the multinomial distribution, a multinomial random sampling is performed on each electric vehicle in the region, and the traffic use category of that vehicle is output; the proportion of vehicle types is then used as the output. , As another set of multinomial distribution probability parameters, multinomial random sampling is performed again on the same electric vehicle to output the vehicle physical type; the traffic use category and vehicle physical type obtained from the two samplings are combined to determine the complete category of the electric vehicle; the category results of all vehicles are obtained by traversing all electric vehicles. In this embodiment of the invention, the probability sampling method includes the Monte Carlo sampling algorithm; Specifically, the total number of electric vehicles to be modeled within the target area is set. Input the macro-level proportion vector of traffic use. and the macroscopic proportion vector of vehicle physical type .
[0038] For the Generate the first independent uniform random number for each electric vehicle. ,and It follows a uniform distribution on [0,1).
[0039] Based on the cumulative probability interval of traffic use Perform mapping to determine the first The vehicle's HBW, HBO, or NHB traffic use category.
[0040] Generate and for the same electric vehicle Independent second uniform random numbers ,and It follows a uniform distribution on [0,1).
[0041] Based on the cumulative probability interval of vehicle physical type Perform mapping to determine the first The vehicle's physical type: T1, P1, T2, or P2.
[0042] The results of the two samplings are combined to form the first... The complete category label of the vehicle, such as HBW-P1 or NHB-T2. Given limited information with only two sets of marginal proportions, the two samplings are treated as independent.
[0043] make Take 1 to 1 in sequence The above process is repeated to obtain the category results for all vehicles; then, the travel and energy consumption parameters are extracted according to the category of each vehicle, and the initial SOC of a single vehicle entering the network is calculated.
[0044] It should be noted that this invention requires only two types of macroscopic proportion parameters, and completes the category allocation by two independent random samplings. It does not require joint distribution priors, is easy to calculate, and is suitable for scenarios with limited information, providing feasible input for subsequent batch calculation of network-connected SOCs.
[0045] In this embodiment of the invention, after completing steps A1-A4, step S200 also includes steps A5-A9; In A5: Based on the vehicle type of each electric vehicle, the battery capacity of the corresponding vehicle is determined by sampling from the probability distribution of battery capacity of the corresponding vehicle category. In A6: Based on the traffic use type of each electric vehicle, the daily travel distance of the corresponding vehicle is determined by sampling from the probability distribution of daily travel distance of the corresponding category of vehicles. In A7: Based on the vehicle type of each electric vehicle, sample from the probability distribution of energy consumption per unit mileage of the corresponding vehicle category to determine the energy consumption per unit mileage of the corresponding vehicle. In A8: Obtain the initial state of charge of the corresponding electric vehicle before departure; In A9: Based on battery capacity, daily travel distance, energy consumption per unit mileage, and initial state of charge before travel, the initial state of charge of the corresponding electric vehicle after completing its trip is calculated according to the linear relationship between state of charge consumption and travel distance.
[0046] In this embodiment of the invention, for any single EV, the traffic use type and vehicle type of the EV are determined by using the Monte Carlo sampling algorithm based on the probability distribution of the proportion of each type of EV.
[0047] Based on the type of EV and the probability distribution of daily travel distance and start and end times of EV, the Monte Carlo sampling algorithm is used to determine the daily travel distance, start time and end time of EV.
[0048] The initial State of Charge (SOC) of an EV upon grid connection is affected by daily travel distance. Assuming a linear relationship between the EV's SOC consumption and daily travel distance, the initial SOC value of the EV upon grid connection is expressed as: in, This is the initial SOC value when the EV is connected to the grid. for The State of Charge (SOC) value when disconnected from the power grid follows a uniform distribution of U(0.8, 0.9). It is used to estimate the maximum driving distance of an EV when it is fully charged.
[0049] It should be noted that by obtaining vehicle type and travel parameters through Monte Carlo sampling and combining them with the linear relationship between battery power and mileage to calculate the initial SOC for network access, the system adapts to limited information conditions and improves the realism of cluster modeling and simulation.
[0050] In this embodiment of the invention, step S300 includes the following sub-steps B1-B5; In B1: Set the power boundary thresholds, including the high power availability threshold and the low power safety threshold; In B2: In response to the initial state of charge of electric vehicles being less than the low-charge safety threshold, the corresponding electric vehicles are classified into a power shortage cluster; In B3: In response to the initial state of charge of an electric vehicle being greater than or equal to the low-charge safety threshold and less than or equal to the high-charge availability threshold, the corresponding electric vehicle is classified into a leisure cluster. In B4: In response to the initial state of charge of electric vehicles being greater than the high battery availability threshold, the corresponding electric vehicles are classified into a fully charged cluster. In B5: The number of electric vehicles in each cluster is counted to obtain the initial static cluster distribution.
[0051] In embodiments of the present invention, such as Figure 3 As shown, for electric vehicles that have completed their trips and connected to charging stations in multiple scenarios, the State of Charge (SOC) of the vehicle battery is used as the sole criterion for classification. Combining the safe operating range of lithium batteries with the actual behavioral characteristics of electric vehicle users, two core thresholds are set for four types of vehicle models, including a high battery availability threshold. The target charging capacity expected by vehicle users; low battery safety threshold. This is the upper limit for triggering charging demand for vehicles; below this value, there is a rigid charging demand. High battery availability threshold. A value of 0.80 is acceptable; this is the low battery safety threshold. 0.20 is acceptable.
[0052] Based on the above thresholds, the vehicles entering the network are initially divided into three mutually exclusive basic clusters: charging clusters, idle clusters, and discharging clusters. The division rules include: Power shortage cluster The vehicle's initial battery level upon registration did not reach the low-battery safety threshold, creating a rigid charging need with no other alternatives, thus meeting the requirements. ; Leisure Cluster After the vehicle is registered with the network, it is in the healthy operating range of the battery. Users do not have a rigid need for charging and discharging and can choose to charge or standby without operation. ; Fully charged cluster The vehicle is registered with the network when its battery is in a high charge range, and the user has no charging needs. The user can choose to discharge the battery or remain in standby mode. .
[0053] The initial number of vehicles in the static cluster is represented by summing all types of vehicles: in, The total number of vehicle type categories, , The first The number of vehicles classified into three clusters within the same category, and the total number of vehicles connected to the network, meet the following requirements. .
[0054] It should be noted that the initial static classification is based solely on the initial SOC at the time of network access, without considering the dynamic evolution of the vehicle's state during charging and discharging. Therefore, a Markov state transition model is introduced to complete the secondary dynamic correction based on the four core transition paths.
[0055] In this embodiment of the invention, step S400 includes the following sub-steps C1-C2; In C1: the power-deficient cluster, the leisure cluster, and the fully charged cluster are used as the state space of the Markov state transition model; In C2: Define a set of state transition paths. The state transition paths must include at least the following: transition from a power-deficient cluster to a leisure cluster, a leisure cluster to its own self-sustaining transition, a leisure cluster to a fully charged cluster, a fully charged cluster to its own self-sustaining transition, and a fully charged cluster to a leisure cluster.
[0056] In one optional embodiment, the state transition path can be from the power shortage cluster to the power shortage cluster, obtain the SOC state and charging behavior of each vehicle in the power shortage cluster at the current moment; determine that the electric vehicles in the cluster have not completed charging within the current simulation step and the SOC is still below the low threshold; keep these vehicles in the power shortage cluster; update the transition probability of the power shortage cluster pointing to itself in the Markov transition matrix to complete the state self-holding transition. In another optional embodiment, the state transition path can also be a direct transition from the power shortage cluster to the full charge cluster. This involves obtaining the charging power and current SOC of the vehicles in the power shortage cluster; calculating the SOC value after charging in the current time step; if the SOC is directly higher than the high threshold after charging is completed, skipping the leisure cluster and directly assigning it to the full charge cluster; updating the number of cluster members and updating the corresponding transition matrix elements. In this embodiment of the invention, the state transition path includes a transition from a power-deficient cluster to a leisure cluster, a leisure cluster transitioning to its own self-sustaining state, a leisure cluster transitioning to a fully charged cluster, a fully charged cluster transitioning to its own self-sustaining state, and a fully charged cluster transitioning to a leisure cluster.
[0057] Specifically, the three initially defined cluster types form the state partitioning space for electric vehicle clusters. Based on the core logic of electric vehicle cluster state evolution, four core state transition paths are defined. The shift from power shortage clusters to leisure clusters, i.e. This includes situations where users have a rigid need to charge when the battery is low, there is no self-holding or discharging behavior, and the SOC exceeds the limit after charging. The threshold will inevitably lead to entry into the leisure cluster. ; Leisure clusters shift towards self-sustainability, i.e. This includes situations where users have no charging or discharging needs within the normal range, the SOC remains unchanged, and the cluster remains in a leisure state. The shift from leisure clusters to fully charged clusters, i.e. This includes situations where users actively choose to recharge within the normal charging range, and the State of Charge (SOC) exceeds the limit after charging. Threshold, enter full-charge cluster ; A fully charged cluster transitions to its self-sustaining state, i.e. This includes situations where users have no need to charge or discharge within the fully charged range, the SOC remains unchanged, and the cluster remains in a fully charged state. The shift from fully charged clusters to leisure clusters, i.e. This includes situations where users actively choose to discharge via V2G within the full charge range, and the SOC drops below a certain level after discharge. Threshold, enter leisure cluster .
[0058] In this embodiment of the invention, after completing steps C1-C2, step S400 further includes steps C3-C4; In C3: Based on the current state of charge of each electric vehicle, the transition probability of each state transition path is determined by fitting a regression model; In C4: Construct the Markov state transition matrix based on each state transition path and its corresponding transition probability.
[0059] In one optional embodiment, the regression model can be a multivariate Probit regression model. The sample dataset is organized, and the input features of the samples are selected as follows: vehicle type, network access time, current cluster, initial SOC, and traffic purpose of the electric vehicle; the label is the target cluster state that the vehicle will actually migrate to in the next moment. Latent variables are set to describe the degree of vehicle tendency to transfer to each cluster. Based on the sample data, the multivariate Probit regression model is trained to learn the correspondence between features and transfer tendencies. The features of the electric vehicle clusters to be modeled are input into the trained model, and the model outputs the probability of a vehicle transitioning to a state along each transfer path. The obtained transfer probabilities for each path are then filled into the Markov transition matrix. In another optional embodiment, the regression model can also be an ordered logistic regression. Training samples are constructed, with input features including SOC, vehicle type, traffic purpose, and time information; sample labels are three ordered cluster states: low power, idle, and fully charged. The ordered logistic regression model is trained to learn the impact of each input feature on state upgrades and downgrades, obtaining the regression coefficients corresponding to the features. The features of vehicles in the current cluster are input into the model to obtain the probabilities of vehicles maintaining their original state, transitioning to a cluster with higher power, and transitioning to a cluster with lower power. The probabilities corresponding to each state transition path are extracted and assigned to the corresponding positions in the Markov transition matrix to complete the determination of the transition probabilities. In this embodiment of the invention, the regression model includes a model fitted using Logistic regression; Specifically, based on the four vehicle types T1 / P1 / T2 / T3, a Logistic regression model is used to characterize the probability of user behavior. The current SOC of the vehicle is the only input, and the output probability is strictly constrained to the range of 0 to 1, which naturally meets the normalization requirements of the Markov model.
[0060] Power shortage cluster The transition probability includes the fact that, under power shortage conditions, users have a rigid need for charging and have no other optional behavior, so they will inevitably enter the leisure cluster by charging. Therefore, the transition probability is a fixed value and does not require additional fitting, and is expressed as: in, For the vehicle from state Transition to state after one computational step The probability, For the vehicle from state Transition to state after one computational step The probability, For the vehicle from state Transition to state after one computational step The probability of; Leisure Cluster The transition probability includes the following: within the normal range, users' willingness to charge decreases exponentially as the State of Charge (SOC) increases; the closer the SOC is to the desired state, the lower the willingness to charge. The lower limit is the stronger the willingness to replenish energy; the closer the SOC is to the lower limit. The higher the upper limit, the weaker the willingness to charge. The charging transition probability is characterized by an inverse logistic fitting model, expressed as: in, A negative slope coefficient indicates that the higher the SOC, the lower the probability of charging. The inflection point SOC, i.e., the SOC corresponding to a 50% charging probability, has a self-holding probability that satisfies the normalization constraint as follows: Fully charged cluster The transfer probability includes the following: within the full charge range, the user's willingness to discharge increases exponentially with the increase of SOC; the closer the SOC is to 100% full charge, the stronger the V2G discharge willingness; the closer the SOC is to... The lower the threshold, the weaker the willingness to discharge. The discharge transition probability is characterized by a positive logistic fitting model, expressed as: in, The slope coefficient represents the higher the SOC, the higher the discharge probability. The inflection point, i.e., the state of charge (SOC) at which the discharge probability is 50%, and the corresponding self-holding, i.e., the probability of neither charging nor discharging, satisfying the normalization constraint, are expressed as follows: Based on the above legal transition paths and probability quantization results, a 3×3 first-order Markov state transition matrix is constructed. The elements in each row of the matrix satisfy the probability normalization constraint. Specifically, it is expressed as follows: It should be noted that by using SOC as the sole input and combining it with Logistic regression to fit the transition probabilities of each state, a fixed probability setting is made for the power shortage state, which satisfies Markov normalization, efficiently quantifies the behavior of electric vehicle clusters, and reduces the modeling complexity.
[0061] In this embodiment of the invention, step S500 includes the following sub-steps D1-D3; In D1: The initial state vector is constructed using the ratio of the number of vehicles in each type of cluster to the total number of vehicles entering the network in the initial static cluster distribution. In D2: Multiply the initial state vector by the Markov state transition matrix to perform a state transition calculation and obtain the state vector after secondary correction. In D3: Based on the state vector after the second correction and the total number of vehicles entering the network, the number of vehicles in each type of cluster after the second dynamic correction is calculated, and the cluster state distribution after the second dynamic correction is obtained.
[0062] In this embodiment of the invention, the result of the initial static cluster partitioning is used as the initial state vector of the Markov model. Let the number of vehicles in the low-power cluster, the leisure cluster, and the fully charged cluster after the initial partitioning be respectively... Total number of vehicles connected to the network Then the initial state vector is: After a first-order Markov state transition, the dynamically modified cluster state vector is obtained, and the quadratic dynamic cluster partitioning is represented as follows: Substituting the transition matrix into the expansion, we obtain the number of vehicles in the three clusters after the second dynamic partitioning as follows: Ultimately, the cluster distribution ratio of electric vehicles in various scenarios, after secondary dynamic correction, is obtained under limited information conditions, providing accurate dynamic state input for subsequent power distribution network carrying capacity assessment and cluster adjustability calculation: in, As for the proportion of leisure clusters, For the percentage of fully charged clusters, This represents the percentage of clusters experiencing power shortages.
[0063] Power shortage cluster After one transfer, the percentage becomes 0, which aligns with the behavioral logic of rigid charging demand.
[0064] It should be noted that, based on the initial SOC calculation and initial static cluster partitioning of electric vehicles for various scenarios, this invention constructs a pure SOC-driven first-order discrete Markov state transition model to characterize the dynamic evolution of cluster states caused by vehicle charging and discharging behavior and to address the problem that static partitioning cannot reflect real-time changes in vehicle states. This model completes a secondary dynamic cluster partitioning through a single state transition, achieving high-precision dynamic modeling of electric vehicle clusters under limited information conditions. Example 2: The above is an illustrative scheme of a dynamic cluster modeling method for multi-scenario electric vehicles according to this embodiment. It should be noted that the technical solution of this dynamic cluster modeling system for multi-scenario electric vehicles belongs to the same concept as the technical solution of the dynamic cluster modeling method for multi-scenario electric vehicles described above. Details not described in detail in the technical solution of the dynamic cluster modeling system for multi-scenario electric vehicles in this embodiment can be found in the description of the technical solution of the dynamic cluster modeling method for multi-scenario electric vehicles described above.
[0065] This embodiment provides a dynamic cluster modeling system for electric vehicles in multiple scenarios, including: The data acquisition module is used to obtain macro-level percentage parameters of electric vehicles by category within the target area; The initial state of charge calculation module is used to determine the category of each electric vehicle based on the macro-proportion parameters of electric vehicle categories in the target area using a probability sampling method, and to calculate the initial state of charge of each electric vehicle after it completes its trip and is connected to the power grid based on the statistical distribution of the travel patterns and energy consumption characteristics of various types of electric vehicles. The initial static cluster partitioning module is used to partition all electric vehicles connected to the grid according to the initial state of charge and to obtain the initial static cluster distribution. The model building module is used to build a Markov state transition model based on the initial static cluster distribution. The dynamic cluster correction module is used to take the initial static cluster distribution as the initial state vector, input the Markov state transition model to perform a state transition, and obtain the cluster state distribution after secondary dynamic correction, so as to realize dynamic cluster modeling of electric vehicles in multiple scenarios.
[0066] This embodiment also provides a computer device applicable to a dynamic cluster modeling method for electric vehicles in multiple scenarios, including: The system includes a memory and a processor. The memory stores computer-executable instructions, and the processor executes these instructions to implement a dynamic cluster modeling method for electric vehicles in multiple scenarios, as proposed in the above embodiments.
[0067] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements a dynamic cluster modeling method for multi-scenario electric vehicles as proposed in the above embodiments.
[0068] The storage medium proposed in this embodiment and the method for implementing a dynamic cluster modeling method for electric vehicles in multiple scenarios proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.
[0069] Based on the above description of the implementation methods, those skilled in the art will clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computing device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.
[0070] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. 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 be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.
Claims
1. A dynamic cluster modeling method for multi-scenario electric vehicles, characterized in that, include: Obtain macro-level percentage parameters of electric vehicle categories within the target area; Based on the macro-proportion parameters of electric vehicles in the target area, the category of each electric vehicle is determined by probability sampling method, and the initial state of charge of each electric vehicle after completing its trip is calculated based on the statistical distribution of the travel patterns and energy consumption characteristics of various types of electric vehicles. Based on the initial state of charge, all electric vehicles connected to the grid are divided and the initial static cluster distribution is obtained. Based on the initial static cluster distribution, a Markov state transition model is constructed. Using the initial static cluster distribution as the initial state vector, the Markov state transition model is input for a first state transition, resulting in a second dynamically corrected cluster state distribution, thus realizing dynamic cluster modeling of electric vehicles in multiple scenarios.
2. The method of claim 1, wherein, Determining the category of each electric vehicle using probability sampling methods includes: The macro-level percentage parameters of electric vehicles within the target area include the macro-level percentage parameters of electric vehicles used for transportation within the target area and the macro-level percentage parameters of electric vehicle types within the target area; Based on the macro-level proportion of electric vehicles in the target area, the traffic use categories of each electric vehicle are obtained through probability sampling. Based on the macro-proportion parameters of electric vehicle types within the target area, the physical type of each electric vehicle is obtained through probability sampling. The category of each electric vehicle is determined by combining its traffic use category with its physical type.
3. The method of claim 2, wherein, The initial state of charge of each electric vehicle after it completes its trip and is connected to the grid includes: Based on the vehicle type of each electric vehicle, the battery capacity of the corresponding vehicle is determined by sampling from the probability distribution of battery capacity of the corresponding vehicle category. Based on the traffic use type of each electric vehicle, the daily travel distance of the corresponding vehicle is determined by sampling from the probability distribution of daily travel distance of the corresponding vehicle category. Based on the vehicle type of each electric vehicle, the energy consumption per unit mile is determined by sampling from the probability distribution of energy consumption per unit mile of the corresponding vehicle category. Obtain the initial state of charge of the corresponding electric vehicle before travel; Based on battery capacity, daily travel distance, energy consumption per unit mileage, and initial state of charge before travel, the initial state of charge of the corresponding electric vehicle after completing its trip is calculated according to the linear relationship between state of charge consumption and travel distance.
4. The method of claim 3, wherein, All electric vehicles connected to the network were divided into groups, and the initial static cluster distribution was statistically obtained, including: Set power boundary thresholds, including a high power availability threshold and a low power safety threshold; In response to the initial state of charge of electric vehicles being less than the low-charge safety threshold, the corresponding electric vehicles are classified into a power shortage cluster. In response to the fact that the initial state of charge of an electric vehicle is greater than or equal to the low-charge safety threshold and less than or equal to the high-charge availability threshold, the corresponding electric vehicle is classified into a leisure cluster. In response to an electric vehicle's initial state of charge being greater than a high battery availability threshold, the corresponding electric vehicle is classified into a fully charged cluster. The number of electric vehicles in each cluster is counted to obtain the initial static cluster distribution.
5. The method of claim 4, wherein, Constructing a Markov state transition model includes: The state spaces of the Markov state transition model are defined as the power shortage cluster, the leisure cluster, and the fully charged cluster. Define a set of state transition paths, which include at least the following: transition from a power-deficient cluster to a leisure cluster, a leisure cluster to its own self-sustaining transition, a leisure cluster to a fully charged cluster, a fully charged cluster to its own self-sustaining transition, and a fully charged cluster to a leisure cluster.
6. The method of claim 5, wherein, Also includes: Based on the current state of charge of each electric vehicle, the transition probability of each state transition path is determined by fitting a regression model. Construct a Markov state transition matrix based on each state transition path and its corresponding transition probability.
7. The method of claim 6, wherein, Inputting the Markov state transition model and performing a single state transition yields the following cluster state distribution after a second dynamic correction: The initial state vector is constructed using the ratio of the number of vehicles in each type of cluster to the total number of vehicles entering the network in the initial static cluster distribution. Multiply the initial state vector by the Markov state transition matrix and perform a state transition calculation to obtain the state vector after secondary correction. Based on the state vector after the second correction and the total number of vehicles entering the network, the number of vehicles in each type of cluster after the second dynamic correction is calculated, and the cluster state distribution after the second dynamic correction is obtained.
8. A dynamic cluster modeling system for multi-scenario electric vehicles, applying a dynamic cluster modeling method for multi-scenario electric vehicles according to any one of claims 1-7, characterized in that, include: The data acquisition module is used to obtain macro-level percentage parameters of electric vehicles by category within the target area; The initial state of charge calculation module is used to determine the category of each electric vehicle based on the macro-proportion parameters of electric vehicle categories in the target area using a probability sampling method, and to calculate the initial state of charge of each electric vehicle after it completes its trip and is connected to the power grid based on the statistical distribution of the travel patterns and energy consumption characteristics of various types of electric vehicles. The initial static cluster partitioning module is used to partition all electric vehicles connected to the grid according to the initial state of charge and to obtain the initial static cluster distribution. The model building module is used to build a Markov state transition model based on the initial static cluster distribution. The dynamic cluster correction module is used to take the initial static cluster distribution as the initial state vector, input the Markov state transition model to perform a state transition, and obtain the cluster state distribution after secondary dynamic correction, so as to realize dynamic cluster modeling of electric vehicles in multiple scenarios.
9. A computer device, comprising: include: A memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of a dynamic cluster modeling method for multi-scenario electric vehicles according to any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that, It stores a computer program, which, when executed by a processor, implements the steps of a dynamic cluster modeling method for multi-scenario electric vehicles according to any one of claims 1 to 7.
