Regional electric vehicle charging load prediction and management method and device, and storage medium

Abstract

The application discloses a regional electric vehicle charging load prediction and management method and device and a storage medium, and relates to the technical field of electric vehicle load interactive operation management. The method comprises the following steps: constructing a regional traffic road model, a vehicle travel space-time model and an electric vehicle charging model; determining whether an electric vehicle is charging at a node based on the relative state of charge of the electric vehicle, and combining the regional traffic road model, the vehicle travel space-time model and the electric vehicle charging model to construct charging load space-time prediction models for nodes and regions respectively; partitioning a power network and constructing a distributed operation management model based on the partitioning; and solving the distributed operation management model based on charging load prediction results obtained by the charging load space-time prediction model to perform power distribution management. The application improves the accuracy of electric vehicle charging load prediction and reduces the complexity of prediction.

Description

Technical Field

[0001] This invention relates to the field of electric vehicle load interactive operation management technology, specifically to a method, device and storage medium for regional electric vehicle charging load prediction and management. Background Technology

[0003] The rapid development of EVs has also brought some challenges. Firstly, the large-scale charging demand of EVs has led to a surge in grid load. Because existing grid infrastructure was built relatively early, many designs were based on traditional electricity demand and failed to fully consider the charging needs of EVs. This situation means that when concentrated charging occurs at specific times or locations, grid stability may be affected, leading to problems such as voltage fluctuations and insufficient line transmission capacity. In particular, the spatiotemporal uncertainty of EV charging poses a greater challenge to grid operation. When the base load is high, if a large number of EVs simultaneously connect for charging, it may trigger a series of problems, increase the complexity of grid dispatching, reduce its economic efficiency, and even threaten grid security. Therefore, to address the potential risks to grid security posed by concentrated EV charging, effective charging management strategies must be developed. Ensuring the safe operation of the grid while improving the user's charging experience is of paramount importance.

[0004] As the number of EVs increases, the charging load gradually intensifies, especially when a large number of EVs are charged simultaneously within a short period, a problem that is receiving increasing attention. Charging demand fluctuates significantly and is uncertain at different times, particularly at night and during peak hours, posing a serious challenge to the stability and reliability of the power system. Power grid dispatch and planning rely on accurate forecasting of EV charging load; therefore, to achieve orderly charging of large-scale EVs, accurate forecasting of charging load is essential. Effective load forecasting is crucial for optimizing power grid resource allocation, improving grid stability, and reducing energy waste. Globally, the academic community has shifted its research focus to in-depth analysis of the interaction between EVs and the power grid, and has conducted extensive research on EV charging load forecasting. EV charging load forecasting can be categorized from different perspectives, primarily based on time scale, method, and model type. In terms of time scale, short-term forecasts (such as hourly and intraday forecasts) are typically used for real-time power grid dispatch, requiring high timeliness and accuracy; while long-term forecasts (such as daily and weekly forecasts) are mostly used for power grid planning and load management, focusing on analyzing EV charging trends and their long-term impact on the power grid. In terms of prediction methods, statistical methods analyze historical data and combine it with relevant factors (such as weather, season, number of charging stations, etc.) to build mathematical models for prediction. These methods are widely used and mature in charging load forecasting. With the growth of data volume, machine learning methods (such as support vector machines, random forests, neural networks, etc.) have also made significant progress, especially in handling nonlinear characteristics and high-dimensional problems. In addition, some studies employ combined prediction methods, combining multiple models to improve prediction accuracy. Common combinations include combining statistical methods with machine learning methods, or integrating multiple machine learning algorithms, such as Extreme Gradient Boosting (XGBoost) and Long Short-Term Memory (LSTM) models. Regarding model types, behavior-based models focus on analyzing the charging behavior of EV users (such as travel habits and charging time selection), suitable for load forecasting of individual users; while group behavior-based models predict regional load by analyzing the charging habits of large-scale user groups, typically covering group charging patterns and charging station load analysis.

[0005] In summary, current EV charging load forecasting increasingly emphasizes the spatiotemporal distribution of user travel. However, existing EV charging load forecasting methods, especially those employing artificial intelligence and deep learning technologies, typically focus only on short-term predictions. The accuracy of these methods is easily affected by data quality, and they suffer from difficulties in acquiring datasets, incomplete relevant data, difficulty in obtaining training data, and complex data processing. These aspects urgently require further improvement.

[0006] Furthermore, the interactive operation management and control of regional electric vehicles and the power grid can reduce system losses and improve power supply reliability, resulting in significant economic and social benefits. This is of great significance to power system research. Ensuring the quality of operation is crucial for the efficient operation and healthy development of the power grid. Traditional distribution network operation management and control mainly focus on regulating on-load tap-changing transformers, switched capacitor banks, and switch reconfiguration to achieve the goal of energy conservation and consumption reduction. However, due to the formation of electric vehicle charging and discharging loads, the optimization of power system operation becomes extremely complex. The dispersed and fluctuating nature of these electric vehicle loads will bring significant technical challenges to the centralized control of regional interactive operation management.

[0007] Currently, centralized optimization scheduling methods are commonly used in regional energy management systems (EMS) where information and data collected from various terminals are uploaded to a central dispatch center. The dispatch center centrally processes all collected global information and data, then communicates and transmits the results to each terminal to achieve optimal scheduling. However, when prosumers enter the market, they integrate a large number of distributed resources, making system optimization extremely complex. The dispersion of these network components presents significant technical challenges to centralized control. In terms of hardware requirements, traditional centralized optimization management significantly increases the demand for configured computers and storage devices at the optimization center, as well as communication bandwidth requirements, which greatly reduces computational efficiency and increases economic costs. Regarding optimization objectives, traditional centralized optimization management can only set uniform optimization objectives globally, failing to achieve different optimization objectives for each region. Information privacy is also a concern; for security reasons, different stakeholders involved in the operation may not be able to obtain detailed information from the centralized EMS. In terms of data processing, renewable energy generation fluctuates significantly, leading to excessive delays in centralized control. The massive amounts of data from large-scale distributed resources place a heavy burden on centralized optimization in terms of communication and data processing. Finally, centralized EMS suffers from single points of failure in terms of system reliability. If the centralized optimization management center malfunctions or is attacked, the entire regional interactive operation management will collapse, with serious consequences. Because the regional operation mode includes optimization objects such as switch reconstruction involving integer variables, it is actually a multi-cycle mixed integer optimization delegation, making it difficult for traditional centralized optimization methods to obtain a satisfactory global optimum within a reasonable timeframe.

[0008] Furthermore, there are certain technical hurdles to integrating distributed generation (DG) resources and control equipment into actual distribution systems, particularly within the context of traditional centralized operation and management. However, distributed operation schemes have proven effective in overcoming these hurdles, especially in partitioning the network solely based on the distribution network topology. Summary of the Invention

[0009] This invention addresses the problems of accuracy and complexity in current electric vehicle charging load forecasting, as well as the complexity and reliability of centralized optimization scheduling. It provides a regional electric vehicle charging load forecasting and management method, equipment, and storage medium, which improves the accuracy of electric vehicle charging load forecasting and reduces the complexity of forecasting.

[0010] The present invention is achieved through the following technical solution.

[0011] On one hand, the present invention provides a method for predicting and managing regional electric vehicle charging load, the method comprising:

[0012] Based on road network topology and regional information, a regional traffic road model is constructed; based on travel information, a spatiotemporal model of electric vehicle travel with different levels of complexity is constructed; based on parameters of household electric vehicles and charging facilities, an electric vehicle charging model is established.

[0013] Based on the relative state of charge of electric vehicles, it is determined whether electric vehicles are charging at nodes. Combined with the regional traffic road model, the vehicle travel spatiotemporal model, and the electric vehicle charging model, spatiotemporal prediction models for charging load are constructed for nodes and regions respectively.

[0014] The power network is partitioned, and a distributed operation and management model is constructed based on the partitions;

[0015] The distributed operation and management model is solved based on the charging load prediction results obtained from the spatiotemporal prediction model of the charging load to perform power distribution management.

[0016] In some embodiments, determining whether a vehicle is charging at a node based on the vehicle's relative state of charge includes:

[0017] Based on whether the product of the difference between the relative state of charge and the minimum power threshold and the battery capacity is greater than the power consumption required to reach the destination in the current travel chain, it is predicted whether to charge at the current node.

[0018] If the product of the difference between the relative state of charge and the minimum power threshold and the battery capacity is greater than the power consumption required from the end of each trip in the current travel chain, it is predicted whether to charge or not at any node between the end of each trip in the current travel chain.

[0019] If the product of the difference between the relative state of charge and the minimum power threshold and the battery capacity is not greater than the power consumption required to reach the end of each trip in the current travel chain, reduce the last segment of the current travel chain and repeat the steps: predict whether to charge at the current node until the needs of the next trip are met.

[0020] In some embodiments, by combining the regional traffic road model and the vehicle travel spatiotemporal model, spatiotemporal prediction models for charging loads at nodes and regions are constructed, respectively, including:

[0021] Obtain the time weight matrix and distance weight matrix from the regional traffic road model;

[0022] Based on the total number of traffic area divisions, the number of electric vehicles at the current node, the time weight matrix, and the distance weight matrix, for each area division, the travel chain of each vehicle and the corresponding total number of trips and the destination number of each trip segment are selected based on the vehicle travel spatiotemporal model.

[0023] Based on the travel chain and the corresponding total number of trips, the destination number of each trip segment, and whether the electric vehicle needs charging and the stopping time at the end of each trip segment in its travel chain, a spatiotemporal prediction model for charging load is constructed for nodes and regions respectively, and the electric vehicle state parameters are updated. The spatiotemporal prediction model for charging load for nodes is the predicted sum of the charging load of the vehicles held at the current node, and the spatiotemporal prediction model for charging load for regions is the predicted sum of the charging load of the vehicles held at all nodes in the travel chain.

[0024] In some embodiments, partitioning the power network includes:

[0025] Each charging node in the power network is treated as a separate sub-region, and a comprehensive index for each sub-region is calculated.

[0026] Select a sub-region as the first sub-region, select multiple second sub-regions that are different from the first sub-region, calculate the comprehensive index of the sub-regions that are the combination of the first sub-region and each second sub-region, and update the combined sub-regions based on the combination with the largest positive value of the comprehensive index to obtain the updated combined sub-regions.

[0027] Treat the updated combined sub-region as a separate node and repeat the following steps: select a sub-region as the first sub-region, select multiple second sub-regions that are different from the first sub-region, calculate the comprehensive index of the sub-regions combining the first and second sub-regions, and update the combined sub-regions based on the combination with the largest positive comprehensive index value to obtain the updated combined sub-regions. Stop partitioning when the calculated comprehensive index is no longer positive.

[0028] In some embodiments, calculating the comprehensive index of the sub-region includes:

[0029] Based on the road segment delay time weight, active power weight, reactive power weight, and the sub-region to which the node belongs, the active power electrical distance and active power electrical distance between nodes are calculated.

[0030] Based on the number of sub-regions of the power network, the active power demand and reactive power demand of the sub-regions, the active power balance value and reactive power balance value are calculated using the power balance conditions of the boundary nodes and the power balance conditions of the boundary regions.

[0031] The computation time of multiplication and addition / subtraction operations in the algorithm is calculated to determine the computational complexity.

[0032] The comprehensive index is calculated based on the active power electrical distance, the active power electrical distance, the active power balance value, the reactive power balance value, and the computational complexity.

[0033] In some embodiments, the comprehensive index is calculated based on the active power electrical distance, the reactive power electrical distance, the active power balance value, the reactive power balance value, and the computational complexity, including:

[0034] After normalizing the active power electrical distance, the reactive power electrical distance, the active power balance value, the reactive power balance value, and the computational complexity, the comprehensive index is calculated.

[0035] In some embodiments, the composite index is expressed as:

[0036] ;

[0037] in, Electrical distance representing normalized active power; Electrical distance representing normalized reactive power; Represents the normalized active power balance value; This represents the normalized reactive power balance value. This represents the maximum computational complexity of normalization.

[0038] In some embodiments, the distributed distribution network operation and management model is as follows:

[0039] ;

[0040] ;

[0041] ;

[0042] ;

[0043] ;

[0044] ;

[0045] ;

[0046] ;

[0047] ;

[0048] ;

[0049] ;

[0050] ;

[0051] ;

[0052] ;

[0053] ;

[0054] Where T represents the total number of time periods; branch ij represents the positive direction of power flow from node i to node j; E represents the set of all branches in the network; B represents the set of all nodes in the network; r ij This represents the resistance value on branch ij. This represents the current in branch ij during time period t; This represents the reactive power compensation amount of the continuous reactive power compensation device at node j during time period t. This represents the active charging amount of energy storage at node j during time period t; This represents the active power discharge of energy storage at node j during time period t. This represents the lower limit of reactive power compensation for the continuous reactive power compensation device at node j during time period t. This represents the upper limit of reactive power compensation for the continuous reactive power compensation device at node j during time period t. This represents the set of nodes that contain continuous reactive power compensation devices. This represents the reactive power compensation of capacitor bank j at node t during time period t. The number of groups put into operation during time period t is a discrete variable. The compensation power for each CB group; This represents the set of nodes containing the grouped switching capacitor banks; and The 0-1 flags represent the switching operation. If =1, it means that an additional CB is put into operation at node j during time period t. Similarly; This indicates the maximum number of CB (Block Controller) switches that can be made during time period T; This indicates the number of groups put into operation during time period t-1; This represents the upper limit of the number of CB groups that node j can switch to in each iteration; and The 0-1 symbols indicate charging and discharging. =1 indicates that node j is charging during time period t. Similarly; This represents the upper limit of the active power charging capacity of node j. This represents the upper limit of the active power discharge of the energy storage at node j; This represents the set of nodes containing the active power discharge of energy storage. This represents the total energy storage capacity of node j during time period t; This represents the total energy storage capacity of node j during time period t-1; and These represent the charging and discharging efficiencies of node j during time period t, respectively. This represents the lower limit of the total energy storage capacity of node j; This represents the maximum total energy storage capacity of node j; , These represent the active power injection and reactive power injection at node j during time period t, respectively. and This represents the active and reactive power injections at node j in time period t. and This represents the active and reactive loads of node j during time period t. and This represents the net active power exchange and net reactive power exchange of node j after the completion of its local transaction in time period t. This represents the active power sold by entity f; This represents the active power purchased by entity f; Represents the set of regions / nodes of type k; ℓ represents the power factor coefficient; and This represents the active and reactive power flowing from parent node k to current node j during time period t. and This represents the active power and reactive power on branch ij during time period t; This represents the set of upstream parent nodes of the current node j; This represents the set of downstream child nodes of the current node j; , Let represent the conductance and susceptance of node j, respectively; This represents the voltage at node i during time period t; This represents the reactance value on branch ij.

[0055] Secondly, the present invention provides a regional electric vehicle charging load prediction and management device, the device comprising:

[0056] At least one processor; and

[0057] At least one memory is coupled to the at least one processor and stores instructions for execution by the at least one processor, which, when executed by the at least one processor, implement the method described above.

[0058] Thirdly, the present invention provides a computer-readable storage medium, characterized in that the computer-readable storage medium stores instructions that, when executed by a computer, cause the computer to perform the methods described above.

[0059] Compared with existing technologies, this invention has the following advantages and beneficial effects: It establishes a spatiotemporal prediction model of electric vehicle charging load under the dual constraints of regional road network and travel chain; then, it simulates the charging load behavior of electric vehicles and establishes a distributed operation and management model that interacts with the distribution network, thereby improving the stability and security of the power grid operation. Furthermore, based on modular functions, supply and demand balance, and computational complexity, a new power network partitioning method is proposed, and a fast Lagrange multiplier algorithm is used to coordinate power flows between different regions, thereby improving the efficiency of large-scale distribution network operation and management. Attached Figure Description

[0060] To more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be considered as limiting the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.

[0061] Figure 1 This is a flowchart of a regional electric vehicle charging load prediction and management method according to an embodiment of the present invention.

[0062] Figure 2 This is an exploded view of the connection branch according to an embodiment of the present invention.

[0063] Figure 3 This is a schematic diagram illustrating the coordination between two regions according to an embodiment of the present invention.

[0064] Figure 4 This is a schematic diagram of the actual 152-node system topology according to an embodiment of the present invention.

[0065] Figure 5 This is a schematic diagram of the residuals between different regions of an actual 152-node system according to an embodiment of the present invention.

[0066] Figure 6 This is a structural block diagram of a regional electric vehicle charging load prediction and management system according to an embodiment of the present invention.

[0067] Figure 7 This is a schematic diagram of a regional electric vehicle charging load prediction and management device according to an embodiment of the present invention. Detailed Implementation

[0068] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.

[0069] On the one hand, the present invention provides a method for predicting and managing regional electric vehicle charging load. Figure 1 This is a flowchart illustrating a regional electric vehicle charging load prediction and management method according to an embodiment of the present invention. (Reference) Figure 1 The method includes: S10 to S40.

[0070] In S10, a regional traffic road model is constructed based on road network topology information and regional information; a vehicle travel spatiotemporal model with different levels of complexity for electric vehicles is constructed based on travel information; and an electric vehicle charging model is established based on parameters of household electric vehicles and charging facilities.

[0071] In S20, the relative state of charge of electric vehicles determines whether electric vehicles are charging at nodes, and combines regional traffic road models, vehicle travel spatiotemporal models, and electric vehicle charging models to construct spatiotemporal prediction models for charging loads at nodes and regions, respectively.

[0072] In S30, the power network is partitioned and a distributed operation and management model is built based on the partitions.

[0073] In S40, the distributed operation and management model is solved based on the charging load prediction results obtained from the charging load spatiotemporal prediction model in order to carry out power distribution management.

[0074] The contents of S10 to S40 will be described in detail below with reference to the accompanying drawings.

[0075] 1. Spatiotemporal simulation and prediction model of electric vehicle charging load

[0076] First, a regional traffic road model is established based on road network topology and regional information. Then, based on travel information, travel chain models (vehicle travel spatiotemporal models) of varying complexity for electric vehicles are established. An electric vehicle charging model is established based on parameters of household electric vehicles and charging facilities. Based on these three models, travel chains are randomly extracted, and the shortest travel time is used as a constraint. Combined with the regional traffic road model, travel routes and travel times are determined. Based on SOC (State of Charge) and travel purpose, electric vehicle charging locations and charging durations are determined. The charging load of electric vehicles at different times and in different regions is predicted, obtaining the spatiotemporal distribution information of electric vehicle charging load.

[0077] 1.1 Regional Traffic Road Model

[0078] 1.1.1 Road Topology Map

[0079] Assuming all roads in the area are bidirectional, its topology is represented by an undirected graph. The road graph is represented as G = (U, E), where U represents the set of nodes in the undirected graph, i.e., the start and end points or intersections of roads, numbered 1, 2, ..., H; E represents the relationships between nodes. When representing the weighted graph using an adjacency matrix, G = (U, E) corresponds to an H×H adjacency matrix D. ω represents the weight function of the road network, i.e., the road resistance function. The elements d in the adjacency matrix D... ij The assignment rules are as follows:

[0080] (1)

[0081] In the formula: ω ij ω represents the weight between road node i and node j. inf This indicates that there is no direct connection between two nodes.

[0082] 1.1.2 Path Resistance Function Model

[0083] In transportation networks, travel time is primarily limited by road class, traffic lights at intersections, and traffic volume. A Logit-based flow delay function is used to represent the road resistance model. For a directly connected road segment with nodes i and j as endpoints, the time it takes for a vehicle to travel from node i to node j is represented by the total road segment delay function:

[0084] (2)

[0085] In the formula: This represents the total delay time for road segments and intersections; Delays at intersections; The delay time for the road segment is shown in equation (3):

[0086] (3)

[0087] In the formula: This refers to the free-flow travel time of the road segment; Road flow rate, which is the number of cars passing through a road segment per hour, can be obtained through manual observation or traffic flow meters; Indicates the traffic capacity of the road section; This is an adaptive coefficient for different road grades, and it is related to the road grade.

[0088] The delay time at the intersection is mainly determined by the traffic light factor, as shown in equation (4):

[0089] (4)

[0090] In the formula: Free-flow travel time at intersections; To improve the traffic capacity of the road section; This is the adaptive coefficient for the intersection, which is related to whether there are traffic lights.

[0091] 1.1.3 Path Selection

[0092] Path p = (v1, v2, ..., v k The weight of () refers to the distance from the starting point v1 to the destination v. k The sum of the weights of the edges traversed is:

[0093] (5)

[0094] Define the weight of the shortest path from node i to node j. for:

[0095] (6)

[0096] The edge weights are represented using the flow delay function described in Section 1.1.2, i.e. To minimize travel time, Dijkstra's algorithm is used to select the optimal path.

[0097] Assuming the shortest path set between nodes i and j contains n directly connected road segments, the travel time ΔT of the s-th road segment is calculated using a logit-based traffic delay function. s The distance Δl of the s-th directly connected road segment. s Given the quantities, the total travel time ΔT between nodes i and j is ΔT. ij and the total distance traveled Δl ij They are respectively:

[0098] (7)

[0099] (8)

[0100] In reality, users don't always choose the shortest path due to their own reasons. Considering the diversity of user path choices, we define the shortest path as the path the user is most likely to choose, while the probabilities of choosing other paths are equal and uniformly distributed. If there are m (m > 1) possible paths from the starting point to the destination, then the probability of the user choosing the shortest path is P. m The probability of choosing another path is (1-Pm) / (m-1), and this paper takes P. m =0.8.

[0101] 1.2 Spatiotemporal Model of Vehicle Travel

[0102] 1.2.1 Travel Chain Structure

[0103] A trip chain structure is used to represent the travel time, purpose, number of activities, and sequence of events for electric vehicle users. The trip chain length represents the number of trips a user undertakes within one day, with an average of 2.5. During a day's journey, a user may make brief stops, such as picking up or dropping off children during commutes; these are secondary purposes, and the likelihood of them choosing to charge is very low. Considering the primary purpose and ignoring secondary purposes, commutes using private cars are divided into three categories: going home (trip H), working (trip W), and others (trip E). Activities in trip E mainly include shopping, dining, and social leisure. The trip chain structure is divided into two modes: simple chain (HW / EH) and complex chain (HEWH and HWEH).

[0104] 1.2.2 Regional Functional Division

[0105] Based on the travel chain structure and land parcel functions, the area is divided into residential, work, and commercial zones, with each parcel having corresponding road nodes. Taking commuting private cars as the research object, the number of electric vehicles is allocated according to the size of residential areas, and it is assumed that the initial location and final destination of electric vehicles within one day both belong to residential areas. Throughout the area, electric vehicles can travel between the three parcels. It is assumed that charging stations in each parcel are reasonably and adequately planned, and that each parcel contains sufficient charging facilities. The response of charging station location distribution to electric vehicle charging behavior and route selection is not considered.

[0106] 1.2.3 Departure Time

[0107] For the travel chain of residential area—work area—residential area (denoted by HWH, the same below), according to statistics from the Ministry of Transport, this travel chain is fitted with a normal distribution for the initial departure time and return time on weekdays, that is:

[0108] (9)

[0109] In the formula: Let t be the expected value of the starting travel time t of the electric vehicle; Standard deviation; and The value is related to the travel chain structure and the start and end points of the trip.

[0110] Different users, and They are not entirely the same. The mean and standard deviation of the start and return times of the HWH travel chain are statistically analyzed, with the mean of the start time being... H-W A gamma distribution can be used for fitting, i.e.

[0111] (10)

[0112] In the formula: x is a random variable; θ is a shape parameter; α is a scale parameter; Γ(α) is a gamma function, Γ(α) = .

[0113] Standard deviation σ Ｈ-Ｗ A standard normal distribution can be used for fitting, as shown in equation (11):

[0114] (11)

[0115] Mean of return journey time μ W-H The Weibull distribution can be used for fitting, as shown in equation (12):

[0116] (12)

[0117] In the formula: k is the shape parameter; c is the scale parameter.

[0118] Standard deviation σ W-H A normal distribution can be used for fitting. For the trip chain HEH, based on statistical data, a three-parameter Weibull distribution is used to fit the start and return times, as shown in equation (13):

[0119] (13)

[0120] In the formula: λ is the position parameter.

[0121] For the travel chain HEWH, the starting time of the third segment WH is the same as the corresponding part of the simple chain HWH. Considering the fixed commuting time, the starting time t of the second segment EW... E-W The distribution pattern used is the same as that of the simple chain HW, but the starting time t of the first segment HE is different. H-EConstrained by the second segment of the stroke EW, as shown in equation (14):

[0122] (14)

[0123] In the formula: and These are the travel constraint coefficients, all of which are constants.

[0124] For the travel chain HWEH, the starting time of the first segment HW is the same as the corresponding part of the simple chain HWH. Considering the fixed nature of the end of the workday, the starting time of the second segment WE is t. W-E It adopts the same distribution pattern as the simple chain WH, but the starting time t of the third segment EH is... E-H Constrained by the second segment of the stroke WE, as shown in equation (15):

[0125] (15)

[0126] In the formula: As a random variable, considering the randomness of users' activity time in the commercial area, The samples are uniformly distributed and randomly selected.

[0127] 1.3 Charging Load Prediction Model

[0128] 1.3.1 Electric Vehicle Charging Model

[0129] Let n represent the total number of trips under different travel chain structures; n=2 for simple chains and n=3 for complex chains. A complete travel cycle is defined as 1 day. When i=0, the electric vehicle is at the starting position. Considering factors such as battery degradation, assume the SOC (State of Charge) value at the starting position is R. SOC,0 (Relative state of charge) = 0.9. When i = n, the electric car ends its day's journey and returns home. Assume that the electric car chooses to recharge upon returning home to ensure that R is charged at the starting position of each day. SOC,0 =0.9. The SOC at point i is calculated using equation (16):

[0130] (16)

[0131] In the formula: Battery capacity; Power consumption per unit distance; From to Mileage at this point.

[0132] At node i=1,2,...,n-1, the user makes a comprehensive judgment on all subsequent trips based on the path between the starting point and the destination in each segment of the selected travel chain before choosing a charging location. The judgment process includes the following steps 1 and 2.

[0133] Step 1: Predict the SOC based on the mileage of each segment in the trip chain, as shown in Equation (17). If Equation (17) is satisfied, the user can randomly choose to charge at any destination between i and n, or not charge at all; if the charging location chosen by the user does not meet the stay time constraint, then a new charging location can be selected.

[0134] i=1,2,...,n-1 (17)

[0135] In the formula: ξ is the minimum power threshold; Let i be the power consumption per unit distance from point i to point i+1.

[0136] Step 2: If equation (17) is not satisfied, reduce the last segment of the trip in the trip chain, i.e., n=n-1, and continue to repeat step 1 until the user meets the needs of the next segment of the trip. The objective conditions of the battery make it necessary to choose to charge at point i.

[0137] Assuming the user chooses to start charging upon arrival at their destination and continues until fully charged, the charging time at point i is... for:

[0138] (18)

[0139] In the formula: η is the charging efficiency; Let be the charging power at point i.

[0140] Start of charging time for:

[0141] (19)

[0142] In the formula: To randomly select the starting time for starting from point i and heading to the next destination; The time taken to travel from point i-1 to point i is obtained from the road resistance function model.

[0143] Duration of stay for:

[0144] = ( + ) (20)

[0145] At the same time, the dwell time constraint must be considered, as shown in Equation (21), that is, the user's dwell time must not be less than his charging time, otherwise the user will reselect a dwell point that meets the constraint to charge.

[0146] (twenty one)

[0147] 1.3.2 Charging Load Forecast

[0148] Based on regional traffic road models and vehicle travel spatiotemporal models, the travel trajectories, dwell times, and State of Charge (SOC) of electric vehicles are obtained, thus acquiring the spatiotemporal distribution information of electric vehicles. A charging calculation model is established, and Monte Carlo methods are used to predict the charging load of electric vehicles in each functional area. The prediction steps include steps 1 to 6 as follows.

[0149] Step 1: Read the regional traffic network information, calculate the road resistance function, and obtain the time weight matrix T and distance weight matrix D of the regional road topology.

[0150] Step 2: The total number of plots in the area is N, and the number of electric vehicles at location i is N. EV, i The travel chains of each vehicle are randomly selected in a uniform distribution. The total number of trips is n. The destination number of each trip is extracted according to the travel chain.

[0151] Step 3: Determine whether the electric vehicle needs to be charged at the end of a certain journey.

[0152] Step 4: Update the electric vehicle status parameters, calculate the dwell time, and continue to the next leg of the journey until the total number of trips in the travel chain is completed.

[0153] Step 5: Continue to the next plot until the charging load forecast for that area is complete. The forecast charging load at point i is expressed in minutes. and the total predicted charging load of the region As shown in equations (22) and (23).

[0154] (twenty two)

[0155] (twenty three)

[0156] In the formula: Let be the charging power of the j-th vehicle at node i.

[0157] 2. Distribution network operation and management model with electric vehicle participation

[0158] 2.1 Equivalent Model of Network Partitioning

[0159] Assume the network system is divided into N PNon-overlapping sub-regions. To achieve region decoupling, variable partitioning was also used, such as... Figure 2 As shown.

[0160] To improve the resolution of the local model in each sub-region, an auxiliary distributed generator source is introduced at node i or j in each sub-region, such as... Figure 3 As shown.

[0161] By P ij and Q ij The power transfer between adjacent sub-regions (with time slot index t omitted) represents the power injection from the auxiliary generator. The updated power balance of the boundary nodes can be determined using equations (24) and (25).

[0162] (twenty four)

[0163] (25)

[0164] in, This represents the net active power injection at node i (positive for injection, negative for outflow). This represents the active power sold in the local market at node i; This represents the active power injected into node i by the upstream power grid (main grid); This represents the active power (power released to the grid) of the energy storage device charging at node i. This represents the active power (power absorbed from the grid) discharged by the energy storage device at node i. This represents the active power purchased by users / loads on node i in the local market; This represents the active load power (power that the power grid needs to supply) at node i. This represents the active power injected into node i from the adjacent region through branch ij; Represents the set of nodes in a network; This represents the set of all cross-regional branch roads; This represents the net reactive power injection at point i; This represents the reactive power injected into node i by the upstream power grid (main grid); This represents the reactive power provided by the capacitor bank at node i; This represents the reactive power provided by the Static VarCompensator at node i; This represents the reactive load power of node i; This represents the reactive power injected into node i from the adjacent region through branch ij.

[0165] For each boundary line, the above relationship is valid as long as boundary branch losses are ignored. Furthermore, these two relationships are considered consensus constraints, ensuring that the transmission power is equal between any region and the boundary region.

[0166] (26)

[0167] Among them, P 1 Q 1 P represents the active and reactive power at the boundary of region 1. 2 and Q 2 P represents the active and reactive power at the boundary of region 2. L and Q L P represents the active and reactive power of the load. sub and Q sub This indicates the active and reactive power of the substation. E tie-line This represents the set of branches between boundary nodes i and j.

[0168] 2.2 Partition Target Model

[0169] 2.2.1 Electrical Distance

[0170] Based on electrical distance, the following weights are preset:

[0171] (27)

[0172] (28)

[0173] (29)

[0174] Based on (28) and (29), the following modular function is obtained:

[0175] (30)

[0176] (31)

[0177] in, Indicates the weight of road segment delay time; Indicates the active power weight; This represents the active power weight edge weight; Indicates reactive power weight; The weight represents the reactive power weight edge weight; when nodes i and j come from different sub-regions, δ(i,j) = 0, while when nodes i and j come from the same sub-region, δ(i,j) = 1.

[0178] 2.2.2 Supply and Demand Balance

[0179] Suppose the network is divided into N PFor each sub-region, based on the power balance conditions of the boundary nodes and the boundary regions in section 2.1, the power balance can be calculated as follows:

[0180] (32)

[0181] (33)

[0182] (34)

[0183] (35)

[0184] (36)

[0185] (37)

[0186] In the formula, N P It is the number of subregions, P a dem and Q a dem These are the active power demand and reactive power demand in region a, respectively; P a gen and Q a gen These are the active power supply and reactive power supply in region a, respectively; P a W0 and P a P0 These are the active power values ​​of wind power and photovoltaic power in region a, respectively. Q a W and Q a P These are the reactive power values ​​of wind power and photovoltaic power in region a, respectively.

[0187] 2.2.3 Computational Complexity

[0188] To improve computational efficiency, a subsequent model is used to obtain the computational complexity function:

[0189] (38)

[0190] st (39)

[0191] (40)

[0192] (41)

[0193] (42)

[0194] (43)

[0195] In the formula, u ∞ This represents the input variables related to the linear constraints. Represents the lower limit of the input associated with the linear constraint; This represents the upper limit of the input associated with linear constraints; u M Represents the input variables related to the second-order cone constraints of the optimization model; y represents the upper bound of the inputs related to the second-order cone constraint of the optimization model, and u represents all input variables. ∞ It is the set of output variables related to linear constraints; Represents the set of lower bounds for outputs related to linear constraints; y represents the set of upper limits of output associated with linear constraints. M It is the set of input variables related to the second-order cone constraint of the optimization model, where Δ represents the change of these variables.

[0196] To solve linear constraint problems, the constraint problems can be solved in the following matrix form:

[0197] (44)

[0198] The computation time for linear constraint multiplication is determined by the following procedure:

[0199] (45)

[0200] Therefore, an approximate estimate of the computation time required for the multiplication operation of the complete optimization model can be obtained as follows:

[0201] (46)

[0202] The computation time for addition and subtraction operations related to linear constraints is as follows:

[0203] (47)

[0204] Therefore, an approximate estimate of the computation time required for addition and subtraction operations in the entire model can be expressed as follows:

[0205] (48)

[0206] In the formula, x ∞ The dimension ∞ of the input variable matrix Δu is related to the linear constraints, x M The input variable matrix Δu represents the input variable matrix associated with the second-order cone constraint. M The dimension of ', T m(∞) and T n(∞)T represents the number of multiplications and additions related to linear constraints, respectively. m and T n These represent the number of multiplications and additions performed in the entire optimization model, respectively.

[0207] In summary, the computational complexity can be determined as follows:

[0208] (49)

[0209] Where, ω m They are T m The weight, ω n It is T n The weight.

[0210] 2.2.4 Composite Index

[0211] The composite index is formed based on electrical distance, supply and demand balance, and computational complexity. Since these indices differ in magnitude, it is necessary to normalize the range to [0,1], as described below.

[0212] (50)

[0213] (51)

[0214] (52)

[0215] (53)

[0216] (54)

[0217] Therefore, the comprehensive index is:

[0218] (55)

[0219] (56)

[0220] Specifically, the power network system is divided into grids to obtain the network division results, including the following steps: A to D.

[0221] Step A: Treat each charging node in the power network as a separate sub-region and calculate the comprehensive index of the sub-region.

[0222] Step B: Based on region a, form a new sub-region (select the first sub-region), randomly select region b (the second sub-region) from the other regions, calculate the comprehensive index change of the combined regions, and update it based on the combination with the largest positive value to obtain the updated combined sub-region.

[0223] Step C: Treat the updated combined sub-region as a single node, repeat step B to perform the partitioning process and obtain the updated sub-region result.

[0224] Step D: When further merging is not possible (the composite index is not positive), the partitioning process stops, the composite index becomes its highest value, and the current sub-region scheme is considered the optimal partitioning result.

[0225] 2.3 Distributed Operation and Management Model of Distribution Network

[0226] Based on the network partitioning results, dynamic reconfiguration implemented network partitioning and enabled the distribution system operator to manage power flow changes caused by electric vehicle charging loads. Building upon this, the distribution network operator optimized network operation and management, establishing a distributed operation and management model, as follows:

[0227] The goal of the distributed operation and management model is to optimize and reduce network and equipment losses through appropriate measures, while ensuring the safe operation of the distribution network.

[0228] (57)

[0229] (58)

[0230] (59)

[0231] (60)

[0232] (61)

[0233] (62)

[0234] (63)

[0235] (64)

[0236] (65)

[0237] (66)

[0238] (67)

[0239] (68)

[0240] (69)

[0241] (70)

[0242] (71)

[0243] Where T represents the total number of time periods; branch ij represents the positive direction of power flow from node i to node j; E represents the set of all branches in the network; B represents the set of all nodes in the network; r ij This represents the resistance value on branch ij. This represents the current in branch ij during time period t; This represents the reactive power compensation amount of the continuous reactive power compensation device at node j during time period t. This represents the active charging amount of energy storage at node j during time period t; This represents the active power discharge of energy storage at node j during time period t. This represents the lower limit of reactive power compensation for the continuous reactive power compensation device at node j during time period t. This represents the upper limit of reactive power compensation for the continuous reactive power compensation device at node j during time period t. This represents the set of nodes that contain continuous reactive power compensation devices. This represents the reactive power compensation of capacitor bank j at node t during time period t. The number of groups put into operation during time period t is a discrete variable. The compensation power for each CB group; This represents the set of nodes containing the grouped switching capacitor banks; and The 0-1 flags represent the switching operation. If =1, it means that an additional CB is put into operation at node j during time period t. Similarly; This indicates the maximum number of CB (Block Controller) switches that can be made during time period T; This indicates the number of groups put into operation during time period t-1; This represents the upper limit of the number of CB groups that node j can switch to in each iteration; and The 0-1 symbols indicate charging and discharging. =1 indicates that node j is charging during time period t. Similarly; This represents the upper limit of the active power charging capacity of node j. This represents the upper limit of the active power discharge of the energy storage at node j; This represents the set of nodes containing the active power discharge of energy storage. This represents the total energy storage capacity of node j during time period t; This represents the total energy storage capacity of node j during time period t-1; and These represent the charging and discharging efficiencies of node j during time period t, respectively. This represents the lower limit of the total energy storage capacity of node j; This represents the maximum total energy storage capacity of node j; , These represent the active power injection and reactive power injection at node j during time period t, respectively. and This represents the active and reactive power injections at node j in time period t. and This represents the active and reactive loads of node j during time period t. and This represents the net active power exchange and net reactive power exchange of node j after the completion of its local transaction in time period t. This represents the active power sold by entity f; This represents the active power purchased by entity f; Represents the set of regions / nodes of type k; ℓ represents the power factor coefficient; and This represents the active and reactive power flowing from parent node k to current node j during time period t. and This represents the active power and reactive power on branch ij during time period t; This represents the set of upstream parent nodes of the current node j; This represents the set of downstream child nodes of the current node j; , Let represent the conductance and susceptance of node j, respectively; This represents the voltage at node i during time period t; This represents the reactance value on branch ij.

[0244] The aforementioned power flow constraints have been widely applied in power flow models. ℓ represents an invariant value. A fast solution algorithm using Lagrange's two multipliers can be used to coordinate power flows between different regions, thereby improving the efficiency of large-scale distribution network operation and management.

[0245] In this invention, a case study of an improved 152-node system was conducted to demonstrate the effectiveness of the proposed model. A schematic diagram of the system is shown below. Figure 4 As shown in Table 1, the parameters of the system are as follows.

[0246] Table 1 System Parameters

[0247]

[0248] The spatiotemporal distribution of electric vehicle charging load within a weekday in a certain region is known. The application of this regional information is explained below.

[0249] 1. The area is 25km*25km, containing 72 road nodes and 122 roads.

[0250] 2. The area consists of 50 plots, including 18 residential areas, 18 work areas, and 13 other areas, each corresponding to a road node. There is also one area representing a location inaccessible to electric vehicles under objective conditions, which is not included in the travel chain.

[0251] 3. Roads are divided into two levels: dark-colored roads are expressways, and light-colored roads are regular roads. The free-flow speed v of expressways... f =55.6km / h, free-flow velocity on ordinary roads v s =45.8km / h. Traffic lights are located at intersections of ordinary roads.

[0252] 4. The number of private cars used for commuting in the area is 8,450, with the initial and final locations both being residential areas. The number of electric vehicles in each residential area is 200-800.

[0253] ω is related to speed and increases with increasing speed. To study the impact of the regional road network on the charging load of electric vehicles, the simulation scenarios are divided into the following three types.

[0254] Scenario 1: Standard Scenario.

[0255] Scenario 2: The area consists of 34 plots, including 14 residential areas, 13 work areas, and 7 other areas. Other parameters are the same as in Scenario 1.

[0256] Scenario 3: Excluding road class and traffic lights, select the route with the shortest travel distance. Other parameters are the same as in Scenario 1. Analysis of the results for Scenario 1, including the evening rush hour time and load value, with varying the number of Monte Carlo simulations, shows that the results obtained after 1000 days of Monte Carlo simulations better reflect the statistical mean and reduce the impact of random simulation errors.

[0257] The maximum load generated by simultaneous charging of electric vehicles during the evening peak in this area is 27.8 MW. Table 2 shows the timing of the evening peak charging load and its proportion of the maximum load under different scenarios. Under the three scenarios, the peak load is approximately the same, with about 40% of electric vehicles charging in residential areas during the evening peak.

[0258] Table 2 Evening Peak Hours and Maximum Load Percentage under Different Scenarios

[0259]

[0260] Comparing Scenario 2 and Scenario 1, the charging load increases in the work area and other areas in Scenario 2. Because the number of plots in Scenario 2 is reduced, the corresponding travel chain combinations are also reduced. Within the same area, the average distance between plots increases, resulting in higher power consumption to reach the work area or other areas compared to Scenario 1, thus increasing charging demand. The reduction in the number of functional areas within the same area will lead to an increase in charging load during the morning peak.

[0261] Comparing Scenario 3 and Scenario 1, Scenario 3 shows reduced traffic complexity and shorter route selection, resulting in a smaller driving distance and a significant decrease in the maximum charging load for electric vehicles during the morning rush hour, while the evening rush hour occurs slightly earlier. As the complexity of the regional road network increases, the charging load for electric vehicles during the morning rush hour rises.

[0262] In addition, such as Figure 5 As shown, the distributed algorithm with double gaps in each iteration converges within 167 iterations when implemented in this system. Furthermore, the proposed distributed operation-managed partitioning model requires minimal computation time compared to other network partitioning methods in the same system. Therefore, the Lagrange dual relaxation algorithm and the proposed network partitioning technique have the potential to be deployed in larger, practical systems consisting of multiple buses and branches, as shown in Table 3.

[0263] Table 3 Comparison of calculation time for regional division models

[0264]

[0265] In this invention, a spatiotemporal prediction model of electric vehicle charging load is established under the dual constraints of regional road network and travel chain. Then, the charging load behavior of electric vehicles is simulated to establish a distributed operation and management model that interacts with the distribution network, thereby improving the stability and security of the power grid operation. Furthermore, based on modular functions, supply and demand balance, and computational complexity, a new power network partitioning method is proposed, and a fast Lagrange multiplier algorithm is used to coordinate power flows between different regions, thereby improving the efficiency of large-scale distribution network operation and management.

[0266] On the other hand, the present invention provides a regional electric vehicle charging load prediction and management system. Figure 6 This is a structural block diagram of a regional electric vehicle charging load prediction and management system according to an embodiment of the present invention. (Reference) Figure 6 The system includes: an auxiliary model building module, a charging load spatiotemporal prediction model building module, a distributed operation management model building module, and a distributed operation management model solving module.

[0267] The auxiliary model building module is used to: build regional traffic road models based on road network topology information and regional information; build spatiotemporal models of electric vehicle travel with different levels of complexity based on travel information; and establish electric vehicle charging models based on parameters of household electric vehicles and charging facilities.

[0268] The module for constructing a spatiotemporal prediction model for charging load is used to: determine whether an electric vehicle is charging at a node based on its relative state of charge, and to construct spatiotemporal prediction models for charging load for nodes and regions by combining regional traffic road models, vehicle travel spatiotemporal models, and electric vehicle charging models.

[0269] The distributed operation and management model construction module is used to partition the power network and build a distributed operation and management model based on the partitions.

[0270] The distributed operation and management model solving module is used to solve the distributed operation and management model based on the charging load prediction results obtained from the charging load spatiotemporal prediction model, so as to carry out power distribution management.

[0271] Other implementation details of the regional electric vehicle charging load prediction and management system can be found in the relevant content of the above-mentioned regional electric vehicle charging load prediction and management method, and will not be repeated here.

[0272] In implementing the functions of the integrated modules described above in hardware, this embodiment of the invention provides a structure for the regional electric vehicle charging load prediction and management device involved in the above embodiments. Figure 7 This is a schematic diagram of a regional electric vehicle charging load prediction and management device according to an embodiment of the present invention. (Reference) Figure 7 The electric vehicle charging load prediction and management device for this region includes: at least one processor; and at least one memory. The at least one memory is coupled to the at least one processor and stores instructions for execution by the at least one processor, which, when executed by the at least one processor, implement the method described above.

[0273] A processor can be a set of logic blocks, modules, and circuits that implement or execute the various exemplary logic blocks, modules, and circuits described in connection with embodiments of the present invention. The processor can be a central processing unit, a general-purpose processor, a digital signal processor, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in connection with embodiments of the present invention. A processor can also be a combination that implements computational functions, such as a combination of one or more microprocessors, a combination of a digital signal processor and a microprocessor, etc.

[0274] The memory may be read-only memory (ROM) or other types of static storage devices that can store static information and instructions, random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, electrically erasable programmable read-only memory (EEPROM), disk storage media or other magnetic storage devices, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer, but is not limited thereto.

[0275] In one implementation, the memory can exist independently of the processor. The memory can be connected to the processor via a bus and used to store instructions or program code. When the processor calls and executes the instructions or program code stored in the memory, it can implement the method provided in the embodiments of the present invention. In another implementation, the memory can also be integrated with the processor.

[0276] On the other hand, the present invention also provides a computer-readable storage medium (e.g., a non-transitory computer-readable storage medium) storing computer program instructions that, when executed on a computer, cause the computer to perform the method as described in any of the above embodiments.

[0277] Exemplary examples show that the aforementioned computer-readable storage media may include, but are not limited to: magnetic storage devices (e.g., hard disks, floppy disks, or magnetic tapes), optical discs (e.g., compact disks (CDs), digital versatile disks (DVDs), etc.), smart cards, and flash memory devices (e.g., erasable programmable read-only memory (EPROMs), cards, sticks, or key drives, etc.). The various computer-readable storage media described in this invention may represent one or more devices and / or other machine-readable storage media for storing information. The term "machine-readable storage medium" may include, but is not limited to, wireless channels and various other media capable of storing, containing, and / or carrying instructions and / or data.

[0278] This invention provides a computer program that, when run on a computer, causes the computer to perform the method of any of the above embodiments.

[0279] This invention provides a computer program product containing instructions that, when run on a computer, cause the computer to perform the method of any of the above embodiments.

[0280] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for predicting and managing regional electric vehicle charging load, characterized in that, The method includes: Based on road network topology and regional information, a regional traffic road model is constructed; based on travel information, a spatiotemporal model of electric vehicle travel with different levels of complexity is constructed; based on parameters of household electric vehicles and charging facilities, an electric vehicle charging model is established. Based on the relative state of charge of electric vehicles, it is determined whether electric vehicles are charging at nodes. Combined with the regional traffic road model, the vehicle travel spatiotemporal model, and the electric vehicle charging model, spatiotemporal prediction models for charging load are constructed for nodes and regions respectively. The power network is partitioned, and a distributed operation and management model is constructed based on the partitions; The distributed operation and management model is solved based on the charging load prediction results obtained from the spatiotemporal prediction model of the charging load to perform power distribution management.

2. The method according to claim 1, characterized in that, Determining whether a vehicle is charging at a node based on its relative state of charge includes: Based on whether the product of the difference between the relative state of charge and the minimum power threshold and the battery capacity is greater than the power consumption required to reach the destination in the current travel chain, it is predicted whether to charge at the current node. If the product of the difference between the relative state of charge and the minimum power threshold and the battery capacity is greater than the power consumption required from the end of each trip in the current travel chain, it is predicted whether to charge or not at any node between the end of each trip in the current travel chain. If the product of the difference between the relative state of charge and the minimum power threshold and the battery capacity is not greater than the power consumption required to reach the end of each trip in the current travel chain, reduce the last segment of the current travel chain and repeat the steps: predict whether to charge at the current node until the needs of the next trip are met.

3. The method according to claim 1 or 2, characterized in that, Combining the regional traffic road model and the vehicle travel spatiotemporal model, spatiotemporal prediction models for charging load are constructed for nodes and regions respectively, including: Obtain the time weight matrix and distance weight matrix from the regional traffic road model; Based on the total number of traffic area divisions, the number of electric vehicles at the current node, the time weight matrix, and the distance weight matrix, for each area division, the travel chain of each electric vehicle and the corresponding total number of trips and the destination number of each trip segment are selected based on the vehicle travel spatiotemporal model. Based on the travel chain and the corresponding total number of trips, the destination number of each trip segment, and whether the electric vehicle needs charging and the stopping time at the end of each trip segment in its travel chain, a spatiotemporal prediction model for charging load is constructed for nodes and regions respectively, and the electric vehicle state parameters are updated. The spatiotemporal prediction model for charging load for nodes is the predicted sum of the charging load of the vehicles held at the current node, and the spatiotemporal prediction model for charging load for regions is the predicted sum of the charging load of the vehicles held at all nodes in the travel chain.

4. The method according to claim 1, characterized in that, The power grid is divided into zones, including: Each charging node in the power network is treated as a separate sub-region, and a comprehensive index for each sub-region is calculated. Select a sub-region as the first sub-region, select multiple second sub-regions that are different from the first sub-region, calculate the comprehensive index of the sub-regions that are the combination of the first sub-region and each second sub-region, and update the combined sub-regions based on the combination with the largest positive value of the comprehensive index to obtain the updated combined sub-regions. Treat the updated combined sub-region as a separate node and repeat the following steps: select a sub-region as the first sub-region, select multiple second sub-regions that are different from the first sub-region, calculate the comprehensive index of the sub-regions combining the first and second sub-regions, and update the combined sub-regions based on the combination with the largest positive comprehensive index value to obtain the updated combined sub-regions. Stop partitioning when the calculated comprehensive index is no longer positive.

5. The method according to claim 4, characterized in that, Calculating the comprehensive index of the sub-region includes: Based on the road segment delay time weight, active power weight, reactive power weight, and the sub-region to which the node belongs, the active power electrical distance and active power electrical distance between nodes are calculated. Based on the number of sub-regions of the power network, the active power demand and reactive power demand of the sub-regions, the active power balance value and reactive power balance value are calculated using the power balance conditions of the boundary nodes and the power balance conditions of the boundary regions. The computation time of multiplication and addition / subtraction operations in the algorithm is calculated to determine the computational complexity. The comprehensive index is calculated based on the active power electrical distance, the active power electrical distance, the active power balance value, the reactive power balance value, and the computational complexity.

6. The method according to claim 5, characterized in that, Based on the active power electrical distance, the reactive power electrical distance, the active power balance value, the reactive power balance value, and the computational complexity, the comprehensive index is calculated, including: After normalizing the active power electrical distance, the reactive power electrical distance, the active power balance value, the reactive power balance value, and the computational complexity, the comprehensive index is calculated.

7. The method according to claim 6, characterized in that, The comprehensive index is expressed as follows: ; in, Electrical distance representing normalized active power; Electrical distance representing normalized reactive power; Represents the normalized active power balance value; This represents the normalized reactive power balance value. This represents the maximum computational complexity of normalization.

8. The method according to claim 1, characterized in that, The distributed power distribution network operation and management model is as follows: ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; Where T represents the total number of time periods; branch ij represents the positive direction of power flow from node i to node j; E represents the set of all branches in the network; B represents the set of all nodes in the network; r ij This represents the resistance value on branch ij. This represents the current in branch ij during time period t; This represents the reactive power compensation amount of the continuous reactive power compensation device at node j during time period t. This represents the active charging amount of energy storage at node j during time period t; This represents the active power discharge of energy storage at node j during time period t. This represents the lower limit of reactive power compensation for the continuous reactive power compensation device at node j during time period t. This represents the upper limit of reactive power compensation for the continuous reactive power compensation device at node j during time period t. This represents the set of nodes that contain continuous reactive power compensation devices. This represents the reactive power compensation of capacitor bank j at node t during time period t. The number of groups put into operation during time period t is a discrete variable. The compensation power for each CB group; This represents the set of nodes containing the grouped switching capacitor banks; and The 0-1 flags represent the switching operation. If =1, it means that an additional CB is put into operation at node j during time period t. Similarly; This indicates the maximum number of CB (Block Controller) switches that can be made during time period T; This indicates the number of groups put into operation during time period t-1; This represents the upper limit of the number of CB groups that node j can switch to in each iteration; and The 0-1 symbols indicate charging and discharging. =1 indicates that node j is charging during time period t. Similarly; This represents the upper limit of the active power charging capacity of node j. This represents the upper limit of the active power discharge of the energy storage at node j; This represents the set of nodes containing the active power discharge of energy storage. This represents the total energy storage capacity of node j during time period t; This represents the total energy storage capacity of node j during time period t-1; and These represent the charging and discharging efficiencies of node j during time period t, respectively. This represents the lower limit of the total energy storage capacity of node j; This represents the maximum total energy storage capacity of node j; , These represent the active power injection and reactive power injection at node j during time period t, respectively. and This represents the active and reactive power injections at node j in time period t. and This represents the active and reactive loads of node j during time period t. and This represents the net active power exchange and net reactive power exchange of node j after the completion of its local transaction in time period t. This represents the active power sold by entity f; This represents the active power purchased by entity f; Represents the set of regions / nodes of type k; ℓ represents the power factor coefficient; and This represents the active and reactive power flowing from parent node k to current node j during time period t. and This represents the active power and reactive power on branch ij during time period t; This represents the set of upstream parent nodes of the current node j; This represents the set of downstream child nodes of the current node j; , Let represent the conductance and susceptance of node j, respectively; This represents the voltage at node i during time period t; This represents the reactance value on branch ij.

9. A regional electric vehicle charging load prediction and management device, characterized in that, The device includes: At least one processor; and At least one memory coupled to the at least one processor and storing instructions for execution by the at least one processor, the instructions implementing the method of any one of claims 1 to 8 when executed by the at least one processor.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 8.

Images