Charging load probability spatiotemporal distribution prediction method considering random fluctuations of commuting demand

Abstract

This application belongs to the field of power system management, specifically disclosing a method for predicting the spatiotemporal distribution of charging load probabilistically, taking into account the random fluctuations in commuting demand. This application constructs a multi-dimensional travel strategy set to simulate users' comprehensive decisions regarding energy, time, and space. Uncertainty is introduced from the commuting demand side, and Monte Carlo sampling is used to generate demand fluctuation samples, fundamentally characterizing the random fluctuations in load. A utility evaluation model is constructed based on cumulative prospect theory to correct for users' bounded rationality characteristics. A multi-layered nested discrete choice model is adopted, decomposing the decision into three levels: charging, departure, and route, and iterating to traffic network equilibrium using a heuristic algorithm. In the equilibrium state, the strategy selection ratio is combined with the demand sample, input into the dynamic traffic network model, and the spatiotemporal probability distribution of charging load is output. This achieves more accurate prediction of the spatiotemporal distribution of electric vehicle charging load probabilistically during evening peak hours, providing a scientific basis for charging facility planning and grid dispatching.

Description

Technical Field

[0001] This application belongs to the field of power system management, and specifically relates to a method for predicting the spatiotemporal distribution of charging load probabilistically considering the random fluctuations in commuting demand. Background Technology

[0002] With the rapid increase in the number of electric vehicles, the electrification and decarbonization transformation of the transportation sector is accelerating, providing an important carrier and flexible resource for the low-voltage side of the power distribution network to absorb massive distributed renewable energy. Electric vehicles can not only reduce dependence on traditional fossil fuels, but also help smooth out fluctuations in intermittent renewable energy output and improve the overall regulation capacity and operating efficiency of the power grid through orderly charging and even discharging into the grid.

[0003] However, the charging behavior of electric vehicle users exhibits significant uncertainty and randomness. Their charging demand tends to concentrate during peak load periods such as evening and night, and spatially tends towards densely populated areas such as residential, workplace, and commercial districts. If left unguided, this spatiotemporal concentration can exacerbate local power grid load peaks, leading to problems such as line overload and voltage exceeding limits. Meanwhile, as a controllable load resource, electric vehicles inherently possess a certain degree of temporal shifting and spatial transfer potential, but currently, these flexibilityes have not been fully explored and utilized due to limitations in charging convenience, user habits, and electricity pricing mechanisms.

[0004] Therefore, in-depth research into the key factors influencing electric vehicle users' charging decisions is of significant practical importance. By clarifying the spatiotemporal evolution and distribution mechanisms of charging load under the coupled effects of multiple factors, more accurate load forecasting models can be constructed, and matching incentive strategies, electricity pricing mechanisms, and coordinated control methods can be designed. This will not only provide a solid theoretical basis for the proactive regulation and orderly guidance of charging load, but also help the distribution network achieve safe, reliable, economical, efficient, green, and low-carbon operation goals in the context of large-scale electric vehicle access, ultimately promoting the coordinated optimization and sustainable development of the energy and transportation system. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this application aims to provide a method for predicting the probabilistic spatiotemporal distribution of charging load that takes into account the random fluctuations in commuting demand. This method is intended to solve the problem that the current charging load is difficult to predict accurately due to the randomness and uncertainty of electric vehicle users' charging behavior, resulting in poor grid security and reliability.

[0006] The first aspect of this application relates to a method for predicting the probabilistic spatiotemporal distribution of charging load that takes into account the random fluctuations in commuting demand, including: Step S10: For heterogeneous vehicles under multiple origin-destination pairs in the transportation network, generate a set of travel strategies including charging quantity, departure time, and travel route; and construct a dynamic model of the transportation network based at least on commuting supply and demand balance, traffic flow dynamics between roads and charging stations, road travel time, and charging station queuing time; Step S20: Based on the probability distribution of the travel strategy set under commuting demand, construct multiple commuting demand sample points for origin-destination pairs using Monte Carlo sampling; Step S30: Using the commuting demand sample points and the dynamic model of the transportation network, obtain the destination state parameters of each commuting demand sample point under the current travel strategy selection ratio as evaluation indicators, and statistically analyze the discrete probability distribution of the evaluation indicators; Step S40: Construct a travel strategy utility evaluation model based on cumulative prospect theory, and calculate the actual utility of each travel strategy using the discrete probability distribution of the evaluation indicators; wherein, the travel strategy utility evaluation model is based on The cumulative prospect theory, by setting psychological reference points for each endpoint state parameter, compares the evaluation index with the psychological reference points and distinguishes them as perceived gains or losses. Then, it combines the weights after subjective correction of the objective discrete probability distribution to calculate the model of actual utility. Step S50: Based on actual utility, the user's travel strategy selection ratio is updated using a multi-layer nested discrete choice model, and step S30 is executed iteratively based on a heuristic algorithm until the current travel strategy selection ratio converges. Then, the converged travel strategy selection ratio and multiple commuting demand sample points are input into the traffic network dynamic model to statistically analyze and output the probabilistic spatiotemporal distribution of charging load. Among them, the multi-layer nested discrete choice model decomposes the choice of travel strategy into three levels of sequential decision-making: charging amount, departure time, and travel route. It uses actual utility as the decision basis and updates the selection ratio by calculating the product of the conditional probabilities of each level.

[0007] In one embodiment, the generation of the travel strategy set in step S10 specifically includes: classifying the vehicles under each origin-destination pair into fuel vehicles and multiple types of electric vehicles with different initial states of charge; for each type of electric vehicle and the traffic network, constructing a discrete set of charging quantity options, a set of departure time options, and a set of travel routes including those that do not pass through fast charging stations and those that pass through one fast charging station; and combining the set of charging quantity options, the set of departure time options, and the set of travel routes to generate a travel strategy set including charging quantity, departure time, and travel route.

[0008] In one embodiment, the construction of the dynamic transportation network model in step S10 specifically includes: based on the commuting supply and demand balance equation and the current travel strategy selection ratio, allocating the total commuting demand of various types of vehicles from each origin-destination pair to their corresponding travel strategies to obtain the dynamic number of vehicles for each travel strategy; based on the dynamic number of vehicles for each travel strategy, the correlation matrix, and the inbound traffic flow equation, calculating the number of vehicles entering each road and each fast charging station at each time; then, based on the outbound traffic flow equation, calculating the number of vehicles leaving each road and each fast charging station at each time according to the time of entry, dynamic travel time, dynamic queuing time, and charging time; based on the vehicle number balance equation, dynamically updating the number of vehicles in the roads and fast charging stations according to the number of vehicles entering and leaving; based on the updated number of vehicles on the roads, calculating the dynamic road travel time using a preset road impedance function; and based on the updated number of vehicles at the fast charging stations, calculating the dynamic charging queuing time using a preset queuing function.

[0009] In one embodiment, step S20 includes: when the commuting demand of each origin-destination pair for each type of vehicle is a random variable that follows a normal distribution, Monte Carlo sampling is used to generate a commuting demand sample set containing multiple sample points based on the normal distribution, wherein each sample point contains a specific commuting demand value for all origin-destination pairs for each type of vehicle.

[0010] In one embodiment, step S30 includes: taking the data from each commuting demand sample point, combining it with the current travel strategy selection ratio, and using a traffic network dynamic model to calculate the destination state parameter corresponding to each travel strategy at that sample point; traversing all commuting demand sample points to obtain multiple sets of destination state parameter values ​​corresponding to each travel strategy; for each destination state parameter of each travel strategy, placing its multiple sets of values ​​into multiple equally spaced value intervals, and counting the number of sample points falling into each value interval to obtain an objective probability, thereby forming a discrete probability distribution of the destination state parameter.

[0011] In one embodiment, step S40 includes: setting a fixed psychological reference point for the state of charge at the destination; setting a psychological reference point for the total charging cost based on the shortest driving distance and the price of slow charging at night; setting three psychological reference points for the arrival time at the destination: the earliest, the best, and the latest; and setting a psychological reference point for the total travel time based on the reliable travel time. For each destination state parameter, based on the discrete probability distribution corresponding to the destination state parameter and the psychological reference point, it is determined whether the corresponding value interval belongs to the perceived benefit interval or the perceived loss interval. For the perceived benefit interval and the perceived loss interval, different preset probability weight functions are used to perform nonlinear transformation on the objective probabilities corresponding to each interval in the discrete probability distribution to obtain subjective probability weights. For each interval, based on the difference between the representative value of the interval and the psychological reference point, combined with a preset value function, the perceived value of the interval is calculated. The value function has different convexity and concavity and amplification coefficient for losses in the benefit interval and the loss interval. The subjective probability weights of each interval are multiplied by the corresponding perceived values ​​and then summed to obtain the actual utility of each travel strategy.

[0012] In one embodiment, updating the user's travel strategy selection ratio in step S50 includes: obtaining the user's perceived utility for charging decisions, arrival decisions, and travel decisions based on the actual utility of each travel strategy; wherein, the perceived utility is composed of the actual utility plus a random term following a preset extreme value distribution; according to the decision order of first selecting the charging amount, then selecting the departure time, and finally selecting the travel route, in a multi-layer nested discrete selection model, calculating the expected maximum perceived charging utility when selecting a certain charging amount from the set of available charging amounts, and the expected maximum perceived charging utility when selecting a certain charging amount from the set of available departure times. The expected maximum perceived utility of travel time is calculated. The expected maximum perceived utility of charging and the expected maximum perceived utility of travel time are used as satisfaction indicators at the decision-making level. Based on a multi-layer nested discrete choice model, the probability of choosing a certain charging amount from the set of charging amount options, the conditional probability of choosing a certain departure time from the set of departure time options under the selected charging amount, and the conditional probability of choosing a certain route from the set of travel route options under the selected charging amount and the selected departure time are calculated sequentially. The calculated probabilities and conditional probabilities are then multiplied together to obtain the updated user's travel strategy selection ratio.

[0013] In one embodiment, statistically analyzing and outputting the probabilistic spatiotemporal distribution of charging load includes: taking data from commuting demand sample points, and using a traffic network dynamic model based on the convergent travel strategy selection ratio, to calculate the charging load at each sample point, at each time, and at each fast charging station; for each time and each fast charging station, placing the charging load values ​​of all sample points into multiple equally spaced value intervals, and counting the number of sample points falling into each value interval to obtain the probability, thereby forming the discrete probability distribution of charging load at that time and for that fast charging station; and aggregating the discrete probability distributions of charging load at all times and for all fast charging stations, and outputting the probabilistic spatiotemporal distribution of charging load.

[0014] In a second aspect, this application provides an electronic device, comprising: at least one memory for storing a program; and at least one processor for executing the program stored in the memory, wherein when the program stored in the memory is executed, the processor is configured to execute the method described in the first aspect or any possible implementation thereof.

[0015] Thirdly, this application provides a computer-readable storage medium storing a computer program that, when run on a processor, causes the processor to perform the method described in the first aspect or any possible implementation thereof.

[0016] Fourthly, this application provides a computer program product that, when run on a processor, causes the processor to perform the method described in the first aspect or any possible implementation thereof.

[0017] Overall, the technical solutions conceived in this application have the following beneficial effects compared with the prior art: This application first constructs a multi-dimensional travel strategy set that simultaneously includes charging amount, departure time and travel route. This surpasses the limitation of traditional models that only consider a single decision dimension, enabling the model to more realistically simulate users' comprehensive decision-making behavior in terms of energy, time and space.

[0018] Secondly, the solution innovatively introduces uncertainty from the commuting demand side, using Monte Carlo sampling to generate a large number of demand fluctuation samples, rather than directly assuming the probability distribution of travel or queuing times. This fundamentally characterizes the deep-seated causes of random load fluctuations. Next, the solution constructs a utility evaluation model based on cumulative prospect theory. By setting psychological reference points and subjectively modifying objective probabilities, it cleverly characterizes the bounded rationality characteristics of users in risk decision-making, such as loss aversion and probability perception bias. This overcomes the prediction bias caused by the traditional assumption of perfect user rationality.

[0019] The proposed solution employs a multi-layered nested discrete choice model, decomposing complex joint travel decisions into sequential choices at three levels: charging, departure, and route. This is then iteratively applied using a heuristic algorithm until the transportation network reaches equilibrium. This design allows the model to precisely simulate users' trade-offs of different utility indicators at each decision level and to iteratively approximate the true stable state. Finally, once user equilibrium is achieved, the determined proportions of travel strategy choices are combined with random commuting demand samples and input into the dynamic transportation network model for statistical analysis, thereby outputting the temporal and spatial probability distribution of charging load.

[0020] Compared to existing technologies, this solution achieves more accurate predictions of the probabilistic spatiotemporal distribution of electric vehicle charging load during evening peak hours by modeling uncertainty at the source of demand and meticulously depicting the hierarchical decision-making process of boundedly rational users in uncertain environments. This provides a more scientific and reliable basis for the optimized planning of charging infrastructure and the safe and economical dispatch of the power grid. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating the method for predicting the spatiotemporal distribution of charging load probabilistics that takes into account random fluctuations in commuting demand, provided in an embodiment of this application. Figure 2 This is a topology diagram of a 13-node traffic system provided in an embodiment of this application; Figure 3 This is the charging load timing curve of the first fast charging station FCS1 provided in the embodiments of this application; Figure 4 This is the charging load timing curve of the second fast charging station FCS2 provided in the embodiments of this application; Figure 5 This is the charging load timing curve of the third fast charging station FCS3 provided in the embodiments of this application; Figure 6 This is the charging load timing curve of the fourth fast charging station FCS4 provided in the embodiments of this application; Figure 7 This is the charging load timing curve of the fifth fast charging station FCS5 provided in the embodiments of this application; Figure 8 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0023] In this application, the term "and / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent three cases: A existing alone, A and B existing simultaneously, and B existing alone. In this application, the symbol " / " indicates that the related objects are in an "or" relationship, for example, A / B means A or B.

[0024] In this application, the terms "first" and "second," etc., are used to distinguish different objects, not to describe a specific order of objects. For example, "first response message" and "second response message," etc., are used to distinguish different response messages, not to describe a specific order of response messages.

[0025] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0026] In the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more, for example, multiple processing units means two or more processing units, multiple elements means two or more elements, etc.

[0027] With the rapid growth of electric vehicle ownership, the transportation sector is transitioning towards electrification and decarbonization, creating opportunities for the absorption of massive distributed renewable energy sources on the low-voltage side of the power distribution network. However, the charging behavior of electric vehicle users is highly random, making it difficult to fully utilize the flexibility of charging load shifting in time and space.

[0028] Based on this, this application proposes a method for predicting the probabilistic spatiotemporal distribution of charging load that takes into account the random fluctuations in commuting demand. Please refer to... Figure 1 , Figure 1 This is a flowchart illustrating the method for predicting the spatiotemporal distribution of charging load probability taking into account random fluctuations in commuting demand, provided in an embodiment of this application.

[0029] In this embodiment, the method includes steps S10 to S50.

[0030] Step S10: For heterogeneous vehicles with multiple origin-endpoint pairs in the transportation network, generate a set of travel strategies including charging amount, departure time and travel route; and construct a dynamic model of the transportation network based at least on the commuting supply and demand balance, traffic flow dynamics of roads and charging stations, road travel time and charging station queuing time.

[0031] Step S20: Based on the probability distribution of the travel strategy set under commuting demand, construct multiple commuting demand sample points for origin-destination pairs using Monte Carlo sampling.

[0032] Step S30: Using commuting demand sample points and a dynamic transportation network model, obtain the endpoint state parameters of each commuting demand sample point under the current travel strategy selection ratio as evaluation indicators, and statistically analyze the discrete probability distribution of the evaluation indicators.

[0033] Step S40: Construct a travel strategy utility evaluation model based on cumulative prospect theory, and calculate the actual utility of each travel strategy using the discrete probability distribution of the evaluation indicators.

[0034] Step S50: Based on actual utility, update the user's travel strategy selection ratio using a multi-layer nested discrete choice model, and iterate by jumping to step S30 based on a heuristic algorithm until the current travel strategy selection ratio converges; then input the converged travel strategy selection ratio and multiple commuting demand sample points into the traffic network dynamic model, and statistically output the probabilistic spatiotemporal distribution of charging load.

[0035] Specifically, step S10 is the foundational data generation and system modeling stage of the prediction method. Its purpose is to provide input strategies and characterize the dynamic environment in which these strategies are executed for subsequent simulation and prediction. This step specifically includes two core tasks: first, generating a set of travel strategies representing the possible behaviors of different vehicles; and second, constructing a dynamic model of the transportation network capable of evaluating the effectiveness of these strategies.

[0036] It should be noted that heterogeneous vehicles here refer to a group of electric vehicles with different parameters such as battery capacity, energy consumption per 100 kilometers, and charging power acceptance. Origin-Destination (OD) pairs are a fundamental concept in transportation planning, representing the spatial distribution of travel demand, such as a journey from a residential area to a business district. Generating a travel strategy set refers to, for each type of heterogeneous vehicle under each type of origin-destination pair, using algorithms to enumerate or sample a set of possible alternative routes that the vehicle might take during a complete trip.

[0037] The charging amount is determined by the vehicle's initial battery level, travel distance, target battery level, and charging decisions along the way; the departure time is a random variable input that follows a specific probability distribution; and the travel route can be generated by a path planning algorithm to provide multiple alternative routes.

[0038] It should be noted that the task involves constructing a dynamic traffic network model based on at least the commuting supply and demand balance, traffic flow dynamics at roads and charging stations, road travel time, and charging station queuing time. This task aims to establish a digital twin environment capable of simulating the time-varying state of a traffic system. In this model, the commuting supply and demand balance is represented by the total travel demand from each origin to each destination within a specific time period. This, along with the capacity of each road segment in the traffic network and the number of charging piles at each charging station, constitutes a supply constraint, jointly determining the degree of congestion in the system. The traffic flow dynamics at roads and charging stations are the core dynamic process of the model. This can be simulated using difference equations from macroscopic traffic flow theory or rules based on cellular transport models to depict the convergence, dissipation, and transfer of traffic flow within the road network. Simultaneously, charging stations can be modeled as nodes with queuing service characteristics; vehicles entering begin charging services, and leaving release resources.

[0039] It should be noted that road travel time is typically calculated dynamically using a function dependent on traffic flow, meaning that travel time increases with traffic density on the road segment. Charging station queuing time, on the other hand, can be calculated using queuing theory models, depending on vehicle arrival rate and charging station service rate. The entire dynamic model can be constructed by programming and calling the interface of professional traffic simulation software, or by directly developing it using an agent-based modeling framework. The output of this dynamic model is the state of each road segment and each charging station in the road network at each simulation time step, providing a computational environment for evaluating the actual time and energy consumption of each travel strategy.

[0040] Understandably, the key decision variable is the proportion of vehicles choosing each travel strategy. Therefore, the traffic network dynamics model can be viewed as a model for calculating the destination state parameters under a given proportion of travel strategy choices; this proportion describes the percentage of vehicles choosing each specific travel strategy within a heterogeneous vehicle population. The model's internal operating mechanism—including the listed dynamic processes of commuting supply and demand balance, traffic flow dynamics, and calculation of travel and queuing times—essentially simulates the interactions and congestion evolution within the traffic network and charging infrastructure when a large number of vehicles execute their respective strategies according to a given selection proportion. Its final output, the destination state parameters, is a specific indicator for quantitatively evaluating the actual effectiveness of each strategy.

[0041] Among them, the destination status parameters are the destination charge status, total charging cost, destination arrival time, and total travel time after the vehicle selects a specific travel strategy to complete the trip.

[0042] Specifically, step S20 is a crucial sampling step in the prediction method that connects probabilistic travel demand with specific traffic simulation. Its purpose is to transform the randomness of commuting demand during the evening peak hours into a series of deterministic input scenarios that can be processed by the dynamic traffic network model, thus laying the foundation for subsequent probabilistic load forecasting. The core operation of this step is to perform Monte Carlo sampling based on the probability distribution of travel strategies to generate sample points representing different probabilities of commuting demand.

[0043] It is important to clarify the specific meaning of the probability distribution of the travel strategy set. In the travel strategy set generated in step S10, each strategy is not chosen with equal probability for a certain type of heterogeneous vehicle traveling from a specific origin to a destination. This probability distribution describes the likelihood of this type of vehicle choosing each alternative strategy under the evening rush hour commuting scenario. This distribution can be obtained through statistical fitting based on historical travel survey data and large-scale vehicle trajectory data, or it can be set by planners as model input parameters. This probability distribution is the direct basis for Monte Carlo sampling.

[0044] It's important to note that Monte Carlo sampling is a statistical method that obtains numerical results through repeated random sampling. In its specific implementation here, it involves using a computer's pseudo-random number generator to conduct a large number of independent, repeated random samplings for each heterogeneous vehicle type under each origin-destination pair, based on the probability distribution of its travel strategy. Each sampling constitutes an experiment, and the result is the random selection of a specific travel strategy for that vehicle type. Summarizing the sampling results for all vehicle types under all origin-destination pairs in a single experiment constitutes a complete commuting demand sample point. This sample point represents a specific combination scenario of all probabilistically possible vehicle travel strategies. By conducting thousands of such independent experiments, an equivalent set of commuting demand sample points can be constructed, thus comprehensively covering all possible random states of evening rush hour commuting demand.

[0045] Specifically, step S30 is the core step in the prediction method involving simulation calculations and probabilistic statistical analysis. Its purpose is to utilize the established dynamic transportation network model to simulate a large number of randomly generated travel scenarios and quantitatively evaluate the statistical characteristics of the results, thereby ultimately transforming the random fluctuations in travel behavior into a measurable risk distribution. This step specifically includes two sequential operations: first, performing batch simulations using the dynamic transportation network model to obtain evaluation indicators; and second, performing statistical analysis on these indicator results to obtain their probability distribution.

[0046] It should be noted that the endpoint state parameters of each commuting demand sample point under the current travel strategy selection ratio are obtained as evaluation indicators through commuting demand sample points and a dynamic transportation network model. This operation is an automated batch simulation process. Each commuting demand sample point generated in step S20 represents a complete combination of all specific travel strategies that are probabilistically possible, and this combination itself implicitly contains the current travel strategy selection ratio. Using this sample point as input, the dynamic transportation network model constructed in step S10 is driven to run, simulating the dynamic interaction process of vehicles in the road network and charging network under this specific strategy combination. After the model runs, the endpoint state parameters of each vehicle in the sample points can be extracted. These parameters quantify the execution effect of the travel strategy in this random scenario from dimensions such as user cost and trip efficiency, and are therefore defined as evaluation indicators. By sequentially performing this simulation-extraction process on a large number of commuting demand sample points, a large-scale sample set of evaluation indicators can be obtained.

[0047] It's important to clarify the discrete probability distribution of the statistical evaluation indicators. This operation involves performing probabilistic statistical analysis on the large-scale simulation results described above. Since the input commuting demand sample points are randomly generated based on a probability distribution, the output evaluation indicators must also be random variables. Statistically analyzing their discrete probability distribution involves performing statistical processing on all sample values ​​for each evaluation indicator to characterize the probability of that indicator taking different values. In practice, this can be achieved by plotting empirical probability density function curves or empirical cumulative distribution function curves; dividing the numerical range into several intervals and generating histograms to display the frequency of each interval; and calculating and outputting key statistics such as mean, variance, and quantiles. This probability distribution reveals the uncertainty range of the evaluation indicators in an intuitive way; for example, it can clearly indicate that the probability of the total travel time exceeding 60 minutes is 10%.

[0048] Specifically, step S40 is a key step in the prediction method that introduces behavioral decision theory to quantify users' subjective preferences. Its purpose is to go beyond traditional expected value calculations and, based on cumulative prospect theory, construct a utility evaluation model that better reflects the psychological behavior of real users under risky decision-making. Using the probability distribution of evaluation indicators obtained in the previous step, it calculates the actual utility of each travel strategy at the user's subjective perception level, thus providing a basis for subsequent strategy optimization or demand response incentives.

[0049] It should be noted that the travel strategy utility evaluation model is constructed based on cumulative prospects theory. Cumulative prospects theory is a behavioral economics model that describes how people make decisions under uncertain conditions. It modifies the classic expected utility theory and is more in line with actual human decision-making behavior.

[0050] In this application scenario, the travel strategy utility evaluation model refers to the mathematical calculation model built upon this theoretical framework. The core components of this model include a value function and a probability weighting function. The value function measures a user's subjective perception of the value of a particular travel outcome. Its typical characteristic is that it is more sensitive to losses than gains; the function curve inverts at a reference point, and it is usually convex with respect to losses and concave with respect to gains.

[0051] The probability weighting function, on the other hand, transforms objective probabilities into decision weights, systematically overestimating low-probability events and underestimating medium- to high-probability events. The specific implementation of this model requires programming these functions onto a computing device.

[0052] It should be noted that the discrete probability distribution of the evaluation indicators is used to calculate the actual utility of each travel strategy. This operation is the specific calculation process of the utility evaluation model. The discrete probability distribution of the evaluation indicators obtained in step S30 provides a key input for each travel strategy: that is, the probability of each evaluation indicator taking each possible value after the strategy is executed. When calculating the actual utility, a reference point must first be set for each evaluation indicator. Then, for each possible outcome value, the difference between it and the reference point is calculated as a gain or loss, and converted into a subjective value through a value function. Next, the objective probability corresponding to the outcome is converted into a decision weight using a probability weight function. Finally, based on the calculation framework of cumulative prospect theory, the subjective values ​​of all possible outcomes are multiplied by their decision weights and summed to obtain the comprehensive prospect value of the travel strategy, which is the actual utility.

[0053] Therefore, the travel strategy utility evaluation model is based on cumulative prospect theory. It sets psychological reference points for each destination state parameter, compares the evaluation index with the psychological reference points and distinguishes them as perceived gains or losses, and then combines the weights after subjective correction of the objective discrete probability distribution to calculate the actual utility model.

[0054] Specifically, step S50 is the core iterative optimization and result generation step in the prediction method, which realizes the behavioral feedback loop and the final probabilistic output. Its purpose is to feed back the actual utility of the strategy calculated based on the behavioral model to the travel selection process, dynamically update the user's strategy selection preferences, and through iterative simulation, achieve a self-consistent and stable equilibrium between the simulated system state and the user's behavioral response. Finally, based on this equilibrium state, a probabilistic spatiotemporal distribution of charging load is output. This step specifically includes an iterative optimization loop and a final statistical output process.

[0055] It should be noted that, based on actual utility, a multi-layered nested discrete choice model is used to update the user's travel strategy selection ratio, and a heuristic algorithm is used to jump to step S30 for iteration until the current travel strategy selection ratio converges. This operation constructs a feedback and equilibrium solution mechanism for a behavior-traffic coupled system. Specifically, after obtaining the actual utility of each travel strategy, it is used as a key input into the multi-layered nested discrete choice model. This model is a classic econometric model used to simulate how users make decisions when faced with a set of hierarchical choices. Specifically, the multi-layered nested discrete choice model decomposes the choice of travel strategy into three levels of sequential decisions: charging capacity, departure time, and travel route. Actual utility is used as the decision-making basis, and the selection ratio is updated by calculating the product of conditional probabilities at each level. Based on the principle of utility maximization, the model uses mathematical formulas to calculate the updated probability of choosing each travel strategy, i.e., the updated travel strategy selection ratio.

[0056] It should be noted that the updated selection ratio is used as the new input, and the process jumps back to step S30 to re-perform traffic dynamic simulation and evaluation index statistics. This process constitutes one iteration. Since the new selection ratio will change the congestion situation of the road network and charging stations, and thus affect the endpoint state parameters and actual utility through the dynamic model, this iterative process needs to be repeated multiple times.

[0057] It should be noted that after each iteration, the travel strategy selection ratios obtained from the previous and subsequent iterations are compared. When the change is less than a preset minimum threshold, the system is considered to have reached a stable state, and the selection ratio has converged. Since this equilibrium problem may be non-convex, direct computation is difficult. Therefore, heuristic algorithms, such as genetic algorithms, simulated annealing algorithms, or particle swarm optimization algorithms, are often used to guide the update direction of the selection ratio or search for the convergence point, in order to improve solution efficiency and global optimization capabilities.

[0058] It should be noted that the converged travel strategy selection ratio and multiple commuting demand sample points are then input into the traffic network dynamic model to statistically analyze and output the probabilistic spatiotemporal distribution of charging load. This operation is the final output stage of the method. When the selection ratio converges, the system reaches a stable behavior-flow equilibrium state. At this point, the converged travel strategy selection ratio is used to weight or resample the multiple commuting demand sample points generated in step S20 to ensure that the strategy distribution in the sample points is consistent with the equilibrium ratio. These sample points, representing the final stable state of the system, are then input into the traffic network dynamic model in a final batch for simulation. In the simulation, the charging power of each charging pile or charging station at each moment can be accurately recorded. By performing spatiotemporal aggregation and probability statistics on the charging power data of all sample points, the probabilistic spatiotemporal distribution of charging load can be output.

[0059] In this embodiment, compared to the traditional model's travel strategy dimension, which does not comprehensively consider the unified decision-making of charging volume, departure time, and travel route, this application takes a more comprehensive approach, has a stronger ability to simulate user decisions in real-world situations, and thus improves the accuracy of load forecasting.

[0060] In this embodiment, compared to directly assuming that road commuting time and charging station queuing time are probabilistically distributed without investigating the underlying reasons for the probability distribution of time, this application introduces uncertainty on the commuting demand side and combines it with a dynamic loading model of the transportation network to deduce the probability distribution of time from the source and clarify its underlying formation process.

[0061] In this embodiment, compared to traditional methods that often assume perfect rationality and fail to reflect bounded rationality behaviors such as risk aversion and overestimation of low-probability events in uncertain environments, this patent modifies the utility evaluation system for electric vehicle user travel strategies based on prospect theory, fully characterizing users' bounded rationality when facing uncertainties.

[0062] Based on the above, this application provides a feasible implementation method. Please refer to... Figure 2 , Figure 2 This is a topology diagram of a 13-node traffic system provided in an embodiment of this application.

[0063] In this embodiment, the probabilistic spatiotemporal distribution prediction method for charging load, which takes into account the random fluctuations in commuting demand, was tested in a traffic system consisting of 13 nodes, 19 road segments, and 5 fast charging stations. This patent sets the OD pair as node 1-node 2, the electric vehicle battery capacity is 100kWh, and the electric vehicles are divided into four categories according to the initial battery capacity of 20kWh, 30kWh, 40kWh and 50kWh. The commuting demand from 17:00 to 22:00 is 200, 350, 500 and 300 vehicles respectively, and the coefficient of variation of commuting demand is 0.1. There are five optional charging capacities for electric vehicles at fast charging stations: 10kWh, 20kWh, 30kWh, 40kWh and 50kWh, with a rated charging power of 50kW. The charging price at fast charging stations is ¥0.6 / kWh, and the charging price at home slow charging piles is ¥0.3 / kWh. There are five optional departure times: 17:00, 17:30, 18:00, 18:30 and 19:00.

[0064] In this embodiment, the generation of the travel strategy set in step S10 specifically includes: classifying the vehicles under each origin-destination pair into fuel vehicles and multiple types of electric vehicles with different initial states of charge; for each type of electric vehicle and the traffic network, constructing a discrete set of optional charging amounts, a set of optional departure times, and a set of optional travel routes including those that do not pass through fast charging stations and those that pass through one fast charging station; and combining the set of optional charging amounts, the set of optional departure times, and the set of optional travel routes to generate a travel strategy set including charging amounts, departure times, and travel routes.

[0065] First, the vehicles under each OD pair are broadly divided into two categories: Gasline Vehicles (GVs) and Electric Vehicles (EVs). Electric vehicles are further subdivided into M categories based on their initial State of Charge (SOC). The higher the model's granularity, the smaller the classification. For example, they can be divided into eight categories based on 20%, 30%, 40%, 50%, 60%, 70%, 80%, and 90%. This patent ignores the heterogeneity of electric vehicle battery capacity and converts the initial state of charge into initial charge capacity.

[0066] Therefore, each OD pair contains M+1 types of vehicles. Construct a heterogeneous vehicle set. The first element 0 represents a gasoline-powered vehicle, which does not need to be charged on its way home. This patent also assumes that the vehicle has sufficient fuel and does not need to go to a gas station for refueling. The following M elements represent electric vehicles with different starting battery levels. In order to alleviate the user's "range anxiety", they may need to charge at fast charging stations (FCS) along the way. This application assumes that the electric vehicle will choose at most one fast charging station along the way to charge before reaching its destination.

[0067] Specifically, the charging strategy set for heterogeneous vehicles is as follows: (1); (2).

[0068] Equation (1) represents the set of selectable charging amounts after discretizing the charging decision of electric vehicles at fast charging stations. ; NE represents the nth optional charge amount; NE represents the total number of optional charge amounts; od represents any OD pair; This is the set of all OD pairs. Equation (2) represents the charging strategy set for each type of vehicle; Let m be the set of charging strategies for vehicles of class m.

[0069] Specifically, the departure time strategy set is as follows: (3).

[0070] Equation (3) represents the departure time strategy set after discretizing the vehicle's departure time decision at the starting point. ; Let NT be the nth optional departure time, and NT be the total number of optional departure times.

[0071] Specifically, the path strategy set is as follows: (4); (5); (6).

[0072] Equation (4) represents the path set of OD to od that does not pass through a fast charging station. ; Let OD be the nth path that does not pass through a fast charging station. Let be the number of paths for OD to od that do not pass through a fast charging station. Equation (5) represents the set of paths for OD to od that pass through a fast charging station once. ; Let OD be the nth path that OD takes once through a fast charging station. Let be the number of paths taken by OD to OD through a fast charging station once. Equation (6) represents the path strategy set for OD to OD for the m-th type of vehicle. .

[0073] Specifically, the sets of available charging amounts, departure times, and travel routes are combined to generate a set of travel strategies that include charging amount, departure time, and travel route, as follows: (7).

[0074] Where e, k, and p represent any charge amount e, any departure time k, and any path p, respectively.

[0075] In this embodiment, the construction of the dynamic transportation network model in step S10 specifically includes: based on the commuting supply and demand balance equation and the current travel strategy selection ratio, allocating the total commuting demand of various types of vehicles from each origin-destination pair to their corresponding travel strategies to obtain the dynamic number of vehicles for each travel strategy; based on the dynamic number of vehicles for each travel strategy, the correlation matrix, and the inbound traffic flow equation, calculating the number of vehicles entering each road and each fast charging station at each time; then, based on the outbound traffic flow equation, calculating the number of vehicles leaving each road and each fast charging station at each time according to the time of entry, dynamic travel time, dynamic queuing time, and charging time; based on the vehicle number balance equation, dynamically updating the number of vehicles in the roads and fast charging stations according to the number of vehicles entering and leaving; based on the updated number of vehicles on the roads, calculating the dynamic road travel time using a preset road impedance function; and based on the updated number of vehicles at the fast charging stations, calculating the dynamic charging queuing time using a preset queuing function.

[0076] Specifically, the dynamic model of the transportation network includes the following equations (8) to (24): (8); Equation (8) represents the supply and demand balance of commuting; For the period under study, OD represents the total commuting demand for vehicle class m in OD. Let OD be the number of trips for vehicle class m in the OD selection strategy s.

[0077] (9); Wherein, Equation (9) represents the number of various types of vehicles entering each road at each time; For the m-th type of vehicle choosing the travel strategy s under OD, in The number of vehicles entering road a at any given time; if vehicles choose strategy s for travel, If the time of arrival at the entrance of road a is [time], then the elements of the correlation matrix are [element]. Otherwise ; For OD pairs, For time period set, For roads.

[0078] (10); Wherein, Equation (10) represents the number of various types of vehicles entering each fast charging station at each time; For the m-th type of vehicle choosing the travel strategy s under OD, in The number of vehicles entering fast charging station b at any given time; if the vehicle chooses strategy s for travel, If the time of arrival at fast charging station entrance b is such that the correlation matrix elements are... ,otherwise ; For fast charging stations.

[0079] (11); Wherein, Equation (11) represents the number of various types of vehicles leaving each road at each time; For the m-th type of vehicle choosing the travel strategy s under OD, in The number of vehicles that leave road a at any given time.

[0080] (12); Wherein, equation (12) represents the travel time of each road at each time point; for The constantly observed travel time of road a. This represents the number of time periods corresponding to the passage time (divided into 5-minute intervals).

[0081] (13); Wherein, Equation (13) represents the number of various types of vehicles leaving each fast charging station at each time; For the m-th type of vehicle choosing the travel strategy s under OD, in The number of vehicles leaving fast charging station b at any given time.

[0082] (14); Wherein, Equation (14) represents the queuing time of each fast charging station at each time; In order to be in The queuing time at fast charging station b was observed in real time. This represents the number of time slots corresponding to the queuing time.

[0083] (15); Wherein, equation (15) represents the charging time; Choose a strategy for your vehicle to determine charging time during your trip. This represents the number of time periods corresponding to the charging time.

[0084] (16); (17); (18); (19); Wherein, equation (16) represents the total number of vehicles entering each road at each time; express The total number of vehicles entering road a at each time point. Equation (17) represents the total number of vehicles leaving each road at each time point; express The total number of vehicles leaving road a at any given time. Equation (18) represents the total number of vehicles entering each fast charging station at each given time. express The total number of vehicles entering fast charging station b at any given time. Equation (19) represents the total number of vehicles leaving each fast charging station at any given time; express The total number of vehicles leaving fast charging station b at any given time.

[0085] (20); (twenty one); Wherein, equation (20) represents the balance of the number of vehicles on the road; for The number of vehicles on road a at time a. Equation (21) represents the balance of the number of vehicles at the fast charging station; for The number of vehicles at fast charging station b at any given time.

[0086] (twenty two); (twenty three); (twenty four).

[0087] Equation (22) represents the road travel time (BPR function). Let be the free passage time for road a. Let be the capacity of road a. Equation (23) represents the queuing time at the charging station (Davidson queuing function). and An empirical coefficient related to queuing time. Let b be the capacity of charging station b. Equation (24) represents the charging time; and The charging time (min) and charging amount (kWh) are respectively for the vehicle choosing strategy s for travel. Rated charging power (kW) for electric vehicles.

[0088] It should be noted that the indicators corresponding to the travel strategy are calculated as follows: (25); Wherein, Equation (25) represents the final state of charge of each OD for each type of vehicle (the final state of charge of a fuel vehicle is denoted as 1). For OD, the destination charge state of vehicle s during its journey is determined by the selection strategy for vehicle m-th class od. Let be the initial battery level of the m-th type of vehicle. and Choose a strategy for your vehicle based on charging frequency and driving range during your trip. Electricity consumption per unit distance.

[0089] (26); Wherein, Equation (26) represents the charging cost of each OD for each type of vehicle. The charging cost of fuel vehicles is 0, and the charging cost of electric vehicles consists of two parts: the first part is the fast charging cost on the way home, and the second part is the slow charging cost to replenish the state of charge to 0.8 after arriving home. The charging cost during the trip for vehicle class m in the OD (Operational Development) strategy. for Charging price at fast charging station b during peak hours (¥ / kWh) The amount of charge within a unit time interval (5 minutes); The charging price for slow charging stations at night.

[0090] (27); Wherein, Equation (27) represents the arrival time of the destination for each type of vehicle under each OD; and For OD, the destination arrival time and travel time of vehicle s when choosing the strategy for vehicle m in the OD trip are calculated. The departure time is the starting point of travel strategy s.

[0091] (28); Equation (28) represents the travel time of each OD for various types of vehicles, which consists of three parts: road travel time, fast charging station queuing time, and charging time. Let OD be the travel time for the m-th type of vehicle when choosing the travel strategy s.

[0092] In this embodiment, step S20 includes: when the commuting demand of each origin-destination pair for each type of vehicle is a random variable that follows a normal distribution, Monte Carlo sampling is used to generate a commuting demand sample set containing multiple sample points based on the normal distribution, wherein each sample point contains a specific commuting demand value for all origin-destination pairs for each type of vehicle.

[0093] Understandably, the OD (Original Design Location) sample points for commuting demand are as follows: (29); (30).

[0094] Equation (29) indicates that the ODs for various types of car commuting demand follow a normal distribution; Let OD be the random variable corresponding to the commuting demand of vehicle class m in relation to OD, and its expectation is... The standard deviation is The standard deviation and the expected value have a linear relationship. Let be the coefficient of variation of OD for each type of vehicle. Equation (30) represents the commuting demand of each OD for each type of vehicle in each sample point, where I is the total number of sample points. ; Let i be the set of commuting demands from OD to OD at sample point i. Let represent the commuting demand of vehicle m-th class OD at sample point i.

[0095] In this embodiment, step S30 includes: taking the data from each commuting demand sample point, combining it with the current travel strategy selection ratio, and using a traffic network dynamic model to calculate the destination state parameter corresponding to each travel strategy at that sample point; traversing all commuting demand sample points to obtain multiple sets of destination state parameter values ​​corresponding to each travel strategy; for each destination state parameter of each travel strategy, placing its multiple sets of values ​​into multiple equally spaced value intervals, and counting the number of sample points falling into each value interval to obtain an objective probability, thereby forming a discrete probability distribution of the destination state parameter.

[0096] Understandably, this is first based on the sample points. i Data Assign values ​​to the commuting demand of various types of vehicles under each OD pair; then set the initial value of the travel strategy selection ratio of each type of vehicle in each OD pair. In the initial state, the strategies in the travel strategy set can be selected proportionally to allocate commuting demand, as shown in equation (31). For sample points i OD in China od Next selection strategy s The first day of the trip m Number of vehicles of each type: (31).

[0097] It should be noted that the allocated commuting demand is injected into the dynamic model of the transportation network, and the index values ​​corresponding to each travel strategy under the sample point are calculated, referring to the above formulas (25) to (28).

[0098] It should be noted that by traversing sample point 1 to sample point I, the index values ​​corresponding to each travel strategy under each sample point are obtained, as shown in equation (32): (32).

[0099] in, , , and Let i represent the destination state of charge, charging cost, destination arrival time, and travel time for vehicle m at sample point i and OD respectively when choosing strategy s for vehicle m.

[0100] Finally, the index values ​​corresponding to each travel strategy under sample point 1 to sample point I are placed into equally spaced intervals. Within each interval, the number of sample points in each interval is counted to obtain the discrete probability distribution of the index values ​​corresponding to each travel strategy, as shown in equation (33): (33)

[0101] in, For the set of interval numbers, , , and Let J represent the representative values ​​of the destination state of charge, charging cost, destination arrival time, and travel time in interval j for OD to select the m-th vehicle type s during the trip. , , and These represent the probabilities corresponding to the representative value.

[0102] In this embodiment, step S40 includes: setting a fixed psychological reference point for the state of charge at the destination; setting a psychological reference point for the total charging cost based on the shortest driving distance and the price of slow charging at night; setting three psychological reference points for the arrival time at the destination: the earliest, the best, and the latest; and setting a psychological reference point for the total travel time based on the reliable travel time. For each destination state parameter, based on the discrete probability distribution corresponding to the destination state parameter and the psychological reference point, it is determined whether the corresponding value interval belongs to the perceived benefit interval or the perceived loss interval. For the perceived benefit interval and the perceived loss interval, different preset probability weight functions are used to perform nonlinear transformation on the objective probabilities corresponding to each interval in the discrete probability distribution to obtain subjective probability weights. For each interval, based on the difference between the representative value of the interval and the psychological reference point, combined with a preset value function, the perceived value of the interval is calculated. The value function has different convexity and concavity and amplification coefficient for losses in the benefit interval and the loss interval. The subjective probability weights of each interval are multiplied by the corresponding perceived values ​​and then summed to obtain the actual utility of each travel strategy.

[0103] It should be noted that the evaluation system for the utility of travel strategies based on the cumulative prospect theory is constructed and the utility is calculated as follows (34) to (52).

[0104] (34); (35); Equation (34) represents the subjective probability that the travel strategy results in a benefit; Let p be a distortion function with respect to the objective probability p. For empirical coefficients. Equation (35) represents the subjective probability that the travel strategy results in a loss; Let p be a distortion function with respect to the objective probability p. This is an empirical coefficient.

[0105] (36); (37); (38); (39); Wherein, equation (36) represents the reference point of the final state of charge and the interval number of the reference point; The reference point for the final state of charge is set to 0.3 in this application; Let be the interval number of the reference point for the destination state of charge when OD selects strategy s for vehicle m of class m. Equation (37) represents the value calculation of the destination state of charge; and For OD to select strategy s for vehicle class m, the representative value and value of the destination charge state in interval j when traveling. and For risk preference parameters, This is the loss aversion parameter.

[0106] Among them, Equation (38) represents the subjective probability weight calculation of the final state of charge in each interval; and For OD to choose strategy s for vehicle class m, the objective probability and subjective probability weight of the destination state of charge in interval j are given. Equation (39) represents the utility calculation of the destination state of charge; Let OD be the terminal state-of-charge utility of the choice strategy s for vehicle class m of OD during the trip.

[0107] (40); (41); (42); (43); (44); Wherein, equation (40) represents the charging cost reference point and the interval number of the reference point; Let OD be the shortest driving distance for vehicle of class m in the OD category. Choose a strategy for the vehicle to cover the mileage during the trip. and Let OD be the reference point for charging costs and its corresponding interval number when the travel strategy of vehicle m is s.

[0108] It should be noted that traditional methods use the lowest charging cost for fast charging along the route and high electricity prices as reference points. This application comprehensively considers the charging price for slow charging after returning home, low electricity prices, and the initial state-of-charge demand on the second day to formulate a comprehensive reference point for charging costs. This fully depicts the user's trade-offs regarding charging costs when alleviating "range anxiety" through fast charging along the route, providing a refined representation of user behavior characteristics, thereby improving the accuracy of load forecasting.

[0109] Equation (41) represents the calculation of the value of charging costs; and For OD to choose strategy s for vehicle class m, the representative value and value of charging cost in interval j are given. Equation (42) represents the subjective probability weight calculation of charging cost in each interval; and For OD to choose strategy s for vehicle class m, the objective probability and subjective probability weight of charging cost in interval j when traveling.

[0110] Equation (43) represents the utility calculation of charging costs; Let OD be the utility of charging cost when choosing strategy s for vehicle m. Equation (44) represents the total utility calculation of charging amount decision, which is the weighted sum of the utility of the final state of charge and the utility of charging cost; The charging utility of OD for the m-th type of vehicle during the trip. These are the weighting coefficients.

[0111] (45); (46); (47); (48); Wherein, equation (45) represents the reference point for the arrival time of the destination and the interval number of the reference point; and The earliest acceptable arrival time for the user and its corresponding interval number. and The optimal arrival time and its corresponding interval number. and This represents the latest arrival time acceptable to the user and its corresponding interval number.

[0112] Equation (46) represents the value calculation at the time of arrival at the destination; and For OD to choose strategy s for vehicle class m, the representative value and value of the destination arrival time in interval j. Equation (47) represents the subjective probability weight calculation of the destination arrival time in each interval; and Let OD be the objective probability and its subjective probability weight of the destination arrival time falling within interval j when choosing the m-th type of vehicle s for travel.

[0113] Equation (48) represents the utility calculation at the time of arrival at the destination; Let OD be the arrival utility of the m-th vehicle selection strategy s when traveling.

[0114] (49); (50); (51); (52); Wherein, equation (49) represents the reference point for travel time and the interval number of the reference point; For OD to select strategy s for the m-th type of vehicle od, the reliable travel time is given when the confidence level is 0.8. and Let OD be the reference point for the travel time of vehicle m-th class OD when choosing strategy s for travel, and its corresponding interval number.

[0115] It should be noted that, once the discrete probability distribution of travel time is determined, traditional methods use a weighted average method to select reference points. This application selects reference points based on confidence levels, which better reflects the heterogeneous psychology of users regarding travel time reliability: for users with higher travel time reliability, the higher the confidence level, the larger the value of the time reference point.

[0116] Equation (50) represents the value calculation of travel time; and For OD to choose strategy s for vehicle class m, the representative value and value of travel time in interval j. Equation (51) represents the subjective probability weight calculation of travel time in each interval; and Let OD be the objective probability and subjective probability weight of the travel time within the interval j when choosing strategy s for vehicle m of class OD.

[0117] Equation (52) represents the utility calculation of travel time; Let OD be the travel utility of vehicle class m when choosing travel strategy s for OD.

[0118] In this embodiment, the update of the user's travel strategy selection ratio in step S50 includes: obtaining the user's perceived utility for charging decisions, arrival decisions, and travel decisions based on the actual utility of each travel strategy; wherein, the perceived utility is composed of the actual utility plus a random term following a preset extreme value distribution; according to the decision order of first selecting the charging amount, then selecting the departure time, and finally selecting the travel route, in a multi-layer nested discrete selection model, calculating the expected maximum perceived charging utility when selecting a certain charging amount from the set of available charging amounts, and the expected maximum perceived charging utility when selecting a certain charging amount from the set of available departure times. The expected maximum perceived utility of travel time is calculated. The expected maximum perceived utility of charging and the expected maximum perceived utility of travel time are used as satisfaction indicators at the decision-making level. Based on a multi-layer nested discrete choice model, the probability of choosing a certain charging amount from the set of charging amount options, the conditional probability of choosing a certain departure time from the set of departure time options under the selected charging amount, and the conditional probability of choosing a certain route from the set of travel route options under the selected charging amount and the selected departure time are calculated sequentially. The calculated probabilities and conditional probabilities are then multiplied together to obtain the updated user's travel strategy selection ratio.

[0119] It should be noted that the method for updating the travel strategy selection ratio based on the multi-layer nested logit selection model is as follows (53) to (64).

[0120] (53); (54); Equation (53) represents the set of travel strategies with consistent charging capacity; Let x be the set of travel strategies for the m-th class of OD vehicles. Equation (54) represents the set of travel strategies where the charging amount is consistent with the departure time; Let x be the total charging amount for the OD's travel strategy for vehicle m, and let the departure time be... The set of strategies. Set as departure time.

[0121] (55); (56); Wherein, equation (55) represents the charging perception utility; , and These are the perceived utility, actual utility, and perceived bias of charging during the trip for vehicle class m (OD) using the OD's choice of strategy s; where the perceived bias... It follows an independent and identically distributed Gumbel distribution with an expected value of 0. Equation (56) represents the charging satisfaction. Let OD represent the charging satisfaction of vehicle m in class OD when the amount of charging along the route is x. This refers to the user's perception sensitivity to charging efficiency.

[0122] (57); (58); Equation (57) represents the perceived utility of arrival; , and These represent the perceived arrival utility, actual utility, and perceived bias of OD for the m-th vehicle selection strategy s during the trip; where the perceived bias... It follows an independent and identically distributed Gumbel distribution with an expected value of 0. Equation (58) represents the satisfaction level reached; Let OD be the arrival satisfaction rate of vehicle m in class m along the route when the charging amount is x, at departure time y. This refers to the user's perception sensitivity to the effectiveness of the service.

[0123] (59); (60); Equation (59) represents the accessibility perception utility; , and These represent the perceived utility, actual utility, and perceived bias of OD for the m-th vehicle selection strategy s during travel; where the perceived bias... It follows an independent and identically distributed Gumbel distribution with an expected value of 0. Equation (60) represents the satisfaction level with passage. Let x be the charging amount along the route for vehicle m of type OD, and y be the departure time. Let y be the traffic satisfaction along route p. This refers to the user's sensitivity to the perceived utility of passage.

[0124] (61); (62); (63); Wherein, equation (61) represents the charging selection probability; For OD to charge class m vehicles from the charging strategy set The probability of choosing the charging amount x. Equation (62) represents the probability of choosing the departure time; For OD, when the charging amount along the route of vehicle of class m is x, the strategy set from the departure time... The probability of choosing time y; Equation (63) represents the probability of choosing a travel path; For an OD vehicle of class m with a charging amount of x along the route and a departure time of y, from the path strategy set... The probability of choosing path z.

[0125] (64); Wherein, equation (64) represents the proportion of travel strategy selection; Let OD allocate the commuting demand of vehicle class m to travel strategy s, where the charging amount for this strategy is... Departure time is The route is .

[0126] Understandably, after setting an initial travel strategy selection ratio, proceed to step S30 and iteratively update the travel strategy selection ratio using a heuristic algorithm until convergence, i.e., the transportation network reaches equilibrium. The iterative update steps for the travel strategy selection ratio based on the heuristic algorithm are as follows: First, initialize the number of iterations. ; Jump to step S30, where equation (31) is replaced by equation (65): (65)

[0127] in, For the d-th iteration, the proportion of commuting demand of vehicle class m in od allocated to travel strategy s by OD. This represents the iteration step size in the heuristic algorithm.

[0128] Then, Update the proportion of travel strategy selections and record it as shown in equation (66): (66).

[0129] It should be noted that if the iterative convergence condition as shown in equation (67) is met, then the iteration stops. (where is the convergence coefficient), and the proportion of travel strategy selection at this time is denoted as . Otherwise, continue iterating.

[0130] (67)

[0131] In this embodiment, the statistical analysis and output of the probabilistic spatiotemporal distribution of charging load includes: taking the data from commuting demand sample points, and based on the convergent travel strategy selection ratio, calculating the charging load at each sample point, at each time, and at each fast charging station using a traffic network dynamic model; for each time and each fast charging station, placing the charging load values ​​of all sample points into multiple equally spaced value intervals, counting the number of sample points falling into each value interval to obtain the probability, thereby forming the discrete probability distribution of the charging load at that time and for that fast charging station; and aggregating the discrete probability distributions of the charging load at all times and for all fast charging stations, outputting the probabilistic spatiotemporal distribution of the charging load.

[0132] It should be noted that the steps for constructing the probabilistic spatiotemporal distribution of charging load are as follows: First, set the proportion of travel strategy selection to To allocate commuting needs, as shown in equation (68): (68)

[0133] in, In the traffic network equilibrium, the number of times the commuting demand of vehicle m-th type od is assigned to travel strategy s for sample point i; Then, a set of travel strategies associated with charging load is constructed, as shown in equation (69): (69)

[0134] in, The travel strategy s in the context has the following attributes: When OD selects the m-th type of vehicle s for travel, It is still charging at fast charging station b; Next, the charging load at sample point i is calculated, as shown in equation (70): (70)

[0135] in, For sample point i when the transportation network is in equilibrium, The charging load of fast charging station B during the specified time period; It should be noted that by repeating the above steps and traversing from sample point 1 to sample point I, the charging load of each fast charging station at each time point is obtained, as shown in equation (71): (71).

[0136] Finally, the charging load of each fast charging station at each time point from sample point 1 to sample point I is placed into equally spaced intervals. Within each interval, the number of sample points in each interval is counted to obtain the discrete probability distribution of the charging load of each fast charging station at each time point, as shown in equation (72): (72); in, For the set of interval numbers, for The representative value of the charging load of fast charging station b in interval j during the specified time period. This represents the probability corresponding to that representative value.

[0137] Understandably, the introduction of uncertainty into the commuting demand side expands the original deterministic load curve into a strip shape, providing power dispatchers with richer load information as a basis for robust dispatching, thereby improving the safety and stability of power system operation.

[0138] In summary, please refer to Figures 3 to 7 The charging load timing curves of the first fast charging station FCS1 to the fifth fast charging station FCS5 are shown respectively.

[0139] Understandably, the charging load at all fast charging stations exhibits typical evening peak characteristics, with the load starting to rise around 5:00 PM, peaking between 6:00 PM and 6:30 PM, then gradually declining, reaching a low point around 8:00 PM. This reflects the concentrated charging behavior of commuters after get off work.

[0140] Understandably, the load capacity of different fast charging stations varies significantly. For example, the peak power of FCS3 is as high as about 40,000 kW, while the peak power of FCS5 is only about 250 kW, indicating that the charging demand is extremely unevenly distributed in different geographical locations and exhibits spatial heterogeneity due to factors such as user density and traffic flow.

[0141] Understandably, each curve depicts the probability distribution of load using three curves: the 5th percentile, the expected value, and the 95th percentile. The expected value curve represents the average forecast level, while the 5th and 95th percentile curves provide the lower and upper limits of load fluctuations, respectively, visually illustrating the load uncertainty caused by the randomness of commuting demand.

[0142] Understandably, these curves are the output results of the modeling scheme proposed in this application, which integrates multidimensional decision-making, Monte Carlo sampling, cumulative prospect theory and nested selection model. They confirm that the scheme can accurately predict the probability distribution of charging load in time and space, and provide data support for charging facility planning and grid dispatch.

[0143] Based on the methods in the above embodiments, please refer to Figure 8This application provides an electronic device that may include a processor, a communications interface, a memory, and a communication bus. The processor, communications interface, and memory communicate with each other via the communication bus. The processor can invoke logical instructions stored in the memory to execute the methods described in the above embodiments.

[0144] Furthermore, the logical instructions in the aforementioned memory can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application.

[0145] Based on the methods in the above embodiments, this application provides a computer-readable storage medium storing a computer program that, when run on a processor, causes the processor to execute the methods in the above embodiments.

[0146] Based on the methods in the above embodiments, this application provides a computer program product that, when run on a processor, causes the processor to execute the methods in the above embodiments.

[0147] It is understood that the processor in the embodiments of this application can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. A general-purpose processor can be a microprocessor or any conventional processor.

[0148] The method steps in this application embodiment can be implemented in hardware or by a processor executing software instructions. The software instructions can consist of corresponding software modules, which can be stored in random access memory (RAM), flash memory, read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), registers, hard disks, portable hard disks, CD-ROMs, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor, enabling the processor to read information from and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can reside in an ASIC.

[0149] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted through the computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic) or wireless means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)).

[0150] It is understood that the various numerical designations used in the embodiments of this application are merely for descriptive convenience and are not intended to limit the scope of the embodiments of this application. Those skilled in the art will readily understand that the above descriptions are merely preferred embodiments of this application and are not intended to limit this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for predicting the probabilistic spatiotemporal distribution of charging load considering stochastic fluctuations in commuting demand, characterized in that, include: Step S10: For heterogeneous vehicles with multiple origin-endpoint pairs in the transportation network, generate a set of travel strategies including charging amount, departure time and travel route; and construct a dynamic model of the transportation network based at least on the commuting supply and demand balance, traffic flow dynamics of roads and charging stations, road travel time and charging station queuing time. Step S20: Based on the probability distribution of the travel strategy set under commuting demand, construct multiple commuting demand sample points for origin-destination pairs using Monte Carlo sampling. Step S30: Using the commuting demand sample points and the transportation network dynamic model, obtain the destination state parameters of each commuting demand sample point under the current travel strategy selection ratio as an evaluation index, and statistically analyze the discrete probability distribution of the evaluation index. Step S40: Construct a travel strategy utility evaluation model based on cumulative prospect theory, and calculate the actual utility of each travel strategy using the discrete probability distribution of the evaluation index. The travel strategy utility evaluation model is based on cumulative prospect theory. It sets psychological reference points for each destination state parameter, compares the evaluation index with the psychological reference points and distinguishes them as perceived gains or losses, and then combines the weights after subjective correction of the objective discrete probability distribution to calculate the actual utility model. Step S50: Based on the actual utility, update the user's travel strategy selection ratio using a multi-layer nested discrete choice model, and iterate by jumping to step S30 based on a heuristic algorithm until the current travel strategy selection ratio converges; then input the converged travel strategy selection ratio and multiple commuting demand sample points into the traffic network dynamic model, and statistically output the probability spatiotemporal distribution of charging load. The multi-layered nested discrete choice model decomposes the choice of travel strategy into three levels of sequential decision-making: charging amount, departure time, and travel route. It uses actual utility as the basis for decision-making and updates the selection ratio by calculating the product of conditional probabilities at each level.

2. The method for predicting the probabilistic spatiotemporal distribution of charging load considering random fluctuations in commuting demand as described in claim 1, characterized in that, The generation of the travel strategy set in step S10 specifically includes: The vehicles at each starting point-to-end point pair are divided into fuel-powered vehicles and multiple types of electric vehicles with different initial states of charge. For each type of electric vehicle and transportation network, a discrete set of charging quantity options, a set of departure time options, and a set of travel path options including those that do not pass through a fast charging station and those that pass through a fast charging station are constructed respectively. The charging quantity selection set, departure time selection set, and travel route selection set are combined to generate a travel strategy set that includes charging quantity, departure time, and travel route.

3. The method for predicting the probabilistic spatiotemporal distribution of charging load considering random fluctuations in commuting demand as described in claim 1, characterized in that, Step S10, which involves constructing a dynamic model of the transportation network, specifically includes: Based on the commuting supply and demand balance equation and the current travel strategy selection ratio, the total commuting demand of each origin-destination pair of various types of vehicles is allocated to their corresponding travel strategies to obtain the dynamic number of vehicles for each travel strategy. Based on the dynamic vehicle count, correlation matrix, and inbound traffic flow equation for each travel strategy, the number of vehicles entering each road and each fast charging station at each time is calculated; then, based on the outbound traffic flow equation, the number of vehicles leaving each road and each fast charging station at each time is calculated according to the time of entry of the vehicles, dynamic passage time, dynamic queuing time, and charging time. Based on the vehicle count balance equation, the number of vehicles on the road and in the fast charging station is dynamically updated according to the number of vehicles entering and leaving. Based on the updated number of vehicles on the road, the dynamic road travel time is calculated using a preset road impedance function; and based on the updated number of vehicles at fast charging stations, the dynamic charging queuing time is calculated using a preset queuing function.

4. The method for predicting the probabilistic spatiotemporal distribution of charging load considering random fluctuations in commuting demand as described in claim 1, characterized in that, Step S20 includes: When the commuting demand of each origin-destination pair for each type of vehicle is a random variable that follows a normal distribution, Monte Carlo sampling is used to generate a commuting demand sample set containing multiple sample points based on the normal distribution. Each sample point contains a specific commuting demand value for each origin-destination pair for each type of vehicle.

5. The method for predicting the probabilistic spatiotemporal distribution of charging load considering random fluctuations in commuting demand as described in claim 1, characterized in that, Step S30 includes: By combining the data from each commuting demand sample point with the current travel strategy selection ratio, the destination state parameters corresponding to each travel strategy under that sample point are calculated through the dynamic model of the transportation network. Iterate through all commuting demand sample points to obtain multiple sets of destination state parameter values ​​for each travel strategy; For each destination state parameter of each travel strategy, its multiple sets of values ​​are placed into multiple value intervals with equal intervals. The number of sample points falling into each value interval is counted to obtain the objective probability, thus forming the discrete probability distribution of the destination state parameter.

6. The method for predicting the probabilistic spatiotemporal distribution of charging load considering random fluctuations in commuting demand as described in claim 1, characterized in that, Step S40 includes: Set a fixed mental reference point for the state of charge at the destination, set a mental reference point for the total charging cost based on the shortest driving distance and the price of slow charging at night, set three mental reference points for the arrival time at the destination: the earliest, the best, and the latest, and set a mental reference point for the total travel time based on the reliable travel time. For each endpoint state parameter, based on the discrete probability distribution and psychological reference point corresponding to the endpoint state parameter, it is determined whether the corresponding value range belongs to the perceived gain range or the perceived loss range. For the perceived gain interval and the perceived loss interval, different preset probability weight functions are used to perform nonlinear transformation on the objective probabilities corresponding to each interval in the discrete probability distribution to obtain subjective probability weights. For each interval, the perceived value of that interval is calculated based on the difference between the interval representative value and the psychological reference point, combined with a preset value function. The value function has different convexity and concavity and amplification factor for loss in the gain interval and loss interval. The actual utility of each travel strategy is obtained by multiplying the subjective probability weight of each interval by the corresponding perceived value and then summing the results.

7. The method for predicting the probabilistic spatiotemporal distribution of charging load considering random fluctuations in commuting demand as described in claim 1, characterized in that, The updated user's travel strategy selection ratio in step S50 includes: Based on the actual utility of each travel strategy, the perceived utility of users' charging decisions, arrival decisions, and travel decisions is obtained respectively; wherein, the perceived utility is composed of the actual utility plus a random term that follows a preset extreme value distribution; Based on the decision-making order of first selecting the charging amount, then selecting the departure time, and finally selecting the travel route, in the multi-layer nested discrete selection model, the expected maximum charging perceived utility is calculated when a certain charging amount is selected from the set of available charging amounts, and the expected maximum travel perceived utility is calculated when a certain charging amount is selected and a certain departure time is selected from the set of available departure times. The expected maximum charging perceived utility and the expected maximum passage perceived utility are used as satisfaction indicators at the decision level. Based on a multi-layer nested discrete choice model, the probability of selecting a certain charging amount in the charging amount selection set, the conditional probability of selecting a certain departure time in the departure time selection set under the selected charging amount, and the conditional probability of selecting a certain passage path in the passage path selection set under the selected charging amount and the selected departure time are calculated sequentially using the satisfaction indicators. The calculated probability is multiplied sequentially by the conditional probability to obtain the updated user's travel strategy selection ratio.

8. The method for predicting the probabilistic spatiotemporal distribution of charging load considering random fluctuations in commuting demand as described in claim 1, characterized in that, The statistical analysis and output of the probabilistic spatiotemporal distribution of charging load includes: Based on the convergent travel strategy selection ratio, the data from the commuting demand sample points are used to calculate the charging load of each sample point, at each time, and at each fast charging station through the dynamic model of the transportation network. For each time point and each fast charging station, the charging load values ​​of all sample points are placed into multiple equally spaced value intervals. The number of sample points falling into each value interval is counted to obtain the probability, thereby forming the discrete probability distribution of the charging load of that time point and that fast charging station. It gathers the discrete probability distribution of charging load at all times and all fast charging stations, and outputs the spatiotemporal probability distribution of charging load.

9. An electronic device, characterized in that, The device includes a memory and one or more processors; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code including computer instructions; the one or more processors invoke the computer instructions to cause the electronic device to perform the method as described in any one of claims 1 to 8.

10. A computer-readable storage medium comprising instructions, characterized in that, When the instructions are executed on an electronic device, the electronic device causes the electronic device to perform the method as described in any one of claims 1 to 8.

Images