Electric vehicle charging load forecasting method based on dynamic energy consumption and user psychology

Abstract

The present invention relates to a method for predicting electric vehicle charging load based on dynamic energy consumption and user psychology. The method comprises the following steps: setting the number of electric vehicles within a predicted area and a starting state of charge (SOC); constructing an electric vehicle travel model and an actual power consumption model; obtaining the optimal strategy for each electric vehicle based on the travel model using a Markov dynamic path decision model, and simulating each electric vehicle's travel process according to the corresponding optimal strategy; during the simulation, for each travel node of each electric vehicle, pre-calculating the power consumption of the next travel segment and the current SOC of the electric vehicle based on the actual power consumption model and the starting state of charge (SOC), determining the charging demand of the electric vehicle at the current travel node, and accumulating the charging demand of all electric vehicles at each travel node to obtain the spatiotemporal distribution of electric vehicle charging demand within the predicted area. Compared with existing technologies, the present invention has the advantages of high accuracy and strong reliability.

Description

Technical Field

[0001] The present invention relates to the field of power systems, and in particular to a method for predicting charging load of electric vehicles based on dynamic energy consumption and user psychology. Background Art

[0002] Due to the low-carbon and environmentally friendly advantages of electric vehicles, the transformation of the automotive industry is an important way to achieve energy conservation and emission reduction. Promoting the development of the new energy vehicle industry is a strategic measure to address climate change and promote green development. Large-scale electric vehicle access will affect the reliability of the distribution network. A refined electric vehicle charging load model can ensure the accuracy of reliability assessment. As a special transferable load and energy storage device, electric vehicles are the carrier connecting road network traffic and urban distribution networks. Their inherent mobility and randomness of spatial transfer are affected by both user travel characteristics and the objective urban road structure. Therefore, establishing a refined electric vehicle charging load prediction model to analyze the spatiotemporal distribution characteristics of the charging load is the key to studying the impact of electric vehicle charging load on the distribution network.

[0003] Based on the analysis of the existing research status, there are three deficiencies in the prediction of electric vehicle charging load:

[0004] First, the prediction model often treats battery capacity as a fixed parameter and ignores the changes in EV energy consumption with the environment, which introduces errors in the charging load prediction results.

[0005] Second, there is a lack of detailed research on user subjective intentions. During actual travel, users are influenced by both real-time road conditions and user psychology, and their route selection may not always follow the established shortest route. Therefore, route selection for electric vehicle travel should not be limited to the shortest path algorithm; further detailed analysis of user route decisions is needed.

[0006] 3. It focuses on the user's subjective willingness to charge, weakens the influence of the user's psychology, does not consider the user's own economic level, consumption capacity and other user attributes, and lacks quantitative analysis. Summary of the Invention

[0007] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and provide an electric vehicle charging load prediction method based on dynamic energy consumption and user psychology, which has high accuracy and strong reliability.

[0008] The purpose of the present invention can be achieved by the following technical solutions:

[0009] A method for predicting electric vehicle charging load based on dynamic energy consumption and user psychology, comprising:

[0010] Set the number of electric vehicles in the area to be predicted and the starting SOC of each electric vehicle;

[0011] Build a travel model and actual power consumption model for electric vehicles;

[0012] According to the travel model of electric vehicles, the optimal strategy of each electric vehicle is obtained through the Markov dynamic path decision model, and the travel process of each electric vehicle is simulated according to the corresponding optimal strategy;

[0013] During the simulation process, for each travel road node of each electric vehicle, the power consumption of the electric vehicle in the next travel process and the current SOC are calculated in advance based on the actual power consumption model and the starting SOC. The charging demand of the electric vehicle at the current travel road node is determined based on the power consumption of the next travel process and the current SOC. The charging demand of all electric vehicles at each travel road node is accumulated to obtain the spatiotemporal distribution of the charging demand of electric vehicles in the predicted area.

[0014] Furthermore, the actual power consumption model is expressed as:

[0015]

[0016] Among them, E all,T is the total energy consumption of electric vehicles at ambient temperature T, K ac is the start / stop state of the vehicle air conditioner of the electric vehicle under the ambient temperature T, and the K ac The value of K is 0 or 1. ac =1 means the vehicle air conditioner is turned on, the K ac =0 means the car air conditioner is off, P ac is the onboard air conditioning power of the electric vehicle at ambient temperature T, n is the total number of segments of the electric vehicle’s travel process, t k and D k are the driving time and mileage during the k-th trip, ξ t is the driving time correction coefficient, E v E is the energy consumption per unit mileage generated by an electric vehicle traveling at speed v. o is the initial energy consumption per unit mileage of electric vehicles;

[0017] Furthermore, the vehicle air conditioner start / stop state K ac The calculation process includes:

[0018] Set the cold threshold T c , thermal threshold T h And the comfortable temperature range, the T h >T c , the comfortable temperature range is between T c and T hThe normal distribution function of air conditioner startup is obtained by normal distribution fitting;

[0019] When the ambient temperature T is lower than the cold threshold T c or greater than the thermal threshold T h If so, let K ac =1, otherwise let K ac =0, when the ambient temperature T is between T c and the comfortable temperature range, or between the comfortable temperature range and T h When the air conditioner starts normal distribution function, K is determined ac Get the value.

[0020] Furthermore, the vehicle air conditioning power P ac The calculation formula is:

[0021] P ac (T) = ω3T 3 +ω210 -5 T 2 +ω1T+ω0

[0022] Where T is the ambient temperature, and ω0, ω1, ω2, and ω3 are function fitting coefficients.

[0023] Furthermore, the driving time correction coefficient ξ t The acquisition process includes:

[0024] Collect the traffic congestion index TPI of each road in the predicted area, and obtain ξ based on TPI t ;

[0025] Among them, ξ t ∈[0,1], the higher the TPI, the t The bigger.

[0026] Furthermore, the energy consumption per unit mileage E v The acquisition process includes:

[0027] The roads in the predicted area are divided into three levels: expressways, main roads and secondary roads. Each level of road corresponds to an average driving speed.

[0028] When simulating an electric vehicle traveling on a road, the electric vehicle travels at an average speed corresponding to the grade of the road.

[0029] Furthermore, it is characterized in that the electric vehicles are divided into household electric vehicles and rental electric vehicles;

[0030] The process of constructing the travel model of the household electric vehicle includes:

[0031] Based on the travel chain theory, a travel chain of household electric vehicles is constructed. According to the travel data in the NHTS2017 database, the variation pattern of the characteristic quantities of each node in the travel chain is obtained through probability distribution fitting analysis, thus forming a travel model for household electric vehicles.

[0032] The construction process of the travel model for renting electric vehicles includes:

[0033] Based on graph theory, roads are abstracted into a road network topology graph. According to travel order data, the OD analysis method is used to obtain the travel starting point distribution and probability transfer matrix of rental electric vehicles, thus forming a travel model for rental electric vehicles.

[0034] Furthermore, the charging demand determination process includes the following steps:

[0035] S801, judging whether the current SOC of the electric vehicle meets the travel demand based on the power consumption of the next travel process, if so, executing step S802, otherwise executing step S803;

[0036] S802: Determine whether the parking time of the electric vehicle at the current travel road node is longer than the time required for slow charging. If so, it is determined that the electric vehicle has a slow charging requirement; otherwise, the electric vehicle has a fast charging requirement, and the process ends.

[0037] S803: Determine whether the expected electricity price of the electric vehicle user is higher than the real-time electricity price. If so, determine that the electric vehicle has a slow charging demand; otherwise, determine that the electric vehicle has no charging demand.

[0038] Furthermore, the calculation formula for the battery capacity of the electric vehicle is:

[0039] C R (T) = η3T 3 +η210 -5 T 2 +η1T+η0

[0040] Where T is the ambient temperature, C R (T) is the relative capacity percentage of the battery of the electric vehicle when the ambient temperature is T, and η0, η1, η2 and η3 are fitting coefficients.

[0041] Furthermore, it is assumed that the initial SOC of all electric vehicles obeys a normal distribution.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] (1) The present invention determines the charging demand of the electric vehicle at the current travel road node based on the power consumption of the next travel process and the current SOC. The actual power consumption model of the electric vehicle comprehensively considers the impact of ambient temperature on battery capacity, the impact of ambient temperature on the energy consumption of vehicle air conditioning start-stop, and the impact of traffic conditions on the mileage energy consumption of the electric vehicle, making the electric vehicle travel simulation process more realistic. A Markov dynamic path decision model based on the optimal strategy is established, so that the path selection of electric vehicle users is not limited to a single shortest path, and the charging demand of electric vehicles under dynamic commuting can be more accurately simulated, thereby improving the accuracy and reliability of electric vehicle charging load prediction;

[0044] (2) The present invention establishes a corresponding air-conditioning activation decision model by analyzing the vehicle air-conditioning usage rate curve, performs normal distribution fitting based on the vehicle air-conditioning usage rate curve, sets cold thresholds, hot thresholds, and comfortable temperature ranges, and quantifies the impact of intraday temperature changes on air-conditioning energy consumption in charging load simulation through a piecewise polynomial model. This model can accurately reflect the impact of different ambient temperatures on vehicle air-conditioning energy consumption, improve the accuracy of energy consumption simulation during electric vehicle travel, and thus improve the accuracy and reliability of electric vehicle charging load prediction.

[0045] (3) The present invention introduces a driving duration correction coefficient ξ t Adjust driving time under different traffic conditions, driving time correction coefficient ξ t According to the traffic congestion index TPI, the higher the TPI, the higher the ξ t The larger the value, the traffic congestion index TPI uses numerical methods to quantitatively describe the road traffic operation status, improve the accuracy of the expression of road traffic operation status, take into account the impact of traffic congestion on user choice intention, make the simulation of electric vehicle travel more realistic, and thus improve the accuracy and reliability of electric vehicle charging load prediction;

[0046] (4) The present invention divides the roads in the predicted area into three levels: expressways, main roads, and secondary roads. Each level of road corresponds to an average driving speed. The energy consumption per unit mileage of electric vehicles takes into account the influence of road level, which is more realistic.

[0047] (5) The present invention divides electric vehicles into household electric vehicles and rental electric vehicles. The starting location of household electric vehicles is relatively fixed and the travel has a clear purpose, which is suitable for the travel chain theory. The travel destination and the number of travel orders of rental electric vehicles are determined by the travel needs of passengers. Therefore, based on the travel order data, the OD analysis method is used to conduct data mining on the travel space location distribution of rental electric vehicles, thereby improving the accuracy and reliability of the electric vehicle charging load prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1A graph showing the energy consumption and usage rate of the vehicle air conditioner relative to temperature;

[0049] Figure 2 Schematic diagram of the user's charging decision-making process;

[0050] Figure 3 Calculate the flow chart for charging demand;

[0051] Figure 4 Provide a spatiotemporal distribution map of electric vehicle charging load in residential areas;

[0052] Figure 5 Provide a spatiotemporal distribution map of the electric vehicle charging load in the work area;

[0053] Figure 6 Provide a spatiotemporal distribution map of electric vehicle charging load in the commercial area;

[0054] Figure 7 It is the time distribution diagram of the total charging load of electric vehicles;

[0055] Figure 8 This is the time distribution diagram of electric vehicle charging load based on the DEC model in summer;

[0056] Figure 9 This is the time distribution diagram of electric vehicle charging load based on the DEC model in winter;

[0057] Figure 10 This is the time distribution graph of daily order quantity;

[0058] Figure 11 This is a graph showing the change in relative capacity of lithium batteries with temperature;

[0059] Figure 12 This is a typical temperature change curve for summer and winter. DETAILED DESCRIPTION

[0060] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0061] A method for predicting electric vehicle charging load based on dynamic energy consumption and user psychology, comprising:

[0062] Firstly, the travel characteristics of household electric vehicles and taxis are analyzed, and the travel processes of different types of electric vehicles are simulated using the Monte Carlo method.

[0063] A penalty coefficient is introduced according to the traffic congestion index, and a Markov dynamic path decision model based on the optimal strategy is established, so that the user's path selection is not limited to a single shortest path;

[0064] Establish a travel mileage energy consumption model for electric vehicles under different temperatures and road conditions, and calculate the real-time dynamic energy consumption per unit mileage under different scenarios;

[0065] Introducing the anchoring effect to analyze user psychology and establishing a user charging decision model based on the anchoring effect;

[0066] The constructed model is verified through the urban road network to predict the spatiotemporal distribution characteristics of electric vehicle charging demand.

[0067] 1. Spatial and temporal characteristics of electric vehicle travel

[0068] 11) Electric vehicle travel model

[0069] The random nature of electric vehicle travel and charging behavior has an uncertain impact on the spatiotemporal characteristics of charging loads. Furthermore, vehicle travel paths also exhibit dynamic uncertainty, further complicating the interactions between these variables. However, electric vehicle travel flows exhibit a certain regularity and periodicity. By considering urban attributes and travel demand in different spatiotemporal scenarios, the temporal and spatial correlation of the probability distribution function of vehicle travel-related variables can be improved. Based on the functional characteristics and geographic information of different urban areas, cities are divided into residential areas (H), work areas (W), commercial areas (C), and other areas (O).

[0070] Household electric vehicles have a relatively fixed starting point and a clear purpose for their trips. Connecting different trip purposes in chronological order forms a household electric vehicle trip chain. Based on trip chain theory, we modeled users' daily travel patterns and categorized the travel chain pattern into simple and complex chains based on the number of stops during travel. Each node in the travel chain contains travel characteristics such as the start time, stop duration, and stop time. Information between two nodes includes the user's mileage and travel time during the trip. By fitting the travel data in the NHTS2017 database with probability distributions, we selected the best-fitting probability function to represent the variation pattern of the characteristic quantity. The probability distribution and parameter fitting results of the characteristic quantity are shown in Table 1:

[0071] Table 1 Electric vehicle travel characteristic parameter settings

[0072]

[0073] Based on the desensitized travel data provided by the GAIA program, the travel characteristics of electric vehicle rental are analyzed. Since the purpose of electric vehicle rental is to provide convenient travel services for urban residents, the travel destination and the number of travel orders are determined by the travel needs of passengers, such as Figure 10As shown in the figure, we select the order data within the third ring road of Chengdu, China to analyze the demand for taxi orders and their time distribution within a day.

[0074] Data mining is conducted on the spatial location distribution of rental electric vehicle travel. The starting and ending locations are screened from the rental electric vehicle travel order data. The OD analysis method is used to obtain the distribution of the starting point and the probability transfer matrix of the rental electric vehicle travel. Based on the graph theory method, the real road is abstracted into a road network topology map. The OD analysis method is used to determine the distribution of the starting and ending points of the rental electric vehicle travel. The user travel OD probability transfer matrix is ​​obtained by the pick-up and drop-off locations and matching time periods in the actual order itinerary. The corresponding electric vehicle road speed is obtained according to the traffic conditions in different time periods. The time required for the rental electric vehicle to travel the route is calculated. The road weight in the road adjacency matrix is ​​replaced by the travel time, as shown in Formula (1) and Formula (2):

[0075]

[0076]

[0077] Among them, P(G) is the set of traffic nodes in the road network, the number of nodes is k, and E(G) is the relationship between any two nodes p in the road network. i and p j The set of driving roads, ψ G is the road travel time adjacency matrix, which expresses the travel time t of each node and road section ij The correspondence between them.

[0078] 12) Markov dynamic path decision model based on optimal strategy

[0079] The Markov decision model consists of five key elements: a set of decision moments Γ, a set of states S, a set of actions A, state transition probabilities p, and a payoff objective function R. At a given decision moment t (t∈Γ), the decision maker takes action a (a∈A(i)) in state i (i∈Γ). Choosing action a yields a payoff R(i,a). At the next decision moment t+1 (t+1∈Γ), the state of the system is determined by the probability distribution p(j|i,a). These five sets {Γ, S, A(i), p(j|i,a), R(i,a)} are denoted as a Markov decision process, where the transition probabilities and payoffs depend solely on the current state and the actions taken by the decision maker, and are independent of past history.

[0080] For the Markov decision problem with a finite number of discrete-time decision moments, the decision moment set Γ = {0, 1, ..., N-1} is selected, 0 < N < ∞. No decision behavior is required in the last decision stage N. The sequence of decision rules adopted when making decisions is recorded as strategy π. The set of all strategies is called the strategy set, recorded as Π. The decision maker will obtain a series of benefits when making decisions at each moment. The accumulated benefits are recorded as the utility function V of the decision model N , as shown in formula (3):

[0081]

[0082] Among them, S(t) and A(t) are the state and action at the decision time t, respectively. R(S(t), A(t)) is the benefit obtained at time t. R(S(N), A(N)) is the terminal benefit of the process. Although the decision process has ended and the decision maker does not need to make any more decisions, the system surplus value can still be obtained.

[0083] If a strategy π is selected from the strategy set π so that the utility function V N is the optimal function, then the strategy π is selected as the optimal strategy solution. Based on the Bellman optimization principle, it can be seen that when any action is selected from the optimal action set in any state, the strategy set composed of the remaining decision rule sequence is still the optimal strategy for the next decision moment. Therefore, the utility function V of the decision model can be obtained by backward recursion of the benefits. N , its recursive function is shown in formula (4):

[0084]

[0085] Among them, j is the state at the next decision time t+1.

[0086] When choosing a road, users give priority to choosing sections with smooth traffic. For sections with higher traffic congestion index values, the lower the driving speed and the longer the driving time, the lower the user's willingness to choose. In order to ensure that users can avoid congested sections when traveling, and considering that local information asymmetry, unfamiliarity with road conditions, and psychological factors may affect the user's decision-making process, the penalty coefficient λ is introduced so that the user's decision is not limited to a single shortest path, as shown in formula (5):

[0087]

[0088] Among them, λ k,t is the penalty coefficient of road k at decision time t, I tpk,t is the traffic congestion index at decision time t, I tp,max is the maximum value of the traffic congestion index of the reference road section.

[0089] The steps for users to make travel decisions are as follows:

[0090] Before setting off, the user selects the travel destination and records the planned shortest travel path as π d , each section in the shortest travel path is recorded as a reference section l d , the set of paths that are not selected in the network nodes is selected as the backup path set, denoted as A k ;

[0091] When reaching decision state i, i.e. road node i, the user will take reference road segment l d The length of the road section L d , Traffic Congestion Index I tp,d and the penalty coefficient λ t,d For reference, the traffic congestion index of the alternative paths is higher than I tp,d road sections;

[0092] The alternative sections in the alternative path set are divided into sections with a length of L k and the reference section length L d Compare and calculate the impact of each road section's penalty coefficient on the state transition probability of the road section.

[0093] The state transition probability p of each road section t As shown in formula (6):

[0094]

[0095] Among them, j k =l d Indicates the state transition probability of the decision action to select the shortest path according to the original plan, j k ∈A k Indicates the state transition probability of selecting an alternative road segment for decision action.

[0096] (2) Dynamic Energy Consumption (DEC) Model of Electric Vehicles

[0097] 21) Battery capacity model

[0098] During normal driving, electric vehicles are affected by external factors such as road grade, traffic conditions, speed, and temperature, which in turn influence charging load requirements and daily conditions. By incorporating this information into the impact of EV range energy consumption, a refined EV energy consumption per mile model was developed. The capacity and charge-discharge characteristics of the same battery vary significantly at different temperatures. Therefore, using experimental data from currently mainstream lithium batteries as an example, a relationship between temperature and battery capacity was established.

[0099] The lithium battery test data from NASA's Ames Research Center is used to analyze the effects of different temperatures and charge-discharge cycles on battery capacity. In the lithium iron phosphate battery test, the battery capacity at 25°C is used as the benchmark to analyze the relative capacity percentage (CR%) of the battery at different temperatures. Figure 11 As shown. In a high temperature environment, CR% increases slightly, but the increase is not significant. At 40°C, CR% is 106%. When the ambient temperature is greater than 55°C, the CR% upward trend turns to a downward trend. In a low temperature environment, CR% decreases significantly relative to the baseline capacity. When the temperature reaches -20°C, CR% is only 43%. In order to quantify the effect of temperature on battery capacity during the charging load simulation, a polynomial function model is used to fit the relationship between temperature and CR%, as shown in formula (7):

[0100] C R (T) = η3T 3 +η210 -5 T 2 +η1T+η0 (6)

[0101] Where: T is the ambient temperature (℃), η0, η1, η2, η3 are the function model fitting coefficients, C R It is the percentage of relative capacity of the battery.

[0102] 22) Energy consumption model of vehicle air conditioning start-stop

[0103] As the main energy-consuming appliance of electric vehicles, the vehicle air conditioning system (including cooling and heating) directly affects the charging demand and driving range of electric vehicles. The start and stop and energy consumption values ​​of the air conditioning system will also vary under different weather conditions such as ambient temperature and humidity. Figure 1 As shown in Figure 3, the energy consumption of the vehicle air conditioner under different ambient temperatures T is analyzed based on the experimental data, and the energy consumption data when the vehicle air conditioner is turned on is fitted to obtain the changes in air conditioner energy consumption and air conditioner usage rate with ambient temperature.

[0104] The corresponding air-conditioning activation decision model is established by analyzing the vehicle air-conditioning usage rate curve, and a normal distribution fitting is performed according to the vehicle air-conditioning usage rate curve, and the cold threshold T is set. c , thermal threshold T h As well as the comfortable temperature range, the segmented curve between the threshold and the comfortable temperature is fitted with the normal distribution according to the cftool toolbox in MATLAB to obtain the corresponding fitting parameters and the corresponding air conditioning startup normal distribution function, as shown in formula (8):

[0105]

[0106] Among them, K open (T) is the probability of air conditioning turning on at temperature T, μ c , δ c is the mean and variance of the heating startup parameters, μ h , δ h is the mean and variance of the cooling startup parameters. When the temperature is lower than the cold threshold T c or greater than the thermal threshold T h When the temperature is within the comfortable temperature range, the probability of turning on the air conditioner is 1. When the temperature is within the comfortable temperature range, the probability of turning on the air conditioner is 0. The parameter value settings are shown in Table 2.

[0107] Table 2 Fitting parameters of vehicle air conditioning on probability

[0108]

[0109] In order to quantify the impact of intraday temperature changes on air conditioning energy consumption in charging load simulation, a piecewise polynomial model can be used to fit the relationship between temperature and air conditioning power, as shown in formula (9):

[0110] P ac (T) = ω3T 3 +ω210 -5 T 2 +ω1T+ω0 (9)

[0111] Among them, P ac is the air conditioning power, ω0, ω1, ω2 and ω3 are function fitting coefficients.

[0112] 23)2.3 Mileage energy consumption model based on traffic conditions

[0113] During normal driving, the energy consumption per mile of electric vehicles is most significantly affected by different road grades and traffic conditions. To address the inconsistency between vehicle speeds and perceived traffic congestion across various road networks, and to fully incorporate people's tolerance for different levels of road congestion, the Transport Performance Index (TPI) was introduced to numerically quantify road traffic operating conditions and improve the accuracy of this representation. Because vehicles travel at uneconomical speeds, traffic congestion increases driving duration and air conditioning service time, reducing driving efficiency. This significantly impacts energy consumption and charging requirements, and thus the driving duration correction factor ζ was introduced. t Adjust the driving time under different traffic conditions. When choosing a road, users give priority to sections with smooth traffic. For sections with higher traffic congestion index values, the lower the driving speed and the longer the driving time, the lower the user's willingness to choose, as shown in Table 3:

[0114] Table 3 Classification of road traffic congestion index

[0115]

[0116] Roads are divided into three levels: expressways, main roads, and secondary roads. The overall traffic condition changes over time are obtained through real-time traffic condition surveys. The corresponding average speed is selected according to the TPI in different time periods and substituted into the traffic energy consumption factor model, as shown in formula (10):

[0117]

[0118] Among them, E v It represents the energy consumption per unit mileage of an electric vehicle when it travels at a speed v on roads of different grades. The unit is kWh / km. v1, v2 and v3 are the driving speeds of three road grades: expressway, main road and secondary road respectively. The speed is the average driving speed on each road section. The unit is km / h.

[0119] The selection of driving speeds for different levels of road under different traffic congestion levels is shown in Table 4:

[0120] Table 4 Driving speeds on different levels of roads at different levels of traffic congestion

[0121]

[0122] In summary, the actual power consumption of electric vehicles under different ambient temperatures T and traffic conditions can be accurately calculated by the battery energy consumption model, as shown in formula (11):

[0123]

[0124] Among them, E all,T is the total energy consumption of electric vehicles at ambient temperature T, K ac The air conditioner start / stop status under the ambient temperature T, with a value of 1 or 0 (indicating that the car is in the air conditioner on / off state), t k 、D k is the driving time and mileage during the k-th trip, E o is the initial energy consumption per unit mileage of electric vehicles, ξ t is the driving time correction coefficient, and its value is shown in Table 3, indicating that the driving time and speed of electric vehicles are corrected when they are affected by traffic congestion.

[0125] (3) Electric vehicle charging demand model based on user psychology

[0126] 31) User charging decision based on anchoring effect

[0127] The anchoring effect refers to the deviation caused by the final estimated result of an individual making a judgment in an uncertain situation approaching the initial value. When users make charging decisions, the time-of-use electricity price will become an "anchor" and affect the user's charging decision behavior. Different anchor values ​​will have different effects on the user's charging willingness

[20] . Since the user's charging behavior is regulated by the grid's time-of-use electricity price, the charging load of electric vehicles is affected by increasing the peak electricity price and reducing the valley electricity price, achieving the effect of "peak shaving and valley filling". The anchoring effect judges the user's acceptance of the electricity price by defining the concepts of "high anchor" and "low anchor". Among them, "high anchor" means that the user's expectation of the electricity price is higher than the actual price, and "low anchor" means that the user's expectation of the electricity price is lower than the actual price. In the charging decision process, different "anchors" will have different effects on the user's charging willingness. Compared with low anchors, high anchors can generate higher charging willingness. Consumer surplus is the difference between the highest electricity price willing to pay and the actual time-of-use electricity price. During the charging decision-making process, users compare the real-time electricity price with the expected electricity price and make charging decisions based on their perception of consumer surplus. The user charging decision-making process is as follows: Figure 2 shown.

[0128] When the EV's state of charge (SOC) cannot meet the power requirements for the next trip, users will select a charging mode based on the length of parking time. Users with ample parking time are defined as "normal rigid users" and choose slow charging. Users whose slow charging cannot meet their next trip needs within the parking time are defined as "emergency rigid users" and choose fast charging. When the EV's SOC is sufficient and meets the power requirements for the next trip, the charging decision is determined by the consumer surplus generated by different anchor value settings. Users whose maximum willingness to pay is higher than the real-time electricity price are defined as high-anchor elastic users, and based on their consumer surplus, they will charge. Users whose real-time electricity price is higher than the user's expected electricity price are defined as low-anchor elastic users, and since they have no surplus, they will not charge. For battery life considerations, elastic users prefer slow charging when charging at home.

[0129] 32) Charging demand calculation

[0130] Based on the travel characteristics of different user types and vehicle types, the charging behavior of electric vehicles is judged at each road node, and the charging demand in different functional areas is superimposed to obtain the spatiotemporal distribution of electric vehicle charging demand, such as Figure 3 As shown in Figure 2, the specific simulation calculation process is as follows:

[0131] Set parameters related to temperature, traffic conditions, and different types of electric vehicle parameters;

[0132] Calculate and update battery capacity based on the number of EV charge and discharge cycles and ambient temperature;

[0133] Based on the trip chain theory, the travel of household electric vehicles is simulated, and the Monte Carlo method is used to randomly extract the travel chain type, time parameters, and travel destinations. According to the spatiotemporal distribution of the daily order quantity and the OD probability transfer matrix, the Monte Carlo method is used to randomly extract the daily trip frequency and trip origin and destination points.

[0134] The driving route is determined through the MDP path decision model with the optimal strategy. At each travel road node, the power consumption generated by the next travel process is pre-calculated, and the user's charging decision is judged based on whether the remaining power of the electric vehicle can meet the power demand of the next travel segment.

[0135] The charging demand in each region is accumulated according to the number of electric vehicles to obtain the total spatiotemporal distribution of electric vehicle charging demand.

[0136] (IV) Example analysis

[0137] 41) Parameter setting

[0138] The number of electric vehicles of various types and the proportion of different types of household electric vehicles and rental electric vehicles are set as shown in Table 5:

[0139] Table 5 Number and parameters of various types of electric vehicles

[0140]

[0141] The SOC of electric vehicles at the start of travel follows a normal distribution N(0.8, 0.12). The charging modes of the charging station are fast charging and slow charging. According to the types and location distribution of electric vehicle charging piles promoted by the country, the charging mode and charging power of different functional areas are set. The charging mode and charging power of each area are shown in Table 6:

[0142] Table 6 Charging mode and charging power in each area

[0143]

[0144] The types and proportions of household electric vehicle travel chains are shown in Table 7:

[0145] Table 7 Types and proportions of travel activities in the travel chain of household electric vehicles

[0146]

[0147]

[0148] The electric vehicle travel road network model is formed with reference to the road network within the Chengdu Third Ring Road area.

[0149] 42) Simulation results

[0150] 421) Temporal and spatial distribution of electric vehicle charging load

[0151] like Figure 4 、 Figure 5 and Figure 6 As shown, the charging load characteristics of electric vehicles vary considerably across functional areas. Taking weekday charging load as an example, during the day, the charging load is primarily concentrated in work areas, with a high proportion of load, peaking at 10:27 AM and reaching a peak of 8.91 MW. At night, the charging load is concentrated in residential areas, with a higher concentration between 4:00 PM and 8:00 PM, reaching a peak of 4.02 MW. Charging load in commercial areas is generally lower on weekdays, with a peak load of 0.23 MW occurring between 12:00 PM and 6:00 PM.

[0152] Comparing the total charging load curves on weekdays and weekends, the total charging load time distribution of electric vehicles in each scenario is as follows: Figure 7 As shown. The total charging load on weekdays gradually increases from 5:18 and reaches a peak of 10.46MW at 12:21. Compared with weekdays, the charging load is more concentrated on weekends, with the peak load delayed to 14:16, reaching a peak load of 15.61MW. The peak load increased by 49.25% compared with weekdays. Since the travel peak on weekdays occurs during the rush hour, while the travel peak on weekends is concentrated between 12:00 and 17:00, the time periods for charging load increase are different. Users travel more frequently on weekends and their travel is concentrated during the travel peak. At the same time, the number of rental car orders increases on weekends, the daytime travel volume increases, and the overall charging demand increases.

[0153] 422) Impact of dynamic energy consumption model on charging load

[0154] like Figure 12 As shown in Figure 2, the typical temperature change curves of Chengdu in summer and winter are selected. Based on the influence of traffic conditions and temperature factors, the results of electric vehicle charging load in summer and winter before and after the introduction of the DEC model are determined, and the energy consumption proportion caused by different factors is obtained, as shown in Figure 2. Figure 8 and Figure 9 As shown in Table 8, the impact of the dynamic energy consumption model on the charging demand of electric vehicles is compared and analyzed. The specific values ​​of the additional energy consumption caused by each factor in the DEC model are shown in Table 8:

[0155] Table 8 Changes in electric vehicle charging load after the introduction of the DEC model

[0156]

[0157]

[0158] Combine Figure 8 、 Figure 9 As shown in Table 8, road conditions and temperature significantly impact EV charging load in hot summer or cold winter environments. For example, during summer weekdays, peak travel time is from 8:00 AM to 10:00 AM, when traffic congestion is high. This increases EV mileage energy consumption, leading to a surge in EV charging demand. During the nighttime, when traffic is smooth, energy consumption due to traffic conditions is significantly lower. After introducing the DEC model, the additional charging load caused by traffic conditions during peak travel time accounts for 28.17% of the total charging load. As summer temperatures gradually increase, energy consumption due to temperature increases during the hot weather period from 10:00 AM to 3:00 PM, with a maximum increase of 13.09%. Analyzing different seasons, the peak charging load and total charging demand in winter are higher than in summer due to higher vehicle air conditioning usage and reduced battery capacity in cold weather, resulting in more frequent charging. The daily charging load in winter due to temperature increases accounts for 21.17% of the total charging load. Simulation results validate the DEC model and align with actual conditions.

[0159] (V) Conclusion

[0160] The spatiotemporal distribution of electric vehicle charging load is affected by multiple objective factors and user psychology. This example proposes an electric vehicle charging load prediction model based on a dynamic energy consumption model and user psychology. The spatiotemporal distribution of electric vehicle charging load is predicted based on the actual road network traffic within the Third Ring Road of Chengdu. The simulation results show that:

[0161] 51) Charging region, charging preferences, and time scenario type affect the size, peak timing, and shape of the charging load curve. The temporal distribution of EV charging loads varies significantly across regions. The charging load curve is relatively flat on weekdays, while it is more concentrated on weekends, with peak loads increasing by 49.25% compared to weekdays.

[0162] 52) The characteristics and amplitudes of the electric vehicle charging load curves differ significantly when factoring in external factors such as road grade, traffic conditions, vehicle speed, and temperature. The peak load difference is as high as 21.17%, and the peak timing shifts. Battery capacity decreases significantly in winter due to temperature. The impact of low temperatures on charging load is even greater in winter, and the frequent use of vehicle air conditioning leads to a continuous increase in peak load and a longer duration of peak load. The total charging demand in winter is much higher than in summer.

[0163] The simulation results verify the effectiveness of the model and the reliability of the method. The dynamic energy consumption model can accurately obtain the actual energy consumption of electric vehicles under different environmental conditions, improve the accuracy of load forecasting results, and provide a reasonable planning basis for the layout and configuration of the distribution network.

[0164] This embodiment proposes a method for predicting electric vehicle charging load based on dynamic energy consumption and user psychology. First, the travel characteristics of household electric vehicles and electric vehicles are analyzed, and the travel processes of different types of electric vehicles are simulated using the Monte Carlo method. A penalty coefficient is introduced based on the traffic congestion index, and a Markov dynamic path decision model based on the optimal strategy is established, so that user path selection is not limited to a single shortest path. A travel mileage energy consumption model is established based on the energy consumption of electric vehicles under different temperatures and road conditions, and the real-time dynamic energy consumption per unit mileage under different scenarios is calculated. The anchoring effect is introduced to analyze user psychology, and a user charging decision model based on the anchoring effect is established. Finally, the constructed model is verified using the urban road network to predict the spatiotemporal distribution characteristics of electric vehicle charging demand.

[0165] The above describes in detail the preferred embodiments of the present invention. It should be understood that those skilled in the art can make numerous modifications and variations based on the concepts of the present invention without inventive effort. Therefore, any technical solutions that can be derived by those skilled in the art through logical analysis, reasoning, or limited experimentation based on the concepts of the present invention and the prior art should be within the scope of protection defined by the claims.

Claims

1. A method for predicting electric vehicle charging load based on dynamic energy consumption and user psychology, characterized in that: include: Set the number of electric vehicles in the area to be predicted and the starting SOC of each electric vehicle; Build a travel model and actual power consumption model for electric vehicles; According to the travel model of electric vehicles, the optimal strategy of each electric vehicle is obtained through the Markov dynamic path decision model, and the travel process of each electric vehicle is simulated according to the corresponding optimal strategy; During the simulation, for each travel node of each electric vehicle, the power consumption of the next travel segment and the current SOC of the electric vehicle are pre-calculated based on the actual power consumption model and the starting SOC. The charging demand of the electric vehicle at the current travel node is determined based on the power consumption of the next travel segment and the current SOC. The charging demand of all electric vehicles at each travel node is accumulated to obtain the spatiotemporal distribution of the charging demand of electric vehicles in the predicted area. The actual power consumption model is expressed as: Among them, E all,T is the total energy consumption of electric vehicles at ambient temperature T, K ac is the start / stop state of the vehicle air conditioner of the electric vehicle under the ambient temperature T, and the K ac The value of K is 0 or 1. ac =1 means the vehicle air conditioner is turned on, the K ac =0 means the car air conditioner is off, P ac is the onboard air conditioning power of the electric vehicle at ambient temperature T, n is the total number of segments of the electric vehicle’s travel process, t k and D k are the driving time and mileage during the k-th trip, ξ t is the driving time correction coefficient, E v E is the energy consumption per unit mileage generated by an electric vehicle traveling at speed v. o is the initial energy consumption per unit mileage of electric vehicles.

2. According to the method for predicting electric vehicle charging load based on dynamic energy consumption and user psychology in claim 1, the vehicle air conditioner start-stop state K ac The calculation process includes: Set the cold threshold T c , thermal threshold T h And the comfortable temperature range, the T h >T c , the comfortable temperature range is between T c and T h The normal distribution function of air conditioner startup is obtained by normal distribution fitting; When the ambient temperature T is lower than the cold threshold T c or greater than the thermal threshold T h If so, let K ac =1, otherwise let K ac =0, when the ambient temperature T is between T c and the comfortable temperature range, or between the comfortable temperature range and T h When the air conditioner starts normal distribution function, K is determined ac Get the value.

3. According to the method for predicting electric vehicle charging load based on dynamic energy consumption and user psychology in claim 1, the vehicle air conditioner power P ac The calculation formula is: P ac (T)=ω3T 3 +ω210 -5 T 2 +ω1T+ω0 in, T is the ambient temperature, ω0, ω1, ω2 and ω3 are function fitting coefficients.

4. According to the method for predicting charging load of electric vehicles based on dynamic energy consumption and user psychology in claim 1, the driving time correction coefficient ξ t The acquisition process includes: Collect the traffic congestion index TPI of each road in the predicted area, and obtain ξ based on TPI t ; Among them, ξ t ∈[0,1], the higher the TPI, the t The bigger.

5. According to the method for predicting charging load of electric vehicles based on dynamic energy consumption and user psychology in claim 1, the energy consumption per unit mileage E v The acquisition process includes: The roads in the predicted area are divided into three levels: expressways, main roads and secondary roads. Each level of road corresponds to an average driving speed. When simulating an electric vehicle traveling on a road, the electric vehicle travels at an average speed corresponding to the grade of the road.

6. The electric vehicle charging load prediction method based on dynamic energy consumption and user psychology according to claim 1 is characterized in that: The electric vehicles are divided into household electric vehicles and rental electric vehicles; The process of constructing the travel model of the household electric vehicle includes: Based on the travel chain theory, a travel chain of household electric vehicles is constructed. According to the travel data in the NHTS2017 database, the variation pattern of the characteristic quantities of each node in the travel chain is obtained through probability distribution fitting analysis, thus forming a travel model for household electric vehicles. The construction process of the travel model for renting electric vehicles includes: Based on graph theory, roads are abstracted into a road network topology graph. According to travel order data, the OD analysis method is used to obtain the travel starting point distribution and probability transfer matrix of rental electric vehicles, thus forming a travel model for rental electric vehicles.

7. The method for predicting electric vehicle charging load based on dynamic energy consumption and user psychology according to claim 1, wherein the process of determining charging demand comprises the following steps: S801, judging whether the current SOC of the electric vehicle meets the travel demand based on the power consumption of the next travel process, if so, executing step S802, otherwise executing step S803; S802: Determine whether the parking time of the electric vehicle at the current travel road node is longer than the time required for slow charging. If so, it is determined that the electric vehicle has a slow charging requirement; otherwise, the electric vehicle has a fast charging requirement, and the process ends. S803: Determine whether the expected electricity price of the electric vehicle user is higher than the real-time electricity price. If so, determine that the electric vehicle has a slow charging demand; otherwise, determine that the electric vehicle has no charging demand.

8. The electric vehicle charging load prediction method based on dynamic energy consumption and user psychology according to claim 1 is characterized in that: The calculation formula of the battery capacity of the electric vehicle is: C R (T)=η3T 3 +η210 -5 T 2 +η1T+η0 Where T is the ambient temperature, C R (T) is the relative capacity percentage of the battery of the electric vehicle when the ambient temperature is T, and η0, η1, η2 and η3 are fitting coefficients.

9. The method for predicting electric vehicle charging load based on dynamic energy consumption and user psychology according to claim 1, wherein the initial SOC of all electric vehicles is set to obey a normal distribution.

Images