Power distribution network electric vehicle carrying capacity evaluation method and system

Abstract

The application discloses a power distribution network electric vehicle carrying capacity evaluation method and system. The method comprises the following steps: for a target power distribution network, analyzing interest points and electric vehicle parameters, using an agent-cellular automaton model to simulate charging behavior to generate diversified charging scenarios. Actual operation data is collected through a smart meter, a DistFlow model is used to calculate the node power of each charging point, and the model is solved in different time intervals, the Kalman filter algorithm is used to process the power flow relationship, and the voltage and line load are checked to determine whether the constraint condition is met to determine the operation feasibility. Then, an optimized random forest model is established, key features are extracted and trained, a mapping between the features and the scene feasibility is established, the carrying capacity of the power distribution network for electric vehicle charging load is evaluated, and the method of the application can not only reduce the dependence on a large number of simulation scenarios, reduce the calculation complexity, better handle uncertain factors, and meet the demand of real-time scheduling of the power grid.

Description

Technical Field

[0001] This invention relates to the field of smart grid technology, and in particular to a method and system for assessing the carrying capacity of electric vehicles in distribution networks. Background Technology

[0002] As the global number of electric vehicles continues to rise, their charging behavior via the power distribution network is becoming a key factor affecting grid stability and power quality. High-density charging demand, especially concentrated in specific areas, not only exacerbates peak-valley fluctuations in grid load but can also lead to technical problems such as voltage exceeding limits and line overload. Therefore, accurately assessing the power distribution network's capacity to support electric vehicles is crucial for ensuring the safe operation of the grid and improving service quality. However, traditional assessment methods are typically based on static scenario analysis, failing to fully consider the impact of user behavior and traffic conditions on electric vehicle charging patterns, thus limiting the accuracy and practicality of the assessment results.

[0003] In practical applications, the travel habits and charging choices of electric vehicle users exhibit significant spatial and temporal diversity, which directly relates to the distribution of charging demand in different regions. Traditional deterministic simulation methods, due to their fixed parameter settings, struggle to reflect this diversity and dynamic changes, leading to discrepancies between predicted results and actual conditions. Furthermore, traditional methods employing static power flow calculations fail to effectively capture the spatiotemporal evolution of electric vehicle charging behavior, neglecting the impact of factors such as traffic flow and congestion on charging demand. This results in a lack of sufficient flexibility and foresight in power grid planning and dispatch decisions.

[0004] Therefore, it is necessary to design a new method that utilizes artificial intelligence technology to deeply mine useful information contained in historical data, establish complex mapping relationships between multi-dimensional features and power grid operating parameters, and thus achieve more accurate carrying capacity assessment. This approach not only reduces reliance on numerous simulation scenarios and lowers computational complexity, but also better handles uncertainties, meets the needs of real-time power grid dispatching, and provides more scientific and reasonable planning recommendations. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method and system for evaluating the load-bearing capacity of electric vehicles in power distribution networks.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a method for assessing the carrying capacity of electric vehicles in power distribution networks, comprising:

[0007] For the target distribution network, we analyze the points of interest and electric vehicle parameters within the target area, and use an agent-cellular automaton model to simulate charging behavior to generate diverse electric vehicle charging scenarios.

[0008] Using the electric vehicle charging scenario, smart meters are used to collect actual power grid operation data, the DistFlow model is used to characterize the power flow relationship of the target distribution network, the node power of each electric vehicle charging connection point is calculated, and the DistFlow model is solved in different time intervals. Based on the solution results, the unscented Kalman filter algorithm is used to process the power flow relationship of the target distribution network, the feasibility problem is defined and solved, and the operational feasibility under each scenario is judged by checking whether the voltage level and line load meet the constraints.

[0009] An optimized random forest model is established to extract key features of the electric vehicle charging scenario. The optimized random forest model is then used to train the key features to establish a mapping between the features and the scenario feasibility, and to evaluate the load-bearing capacity of the power distribution network for electric vehicle charging.

[0010] This invention also provides a power distribution network electric vehicle carrying capacity assessment system, comprising:

[0011] The scene generation unit is used to analyze points of interest and electric vehicle parameters within the target area for the target power distribution network, and to simulate charging behavior using an agent-cellular automaton model to generate diverse electric vehicle charging scenarios.

[0012] The judgment unit is used to collect actual power grid operation data using smart meters in the electric vehicle charging scenario, use the DistFlow model to characterize the power flow relationship of the target distribution network, calculate the node power of each electric vehicle charging connection point, solve the DistFlow model in different time intervals, use the unscented Kalman filter algorithm to process the power flow relationship of the target distribution network based on the solution results, define and solve the feasibility problem, and judge the operational feasibility under each scenario by checking whether the voltage level and line load meet the constraints.

[0013] The prediction unit is used to establish an optimized random forest model, extract key features of the electric vehicle charging scenario, train the optimized random forest model on the key features, establish a mapping between features and scenario feasibility, and evaluate the carrying capacity of the power distribution network for electric vehicle charging load.

[0014] The advantages of this invention compared to existing technologies are as follows: This invention integrates an agent-cellular automaton model, a DistFlow model, and an unscented Kalman filter algorithm, combined with an optimized random forest model, to leverage artificial intelligence technology to deeply mine useful information from historical data. The method first generates diverse charging scenarios based on points of interest and electric vehicle parameters within the target area. Then, using data collected from smart meters, it calculates the node power of each charging point using the DistFlow model, solves the power flow relationship in different time intervals, and uses an unscented Kalman filter algorithm to ensure operational feasibility. Next, key features are extracted to train and optimize the random forest model, establishing a mapping between features and scenario feasibility, thereby accurately assessing the carrying capacity of the distribution network. This method not only reduces reliance on a large number of simulated scenarios and lowers computational complexity but also effectively addresses uncertainties, meets the real-time scheduling needs of the power grid, and provides more scientific and reasonable planning suggestions.

[0015] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. Attached Figure Description

[0016] 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 some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A schematic flowchart of the method for assessing the carrying capacity of electric vehicles in a power distribution network provided in an embodiment of the present invention;

[0018] Figure 2 A power grid topology diagram provided for an embodiment of the present invention;

[0019] Figure 3 This is a schematic block diagram of a power distribution network electric vehicle load-bearing capacity assessment system provided in an embodiment of the present invention;

[0020] Figure 4 A schematic block diagram of a computer device provided for an embodiment of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0023] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0024] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0025] Please see Figure 1 , Figure 1 This is a flowchart illustrating the method for assessing the electric vehicle carrying capacity of a distribution network according to an embodiment of the present invention. This method is applied in a server. It utilizes artificial intelligence techniques, such as an agent-cellular automaton model, to simulate electric vehicle charging behavior and generate diverse charging scenarios. The DistFlow model, combined with actual operating data, characterizes the power flow relationship of the distribution network and calculates node power. Then, an unscented Kalman filter algorithm is used to process the power flow relationship, checking whether voltage levels and line loads meet constraints to determine operational feasibility. Furthermore, an optimized random forest model is established to extract key features and is trained to establish a mapping between features and scenario feasibility, thereby assessing the distribution network's carrying capacity for electric vehicle charging loads. This method not only deeply mines useful information from historical data and establishes a complex mapping relationship between multi-dimensional features and grid operating parameters, achieving a more accurate carrying capacity assessment, but also reduces reliance on a large number of simulated scenarios, lowers computational complexity, better handles uncertainties, meets the needs of real-time grid dispatching, and provides more scientific and reasonable planning suggestions.

[0026] Figure 1 This is a flowchart illustrating the method for assessing the carrying capacity of electric vehicles in a power distribution network, as provided in an embodiment of the present invention. Figure 1 As shown, the method includes the following steps S110 to S130.

[0027] S110. For the target distribution network, analyze the points of interest and electric vehicle parameters within the target area, and use an agent-cellular automaton model to simulate charging behavior to generate diverse electric vehicle charging scenarios.

[0028] In this embodiment, the electric vehicle charging scenario refers to simulating the charging behavior of electric vehicles in different functional areas using an agent-cellular automaton model to generate diverse charging demand scenarios, including feasible, critical, and infeasible states, in order to assess the carrying capacity of the power distribution network.

[0029] In one embodiment, step S110 described above may include steps S111 to S113.

[0030] S111. For the target distribution network, analyze the distribution of points of interest within the target distribution network area to classify the types of areas and determine the boundary range where electric vehicles are concentrated.

[0031] In this phase, the first step is to grid the geographical area where the target distribution network is located. Each grid represents a small area, which can be categorized as residential areas (such as apartment complexes), commercial areas (shopping malls, office buildings), or public service areas (hospitals, government agencies). This zoning helps identify different types of usage patterns and their impact on electricity demand.

[0032] In one embodiment, step S111 described above may include steps S1111 to S1112.

[0033] S1111. Divide the target power distribution network area according to a custom scale, covering residential areas, commercial areas and public service areas;

[0034] S1112. Calculate the frequency density ratio of different functional areas based on the number and weight of points of interest. If the frequency density ratio of a certain type of point of interest exceeds the threshold, define the area as the corresponding functional area. The functional areas include residential areas, commercial areas, or public service areas.

[0035] Specifically, the target distribution network area is divided according to a custom scale, covering residential areas, commercial areas, and public service areas. For each grid, the frequency density of points of interest (POIs) is calculated. First, the number of POIs in the three types of areas is counted: POIs in residential areas include neighborhoods and apartments; POIs in commercial and financial areas include shopping malls and office buildings; and POIs in public service areas include hospitals and government agencies. The proportion of the frequency density of the i-th type of POI is calculated using the following formula: ; ,in, Frequency density; It is the sum of the frequency densities of the three types of interest points; Weights for interest points; Number of points of interest; The network area is defined as follows: if the frequency density of a certain type of interest point accounts for more than 50%, then the network can be classified into the corresponding functional area.

[0036] This method not only helps to accurately locate high-demand areas for electric vehicles, but also provides basic data support for subsequent simulation of charging behavior using agent-cellular automata models, making the generated electric vehicle charging scenarios closer to reality, thereby improving the accuracy and reliability of the entire evaluation process.

[0037] S112. Based on historical data, randomly select samples to obtain the basic parameters of electric vehicles, abstract them into intelligent agents, set attribute and behavior rules, and construct a cellular automaton environment to obtain an intelligent agent-cellular automaton model.

[0038] In this embodiment, the agent-cellular automaton model refers to a computational model that uses randomly selected samples from historical data to obtain the basic parameters of electric vehicles, abstracts them into agents, sets attributes and behavioral rules, and then constructs a cellular automaton environment to simulate the driving and charging behavior of electric vehicles in different functional areas. This model aims to generate diverse electric vehicle charging scenarios to assess the power distribution network's capacity to meet these demands.

[0039] In one embodiment, step S112 described above may include steps S1121 to S1125.

[0040] S1121. Abstract each electric vehicle as an intelligent agent, and define the basic and dynamic attributes of the electric vehicle as an intelligent agent. The basic attributes include the starting point, the ending point, the initial SOC, the battery capacity C, and the rated driving range R; the dynamic attributes include the current position, speed, remaining SOC, and charging demand status.

[0041] S1122. Simulate the selection of vehicle departure and destination using historical data.

[0042] In this embodiment, historical data is used to simulate the vehicle's origin and destination selection process. Random sampling is performed from this data using probability statistical distributions to reflect the randomness characteristics in actual traffic travel patterns.

[0043] S1123. Formulate driving rules and charging decision rules for the intelligent agent. If the remaining power is less than the minimum limit and the destination cannot be reached, the nearest charging station will be selected for charging. Otherwise, the agent will be charged after reaching the destination. The driving rules include acceleration, deceleration, constant speed, and random adjustment.

[0044] In this embodiment, a charging decision rule is defined when the remaining battery power is below a minimum limit. That is, if the destination cannot be reached, the nearest charging station is selected for charging; otherwise, charging is performed after reaching the destination. Driving rules are also set, including acceleration, deceleration, constant speed, and speed changes that are randomly adjusted according to actual conditions.

[0045] S1124. Divide the road network into cells, define the neighborhood relationships of the cells, adopt an extended Moore-type neighborhood structure to support the movement of intelligent agents between cells, and record the current number of vehicles and the degree of congestion for each cell.

[0046] S1125. Set the behavior parameters of the electric vehicle according to the characteristics of the functional area, including the maximum vehicle speed, random deceleration probability and energy consumption in different areas, so as to obtain the agent-cellular automaton model.

[0047] In this embodiment, each electric vehicle is abstracted as an intelligent agent. First, its intelligent agent attributes are divided into basic attributes and dynamic attributes. The basic attributes include the starting point, the destination, the initial SOC, the battery capacity C, and the rated driving range R. The dynamic attributes include the current location, speed, remaining SOC, and charging demand status. A traffic demand prediction model based on intelligent agents is constructed. Considering the randomness of the behavior of large-scale electric vehicles and the characteristics of functional areas, when random vehicles simulate the origin and destination, the model is randomly sampled from the probability distribution based on the statistical historical charging data.

[0048] Then, behavioral rules are set, including charging decisions and driving rules. If the remaining battery power is less than the minimum limit and the destination cannot be reached, the nearest charging station will be selected; otherwise, charging will occur upon reaching the destination. Driving rules include acceleration, deceleration, constant speed, and random adjustment, which are given by the following formulas: ,in, Let n be the speed of vehicle n at time t; Maximum speed; Let n be the distance between the vehicle and the vehicle in front.

[0049] The road network is divided into cells, and the neighborhood relationships of these cells are defined. An extended Moore-type neighborhood is adopted, which consists of the eight neighboring cells of the central cell, allowing the agent to move between cells. Each cell needs to record the current number of vehicles and the degree of congestion. At the same time, the behavioral parameters of electric vehicles are set according to the characteristics of functional areas, including the maximum speed, random deceleration probability, and energy consumption in different areas.

[0050] Through the above steps, the spatiotemporal distribution of electric vehicles in different urban areas can be simulated, thereby generating diverse charging scenarios including feasible, critical, and infeasible states, providing a basis for subsequent analysis of the carrying capacity of the power distribution network.

[0051] S113. Use an agent-cellular automaton model to simulate charging behavior in order to generate diverse electric vehicle charging scenarios.

[0052] In one embodiment, step S113 described above may include steps S1131 to S1134.

[0053] S1131, Using the agent-cellular automaton model with A-based... The algorithm's shortest path planning method determines the travel route for each agent.

[0054] In this embodiment, an agent-cellular automaton model is used to abstract each electric vehicle as an agent, employing an A-based... The algorithm's shortest path planning method determines its travel route. A The algorithm is a heuristic search algorithm that calculates the cost from the starting point to the ending point and the estimated cost (i.e., the overall priority) to select the next node to be traversed in order to find the optimal path to the destination.

[0055] Adopting A The shortest path planning method is used to set the path for each agent, and the priority of each node is calculated by the following formula: ,in, This indicates the overall priority of node x. When selecting the next node to traverse, the node with the highest overall priority should be selected. The cost from the starting point to the destination; This is the estimated cost from the starting point to the end point.

[0056] S1132. Adjust the vehicle's speed according to the speed limit of the area and update the vehicle's position at a certain time.

[0057] In this embodiment, the vehicle's speed is adjusted according to the speed limits of the area, and the vehicle's location at a given time is updated. For example, there may be lower maximum speed limits in residential areas, while there may be different speed limits in commercial areas or public service areas. This step not only considers the actual speed limits on the road but also incorporates factors such as traffic conditions to dynamically adjust the vehicle speed and update the vehicle's location information accordingly.

[0058] Adjust the speed according to the speed limit of the area and update the current location, as shown in the following formula: ,in, Let n be the position of vehicle n at time t; Let be the speed of the vehicle at time t+1. Let n be the position of vehicle n at time t+1.

[0059] S1133. Update the remaining battery power of the electric vehicle based on the distance traveled and the corresponding energy consumption. When the remaining battery power is detected to be lower than the set minimum limit, guide the vehicle to the nearest charging station for charging.

[0060] In this embodiment, the remaining battery power of the electric vehicle is updated based on the distance traveled and the corresponding energy consumption. If the remaining battery power is detected to be below a set minimum limit, the system will guide the vehicle to the nearest charging station for charging. This process involves the linear discharge assumption of the vehicle battery, that is, assuming that the battery discharges linearly with the distance traveled and is completely depleted after exceeding the rated driving range. When charging is required, the vehicle will select the nearest charging station to ensure that it can continue its journey.

[0061] The remaining SOC of the electric vehicle is updated based on the driving distance and energy consumption. If the remaining SOC is less than the minimum limit, the nearest charging station is selected for charging, and the value is updated using the following formula: ,in, Let be the rated driving range of the electric vehicle, assuming that the battery discharges linearly with the driving distance and is completely depleted after exceeding the rated driving range; k is the driving distance.

[0062] S1134. Statistically count the number of electric vehicles and their charging demand in each cell, and record the changes in charging load at different time points. Based on the spatiotemporal distribution evolution model of charging demand, analyze and obtain the distribution of electric vehicle charging demand in different time periods and different functional areas to obtain the electric vehicle charging scenario.

[0063] In this embodiment, the number of electric vehicles and their charging needs within each cell are statistically analyzed, and the changes in charging load at different time points are recorded. Based on the spatiotemporal distribution evolution model of charging demand, the distribution of electric vehicle charging demand in different time periods and functional areas is analyzed and obtained. This process helps to understand the charging patterns and demands of electric vehicles in different functional areas (such as residential areas, commercial areas, and public service areas) within a set time period, thereby generating diverse electric vehicle charging scenarios that cover feasible, critical, and infeasible states.

[0064] By following the steps described above, the driving and charging behavior of electric vehicles in different urban areas can be simulated, thereby assessing the impact of these behaviors on the power distribution network. This is particularly crucial for evaluating the carrying capacity of electric vehicle charging loads. This simulation not only helps identify potential problem areas but also provides data support for optimizing the layout of charging infrastructure.

[0065] S120. Using the electric vehicle charging scenario, smart meters are used to collect actual power grid operation data. The DistFlow model is used to characterize the power flow relationship of the target distribution network. The node power of each electric vehicle charging connection point is calculated, and the DistFlow model is solved in different time intervals. Based on the solution results, the unscented Kalman filter algorithm is used to process the power flow relationship of the target distribution network, and the feasibility problem is defined and solved. By checking whether the voltage level and line load meet the constraints, the operational feasibility under each scenario is judged.

[0066] In one embodiment, the above-mentioned use of the DistFlow model to characterize the power flow relationships of the target distribution network includes:

[0067] Obtain the topology model of the target distribution network N=(V,E), where V is the set of nodes. Given a set of lines, and considering the electric vehicle charging scenario, in each cell, the electric vehicle will select the nearest charging station for charging. The load demand of the charging station is obtained by accumulating the charging requests of all electric vehicles in its cell and mapping it to each node of the distribution network.

[0068] Based on advanced measurement technology, smart meters are used to measure actual power grid operating data, including active and reactive power injection at nodes and the magnitude of node voltage. Describe the DistFlow model, where, p and q represent the active and reactive power injections at the node, respectively; v is the square of the node voltage amplitude; P and Q represent the active and reactive power flow of the line, respectively. and The lines are respectively The resistance and reactance; constraints include active / reactive power flow and voltage limits on the line, denoted as: .

[0069] In one embodiment, the above-mentioned calculation of the nodal power of each electric vehicle charging connection point and solving the DistFlow model in different time intervals includes:

[0070] Electric vehicle candidate access locations are Scene This represents the electric vehicle resources accessed at location L and their net injection or consumption at each node during time period T. The net power curve is ;

[0071] For each electric vehicle scenario ψ, including location, quantity, and charging behavior, through... The model within the corresponding time step t is obtained, where, These represent the active and reactive power flow and the square of the voltage amplitude of the line within the corresponding time step t, respectively.

[0072] Assuming the electric vehicle operates in constant power charging mode and maintains a fixed power factor η through simple reactive power control, combined with the baseline load { , } and electric vehicle charging load{ },pass , Net power injection into compute nodes , , where the matrix It is the adjacency matrix from the electric vehicle's location to the bus, that is, when the j-th electric vehicle is located on bus i... ,otherwise .

[0073] In one embodiment, the above-mentioned unscented Kalman filter algorithm is used to process the power flow relationship of the target distribution network based on the solution results, define and solve the feasibility problem, and determine the operational feasibility under each scenario by checking whether the voltage level and line load meet the constraints, including:

[0074] Based on the DistFlow model corresponding to the solution results, the state and observation variables for power flow calculation of the distribution network are set, the no-load state and its uncertainty matrix are initialized, and the UKF algorithm parameters are configured according to the characteristics of the distribution network.

[0075] Sigma points are generated using the current state and covariance to represent the current state distribution. The future state changes of the Sigma points are predicted using the DistFlow model to obtain the mean and covariance of the predicted states, thus forming the prediction results.

[0076] The prediction results are transformed into the observation space to calculate the observation mean and covariance and determine the Kalman gain. The system state estimate is then updated using actual measurement data.

[0077] Define a feasibility problem to check whether voltage and line load constraints are met. For each node within the time step, set its upper and lower voltage limits, and specify active and reactive power flow and apparent power limits for each line;

[0078] The scenario is constructed and divided into two cases: feasible and infeasible. The overall feasibility of the scenario is evaluated based on the upper limit of the number of times an infeasible solution can occur in several time steps. If the voltage and line load constraints are met in all time steps, the scenario is considered feasible; otherwise, it is infeasible.

[0079] Specifically, based on the obtained DistFlow model, the unscented Kalman filter algorithm is used to process the power flow relationship of the distribution network, as follows:

[0080] Based on the obtained DistFlow model, state variables and observation variables are defined, and the state variables of the distribution network power flow are defined as follows: The observed variable is ,in This includes data collected by the measuring equipment. Initial state. Take the DistFlow solution of the distribution network under no-load operation, initial covariance This is a diagonal matrix, where the diagonal elements represent the uncertainties of the initial state. The UKF parameters are initialized based on the characteristics of the target distribution network, and the parameters are set as follows: , , , which are the Sigma point distribution range, Gaussian fit degree, and minor scaling parameter, respectively;

[0081] In the prediction phase, based on the current state Covariance Generate 2N+1 Sigma points, calculated using the following formula: ; ,in, It is the scaling factor; N is the state dimension;

[0082] Then, for each Sigma point The proposed DistFlow model state transition relationship is used to calculate the predicted Sigma point. For the line The transfer of active power, reactive power, and voltage is performed using the following formula: ;

[0083] The mean value of the predicted state is calculated using the following formula. With covariance : ; ; ,in, and These are the mean weight and covariance weight, respectively, and Q is the process noise covariance;

[0084] During the update phase, the Sigma points are transformed into the observation space, mapping the predicted Sigma points to the measurement dimension, corresponding to the observed variable Z, as shown in the following equation: In the formula, The corresponding measurement items are the predicted voltage square at the Sigma point, and the active and reactive power flow of the line.

[0085] Then the observed mean is calculated using the following formula. With covariance : In the formula, R is the observation noise covariance; the cross covariance is calculated using the following formula. With Kalman gain : ; ;

[0086] Finally, update the state using the following formula. Covariance : ; In the formula, This is the current measurement data;

[0087] Repeat the above steps to filter the measurement data for each time step and continuously output unbiased active and reactive power flow and voltage values ​​to eliminate measurement uncertainty.

[0088] In addition, for defining feasibility issues, the voltage and line load constraints are checked to determine whether feasibility is possible, specifically:

[0089] First, within a time step t, the upper and lower limits of node voltage and the line load rate constraints are given by the following formula: In the formula, and These are the upper and lower limits of the square of the node voltage magnitude, respectively; and The lines within time step t are respectively The trend of meritorious and ineffective actions; For the line The upper limit of apparent power; ;

[0090] Then, the distribution network carrying capacity analysis is constructed as a feasibility problem, for each time period. Scenario ψ is divided into two states: feasible and infeasible. Feasible is denoted as ψfeasible. Infeasible is marked as If all time If the voltage and line load constraints are met, scenario ψ is feasible; otherwise, it is not. The evaluation function for scenario ψ is defined by the following formula: In the formula, It is the upper limit of the number of times an infeasible solution is allowed to occur within T time steps.

[0091] In this embodiment, step S120 details how to utilize the electric vehicle charging scenario, collect actual power grid operation data through smart meters, and use the DistFlow model to characterize the power flow relationship of the target distribution network. Furthermore, it explains how to calculate the nodal power of each electric vehicle charging connection point and solve the DistFlow model in different time intervals. Finally, based on the solution results, the unscented Kalman filter (UKF) algorithm is used to process the power flow relationship of the distribution network, define and solve feasibility issues, and check whether the voltage level and line load meet the constraints to determine the operational feasibility under each scenario.

[0092] First, obtain the topology model of the target distribution network. N =( V , E ),in V Represents a set of nodes. EThis represents the set of lines. Based on the spatiotemporal distribution of electric vehicle charging demand obtained above, in each cell, an electric vehicle will select the nearest charging station to charge. The load demand of a charging station is obtained by summing the charging requests of all electric vehicles within its cell and mapping it to each node of the distribution network.

[0093] Based on advanced measurement technology, smart meters are used to measure actual power grid operating data, including active and reactive power injection at nodes and the magnitude of node voltage. This data is represented by the DistFlow model, which takes into account active / reactive power flow and voltage limitations on the lines.

[0094] For each electric vehicle scenario ψ, including location, quantity, and charging behavior, a model within the corresponding time step t can be obtained by setting a formula, which includes the active and reactive power flow of the line and the square of the voltage amplitude. It is assumed that the electric vehicles operate in a constant power charging mode and maintain a fixed power factor η through simple reactive power control. Combining the baseline load and the electric vehicle charging load, the net power injection at the nodes is calculated using a formula.

[0095] Next, the unscented Kalman filter algorithm is applied to handle power flow relationships:

[0096] Definition of state variables and observation variables: Based on the obtained DistFlow model, define the state variables and observation variables of the power flow in the distribution network.

[0097] Initialization: The initial state is taken from the DistFlow solution of the no-load operation state of the distribution network. The initial covariance is a diagonal matrix, where the diagonal elements represent the uncertainty of the initial state.

[0098] Prediction phase: Generate Sigma points based on the current state and covariance, and use the state transition relationship of the DistFlow model to calculate the predicted Sigma points.

[0099] Update phase: The predicted Sigma points are transformed into the observation space, the observation mean and covariance are calculated, the Kalman gain is determined, and the system state estimate is updated using the actual measurement data.

[0100] Define and resolve the feasibility problem:

[0101] Set constraints: Within time step t, provide upper and lower limits for node voltage and line load rate constraints.

[0102] Assessing scenario feasibility: Construct scenarios, categorizing them as feasible or infeasible. If the voltage and line load constraints are met at all time steps, the scenario is considered feasible; otherwise, it is infeasible. An upper limit on the number of infeasible solutions allowed is used to evaluate the overall feasibility of the scenario.

[0103] This series of steps effectively analyzes and evaluates the carrying capacity of the target distribution network under different electric vehicle charging scenarios, ensuring maximum utilization of existing grid resources while meeting voltage and line load constraints. This approach not only helps optimize the layout of electric vehicle charging infrastructure but also improves the stability and reliability of the power system.

[0104] S130. Establish an optimized random forest model, extract key features of the electric vehicle charging scenario, and train the optimized random forest model on the key features to establish a mapping between features and scenario feasibility, and evaluate the carrying capacity of the power distribution network for electric vehicle charging load.

[0105] In one embodiment, step S130 described above may include steps S131 to S135.

[0106] S131. Extract quantitative features that significantly affect feasibility from the electric vehicle charging scenario, including the number and proportion of electric vehicles, the spatiotemporal distribution of charging demand, access location, and net power injection of distribution network nodes, to form a feature vector.

[0107] S132. The actual feasibility of each electric vehicle charging scenario is determined by simulating the voltage and line load constraints within a set time period, and a label value is set for each scenario accordingly to construct an initial training set.

[0108] S133. Using a resampling method, samples are repeatedly randomly selected from the initial training set as training subsets of the decision tree. For each node, the optimal splitting strategy is applied to the randomly selected variables until they cannot be split anymore, forming a random forest structure composed of multiple decision trees. The remaining samples are used to perform regression prediction on each decision tree in the random forest to verify the model performance.

[0109] S134. Optimize the hyperparameters of the AdaBoost model through grid search, and iteratively optimize the performance of the entire training tree model based on the AdaBoost idea. Enhance the performance of the learner by weighted training of each decision tree. After each iteration, adjust the weights of the training data according to the prediction error of the previous decision tree and retrain a new decision tree. Use mean absolute error and root mean square error as evaluation indicators to verify the model's prediction accuracy.

[0110] S135. The prediction results of all decision trees are weighted and summed to obtain the final prediction result. This is used to determine whether the voltage / line load constraints of the target area are met, and thus determine the feasibility of the scenario. In this process, by analyzing the characteristic distribution pattern of feasible scenarios, the maximum node power injection and voltage values ​​that meet the constraints are obtained, and the carrying capacity of electric vehicle charging load in the distribution network is determined.

[0111] In this embodiment, firstly, quantitative features that significantly affect "feasibility" are extracted from the scenario, including the number and proportion of electric vehicles in different functional areas related to electric vehicles, the spatiotemporal distribution of charging demand and the location of electric vehicle access, and the net power injection of nodes, the average node voltage, and the line load rate related to the power distribution network. The feature vector is constructed as follows: In the formula, Location-related features; Time-related features; Power-related characteristics; To constrain relevant features;

[0112] The feasibility of the scenarios is determined by examining voltage and line load constraints. For each scenario, its simulation is performed over a time set. Does the voltage / line load constraint within the circuit meet the constraint?

[0113] Building the initial training set , and Let k be the feature matrix and label vector respectively. If the constraint is satisfied, the label vector = 1; otherwise, the label vector = 0. Combine the feature matrix with the corresponding label to form training samples. Use the resampling method to repeatedly and randomly select samples from the original training set and replace them as the training subset of the decision tree.

[0114] For each node, apply the optimal splitting method. A random forest structure is formed by randomly selecting variables, without setting the number of splits, and without pruning the trees, until each tree can no longer split, ultimately creating a random forest structure composed of multiple decision trees. The remaining samples are used as test samples to perform regression predictions on each decision tree in the random forest.

[0115] Based on the segmented dataset and the established model, the model is trained on the training set data using appropriate software to verify the feasibility and predictive accuracy of scenario ψ.

[0116] The hyperparameters of the AdaBoost model are optimized using grid search, and iterative optimization is carried out based on the AdaBoost concept to improve the performance of the entire training tree model. Weighted training is performed on each decision tree to enhance the learner's performance. In each iteration, the weights of the training data are adjusted based on the prediction error of the previous decision tree, and then a new decision tree is retrained. MAE and RMSE are used as evaluation metrics to verify the model's prediction accuracy, calculated using the following formula: ; In the formula, n is the number of valid sessions; and These are the predicted and measured values ​​for the feasibility of scenario ψ, respectively.

[0117] The prediction results of all decision trees are weighted and summed to obtain the final prediction result, namely whether the voltage / line load constraints of the target area are met and whether the scenario is feasible. During the training process, by inputting the feature data of the training samples, the distribution law of scenario features that meet the constraints is obtained in the scenarios judged to be feasible, the maximum values ​​of node power injection and voltage are obtained, and the electric vehicle charging load carrying capacity of the distribution network is determined accordingly.

[0118] A well-trained random forest can directly predict the feasibility of unlabeled scenarios, that is, it can analyze the carrying capacity of electric vehicles in the target power distribution network.

[0119] In this embodiment, step S130 describes in detail how to establish an optimized random forest model, extract key features of electric vehicle charging scenarios, and use the model to train these key features to establish a mapping between features and scenario feasibility, thereby evaluating the load-bearing capacity of the power distribution network for electric vehicle charging loads.

[0120] Extract quantitative features that significantly affect feasibility from electric vehicle charging scenarios, including but not limited to:

[0121] The number of electric vehicles and their proportion in different functional areas; the spatiotemporal distribution of charging demand; the location of electric vehicle access; and the net power injection into the distribution network nodes.

[0122] These features together form the feature vector, which is used for subsequent model training and prediction.

[0123] The feasibility of each electric vehicle charging scenario is determined by simulating the voltage and line load constraints within a set time period. For each scenario, a label value is assigned based on whether the voltage and line load constraints are met (e.g., a scenario that meets the constraints is labeled 1, and one that does not is labeled 0). An initial training set is constructed using these label values ​​and the corresponding feature vectors.

[0124] A resampling method (such as Bootstrap) is used to repeatedly and randomly select samples from the initial training set as training subsets for the decision trees. For each node, the optimal splitting strategy is applied to the randomly selected variables until no further splitting is possible, ultimately forming a random forest structure composed of multiple decision trees. The remaining samples are used to perform regression predictions on each decision tree in the random forest to validate the model performance.

[0125] The hyperparameters of the AdaBoost model are optimized using grid search, and the performance of the entire training tree model is iteratively optimized based on the AdaBoost concept. The learner's performance is enhanced by weighted training of each decision tree. After each iteration, the weights of the training data are adjusted based on the prediction error of the previous decision tree, and a new decision tree is retrained. Mean absolute error (MAE) and root mean square error (RMSE) are used as evaluation metrics to verify the model's prediction accuracy.

[0126] The final prediction result is obtained by weighted summation of the prediction results of all decision trees. This result is used to determine whether the voltage / line load constraints of the target area are met, thereby determining the feasibility of the scenario. By analyzing the characteristic distribution patterns of feasible scenarios, the maximum node power injection and voltage values ​​that meet the constraints are obtained, thereby determining the carrying capacity of the electric vehicle charging load in the distribution network.

[0127] By following the steps above, an optimized random forest model can be effectively established to assess the impact of electric vehicle charging load on the power distribution network, ensuring that the network maintains stable operation while supporting electric vehicle charging demand. This approach not only helps optimize the layout of electric vehicle charging infrastructure but also improves the stability and reliability of the power system.

[0128] For example, such as Figure 2 As shown, a radial power grid was selected as the test area, and its power grid topology is as follows. Figure 2 As shown, node 1 has a voltage level of 110 kV, nodes 2 and 3 have a voltage level of 35 kV, nodes 4, 5, and 6 have a voltage level of 10 kV, and the remaining nodes have a voltage level of 380 V. The lower-level nodes of this topology can be divided into three functional areas based on the statistical density ratio of points of interest: residential area, commercial area, and public service area. Points of interest in the residential area include residential communities and apartments, in the commercial area they include shopping malls and office buildings, and in the public service area they include hospitals and government agencies.

[0129] Consider two types of electric vehicles, type 1 with the parameter being battery capacity. =0.07 MWh, charging and discharging power =0.01 MW, Type 2 parameter is battery capacity =0.024 MWh, charging and discharging power =0.004 MW, the initial state of charge of the electric vehicle follows a normal distribution with a mean of 0.5 and a standard deviation of 0.1, and the node voltage limit is set to... The apparent power limit of the line is determined based on the actual capacity of the line.

[0130] The simulation environment is a Windows 64-bit operating system, an Intel(R) Core(TM) i7-7700 CPU @3.6GHz, and 64GB of memory.

[0131] Initialize the scenarios and set up three scenarios with different date types for analysis and comparison. Scenario 1 is set to 9:00 AM on a weekday, Scenario 2 to 6:00 PM on a weekday, and Scenario 3 to 9:00 AM on a weekend. The total number of trips for each scenario is set to 2000. Considering the characteristics of the functional area, when random vehicles simulate departure and destination, Table 1 sets different distribution ratios of departure and destination according to different date types and scenarios.

[0132] Table 1. Distribution ratio of origin and destination

[0133]

[0134] Statistical analysis was performed on the voltage amplitude of each node under different scenarios, and voltage distribution curves were plotted. The results show that at 6 PM on a weekday, i.e., under scenario 2 where electric vehicles are charging intensively, the voltage of some nodes exceeds the limit. Among them, the voltage drop of nodes in commercial and residential areas is significant. After optimizing the charging strategy using the evaluation method of this invention, the node voltage can be kept within the normal range. Analyzing the line load rate, at 6 PM on a weekday, i.e., under scenario 2 where electric vehicles are charging haphazardly, the line load rate of some lines exceeds the upper limit, posing an overload risk. After optimizing the charging mode using the method of this invention, the line load rate can be kept within the safe range.

[0135] Meanwhile, the feasibility of the scenario is judged by using a trained optimized random forest classifier. The carrying capacity assessment results of electric vehicle charging load in the power distribution network are shown in Table 2, which gives the maximum values ​​of node voltage and node power injection for each scenario.

[0136] Table 2. Bearing capacity assessment results

[0137]

[0138] The evaluation results of this embodiment are compared with those using the traditional Monte Carlo method. The errors of the two methods are calculated. The results show that the error between the evaluation method of this invention and the Monte Carlo method is within 5%, demonstrating high evaluation accuracy. Furthermore, comparing the evaluation times shows that this method, when handling the same number of scenarios, takes only 20% of the time of the Monte Carlo method, significantly reducing the evaluation time. This demonstrates that this embodiment can provide a scientific basis for decision-making in distribution network planning and scheduling.

[0139] This embodiment introduces an agent-cellular automaton model, which abstracts electric vehicles as intelligent agents with autonomous decision-making capabilities, defines their attributes and behavioral rules, and constructs a cellular automaton environment. This model can simulate the spatiotemporal distribution and charging behavior of electric vehicles in different functional areas such as residential areas, commercial areas, and public service areas. It covers various states from feasible to critical to infeasible, thus accurately reflecting the dynamic changes of electric vehicles in these areas.

[0140] This embodiment utilizes the DistFlow model to characterize the power flow relationships of the target distribution network. Based on grid data measured by smart meters and combined with the spatiotemporal distribution of electric vehicle charging demand, the model solves the DistFlow model at different time periods to calculate the active and reactive power flow and voltage values ​​at the electric vehicle charging grid connection point. Considering the nonlinear characteristics of distribution network power flow, this method not only improves the accuracy of the calculation but also enhances computational efficiency, making it particularly suitable for handling computational tasks under extreme conditions.

[0141] To further improve the accuracy of distribution network operation state estimation, this embodiment employs an unscented Kalman filter algorithm. Based on the aforementioned DistFlow model, this algorithm initializes parameters according to the core variables of the distribution network and applies the state transition relationship of the DistFlow model to calculate prediction points. By filtering the measurement data at each time step, the algorithm can output unbiased active and reactive power flow and voltage values, effectively eliminating measurement errors, and is particularly suitable for complex distribution network power flow calculation scenarios.

[0142] Finally, this embodiment proposes an optimized random forest algorithm to construct a mapping between quantified features and scenario feasibility. By extracting quantified features that significantly impact feasibility from electric vehicle charging scenarios, and checking voltage and line load constraints based on defined feasibility questions, the feasibility of the target scenario is determined. This method enables rapid processing of a large number of scenarios, significantly improving the efficiency of assessing the distribution network's capacity to support electric vehicle charging loads, and solving the problems of long processing times and computational complexity in traditional methods. This makes the assessment of the electric vehicle carrying capacity of the distribution network more efficient and practical.

[0143] The aforementioned method for assessing the carrying capacity of electric vehicles in power distribution networks integrates an agent-cellular automaton model, a DistFlow model, and an unscented Kalman filter algorithm, combined with an optimized random forest model, to leverage artificial intelligence to deeply mine useful information from historical data. This method first generates diverse charging scenarios based on points of interest and electric vehicle parameters within the target area. Then, using data collected from smart meters, it calculates the nodal power of each charging point using the DistFlow model, solves the power flow relationship across different time intervals, and processes the data using an unscented Kalman filter algorithm to ensure operational feasibility. Next, key features are extracted to train and optimize the random forest model, establishing a mapping between features and scenario feasibility, thereby accurately assessing the carrying capacity of the power distribution network. This method not only reduces reliance on numerous simulated scenarios and lowers computational complexity but also effectively addresses uncertainties, meets the real-time dispatching needs of the power grid, and provides more scientific and reasonable planning recommendations.

[0144] Figure 3 This is a schematic block diagram of a power distribution network electric vehicle load-bearing capacity assessment system 300 provided in an embodiment of the present invention. Figure 3 As shown, corresponding to the above-described method for assessing the carrying capacity of electric vehicles in power distribution networks, the present invention also provides a system 300 for assessing the carrying capacity of electric vehicles in power distribution networks. This system 300 includes a unit for executing the above-described method for assessing the carrying capacity of electric vehicles in power distribution networks, and the system can be configured in a server. Specifically, please refer to... Figure 3 The electric vehicle carrying capacity assessment system 300 for power distribution networks includes a scenario generation unit 301, a judgment unit 302, and a prediction unit 303.

[0145] The scenario generation unit 301 is used to analyze points of interest and electric vehicle parameters within the target area of ​​the target distribution network, and simulate charging behavior using an agent-cellular automaton model to generate diverse electric vehicle charging scenarios. The judgment unit 302 is used to collect actual power grid operation data using smart meters, characterize the power flow relationship of the target distribution network using the DistFlow model, calculate the node power of each electric vehicle charging connection point, solve the DistFlow model in different time intervals, and use the unscented Kalman filter algorithm to process the power flow relationship of the target distribution network based on the solution results, define and solve feasibility problems, and judge the operational feasibility of each scenario by checking whether the voltage level and line load meet the constraints. The prediction unit 303 is used to establish an optimized random forest model, extract the key features of the electric vehicle charging scenario, train the optimized random forest model on the key features, establish a mapping between features and scenario feasibility, and evaluate the carrying capacity of the distribution network for electric vehicle charging load.

[0146] In one embodiment, the scene generation unit 301 includes: a partitioning subunit, used to analyze the distribution of points of interest within the target power distribution network area to determine the type of partitioned area and the boundary range where electric vehicles are concentrated; a construction subunit, used to randomly select samples based on historical data to obtain the basic parameters of electric vehicles, abstract them into intelligent agents, set attribute and behavior rules, and construct a cellular automaton environment to obtain an intelligent agent-cellular automaton model; and a simulation subunit, used to simulate charging behavior using the intelligent agent-cellular automaton model to generate diverse electric vehicle charging scenarios.

[0147] In one embodiment, sub-units are used to divide the target power distribution network area according to a custom scale, covering residential areas, commercial areas and public service areas; the frequency density ratio of different functional areas is calculated according to the number and weight of points of interest. If the frequency density ratio of a certain type of point of interest exceeds a threshold, the area is defined as the corresponding functional area, wherein the functional area includes residential areas, commercial areas or public service areas.

[0148] In one embodiment, the construction subunit is used to abstract each electric vehicle as an intelligent agent, defining the basic and dynamic attributes of the electric vehicle as the intelligent agent. The basic attributes include starting point, destination, initial SOC, battery capacity, and rated driving range; the dynamic attributes include current location, speed, remaining SOC, and charging demand status; simulating the selection of the vehicle's departure and destination using historical data; formulating driving rules and charging decision rules for the intelligent agent, wherein if the remaining battery power is less than a minimum limit and the destination cannot be reached, the nearest charging station is selected for charging; otherwise, charging is performed after reaching the destination; driving rules include acceleration, deceleration, constant speed, and random adjustment; dividing the road network into cells, defining the neighborhood relationships of the cells, and using an extended Moore-type neighborhood structure to support the intelligent agent's movement between cells; recording the current number of vehicles and congestion level for each cell; and setting the behavioral parameters of the electric vehicle according to the characteristics of functional areas, including the maximum speed, random deceleration probability, and energy consumption in different areas, to obtain an intelligent agent-cellular automaton model.

[0149] In one embodiment, a simulation subunit is used to utilize an agent-cellular automaton model based on A... The algorithm's shortest path planning method determines the driving route for each agent; adjusts the vehicle's speed according to the speed limit of the area, and updates the vehicle's position at a certain time; updates the remaining battery power of the electric vehicle based on the distance traveled and the corresponding energy consumption; when the remaining battery power is detected to be lower than the set minimum limit, the vehicle will be guided to the nearest charging station for charging; counts the number of electric vehicles and their charging demand in each cell, and records the changes in charging load at different time points; based on the spatiotemporal distribution evolution model of charging demand, it analyzes and obtains the distribution of electric vehicle charging demand in different time periods and different functional areas to obtain the electric vehicle charging scenario.

[0150] In one embodiment, a determination sub-unit is used to obtain the topology model N=(V,E) of the target distribution network, where V is the set of nodes. For the set of lines, combined with the electric vehicle charging scenario, in each cell, the electric vehicle will select the nearest charging station for charging. The load demand of the charging station is obtained by accumulating the charging requests of all electric vehicles in its cell, and mapped to each node of the distribution network. Based on advanced measurement technology, smart meters are used to measure the actual operating data of the power grid, including the active and reactive power injection of nodes and the amplitude of node voltage. Describe the DistFlow model, where, p and q represent the active and reactive power injections at the node, respectively; v is the square of the node voltage amplitude; P and Q represent the active and reactive power flow of the line, respectively. and The lines are respectively The resistance and reactance; constraints include active / reactive power flow and voltage limits on the line, denoted as: .

[0151] In one embodiment, the determination unit 302 is used for determining the candidate access location of the electric vehicle. Scene This represents the electric vehicle resources accessed at location L and their net injection or consumption at each node during time period T. The net power curve is For each electric vehicle scenario ψ, including location, quantity, and charging behavior, through... The model within the corresponding time step t is obtained, where, These represent the active and reactive power flow and the square of the voltage amplitude of the line within the corresponding time step t; assuming the electric vehicle operates in constant power charging mode and maintains a fixed power factor η through simple reactive power control, combined with the baseline load { , } and electric vehicle charging load{ },pass , Net power injection into compute nodes , , where the matrix It is the adjacency matrix from the electric vehicle's location to the bus, that is, when the j-th electric vehicle is located on bus i... ,otherwise .

[0152] In one embodiment, the judgment unit 302 is used to set the state and observation variables of the power flow calculation of the distribution network based on the DistFlow model corresponding to the solution result, initialize the no-load state and its uncertainty matrix, configure the UKF algorithm parameters according to the characteristics of the distribution network; generate Sigma points to represent the current state distribution using the current state and covariance, predict the future state changes of the Sigma points through the DistFlow model to obtain the predicted state mean and covariance, and form a prediction result; convert the prediction result into the observation space to calculate the observation mean and covariance and determine the Kalman gain, update the system state estimate using actual measurement data; define a feasibility problem to check whether the voltage and line load constraints are met. For each node within the time step, set its voltage upper and lower limits, and specify the active and reactive power flow and apparent power upper limits for each line; construct a scenario, divided into feasible and infeasible cases, and evaluate the overall feasibility of the scenario based on the upper limit of the number of times an infeasible solution is allowed to occur in several time steps. If the voltage and line load constraints are met in all time steps, the scenario is considered feasible; otherwise, it is infeasible.

[0153] In one embodiment, the prediction unit 303 is used to extract quantitative features that significantly affect feasibility from the electric vehicle charging scenario, including the number and proportion of electric vehicles, the spatiotemporal distribution of charging demand, access location, and net power injection of distribution network nodes, forming a feature vector; to determine the actual feasibility of each electric vehicle charging scenario by simulating the voltage and line load constraints within a set time period, and to set a label value for each scenario accordingly, constructing an initial training set; to repeatedly and randomly select samples from the initial training set using a resampling method as a training subset for the decision tree, and for each node, to apply the optimal splitting strategy to the randomly selected variables until no further splitting is possible, forming a random forest structure composed of multiple decision trees, and to use the remaining samples to backsample each decision tree in the random forest. The prediction results are used to verify the model's performance. The hyperparameters of the AdaBoost model are optimized using grid search, and the performance of the entire training tree model is iteratively optimized based on the AdaBoost concept. The learner's performance is enhanced by weighted training of each decision tree. After each iteration, the weights of the training data are adjusted based on the prediction error of the previous decision tree, and a new decision tree is retrained. Mean absolute error and root mean square error are used as evaluation metrics to verify the model's prediction accuracy. The prediction results of all decision trees are weighted and summed to obtain the final prediction result, which is used to determine whether the voltage / line load constraints of the target area are met, thereby determining the feasibility of the scenario. Specifically, by analyzing the characteristic distribution patterns of feasible scenarios, the maximum node power injection and voltage values ​​that meet the constraints are obtained to determine the carrying capacity of the electric vehicle charging load in the distribution network.

[0154] It should be noted that those skilled in the art can clearly understand that the specific implementation process of the above-mentioned power distribution network electric vehicle carrying capacity assessment system 300 and each unit can be referred to the corresponding description in the foregoing method embodiments. For the sake of convenience and brevity, it will not be repeated here.

[0155] The aforementioned power distribution network electric vehicle carrying capacity assessment system 300 can be implemented as a computer program, which can be used in, for example... Figure 4 It runs on the computer device shown.

[0156] Please see Figure 4 , Figure 4 This is a schematic block diagram of a computer device provided in an embodiment of this application. The computer device 500 can be a server, wherein the server can be a standalone server or a server cluster composed of multiple servers.

[0157] See Figure 4The computer device 500 includes a processor 502, a memory, and a network interface 505 connected via a system bus 501. The memory may include a non-volatile storage medium 503 and internal memory 504.

[0158] The non-volatile storage medium 503 may store an operating system 5031 and a computer program 5032. The computer program 5032 includes program instructions that, when executed, cause the processor 502 to perform a method for evaluating the carrying capacity of electric vehicles in a power distribution network.

[0159] The processor 502 provides computing and control capabilities to support the operation of the entire computer device 500.

[0160] The internal memory 504 provides an environment for the operation of the computer program 5032 in the non-volatile storage medium 503. When the computer program 5032 is executed by the processor 502, the processor 502 can execute a method for evaluating the carrying capacity of electric vehicles in a power distribution network.

[0161] This network interface 505 is used for network communication with other devices. Those skilled in the art will understand that... Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device 500 to which the present application is applied. The specific computer device 500 may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0162] The processor 502 is used to run the computer program 5032 stored in the memory to implement all the steps of the method for assessing the carrying capacity of electric vehicles in the power distribution network.

[0163] It should be understood that in the embodiments of this application, the processor 502 may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0164] It will be understood by those skilled in the art that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program includes program instructions and can be stored in a storage medium, which is a computer-readable storage medium. The program instructions are executed by at least one processor in the computer system to implement the process steps of the embodiments of the above methods.

[0165] Therefore, the present invention also provides a storage medium. This storage medium can be a computer-readable storage medium. The storage medium stores a computer program, wherein when executed by a processor, the computer program causes the processor to perform all the steps of the power distribution network electric vehicle carrying capacity assessment method.

[0166] The storage medium can be any computer-readable storage medium capable of storing program code, such as a USB flash drive, portable hard drive, read-only memory (ROM), magnetic disk, or optical disk.

[0167] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0168] In the embodiments provided by this invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For example, the division of each unit is only a logical functional division, and there may be other division methods in actual implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.

[0169] The steps in the method of this invention can be adjusted, merged, or reduced in order according to actual needs. The units in the system of this invention can be merged, divided, or reduced according to actual needs. Furthermore, the functional units in the various embodiments of this invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0170] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a terminal, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention.

[0171] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for assessing the carrying capacity of electric vehicles in a power distribution network, characterized in that, include: For the target distribution network, we analyze the points of interest and electric vehicle parameters within the target area, and use an agent-cellular automaton model to simulate charging behavior to generate diverse electric vehicle charging scenarios. Using the electric vehicle charging scenario, smart meters are used to collect actual power grid operation data, the DistFlow model is used to characterize the power flow relationship of the target distribution network, the node power of each electric vehicle charging connection point is calculated, and the DistFlow model is solved in different time intervals. Based on the solution results, the unscented Kalman filter algorithm is used to process the power flow relationship of the target distribution network, the feasibility problem is defined and solved, and the operational feasibility under each scenario is judged by checking whether the voltage level and line load meet the constraints. An optimized random forest model is established to extract key features of the electric vehicle charging scenario. The optimized random forest model is then used to train the key features to establish a mapping between features and scenario feasibility, and to evaluate the load-bearing capacity of the power distribution network for electric vehicle charging. The solution results are used to process the power flow relationships of the target distribution network using an unscented Kalman filter algorithm, defining and solving feasibility issues. By checking whether the voltage level and line load meet the constraints, the operational feasibility under each scenario is determined, including: Based on the DistFlow model corresponding to the solution results, the state and observation variables for power flow calculation of the distribution network are set, the no-load state and its uncertainty matrix are initialized, and the UKF algorithm parameters are configured according to the characteristics of the distribution network. Sigma points are generated using the current state and covariance to represent the current state distribution. The future state changes of the Sigma points are predicted using the DistFlow model to obtain the mean and covariance of the predicted states, thus forming the prediction results. The prediction results are transformed into the observation space to calculate the observation mean and covariance and determine the Kalman gain. The system state estimate is then updated using actual measurement data. Define a feasibility problem to check whether the voltage and line load constraints are met; for each node within the time step, set its voltage upper and lower limits, and specify active and reactive power flow and apparent power upper limits for each line; The scenario is constructed and divided into two cases: feasible and infeasible. The overall feasibility of the scenario is evaluated based on the upper limit of the number of times an infeasible solution can occur in several time steps. If the voltage and line load constraints are met in all time steps, the scenario is considered feasible; otherwise, it is infeasible.

2. The method for assessing the carrying capacity of electric vehicles in a power distribution network according to claim 1, characterized in that, The method involves analyzing points of interest and electric vehicle parameters within the target area of ​​the target power distribution network, and simulating charging behavior using an agent-cellular automata model to generate diverse electric vehicle charging scenarios, including: For the target distribution network, analyze the distribution of points of interest within the target distribution network area to classify the types of areas and determine the boundary range of concentrated occurrence of electric vehicles; Based on historical data, samples are randomly selected to obtain the basic parameters of electric vehicles, which are then abstracted into intelligent agents. Attributes and behavioral rules are set, and a cellular automaton environment is constructed to obtain the intelligent agent-cellular automaton model. We use an agent-cellular automaton model to simulate charging behavior and generate diverse electric vehicle charging scenarios.

3. The method for assessing the carrying capacity of electric vehicles in a power distribution network according to claim 2, characterized in that, The process of analyzing the distribution of points of interest within the target distribution network area to classify the region and determine the boundary range of concentrated electric vehicle occurrences includes: The target power distribution network area is divided according to a custom scale, covering residential areas, commercial areas and public service areas; The frequency density ratio of different functional areas is calculated based on the number and weight of points of interest. If the frequency density ratio of a certain type of point of interest exceeds the threshold, the area is defined as the corresponding functional area. The functional areas include residential areas, commercial areas, or public service areas.

4. The method for assessing the carrying capacity of electric vehicles in a power distribution network according to claim 2, characterized in that, The process involves randomly selecting samples from historical data to obtain basic parameters of electric vehicles, abstracting these parameters as intelligent agents, setting attribute and behavioral rules, and constructing a cellular automaton environment to obtain an agent-cellular automaton model, including: Each electric vehicle is abstracted as an intelligent agent, and the basic and dynamic attributes of the electric vehicle as an intelligent agent are defined. The basic attributes include the starting point, the destination, the initial SOC, the battery capacity, and the rated driving range; the dynamic attributes include the current position, speed, remaining SOC, and charging demand status. Simulate vehicle departure and destination selection using historical data; Formulate driving rules and charging decision rules for the intelligent agent. If the remaining battery power is less than the minimum limit and the destination cannot be reached, the agent will choose the nearest charging station to charge; otherwise, it will charge after reaching the destination. The driving rules include acceleration, deceleration, constant speed, and random adjustment. The road network is divided into cells, the neighborhood relationship of the cells is defined, and an extended Moore-type neighborhood structure is adopted to support the movement of intelligent agents between cells. The current number of vehicles and the degree of congestion are recorded for each cell. The behavioral parameters of electric vehicles are set according to the characteristics of functional areas, including the maximum vehicle speed, random deceleration probability and energy consumption in different areas, in order to obtain an agent-cellular automaton model.

5. The method for assessing the carrying capacity of electric vehicles in a power distribution network according to claim 2, characterized in that, The method of using an agent-cellular automaton model to simulate charging behavior to generate diverse electric vehicle charging scenarios includes: Using the agent-cellular automaton model with A The algorithm's shortest path planning method determines the travel route for each agent; Adjust the vehicle's speed according to the speed limit of the area and update the vehicle's position at a certain time; The remaining battery level of the electric vehicle is updated based on the distance traveled and the corresponding energy consumption. When the remaining battery level is detected to be lower than the set minimum limit, the vehicle will be guided to the nearest charging station for charging. The number of electric vehicles and their charging demand in each cell are counted, and the changes in charging load at different time points are recorded. Based on the spatiotemporal distribution evolution model of charging demand, the distribution of electric vehicle charging demand in different time periods and different functional areas is analyzed and obtained to obtain the electric vehicle charging scenario.

6. The method for assessing the carrying capacity of electric vehicles in a power distribution network according to claim 1, characterized in that, The method of using the DistFlow model to characterize the power flow relationships of the target distribution network includes: Obtain the topology model of the target distribution network N =( V,E ), V For a set of nodes, Given a set of lines, and considering the electric vehicle charging scenario, in each cell, the electric vehicle will select the nearest charging station for charging. The load demand of the charging station is obtained by accumulating the charging requests of all electric vehicles in its cell and mapping it to each node of the distribution network. Based on advanced measurement technology, smart meters are used to measure actual power grid operating data, including active and reactive power injection at nodes and the magnitude of node voltage. Describe the DistFlow model, where, p and q represent the active and reactive power injections at the node, respectively; v is the square of the node voltage amplitude; P and Q represent the active and reactive power flow of the line, respectively. and The lines are respectively The resistance and reactance; constraints include active / reactive power flow and voltage limits on the line, denoted as: .

7. The method for assessing the carrying capacity of electric vehicles in a power distribution network according to claim 1, characterized in that, The calculation of the nodal power of each electric vehicle charging connection point and the solution of the DistFlow model in different time intervals include: Electric vehicle candidate access locations are Scene This represents the electric vehicle resources accessed at location L and their net injection or consumption at each node during time period T. The net power curve is ; For each electric vehicle scenario ψ, including location, quantity, and charging behavior, through... The model within the corresponding time step t is obtained, where, These represent the active and reactive power flow and the square of the voltage amplitude of the line within the corresponding time step t, respectively. Assuming the electric vehicle operates in constant power charging mode and maintains a fixed power factor η through simple reactive power control, combined with the baseline load { , } and electric vehicle charging load{ },pass , Net power injection into compute nodes , , where the matrix It is the adjacency matrix from the electric vehicle's location to the bus, that is, when the j-th electric vehicle is located on bus i... ,otherwise .

8. The method for assessing the carrying capacity of electric vehicles in a power distribution network according to claim 1, characterized in that, The process involves establishing an optimized random forest model to extract key features of the electric vehicle charging scenario, training the optimized random forest model on these key features, establishing a mapping between features and scenario feasibility, and evaluating the power distribution network's capacity to handle electric vehicle charging loads. This includes: Quantitative features that significantly affect feasibility are extracted from the electric vehicle charging scenario, including the number and proportion of electric vehicles, the spatiotemporal distribution of charging demand, access location, and net power injection of distribution network nodes, to form a feature vector. The feasibility of each electric vehicle charging scenario is determined by simulating the voltage and line load constraints within a set time period, and a label value is set for each scenario accordingly to construct an initial training set. The resampling method is used to repeatedly and randomly select samples from the initial training set as a training subset of the decision tree. For each node, the optimal splitting strategy is applied to the randomly selected variable until it can no longer be split, forming a random forest structure composed of multiple decision trees. The remaining samples are used to perform regression prediction on each decision tree in the random forest to verify the model performance. The hyperparameters of the AdaBoost model are optimized by grid search, and the performance of the entire training tree model is iteratively optimized based on the AdaBoost idea. The performance of the learner is enhanced by weighted training of each decision tree. After each iteration, the weights of the training data are adjusted according to the prediction error of the previous decision tree, and a new decision tree is retrained. The mean absolute error and root mean square error are used as evaluation indicators to verify the prediction accuracy of the model. The final prediction result is obtained by weighted summation of the prediction results of all decision trees. This is used to determine whether the voltage / line load constraints of the target area are met, and thus determine the feasibility of the scenario. In particular, by analyzing the characteristic distribution pattern of feasible scenarios, the maximum node power injection and voltage values ​​that meet the constraints are obtained, and the carrying capacity of electric vehicle charging load of the distribution network is determined.

9. A power distribution network electric vehicle load-bearing capacity assessment system, characterized in that, The system uses the electric vehicle carrying capacity assessment method for power distribution networks as described in any one of claims 1 to 8, including: The scene generation unit is used to analyze points of interest and electric vehicle parameters within the target area for the target power distribution network, and to simulate charging behavior using an agent-cellular automaton model to generate diverse electric vehicle charging scenarios. The judgment unit is used to collect actual power grid operation data using smart meters in the electric vehicle charging scenario, use the DistFlow model to characterize the power flow relationship of the target distribution network, calculate the node power of each electric vehicle charging connection point, solve the DistFlow model in different time intervals, use the unscented Kalman filter algorithm to process the power flow relationship of the target distribution network based on the solution results, define and solve the feasibility problem, and judge the operational feasibility under each scenario by checking whether the voltage level and line load meet the constraints. The prediction unit is used to establish an optimized random forest model, extract key features of the electric vehicle charging scenario, train the optimized random forest model on the key features, establish a mapping between features and scenario feasibility, and evaluate the carrying capacity of the power distribution network for electric vehicle charging load.

Images