Multi-scene urban electric vehicle charging station stochastic planning method and system

Abstract

The invention provides a multi-scene urban electric vehicle charging station stochastic planning method and system based on an improved Huff-weighted Voronoi diagram, and the method comprises the steps: simulating the multi-scene travel behaviors of an electric vehicle in different typical days and urban functional areas through employing a Monte Carlo method, and generating charging demand space-time distribution considering the randomness of pile selection; establishing an improved Huff model fusing three types of attraction of power consumption, charging duration and electricity price, calculating a user station selection probability, constructing an improved weighted Voronoi diagram by taking the probability as a weight, and dividing the service range of each charging station; establishing a comprehensive expected cost minimization objective function, and applying a charging station service range constraint, a charging waiting time constraint and a charging pile number constraint; and solving the model by using an improved particle swarm algorithm combining multi-dimensional chaotic mapping and a linear decline strategy, and outputting an optimal charging station position and an optimal charging pile number. According to the invention, the randomness of user behaviors can be effectively described, and the economy and applicability of a charging station planning scheme are improved.

Description

Technical Field

[0001] This invention relates to the field of urban transportation and energy management technology, and more specifically to a multi-scenario stochastic planning method and system for urban electric vehicle charging stations based on an improved Huff-weighted Voronoi diagram. Background Technology

[0002] With increasing global focus on environmental protection and sustainable development, electric vehicles (EVs), as a key technology driving the green transformation of the transportation sector, have experienced rapid development. Governments worldwide have introduced policies to encourage EV adoption, aiming to reduce fossil fuel consumption and greenhouse gas emissions. Against this backdrop, the construction of Electric Vehicle Charging Stations (EVCSs) has become particularly important, as it directly impacts not only the convenience and adoption of EVs but also the overall energy efficiency of a city. However, traditional EVCS site selection and capacity planning methods face the following challenges: First, existing EVCS planning methods often rely on static historical load data for forecasting, neglecting the randomness and volatility of charging demand. This approach assumes that charging demand is constant across different times and locations. However, in reality, EV charging demand is influenced by various factors, such as seasonal variations, day type (weekday / weekend), and the characteristics of different functional areas within a city (commercial areas, residential areas, etc.). Therefore, planning based on static data may lead to insufficient charging infrastructure in certain times and areas, while resulting in excess capacity at other times, leading to resource waste and a poor user experience.

[0003] Secondly, most EVCS planning only considers the physical distance of geographical locations, failing to fully reflect user preferences. For example, users may prioritize charging stations with lower charging costs and higher charging efficiency, even if they are farther away. Traditional methods lack effective capture of users' station selection intentions, thus failing to flexibly reflect the differences in spatial influence among charging stations.

[0004] Furthermore, current EVCS (Electric Vehicle Control System) plans are mostly limited to charging stations with a single power level, failing to adapt to the diverse charging needs of electric vehicle users. Fast charging and slow charging are of great value to different types of users and different usage scenarios. A single power level plan may not meet the needs of all users, especially those who need emergency charging or want to replenish sufficient power within a short stay.

[0005] Existing EVCS planning often focuses on short-term construction and operation cost control, while neglecting long-term economics and market adaptability. With the rapid development of the EV market, charging stations need to be flexible enough to adapt to changes in future charging demand.

[0006] Current EVCS planning has significant limitations, failing to fully consider the randomness of load across multiple scenarios, user preferences for charging station selection, the diversity of charging infrastructure, and long-term economic benefits. Therefore, there is an urgent need to develop more advanced and comprehensive planning methods to improve EVCS deployment efficiency and service quality. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention provides a multi-scenario stochastic planning method and system for urban electric vehicle charging stations based on an improved Huff-weighted Voronoi diagram. This aims to fill the gap in existing technologies and provide a more accurate and practical solution for EVCS planning.

[0008] The first aspect of this invention provides a stochastic planning method for urban electric vehicle charging stations in multiple scenarios based on an improved Huff-weighted Voronoi diagram, comprising: Step 1: Use the Monte Carlo method to simulate the multi-scenario travel behavior of electric vehicles on different typical days and in urban functional areas, generate the spatiotemporal distribution of charging demand considering the randomness of charging station selection, and obtain the simulated spatiotemporal distribution of charging demand. Step Two: Based on the spatiotemporal distribution of charging demand generated in Step One, an improved Huff attraction model is established to calculate the probability of a user selecting a charging station. This probability is then used as the weight of an improved weighted Voronoi diagram to construct an improved Huff-weighted Voronoi diagram, dynamically dividing the service area of ​​each charging station. The probability of a user selecting a charging station is calculated based on the spatiotemporal distribution of charging demand simulated in Step One. The improved Huff attraction model is jointly determined by the attractiveness of power consumption, the attractiveness of charging duration, and the attractiveness of electricity price. The attractiveness of power consumption is based on the energy cost for a user to travel from a charging demand point to a charging station. The attractiveness of charging duration reflects charging efficiency and service waiting time, and the attractiveness of electricity price considers the time-of-use electricity price differences for different power levels. Step 3: With the goal of minimizing the sum of the expected annual charging cost for electric vehicle users, the expected annual construction and maintenance cost of charging stations, and the expected cost of bilateral penalty losses, a stochastic programming model is established to optimize the service range of charging stations determined in Step 2 and the number of charging piles for each power level. Further, the optimized stochastic programming model is constrained by the service range of charging stations, the charging waiting time, and the number of charging piles to obtain the optimized and constrained stochastic programming model. Step 4: The improved particle swarm optimization algorithm is used to solve the stochastic programming model optimized and constrained in Step 3. The improved particle swarm optimization algorithm combines multidimensional chaotic mapping and linear decreasing strategy, and updates the velocity and position of each particle by synthesizing inertial weights, thereby outputting the optimal charging station service range, location and number of charging piles of each power level.

[0009] Furthermore, step one also includes: by conducting multi-dimensional analysis of historical data, obtaining the average driving mileage, average charging time, and average dwell time for each typical day in different urban functional areas, as well as the corresponding distribution parameters, thereby establishing a statistical model of electric vehicle travel behavior in multiple scenarios to more accurately simulate the spatiotemporal distribution of charging demand.

[0010] Furthermore, step one specifically includes: S11: Conduct multi-dimensional analysis of historical data to obtain the average driving mileage, average charging time, and average dwell time for each typical day in different urban functional areas, as well as the corresponding distribution parameters. This will enable the establishment of a statistical model for electric vehicle travel behavior across multiple scenarios, and the acquisition of a set of EV user travel behavior characteristics distributions under various scenarios. S12: Using the multi-scenario travel behavior feature distribution set of EV users obtained in S11, a stochastic model of EV user charging selection behavior described by a probability density function is established; the model is defined based on driving mileage variables, charging start time variables, and dwell time variables; the maximum likelihood estimation method is used to calibrate the distribution parameters, and the estimated distribution parameters are substituted into the probability density function for goodness of fit verification. S13: Utilizing the multi-scenario travel behavior feature distribution obtained in S11 and the charging selection behavior obtained in S12, a battery state-of-charge model is introduced to trigger charging demand, characterizing the charging behavior of EV users. In the spatiotemporal distribution simulation of charging demand, three power levels—slow charging, fast charging, and ultra-fast charging—are defined, and the choice of charging station depends on charging time constraints. A joint probability density function is constructed based on user dwell time, expected charging amount, and the power level of the charging facility to calculate the probability of selecting a charging station, taking into account the remaining battery power of the EV. SOC end The constraints are met, and three scenarios are used for decision-making: in scenario one, the three power levels are randomly selected according to probability; in scenario two, the probability is normalized after excluding slow charging; and in scenario three, ultra-fast charging is forcibly selected. The system simulates the charging process based on the selected scenario to calculate the load, obtain the charging duration, and uses Monte Carlo iteration and convergence to obtain the load fluctuation coefficient.

[0011] Furthermore, in the improved Huff attraction model described in step two, the weighting coefficients of the power consumption attraction, charging time attraction, and electricity price attraction are configured differently according to the urban characteristics of the target city. Specifically, for first- and second-tier cities, the weighting coefficient of the attractiveness of charging time is increased, while the weighting coefficient of the attractiveness of electricity price is decreased accordingly; for third- and fourth-tier cities, the weighting coefficient of the attractiveness of electricity price is increased, while the weighting coefficient of the attractiveness of charging time is decreased accordingly; wherein, the specific values ​​of each weighting coefficient are determined by regression fitting of historical charging behavior samples of the target city.

[0012] It should be noted that since the station selection probability in step two is calculated based on the spatiotemporal distribution of charging demand in step one, and the spatiotemporal distribution of charging demand changes with time and region, the weights (i.e., station selection probabilities) of the improved weighted Voronoi diagram will be dynamically adjusted with demand. The corresponding service area of ​​the charging station is not fixed, but is dynamically updated with the spatiotemporal changes of charging demand.

[0013] Furthermore, step three specifically includes: S31: Construct a stochastic programming model with the goal of minimizing the overall expected cost, optimize the service range of charging stations, and determine the number of charging piles for each power level; The objective function of the stochastic programming model is expressed as: ; in, C AC The comprehensive expected cost, C EV Expected annual charging costs for electric vehicle users C CS The expected annual construction and operation cost of a charging station. C Loss The expected cost of bilateral penalty losses; C EV , C CS , C Loss All calculations are performed using a multi-scenario probabilistic expected value format. S32: Based on the optimized charging station service range and the number of charging piles at each power level in S31, constraints are imposed in conjunction with the charging station service range, charging waiting time, and the number of charging piles. Specifically, the constraints include the following conditions: Charging station service range constraints: based on road curvature coefficient l cur The distance between adjacent charging stations is constrained to not exceed twice the service radius, and 0-1 decision variables are introduced to exclude special geographical locations such as rivers; Charging wait time constraint: The user queuing wait time is calculated based on the M / M / c queuing theory, and the user queuing wait time is constrained not to exceed the user's maximum acceptable wait time; the maximum acceptable wait time is obtained based on the statistics of users' historical charging wait times in the region; Charging pile quantity constraint: The total number of charging piles in a single charging station is constrained between the maximum and minimum number of charging piles set in a certain area node. The maximum and minimum values ​​are dynamically determined based on the total number of EVs in the planning area, the spatial distribution ratio of vehicles, and the peak and valley values ​​of typical daily charging demand.

[0014] Furthermore, step four specifically includes: S41: Set the particle swarm size, maximum number of iterations, and initialize the position and velocity of each particle; the position corresponds to the site selection and capacity setting scheme of the charging station, and the velocity represents the update direction and magnitude of the scheme; S42: Inertia weight composition calculation: In each iteration, dynamic inertial weights are synthesized for each dimension of each particle. The horizontal dimension is updated using a linear decreasing strategy, while the vertical dimension generates chaotic variables through a Logistic mapping method. The dynamic weights are then updated by iterative decay of the chaotic variables. The initial values ​​for the horizontal and vertical dimensions are set to range from 0.4 to 1.2. The horizontal and vertical inertial weights are treated as components of a two-dimensional vector, and the synthesized inertial weights are calculated using the Pythagorean theorem. S43: Using the composite inertial weights obtained in S42, update the velocity and position of each particle according to the standard particle swarm update formula; S44: Repeat steps S42 to S43 until the maximum number of iterations is reached, and output the solution represented by the globally optimal particle as the optimal charging station service range, location and number of charging piles at each power level.

[0015] Preferably, in step four, S42, the Logistic mapping sets the decay radius. r =3 is used to control the chaotic variation amplitude of the longitudinal component of the inertia weight, accelerate the convergence process of the algorithm, improve the optimization accuracy, and ensure that the global optimal solution for the configuration of charging station location and number of charging piles is obtained.

[0016] A second aspect of the present invention provides a multi-scenario stochastic planning system for urban electric vehicle charging stations based on an improved Huff-weighted Voronoi diagram, comprising: The charging demand spatiotemporal distribution generation module uses the Monte Carlo method to simulate the multi-scenario travel behavior of electric vehicles on different typical days and in urban functional areas, and generates a charging demand spatiotemporal distribution that takes into account the randomness of charging station selection, thus obtaining the simulated charging demand spatiotemporal distribution. The initial charging station service area generation module: Based on the charging demand spatiotemporal distribution generated by the charging demand spatiotemporal distribution generation module, an improved Huff attraction model is established to calculate the probability of users selecting a charging station. This probability is then used as the weight of an improved weighted Voronoi diagram to construct an improved Huff-weighted Voronoi diagram, dynamically dividing the service area of ​​each charging station. The improved Huff attraction model is jointly determined by power consumption attraction, charging duration attraction, and electricity price attraction. Power consumption attraction is based on the energy cost for users to travel from the charging demand point to the charging station. Charging duration attraction reflects charging efficiency and service waiting time, while electricity price attraction considers the time-of-use electricity price differences for different power levels. Model optimization and constraint module: With the goal of minimizing the sum of the expected annual charging cost for electric vehicle users, the expected annual construction and operation cost of charging stations, and the expected cost of bilateral penalty losses, a stochastic programming model is established to optimize the charging station service range determined by the initial charging station service range generation module and the number of charging piles for each power level; further, the optimized stochastic programming model is constrained by the charging station service range, charging waiting time, and the number of charging piles to obtain an optimized and constrained stochastic programming model; Further optimization and output module: An improved particle swarm optimization algorithm is used to solve the stochastic programming model that optimizes the model and constraint modules and applies constraints. The improved particle swarm optimization algorithm combines multidimensional chaotic mapping and linear decreasing strategy, and updates the velocity and position of each particle by synthesizing inertial weights, thereby outputting the optimal charging station service range, location and number of charging piles.

[0017] The beneficial effects of this invention are as follows: The method described in this invention combines an improved Huff-weighted Voronoi diagram stochastic programming model with an intelligent optimization algorithm to effectively address the uncertainty of charging demand in various scenarios, achieving refined planning of charging station layout and charging pile configuration. This method not only improves planning accuracy but also significantly enhances charging station utilization and reduces user charging wait times, playing a crucial role in promoting the development of urban electric vehicles and improving the layout of charging infrastructure. In practical applications, this method can help decision-makers make more scientific and rational decisions regarding charging station construction in complex urban environments, effectively promoting the improvement of green transportation systems. Attached Figure Description

[0018] Figure 1 This is a flowchart of the multi-scenario stochastic planning method for urban electric vehicle charging stations based on an improved Huff-weighted Voronoi diagram, as presented in this invention. Figure 2 These are iterative convergence curves of electric vehicle charging station planning models solved using different particle swarm optimization algorithms. Detailed Implementation

[0019] To make the objectives, technical solutions, beneficial effects, and significant advancements of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings provided in the examples of the present invention. Obviously, all the described embodiments are only some embodiments of the present invention, and not all embodiments. 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.

[0020] In the description of this application, unless otherwise expressly specified and limited, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance; the term "multiple" refers to two or more; unless otherwise specified or explained, the terms "connected," "fixed," etc., should be interpreted broadly. For example, "connected" can be a fixed connection, a detachable connection, an integral connection, or an electrical connection; "connected" can be a direct connection or an indirect connection through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0021] like Figure 1 As shown, a stochastic planning method for urban electric vehicle charging stations in multiple scenarios based on an improved Huff-weighted Voronoi diagram includes: Step 1: Use the Monte Carlo method to simulate the multi-scenario travel behavior of electric vehicles on different typical days and in urban functional areas to characterize the spatiotemporal distribution of charging demand considering the randomness of charging station selection, and obtain the simulated spatiotemporal distribution of charging demand.

[0022] S11: Perform multi-dimensional analysis of historical data to obtain the average driving mileage, average charging time, and average dwell time in different urban functional areas for each typical day, as well as the corresponding distribution parameters, thereby establishing a statistical model of electric vehicle travel behavior in multiple scenarios and obtaining the distribution of EV user travel behavior characteristics in multiple scenarios.

[0023] S111: Data Acquisition and Scene Layering Acquire historical electric vehicle (EV) operation data for the city to be predicted, including vehicle identification number (VIN), trip start / end time, average mileage, average charging time, average stay duration, and the functional zones of the city where the trip originates and ends. Establish a hierarchical framework for the historical EV operation data according to time and spatial dimensions, classifying typical day types (weekdays / weekends / holidays) and seasonal characteristics according to time dimension; and label the city's functional zone attributes according to spatial dimension. Specifically, this includes: Historical data is stratified by time dimension, and the sample size of each stratum is weighted according to the proportion of electric vehicle ownership in the corresponding region / time to ensure that the data proportion of each sub-scenario is consistent with the actual operation scenario; distribution parameters are fitted to the data of each stratum separately, and finally integrated into a multi-scenario statistical model to ensure that the data is representative. Historical data is grouped spatially, specifically by "typical day (weekday / weekend / holiday) and urban functional area (residential / commercial / office area, etc.)". For each group, average mileage, average charging time, and average dwell time are calculated, and parameters (mean, variance, etc.) of the corresponding distribution (e.g., log-normal, normal, exponential distribution) are fitted. Further grouping is performed by urban functional area (residential, commercial, office, industrial, etc.) and time dimension (different seasons, different typical days).

[0024] S112: A dual filtering mechanism of "3σ criterion + scene label removal" is used to handle data anomalies. Single-field outlier removal: Apply the 3σ criterion to continuous variables such as mileage and charging time to remove outlier records that deviate from the mean by more than three standard deviations. Special scenario elimination: Combine meteorological data to mark extreme weather dates (i.e. scenario labels), and eliminate abnormal travel / charging behavior data under special scenarios such as extreme weather (such scenarios do not belong to the regular travel scenarios of "typical days"); recalculate statistical parameters for the dataset after removing outliers to ensure that the model only reflects the behavioral patterns of regular scenarios.

[0025] S113: For missing data, a combination of "scenario-specific mean imputation + K-nearest neighbor (KNN) completion" is used: For a small number of missing values ​​in a single field, the mean of the field in the corresponding scenario of "typical day + urban functional area" is used for imputation; for samples with missing values ​​in multiple fields, the KNN algorithm is used to match the complete data of similar samples based on the correlation features such as "travel origin-destination" and "travel time period" for imputation; after imputation, the dataset is checked for consistency to ensure that the imputed data conforms to the distribution pattern of the corresponding scenario, thereby obtaining a set of travel behavior feature distributions of EV users covering time-space scenarios.

[0026] S12: Describe the charging selection behavior of EV users during travel using a probability density function, and simulate the spatiotemporal distribution of charging demand.

[0027] Using the multi-scenario travel behavior feature distribution set obtained from S11, a stochastic model of EV user charging selection behavior described by a probability density function is established. The model is defined based on variables of driving mileage, charging start time, and dwell time.

[0028] Among them, the driving mileage variable is: It is assumed that the driving mileage of a single EV trip follows a log-normal distribution, and its probability density function is shown in Equation (1); (1) Charging start time variable: the time when the EV is connected to the grid for charging t tr It satisfies a normal distribution, as shown in equation (2); (2) Duration of stay variable: The duration of stay of EV follows an exponential distribution, and its probability density function is shown in equation (3); (3) In the formula: d k Indicates the first k The mileage driven by the EV during this trip. k Take a positive integer; The logarithm of the driving distance ln d k The mean; The logarithm of the driving distance ln d k Standard deviation; Expectations for the charging time before grid connection; The standard deviation of the charging time for grid connection; Δ T stay,L The dwell time of EVs in different urban functional areas; the mean and standard deviation are both . The index is Subscript S , D and L These represent different seasons, days, and land types.

[0029] To determine the specific distribution parameter values ​​in the above probability density function ( , , , , This involves collecting, classifying, and statistically analyzing historical data, then using maximum likelihood estimation to determine the mean and variance of the corresponding distribution of the variable. Specifically, this includes: Sample and distribution matching: For historical data samples of driving mileage (taking its natural logarithm), charging time, and dwell time, likelihood functions in the form of probability density functions of each variable are constructed according to logarithmic normal distribution, normal distribution, and exponential distribution respectively. To simplify the calculation, the natural logarithm of the likelihood function is taken. Then, partial derivatives are taken with respect to the parameters to be estimated for the distribution (such as the mean and standard deviation of log-normal mileage, the mean and standard deviation of charging time in a normal distribution, and the characteristic parameters of dwell time in an exponential distribution). The partial derivatives are set to 0, and the estimated values ​​of the parameters are obtained by solving the likelihood equation. The estimated values ​​of the parameters for the normal distribution are the sample mean and the sample standard deviation; the estimated values ​​of the parameters for the exponential distribution are the reciprocal of the sample mean. Then the estimated parameters , , , , Substitute the probability density functions of their respective distributions and verify the fit with the statistical characteristics of historical samples to ensure that the parameters can accurately represent the distribution patterns of the corresponding variables.

[0030] S13: Based on the parameterized probabilistic model and calibrated distribution parameters established in S12, Monte Carlo simulation is used to obtain EV travel behavior under different typical days and land types, and to obtain the spatiotemporal distribution of charging demand at different charging power levels. S131: By coupling the travel behavior obtained in S11 under multiple scenarios and the charging selection behavior obtained in S12, a battery state of charge (SOC) model is introduced to accurately trigger charging demand and characterize the charging behavior of EV users.

[0031] Calculate the remaining battery power of the EV after each trip. SOC end for: (4) In the formula: SOC init This is the initial charge level of the battery; k end For the number of trips; e S This refers to the power consumption per unit distance, which is affected by weather temperature and varies in different seasons; among which... SOC end It is related to the vehicle's mileage and power consumption per unit distance.

[0032] When the EV has remaining battery power SOC end Below the preset state of charge threshold SOC thr If the EV needs charging, the user will proceed with a decision-making process to select a charging station with multiple power levels.

[0033] S132: Probabilistic Model for Selecting Charging Facilities with Multiple Power Levels In the charging selection decision of multi-power level charging facilities, three power level charging settings are set: slow charging facilities (SCF, hereinafter referred to as slow charging), fast charging facilities (FCF, hereinafter referred to as fast charging), and ultra-fast charging facilities (UCF, hereinafter referred to as ultra-fast charging).

[0034] Charging time constraints: The core constraint for selecting charging stations with different power levels is to meet the EV's desired charge level within the specified dwell time. SOC exp Specifically, as shown in equations (5)-(7).

[0035] (5) (6) (7) In the formula: , and These represent the rated charging power of slow charging facilities, fast charging facilities, and super-fast charging facilities, respectively. SOC exp To charge to the desired level; t S For the season S The charging efficiency is as follows.

[0036] Assume Δ T stay,L Δ represents the dwell time of EVs in different urban functional areas. T stay,L and the k EV mileage during this trip d k These two variables are independent of each other. Combined with constraints (5)-(6), a joint probability density function is constructed based on user dwell time, expected charging amount, and the power level of the charging facility to quantify the probability of selecting a charging station. For the numbered... n The probability of selecting different charging power levels for the EV is shown in equations (8)-(10).

[0037] (8) (9) (10) In the formula: , and They are numbered as followsn The probability of choosing slow charging, fast charging, and super-fast charging for EVs; Let be the probability density function of the dwell time in formula (3); f ( d k ) is the probability density function of EV driving mileage in formula (1), and for k end The summation is performed on the second trip; L j For urban functional areas j The weighting coefficients can be assigned according to the different functional areas of the city.

[0038] S133: Based on the remaining battery power of the EV SOC end The constraints are satisfied, and there are three scenarios for decision-making: Scenario 1: When the remaining power satisfies equation (5), it means that slow charging can meet the charging needs. All three power levels can be selected. The selection probability is calculated according to equations (8)-(10), and SCF, FCF and UCF charging piles are randomly selected for charging.

[0039] Scenario 2: When the remaining battery power does not satisfy equation (5), it means that slow charging cannot meet the charging needs of EV users. In this case, the probability of choosing slow charging is 0, and only equation (9) needs to be adjusted to reduce the probability of choosing fast charging. The probability of choosing ultra-fast charging is calculated according to equation (10). Scenario 3: When the remaining battery power does not satisfy equations (5) and (6), that is, neither slow charging nor fast charging can meet the charging requirements, then the probability of choosing slow charging or fast charging is 0, while the probability of choosing super-fast charging is... .

[0040] S134: Charging process simulation and load calculation If the EV is destined for a certain location for charging, then it is assumed that it will connect to the charging facility the moment it arrives at the location and continue charging until it reaches the desired charge level. SOC exp Or until the maximum stay time. F This indicates different power levels, namely: F ∈{SCF, PCF, UCF}, then the charging time T ch As shown in equation (11).

[0041] (11) Combined with the timing of EV charging upon grid connection t gt With charging time Tch It can calculate the charging end time, which makes it easy to accumulate the charging demand of vehicles at different times with multiple power levels, i.e., the probabilistic load under different scenarios.

[0042] The expected load is calculated as necessary load data for subsequent planning. Assume the number of EVs in the area is... n max Introducing load fluctuation coefficient c it To determine the number of iterations N it Whether it is reasonable is described by the variance of the load forecast values ​​at various time points. The higher the number of iterations, the lower the volatility coefficient, and the more stable the composite curve. If the volatility coefficient is less than the mean... c ave Once the system is stable, the Monte Carlo simulation can be stopped.

[0043] Step Two: Establish an improved Huff attraction model to calculate the probability of users selecting a charging station. Use this probability as the weight of each charging station to construct an improved Huff-weighted Voronoi diagram-based stochastic programming model for urban electric vehicle charging stations across multiple scenarios, dynamically dividing the service area of ​​each charging station. Optimize the location, service area, and number of charging piles at each power level of the charging station, using the comprehensive expected cost (including the annual expected charging cost for EV users, the annual expected construction and maintenance cost of the charging station, and the expected cost of bilateral penalty losses) as the objective. The calculation of the user selection probability is based on the spatiotemporal distribution of charging demand simulated in Step One. Specifically, this includes: S21: Calculate user station selection probability based on improved Huff attraction model The probability of a user choosing a station is calculated using the improved Huff attraction, and the probability of choosing a station is then used. p n.j As weights for improving the weighted Voronoi diagram, the charging stations are determined. a j The service scope is as shown in equation (12).

[0044] (12) In the formula: V ( a j ,p j ) is a charging station a j V-polygon with vertices; p n.j For users n Choose a charging station a j The probability of charging is also the corresponding weight in the improved weighted Voronoi equation, which can be specifically determined by equation (13-14); xIt is any point in the V-polygon graph; d ( x , a j ) represents a point on the plane. x to charging station a j The Euclidean distance between them; N v It represents the number of charging stations on the plane.

[0045] User site selection probability p n.j Calculated by the following formula: (13) (14) In the formula: S j For charging stations a j The utility of this utility is determined by the power consumption and attractiveness. w e,ij Charging time appeal w t,ij and electricity price attractiveness w r,ij The decision is made jointly, as shown in equation (15); For EV users n From the point of charging demand ( x i , y i Head to the charging station a j The normalized power consumption is obtained by normalizing the spatiotemporal distribution of charging demand simulated in step one. For EV users in this area n At the charging station a j Normalized time; This refers to the normalized time-of-use electricity price for different power levels within the charging station.

[0046] Among them, power consumption attraction w e,ij Charging time appeal w t,ij and electricity price attractiveness w r,ij The calculation formula is: (15) In the formula: c and f These are the weighting coefficients for power consumption and the number of charging stations, respectively. x d , x e and x l These are the distance to the station, charging efficiency, and the weighting coefficient of urban functional areas, respectively. f e , f t and f r These are the attraction constants corresponding to each factor; For the season s Normalized unit power consumption; For the season s Normalized charging efficiency; To start from the demand point i to charging station j Normalized distance; j For charging stations j The normalized number of charging piles is determined by step three, S31. Therefore, in this step... These are the variables to be optimized (specific values ​​are not yet provided). Power consumption attractiveness. w e,ij Charging time appeal w t,ij and electricity price attractiveness w r,ij In the utility of charging stations S j The weighting percentage in the formula has been achieved through the "normalization process" in equation (14).

[0047] In this model, the weighting of power consumption, charging time, and electricity price attractiveness needs to be adjusted according to city characteristics: for first- and second-tier cities, users are more sensitive to charging time and less sensitive to electricity price, so the weighting coefficient of charging time attractiveness needs to be increased. φt Reduce the weighting factor for the attractiveness of electricity prices φc For third- and fourth-tier cities, users are more sensitive to electricity prices and have a higher tolerance for charging time, so the weighting coefficient for the attractiveness of electricity prices needs to be increased. φc Reduce the weighting factor of charging time as an attractive factor φt The specific values ​​of the above weighting coefficients need to be determined by fitting historical charging behavior samples of the target city.

[0048] Step 3: With the objective of minimizing the sum of the expected annual charging cost for electric vehicle users, the expected annual construction and maintenance cost of charging stations, and the expected cost of bilateral penalty losses, a stochastic programming model is established to optimize the charging station service area determined in Step 2 and the number of charging piles at each power level. Further, constraints on the charging station service area, charging waiting time, and the number of charging piles are imposed on the optimization results to obtain the optimized and constrained stochastic programming model. Specifically, this includes: S31: Based on the goal of minimizing the comprehensive expected cost of electric vehicle users' annual expected charging cost, the annual expected construction and operation cost of charging stations, and the expected cost of bilateral penalty losses, optimize the service range of charging stations determined in step two and determine the number of charging piles for each power level.

[0049] Based on a regional segmentation strategy, the expected cost is the comprehensive expected cost calculated using the annual expected charging cost for EV users, the annual expected construction and maintenance cost of charging stations, and the expected cost of bilateral penalty losses. C AC With the goal of minimizing, an EVCS stochastic programming model is established to optimize the charging location coordinates and F Number of charging stations of different power levels That is, to solve the site selection and capacity determination problems together, as shown in equation (16).

[0050] (16) In the formula: C EV The expected annual charging cost for EV users is shown in equations (17)-(19); C CS The expected annual construction and maintenance costs for EVCS are shown in equations (20)-(23); C Loss The expected cost of the bilateral expected penalty loss is shown in equations (24)-(26); (17) (18) (19) (20) (twenty one) (twenty two) (twenty three) (twenty four) (25) (26) In the formula:p Ω For each scenario, the probability of combinations of seasons, holidays, and urban functional areas throughout the year is considered. Oh Different values ​​represent different scenarios (i.e., weekdays and weekends in four seasons across three types of urban functional zones, totaling...). Oh tot =24 scenarios); T year The number of days in a year; , EV users n exist Oh Scene: Heading to the charging station j The time and energy costs of charging; t q The charging wait time for EV users can be determined by equation (29); Δ T ij Arrival of the EV at the charging station j Time required; for F The unit price of charging corresponding to the power level; C IC and C OM These represent the annual investment and construction costs and the annual operation and maintenance costs of the EVCS, respectively. , and These represent the unit investment construction costs for slow charging piles, fast charging piles, and super-fast charging piles, respectively. They are directly set based on the industry quotas and publicly available cost data for charging pile construction in the target city (approximately RMB 10,000 for a single slow charging pile, RMB 65,000 for a fast charging pile, and RMB 250,000 for a super-fast charging pile; the specific values ​​can be dynamically adjusted based on the building materials and equipment procurement prices in the target city). , and These represent the nodes respectively. j The number of slow, fast, and super-fast charging piles installed in the charging stations is the optimization variable of this stochastic programming method; , and These represent the annual operation and maintenance costs per unit quantity of slow charging piles, fast charging piles, and super-fast charging piles, respectively. These costs are set based on industry standards and actual operating data for charging pile operation and maintenance in the target city (the annual operation and maintenance cost per unit for a single slow charging pile is approximately 0.1 million yuan, for a fast charging pile approximately 0.5 million yuan, and for a super-fast charging pile approximately 1.2 million yuan. The specific values ​​can be dynamically adjusted based on equipment depreciation and labor maintenance costs). ψ It is an auxiliary variable, as shown in equation (23); iThe discount rate is based on the benchmark discount rate for infrastructure projects in the target city (usually published by the local development and reform commission and finance department), combined with the investment risk level of the charging station project (e.g., the industry average risk is taken as 8%~10%). for F Economic lifespan of charging stations based on power rating; It is the penalty cost for unmet capacity expectations; It is the expected penalty cost for capacity redundancy; β u and β e These represent the penalty coefficients for unsatisfactory demand and excessive capacity, respectively. For the scene Oh exist t Charging stations during the time period j The charging demand can be obtained from the selected station location and the charging demand in step one, according to the corresponding area division in step two. For the scene Oh exist t Time-of-use charging stations a j The capacity can be determined based on the number of charging piles and their corresponding power levels; d t This represents the reasonable redundancy coefficient of the capacity, which is set to 1.2 here. This value is determined by combining the conventional configuration of charging station capacity redundancy in the industry (set to 1.1~1.3) and the fluctuation data of peak charging demand in the target city over the past 3 years. In actual operation, it can be dynamically adjusted according to changes in demand.

[0051] S32: Based on the optimized charging station service range and the number of charging piles of each power level determined in S31, constraints are imposed in combination with geospatial factors (constraining the charging station service range based on road curvature), user experience (limiting charging waiting time based on queuing theory), and resource allocation (determining the number of multi-power charging piles based on multi-scenario charging needs).

[0052] 1) Service range constraints of charging stations To prevent users from having to travel excessive distances to charge their devices and to ensure that the charging range covers the entire planned area, the distance between two adjacent charging stations shall not exceed twice the service radius of the charging station.

[0053] (27) In the formula: D ab For charging stations a to charging station b distance, a , b ∈ j ; R fThe service area of ​​the charging station; l cur The curvature coefficient of the road is the ratio of the actual distance between two points to the Euclidean distance, constraining the distance between adjacent charging stations to not exceed twice the service radius.

[0054] In addition, charging stations cannot be located in special locations such as rivers, so Boolean variables, i.e., 0-1 variables, are introduced as auxiliary constraints. z j ( z j =1 indicates that the selected site location is selected. j Combined with special position indication functions d ( x j , y j ()( d =0 indicates a special position, then it exists: (28) That is, when the station location j Not located in a special location ( d The site will only be selected if the value is 1; otherwise, the result will be 0, which violates the constraint and means the site will not be selected.

[0055] 2) Charging wait time constraint Assuming that the waiting time of EV users at centralized charging stations can be determined by the M / M / c queuing theory, and that the arrival process is determined by the Kolmogorov-Smirnov test (KS test) to be a Poisson distribution, then the waiting time of users is as shown in equations (29)-(31).

[0056] (29) (30) (31) In the formula: t q This indicates the waiting time for EV users. U j For when the vehicle goes to a centralized charging station j The value is 1 when charging is in progress, and 0 otherwise. P 0 indicates no EV has arrived at the charging station. j The probability of; for F Power level charging pile service intensity; meets ; l EV The number of EVs that visit a charging station for service per hour; m nThe number of vehicles that can be charged per charging device in 1 hour; h For index value; t q,max This indicates the maximum acceptable waiting time for EV users to charge; the maximum acceptable waiting time is obtained based on the statistical analysis of historical charging waiting times of users within the region.

[0057] 3) Constraints on the number of charging stations with different power levels: The maximum and minimum number of charging piles at a single charging station are related to the proportion of vehicles in the planned area, the overall charging demand, and the multiple power levels of the charging facilities, as shown in equations (32)-(33).

[0058] (32) (33) Where: min( D Ω,t,j ) and max( D Ω,t,j ) is a typical day Oh Inside t Minimum and maximum charging demand during different time periods; N EV This represents the total number of EVs within the planned area. Nj This represents the total number of charging stations with different power ratings. N j,max and N j,min They are in j The maximum and minimum number of charging piles set up at regional nodes; g j It is an EV vehicle. j The proportion of regional nodes; ceil() is the floor function.

[0059] Step 4: Solve the optimization problem using a particle swarm optimization algorithm that combines multidimensional chaotic mapping and a linear decreasing strategy to determine the optimal configuration of the charging station's service range, location, and number of charging piles. The optimal configuration is determined based on the further optimized charging station service range, location, and number of charging piles at each power level obtained in Step 3. Specifically, this includes: The logistic mapping is a common chaotic mapping that generates chaotic variables through a specific functional relationship, causing each particle to exhibit chaotic changes in all dimensions within its current decay radius, rather than a single linear change. (Worth decay radius) r The determination method is as follows: combining the inertial weight distribution characteristics of the particle swarm in the early stage of the search, referring to the conventional value of the decay radius in similar optimization problems (usually 2-5), and verifying it through preliminary experiments, when rWhen the inertia weight is 3, the decay rate of the inertia weight can ensure the diversity of the early search while avoiding slow convergence in the later stage. Therefore, it is set to... r =3; in practical applications, it can be fine-tuned according to the search requirements of the solution space.

[0060] The improved particle swarm optimization algorithm used in this step includes a dynamic inertia weight adjustment mechanism and a multidimensional chaotic mapping strategy. By iteratively decaying chaotic variables, it significantly accelerates the convergence process of the algorithm, improves the optimization accuracy, and ensures that the global optimal solution for the service range, location, and number of charging piles of charging stations is obtained.

[0061] To ensure that the inertial weight of each particle in each dimension exhibits its own distinct chaos within the current decay radius, a linear decreasing strategy is adopted to update the horizontal dimension (global search capability). oh h Specifically, as shown in equation (34); vertically (local search capability), chaotic variables are generated through Logistic mapping, and then the dynamics are updated through iterative decay of chaotic variables. oh z,new Specifically, as shown in equation (35). The horizontal and vertical inertial weights are considered as components of a two-dimensional vector, and the composite inertial weights are calculated using the Pythagorean theorem. oh sum Specifically, as shown in equation (36).

[0062] (34) (35) In the formula: oh h This is the set initial lateral value; the maximum lateral inertia weight. oh max yes oh h The maximum value is set to 1.2 (referring to the common range of 0.4-1.2 for inertia weight in particle swarm optimization, and considering the search accuracy requirements of this problem, the upper limit is selected to enhance the early global search capability). Iter This is the current iteration number; Iter max This is the maximum number of iterations, set based on the convergence of the objective function after 500 iterations in the preliminary experiment, to ensure that the algorithm achieves a balance between accuracy and efficiency. oh z This is the initial value set for the vertical direction, which is 0.9 (to preserve strong inertia in the early stages of the algorithm and avoid premature convergence to a local optimum). r t This is the decay radius of the particle swarm's inertial weight in the early stages of the search.

[0063] (36) To verify the effectiveness and superiority of the improved particle swarm optimization algorithm proposed in this invention, this embodiment employs the traditional particle swarm optimization algorithm (traditional PSO), the linear decreasing particle swarm optimization algorithm (linear decreasing PSO), the chaotic particle swarm optimization algorithm (chaotic PSO), and the multi-scenario urban electric vehicle charging station stochastic planning method based on the improved Huff-weighted Voronoi diagram described in this invention, respectively, on the same electric vehicle charging station (EVCS) stochastic programming model. Figure 2 The four algorithms were compared and tested using a multidimensional chaotic PSO model. Model parameters and scene settings were kept consistent. Figure 2 As shown by the curve trend, the stochastic programming method described in this invention tends to stabilize and converge to the global optimum (target cost approximately 2.95 × 10⁻⁶) after approximately 60 iterations. Compared to the traditional PSO (which converges in 50 iterations but suffers from local oscillations, with a target cost of approximately 3.1 × 10⁻⁶), this approach is significantly more efficient. Linearly decreasing PSO (converged in 40 iterations, target cost approximately 3.0 × 10⁻⁶) Chaotic PSO (converged in 35 iterations, target cost approximately 3.05 × 10⁻⁶) Although the stochastic programming method described in this invention requires slightly more convergence iterations, it achieves this by fusing multidimensional chaotic mapping with a linear decreasing strategy, employing a dynamic... oh The proposed improved algorithm significantly enhances the optimization effect. The dual-dimensional design avoids the problem of insufficient search capability caused by the single inertial weight in the traditional PSO algorithm and effectively avoids the local optimum problem. The final target cost is reduced by about 5% compared with the traditional PSO algorithm, and there is no significant fluctuation after convergence. It not only ensures the global optimality of the solution result, but also improves the stability of the planning scheme. This verifies the effectiveness and superiority of the proposed improved algorithm in solving the EVCS planning model.

[0064] The stochastic programming method described in this invention ensures the algorithm's ability for refined local search in the later stages through a horizontal linear decreasing strategy, while simultaneously enhancing the diversity of global exploration in the early stages by utilizing a vertical multidimensional chaotic mapping with dynamic perturbation within the decay radius (r=3). This "dual-dimensional" dynamic inertial weight design effectively overcomes the imbalance between exploration and development capabilities caused by the single weight update strategy in traditional algorithms. Thus, with a slight increase in the number of iterations, it significantly avoids the local optimum trap and finds a better charging station site selection and capacity determination scheme.

[0065] The stochastic programming method described in this invention is applicable to the planning of various types of charging stations, including centralized and distributed charging stations. For centralized charging stations, its large-scale, high-load characteristics can be directly matched through "spatial partitioning optimization + capacity configuration." For distributed charging stations, only the parameters of the "service range constraint" (such as reducing the service radius threshold) need to be adjusted to adapt to their small-capacity, distributed layout requirements. Adaptability: The "multi-scenario preference-oriented + multi-power-level configuration" logic of the stochastic programming method described in this invention is compatible with the user needs and capacity characteristics of different types of charging stations, possessing universality.

[0066] It is obvious to those skilled in the art that the modules or steps of the present invention described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. They can be implemented using computer-executable program code, and thus can be stored in a storage device for execution by a computing device. In some cases, the steps shown or described can be performed in a different order than those described herein, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

[0067] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style of the specification is merely for clarity. Those skilled in the art should regard the specification as a whole, and the technical solutions in the embodiments can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A stochastic planning method for urban electric vehicle charging stations in multiple scenarios based on an improved Huff-weighted Voronoi diagram, characterized in that, include: Step 1: Use the Monte Carlo method to simulate the multi-scenario travel behavior of electric vehicles on different typical days and in urban functional areas, generate the spatiotemporal distribution of charging demand considering the randomness of charging station selection, and obtain the simulated spatiotemporal distribution of charging demand. Step Two: Based on the spatiotemporal distribution of charging demand generated in Step One, an improved Huff attraction model is established to calculate the probability of a user selecting a charging station. This probability is then used as the weight of an improved weighted Voronoi diagram to construct an improved Huff-weighted Voronoi diagram, dynamically dividing the service area of ​​each charging station. The probability of a user selecting a charging station is calculated based on the spatiotemporal distribution of charging demand simulated in Step One. The improved Huff attraction model is jointly determined by the attractiveness of power consumption, the attractiveness of charging duration, and the attractiveness of electricity price. The attractiveness of power consumption is based on the energy cost for a user to travel from a charging demand point to a charging station. The attractiveness of charging duration reflects charging efficiency and service waiting time, and the attractiveness of electricity price considers the time-of-use electricity price differences for different power levels. Step 3: With the goal of minimizing the sum of the expected annual charging cost for electric vehicle users, the expected annual construction and maintenance cost of charging stations, and the expected cost of bilateral penalty losses, a stochastic programming model is established to optimize the service range of charging stations determined in Step 2 and the number of charging piles for each power level. Further, the optimized stochastic programming model is constrained by the service range of charging stations, the charging waiting time, and the number of charging piles to obtain the optimized and constrained stochastic programming model. Step 4: The improved particle swarm optimization algorithm is used to solve the stochastic programming model optimized and constrained in Step 3. The improved particle swarm optimization algorithm combines multidimensional chaotic mapping and linear decreasing strategy, and updates the velocity and position of each particle by synthesizing inertial weights, thereby outputting the optimal charging station service range, location and number of charging piles of each power level.

2. The stochastic planning method for multi-scenario urban electric vehicle charging stations according to claim 1, characterized in that, Step one specifically includes: S11: Conduct multi-dimensional analysis of historical data to obtain the average driving mileage, average charging time, and average dwell time for each typical day in different urban functional areas, as well as the corresponding distribution parameters. This will enable the establishment of a statistical model for electric vehicle travel behavior across multiple scenarios, and the acquisition of a set of EV user travel behavior characteristics distributions under various scenarios. S12: Using the multi-scenario travel behavior feature distribution set of EV users obtained in S11, a stochastic model of EV user charging selection behavior described by a probability density function is established; the model is defined based on driving mileage variables, charging start time variables, and dwell time variables; the maximum likelihood estimation method is used to calibrate the distribution parameters, and the estimated distribution parameters are substituted into the probability density function for goodness of fit verification. S13: Utilizing the multi-scenario travel behavior feature distribution obtained in S11 and the charging selection behavior obtained in S12, a battery state-of-charge model is introduced to trigger charging demand, characterizing the charging behavior of EV users. In the spatiotemporal distribution simulation of charging demand, three power levels—slow charging, fast charging, and ultra-fast charging—are defined, and the choice of charging station depends on charging time constraints. A joint probability density function is constructed based on user dwell time, expected charging amount, and the power level of the charging facility to calculate the probability of selecting a charging station, taking into account the remaining battery power of the EV. SOC end The constraints are met, and three scenarios are used for decision-making: in scenario one, the three power levels are randomly selected according to probability; in scenario two, the probability is normalized after excluding slow charging; and in scenario three, ultra-fast charging is forcibly selected. The system simulates the charging process based on the selected scenario to calculate the load, obtain the charging duration, and uses Monte Carlo iteration and convergence to obtain the load fluctuation coefficient.

3. The stochastic planning method for multi-scenario urban electric vehicle charging stations according to claim 1, characterized in that, In the improved Huff attraction model described in step two, the weighting coefficients of the power consumption attraction, charging time attraction, and electricity price attraction are configured differently according to the urban characteristics of the target city. For first- and second-tier cities, the weighting coefficient of the attractiveness of charging time is increased, and the weighting coefficient of the attractiveness of electricity price is decreased accordingly; for third- and fourth-tier cities, the weighting coefficient of the attractiveness of electricity price is increased, and the weighting coefficient of the attractiveness of charging time is decreased accordingly; wherein, the specific values ​​of each weighting coefficient are determined by regression fitting of historical charging behavior samples of the target city.

4. The stochastic planning method for multi-scenario urban electric vehicle charging stations according to claim 1, characterized in that, Step three includes: S31: Construct a stochastic programming model with the goal of minimizing the overall expected cost, optimize the service range of charging stations, and determine the number of charging piles for each power level; The objective function of the stochastic programming model is expressed as: ； in, C AC The comprehensive expected cost, C EV Expected annual charging costs for electric vehicle users C CS The expected annual construction and operation cost of a charging station. C Loss The expected cost of bilateral penalty losses; C EV , C CS , C Loss All calculations are performed using a multi-scenario probabilistic expected value format. S32: Based on the optimized charging station service range and the number of charging piles at each power level in S31, constraints are imposed in conjunction with the charging station service range, charging waiting time, and the number of charging piles. Specifically, the constraints include the following conditions: Charging station service range constraints: based on road curvature coefficient λ cur The distance between adjacent charging stations is constrained to not exceed twice the service radius, and 0-1 decision variables are introduced to exclude the location of rivers; Charging wait time constraint: The user queuing wait time is calculated based on the M / M / c queuing theory, and the user queuing wait time is constrained not to exceed the user's maximum acceptable wait time; the maximum acceptable wait time is obtained based on the statistics of users' historical charging wait times in the region; Charging pile quantity constraint: The total number of charging piles in a single charging station is constrained between the maximum and minimum number of charging piles set in a certain area node. The maximum and minimum values ​​are dynamically determined based on the total number of EVs in the planning area, the spatial distribution ratio of vehicles, and the peak and valley values ​​of typical daily charging demand.

5. The stochastic planning method for multi-scenario urban electric vehicle charging stations according to claim 1, characterized in that, Step four includes: S41: Set the particle swarm size, maximum number of iterations, and initialize the position and velocity of each particle; the position corresponds to the site selection and capacity setting scheme of the charging station, and the velocity represents the update direction and magnitude of the scheme; S42: Inertia weight composition calculation: In each iteration, dynamic inertial weights are synthesized for each dimension of each particle. The horizontal dimension is updated using a linear decreasing strategy, while the vertical dimension generates chaotic variables through a Logistic mapping method. The dynamic weights are then updated by iterative decay of the chaotic variables. The initial values ​​for the horizontal and vertical dimensions are set to range from 0.4 to 1.

2. The horizontal and vertical inertial weights are treated as components of a two-dimensional vector, and the synthesized inertial weights are calculated using the Pythagorean theorem. S43: Using the composite inertial weights obtained in S42, update the velocity and position of each particle according to the standard particle swarm update formula; S44: Repeat steps S42 to S43 until the maximum number of iterations is reached, and output the solution represented by the globally optimal particle as the optimal charging station service range, location and number of charging piles at each power level.

6. The stochastic planning method for multi-scenario urban electric vehicle charging stations according to claim 5, characterized in that, In step S42 of the fourth step, the Logistic mapping sets the attenuation radius. r =3 to control the chaotic variation amplitude of the longitudinal component of the inertial weight.

7. A multi-scenario stochastic planning system for urban electric vehicle charging stations based on an improved Huff-weighted Voronoi diagram, characterized in that, include: The charging demand spatiotemporal distribution generation module uses the Monte Carlo method to simulate the multi-scenario travel behavior of electric vehicles on different typical days and in urban functional areas, and generates a charging demand spatiotemporal distribution that takes into account the randomness of charging station selection, thus obtaining the simulated charging demand spatiotemporal distribution. The initial charging station service area generation module: Based on the charging demand spatiotemporal distribution generated by the charging demand spatiotemporal distribution generation module, an improved Huff attraction model is established to calculate the probability of users selecting a charging station. This probability is then used as the weight of an improved weighted Voronoi diagram to construct an improved Huff-weighted Voronoi diagram, dynamically dividing the service area of ​​each charging station. The improved Huff attraction model is jointly determined by power consumption attraction, charging duration attraction, and electricity price attraction. Power consumption attraction is based on the energy cost for users to travel from the charging demand point to the charging station. Charging duration attraction reflects charging efficiency and service waiting time, while electricity price attraction considers the time-of-use electricity price differences for different power levels. Model optimization and constraint module: With the goal of minimizing the sum of the expected annual charging cost for electric vehicle users, the expected annual construction and operation cost of charging stations, and the expected cost of bilateral penalty losses, a stochastic programming model is established to optimize the charging station service range determined by the initial charging station service range generation module and the number of charging piles for each power level; further, the optimized stochastic programming model is constrained by the charging station service range, charging waiting time, and the number of charging piles to obtain an optimized and constrained stochastic programming model; Further optimization and output module: The improved particle swarm optimization algorithm is used to solve the stochastic programming model that optimizes the model and constraint modules and applies constraints. The improved particle swarm optimization algorithm combines multidimensional chaotic mapping and linear decreasing strategy, and updates the velocity and position of each particle by synthesizing inertial weights, thereby outputting the optimal charging station service range, location and number of charging piles.

Images