An electric vehicle charging load prediction method and terminal

Abstract

The application discloses a kind of electric vehicle charging load prediction method and terminal, obtain user historical travel data, classify and filter historical travel data, obtain historical travel chain data;According to historical travel chain data, user travel prediction is carried out, and predicted travel chain data is generated;According to predicted travel chain data and energy consumption model, travel energy consumption data is predicted, and energy consumption chain data is generated;According to energy consumption chain data, in combination with anxiety model, charging chain data under different charging strategies is predicted and generated;According to charging chain data, charging load change curve under different charging strategies is generated;The application considers the uncertainty of electric vehicle user travel behavior and space-time coupling characteristics, deduces the travel behavior of user by historical travel data, to deduce electric vehicle energy consumption, in combination with anxiety model, obtain charging chain information in line with charging law, obtain charging load curve change under different charging strategies, realize the simulation of electric vehicle charging load.

Description

Technical Field

[0001] This invention relates to the field of electric vehicle charging load prediction technology, and in particular to an electric vehicle charging load prediction method and terminal. Background Technology

[0002] With fossil fuels gradually dwindling and global energy issues intensifying, the pollution from fossil fuel emissions is gaining increasing attention, making the search for clean, pollution-free energy sources urgent and environmental protection imperative. Amidst this growing pollution from vehicle exhaust, electric vehicles have gradually emerged as a solution.

[0003] Therefore, the new energy vehicle industry will continue to grow and expand for a long time to come. Since 2015, China has ranked first in the world in terms of electric vehicle production, sales, and ownership for five consecutive years. As the number of electric vehicles increases, charging infrastructure is also gradually improving. As of December 2022, the cumulative number of charging infrastructure nationwide reached 5.21 million units, an increase of 99.1% year-on-year.

[0004] However, disorderly charging not only hinders the normal operation of charging stations, but also puts enormous pressure on the energy supply of the power distribution network. Therefore, regulating energy distribution and improving the interaction between electric vehicles and charging stations is of great significance for peak shaving and valley filling of the power grid load and improving the stability of the power grid.

[0005] Among these, the prediction of electric vehicle charging load plays an important role in regulating energy distribution. How to better predict electric vehicle charging load is a technical problem that is constantly being studied. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a method and terminal for predicting the charging load of electric vehicles, so as to realize the prediction of the charging load of electric vehicles.

[0007] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0008] A method for predicting electric vehicle charging load includes the following steps:

[0009] S1. Obtain user's historical travel data, classify and filter the historical travel data to obtain historical trip chain data;

[0010] S2. Based on the historical travel chain data, predict user travel and generate predicted travel chain data.

[0011] S3. Based on the predicted travel chain data and energy consumption model, predict travel energy consumption data and generate energy consumption chain data;

[0012] S4. Based on the energy consumption chain data and combined with the anxiety level model, predict and generate charging chain data under different charging strategies.

[0013] S5. Based on the charging chain data, generate charging load change curves under different charging strategies.

[0014] To solve the above-mentioned technical problems, another technical solution adopted by the present invention is as follows:

[0015] An electric vehicle charging load prediction terminal includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps in the electric vehicle charging load prediction method described above.

[0016] The beneficial effects of the present invention are as follows: The electric vehicle charging load prediction method and terminal of the present invention consider the uncertainty and spatiotemporal coupling characteristics of electric vehicle users' travel behavior, deduce and predict users' travel behavior through historical travel data, thereby deduce the energy consumption of electric vehicles, obtain charging chain information that conforms to the charging pattern based on electric vehicle power data and subsequent trips, and combine it with an anxiety model, and obtain the charging load curve changes under different charging strategies according to different charging strategies, thereby realizing the simulation of electric vehicle charging load. Attached Figure Description

[0017] Figure 1 This is a flowchart of an electric vehicle charging load prediction method according to an embodiment of the present invention;

[0018] Figure 2 This is a structural diagram of an electric vehicle charging load prediction terminal according to an embodiment of the present invention;

[0019] Figure 3 This is a flowchart of an electric vehicle charging load prediction method according to an embodiment of the present invention;

[0020] Figure 4 This is a flowchart illustrating the trip chain information generation process of an electric vehicle charging load prediction method according to an embodiment of the present invention.

[0021] Figure 5 This is a flowchart of an unordered charging strategy for an electric vehicle charging load prediction method according to an embodiment of the present invention.

[0022] Figure 6 This is a flowchart of a peak-avoidance charging strategy for an electric vehicle charging load prediction method according to an embodiment of the present invention.

[0023] Label Explanation:

[0024] 1. An electric vehicle charging load prediction terminal; 2. A processor; 3. A memory. Detailed Implementation

[0025] To explain in detail the technical content, objectives, and effects of the present invention, the following description is provided in conjunction with the embodiments and accompanying drawings.

[0026] Please refer to Figure 1 A method for predicting electric vehicle charging load includes the following steps:

[0027] S1. Obtain user's historical travel data, classify and filter the historical travel data to obtain historical trip chain data;

[0028] S2. Based on the historical travel chain data, predict user travel and generate predicted travel chain data.

[0029] S3. Based on the predicted travel chain data and energy consumption model, predict travel energy consumption data and generate energy consumption chain data;

[0030] S4. Based on the energy consumption chain data and combined with the anxiety level model, predict and generate charging chain data under different charging strategies.

[0031] S5. Based on the charging chain data, generate charging load change curves under different charging strategies.

[0032] As can be seen from the above description, the beneficial effects of the present invention are as follows: The electric vehicle charging load prediction method and terminal of the present invention consider the uncertainty and spatiotemporal coupling characteristics of electric vehicle users' travel behavior, deduce and predict users' travel behavior through historical travel data, thereby deduce the energy consumption of electric vehicles, obtain charging chain information that conforms to the charging pattern based on electric vehicle power data and subsequent trips, and combine with the anxiety model, and obtain the charging load curve changes under different charging strategies according to different charging strategies, thereby realizing the simulation of electric vehicle charging load.

[0033] Furthermore, step S1 includes the following steps:

[0034] S11. Filter data based on the number of trips and mode of transportation for each trip chain in the user's historical travel data, and remove abnormal data at the same time;

[0035] S12. Classify the trip chain data according to the pre-classified travel locations;

[0036] The abnormal data includes trip chain data where any of the following conditions exist: trip time is 0, trip distance is 0, parking time is 0, the ratio of trip distance to trip time exceeds a preset threshold, trip start point is uncertain, trip end point is uncertain, trip time is uncertain, and parking time is uncertain.

[0037] As described above, the system filters and removes anomalies from users' historical travel data, identifying travel data using electric vehicles and removing abnormal travel data to ensure data validity. The system also categorizes trip chain data based on travel location for subsequent trip chain prediction.

[0038] Furthermore, step S2 specifically includes:

[0039] Based on the user's trip chain data, the predicted trip chain data is obtained by iteratively processing the data using a Gaussian mixture model.

[0040] The predicted trip chain data includes total number of trips, first trip time, trip destination, trip distance, trip time, and parking time.

[0041] As described above, based on historical travel chain data, travel prediction is achieved through iterative processing using a Gaussian mixture model, which then generates future travel chain data.

[0042] Further, step S2 includes the following steps:

[0043] S21. Use the frequency of occurrence of trips to represent the probability of occurrence, and use the roulette wheel algorithm to obtain the total number of trips;

[0044] S22. Use the GMM formula to fit the first travel time in all data to the model, and then perform random sampling to obtain the required first travel time data;

[0045] S23. Calculation of the end point value of the journey:

[0046] Based on the starting point of the trip s The value of H, the end point of the journey l e Value W, first trip time data t s The value T s And the value N of the total number of trips n:

[0047]

[0048] In the denominator:

[0049] P(l s =H,t s =T s ,n=N)=P(l s =H,n=N)*P(t s =T s |l s =H,n=N);

[0050] Substituting, we get:

[0051]

[0052] In the molecule, due to P(l) s =H,n=N,l e Since the variables in =W) are all discrete values, we can directly count the number of data points corresponding to the conditions and divide them by the total number of data points. We can then use the statistical frequency value as the probability value to obtain the result p1.

[0053] In the molecule, calculate P(t) s =T s |l s =H,n=N,l e When =W), since the probability value is the integral of the probability density value, the probability density function is used to replace the probability in the calculation, and the final result is normalized to obtain the calculated probability value.

[0054] Then we can obtain:

[0055]

[0056] Wherein, P(l s The result p2 is obtained by calculating the statistical frequencies (H, n = N).

[0057] Calculate P(t) s =T s |l s When H = n = N, the probability density function is used instead of probability calculation, and the calculation formula is:

[0058]

[0059] The approximate probability that the destination of the journey is W is:

[0060]

[0061] The approximate probability values ​​for different trip endpoints are calculated sequentially and then normalized to obtain the probability values ​​for each trip endpoint.

[0062] The trip endpoint value is obtained based on the trip endpoint probability value;

[0063] S24. Calculation of travel distance:

[0064] Based on the trip start point and trip end point, historical driving distance data is filtered from the historical trip data, and the filtered historical driving distance data is used as a training set to construct a probability density model, as shown in the following formula:

[0065]

[0066] The travel distance is determined based on the probability density model;

[0067] S25. Divide the travel distance by the average travel speed to obtain the travel time;

[0068] S26. Calculation of parking time:

[0069] Ending point l e With the end time of the trip t e Using the initial variable as an initial variable, the GMM method is used for fitting, where the termination location is l. e The time t is a discrete variable representing the end of the journey. e and parking time t p For continuous variables;

[0070] If the instantaneous probability density is used as the probability value in the calculation, then:

[0071]

[0072] For p(t) p =T,t e =T1|l e =W), using a two-dimensional Gaussian mixture model for fitting, the formula is expressed as:

[0073]

[0074] Wherein, the mean vector and the covariance matrix are respectively:

[0075]

[0076] p(t) in the denominator e =T1|l e =W) is a one-dimensional Gaussian mixture model, and satisfies:

[0077]

[0078] Due to mathematical properties, the conditional distribution of a Gaussian mixture model remains a Gaussian mixture model. It is a one-dimensional Gaussian mixture distribution:

[0079]

[0080] in:

[0081]

[0082]

[0083]

[0084] The parking time was obtained by sampling based on a Gaussian mixture model and a roulette wheel algorithm.

[0085] As described above, the Gaussian mixture model predicts multiple factors, including total number of trips, first trip time, trip end point, trip distance, travel time, and parking time, and then iteratively organizes these factors to obtain trip chain data.

[0086] Furthermore, the energy consumption model includes a power consumption-speed related model and a power consumption-temperature related model;

[0087] The power consumption-speed related model is expressed as follows:

[0088] P=(kv 3 +f rl mgv)*K a ;

[0089] Among them, kv 3 f represents the power loss caused by wind resistance. rl mgv represents the power loss caused by rolling resistance, K a This represents the correction factor generated by neglecting the acceleration and deceleration process.

[0090] The power consumption and temperature-related model is expressed as follows:

[0091]

[0092] Among them, P EV T represents power consumption. amb This indicates the current temperature.

[0093] As described above, energy consumption information can be obtained based on the current ambient temperature and the electric vehicle's speed, provided that the vehicle model parameters are available.

[0094] Furthermore, generating charging chain data by combining the anxiety level model includes the following steps:

[0095] The anxiety level model formula is as follows, which simulates the probability of a user charging based on battery status:

[0096]

[0097] Where k is the gradient at the midpoint of the medium probability model curve. The higher the value, the faster the probability of charging increases as the remaining power decreases. %SOC is the SOC value corresponding to the midpoint of the model probability curve. When SOC = %SOC, the probability of charging is 50%.

[0098] As described above, based on SOC, the anxiety level model predicts user charging behavior; the lower the SOC, the more likely the user is to charge.

[0099] Furthermore, the charging strategy includes an unordered charging strategy:

[0100] S411. Obtain trip chain information, and calculate energy consumption information for each stage and trip start SOC information based on the energy consumption model.

[0101] S412. At the end of each trip, calculate the current SOC and compare it with the SOC consumed in the next trip plus a preset percentage. If the current SOC is less than the SOC consumed in the next trip plus a preset percentage, then charging is required. Otherwise, determine whether the parking time is less than a preset time threshold. If the parking time is less than the time threshold, then charging is not required. Otherwise, determine whether charging is required according to the anxiety level model formula.

[0102] As described above, under the disordered charging strategy, users decide whether to charge based on their travel needs and an anxiety model based on SOC.

[0103] Furthermore, the charging strategy includes a peak-shaving charging strategy:

[0104] S421. Calculate the energy consumption information of each stroke based on the stroke chain and energy consumption chain, and calculate the total SOC required for the entire stroke chain;

[0105] S422. For each parking time, determine whether the current parking time is within a high electricity price period. If it is, determine whether the current SOC meets the needs of the next trip. If it meets the needs of the next trip, charging is not required; otherwise, charging is required.

[0106] S423. If the current parking time is during a low electricity price period or is entirely during a low electricity price period, determine whether the current trip is the last trip in this trip chain. If it is not the last trip in this trip chain, determine whether charging is needed based on whether the current SOC meets the SOC required for the next trip. If it is the last trip, determine whether the current SOC is greater than the SOC required for the next trip chain plus a preset percentage. If so, determine whether charging is needed based on the anxiety level model formula; otherwise, charging is required.

[0107] If charging is required, calculate the available charging time:

[0108] t c =min(T) H ,t s0 )+24-max(T L ,f overnight *24+t e );

[0109]

[0110] Among them, T H With T L These represent the start times of high electricity prices and low electricity prices, respectively; ts0 This indicates the start time of the next day's itinerary chain, replaced by the start time of the current itinerary chain; f overnight Does this indicate whether it exceeds midnight? e This indicates the end time of this trip.

[0111] As described above, under the peak-avoidance charging strategy, users aim to minimize costs. Through the above calculations, they can obtain the duration of low electricity prices at night and charge at low prices, thus achieving maximum cost savings. Compared with disorderly charging, this method is often cheaper and requires fewer charging sessions.

[0112] Please refer to Figure 2 An electric vehicle charging load prediction terminal includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps in the electric vehicle charging load prediction method described above.

[0113] The present invention provides a method and terminal for predicting electric vehicle charging load, which is applicable to scenarios that require prediction of user charging behavior and charging load.

[0114] Please refer to the figure; Embodiment 1 of the present invention is as follows:

[0115] A method for predicting electric vehicle charging load, characterized by comprising the following steps:

[0116] S1. Obtain user's historical travel data, classify and filter the historical travel data to obtain historical trip chain data;

[0117] Step S1 includes the following steps:

[0118] S11. Filter data based on the number of trips and mode of transportation for each trip chain in the user's historical travel data, and remove abnormal data at the same time;

[0119] S12. Classify the trip chain data according to the pre-classified travel locations;

[0120] The abnormal data includes trip chain data where any of the following conditions exist: trip time is 0, trip distance is 0, parking time is 0, the ratio of trip distance to trip time exceeds a preset threshold, trip start point is uncertain, trip end point is uncertain, trip time is uncertain, and parking time is uncertain.

[0121] In this embodiment, reliable travel data is collected, organized, and then categorized and filtered according to the relevant requirements for fitting the travel data. The data classification and filtering steps include:

[0122] (1) Summarize and count the number of trips in each trip chain. This requires calculating the maximum number of trips while maintaining the same user ID. Furthermore, to ensure a more reasonable result, only trips with a count of 10 or less should be counted.

[0123] (2) Filter the required vehicle types. Remove travel categories such as air travel, ship travel, bus travel, and subway travel that do not meet the needs of this paper, and only collect data on car travel as the data source for this paper. Furthermore, group the data and only count trip chains where the entire journey involves car travel.

[0124] (3) Delete obviously unreasonable data such as trip time of 0, trip distance of 0, parking time of 0, and trip distance to trip time ratio that is too high (i.e., driving speed is too high);

[0125] (4) Delete all itinerary chains with incomplete information such as uncertain start and end points, uncertain itinerary time, and uncertain parking time;

[0126] (5) Reclassify travel destinations. The data source provides travel destinations in the form of travel purpose and statistically analyzes the travel destinations of cars. However, there are too many categories and many categories have a small proportion. After summarizing, travel destinations are classified into 6 categories: home, work, shopping, dining, entertainment, and others.

[0127] S2. Based on the historical travel chain data, predict user travel and generate predicted travel chain data.

[0128] Step S2 is as follows:

[0129] Based on the user's trip chain data, the predicted trip chain data is obtained by iteratively processing the data using a Gaussian mixture model.

[0130] The predicted trip chain data includes total number of trips, first trip time, trip destination, trip distance, trip time, and parking time.

[0131] In this embodiment, the model used to predict travel chain information is a Gaussian Mixture Model (hereinafter referred to as GMM), and its implementation principle is as follows:

[0132] For a random variable x in an n-dimensional sample space χ, if x follows a Gaussian distribution and its probability density function is:

[0133]

[0134] Where μ is an n-dimensional mean vector and ∑ is an n×n covariance matrix, the above equation is simplified to p(x|μ,∑), then the GMM is defined as:

[0135]

[0136] Where Θ=(θ1,θ2,...,θ k () represents the parameter set of each Gaussian component. This distribution consists of K mixture components, each of which follows a Gaussian distribution, and π... k The weight of the k-th Gaussian component satisfies:

[0137]

[0138] After calculating the GMM with specific parameters, sample generation can be performed. The main process is to first select Gaussian variables based on the weight π, and then sample the probability density function of the selected Gaussian variables to generate the required samples.

[0139] The following is a brief introduction on how to derive the specific parameters of GMM.

[0140] For the training set D = {x1, x2, x3, ..., x...} m},z j Let π ∈{1,2,...k} i =P(z) j =i), and let the posterior probability be:

[0141] p(z j =i|x j )=γ ji ;

[0142] Through the derivation of Bayes' theorem, we can obtain:

[0143]

[0144] This yields the Gaussian component weight π. i With z j The relationship between posterior probabilities.

[0145] When Gaussian mixture model (GMM) is known, GMM divides the sample set D into k clusters, each sample x j Cluster label λ j Determined by the following formula:

[0146]

[0147] For a given sample set:

[0148] D = {x1, x2, x3, ..., x} m};

[0149] The parameters of the Gaussian mixture clustering model can be solved using the maximum likelihood estimation method:

[0150] {(π i ,μ i ,∑ i|1≤i≤k};

[0151] That is, to maximize the log-likelihood.

[0152]

[0153] Based on this, the EM algorithm is used for iterative optimization and solution. The specific derivation process is not shown here. The final parameter calculation formula is as follows:

[0154]

[0155]

[0156]

[0157] Most of the GMM models used in this embodiment are one-dimensional models, and their corresponding one-dimensional GMM formulas are as follows:

[0158]

[0159] In the formula: Θ=(θ1,θ2,...,θ k ) represents the parameter set of each Gaussian component; It is a specific parameter of the k-th Gaussian component, π k It is the weight of the k-th Gaussian component, satisfying μ k , These are the mean and variance of the k-th Gaussian distribution.

[0160] The specific process will be further explained below.

[0161] In establishing the travel chain, each travel variable needs to be simulated separately. The simulation methods are as follows:

[0162] (1) Total number of trips: The frequency of trips is used to represent the probability of occurrence. The total number of trips is obtained using the roulette wheel algorithm.

[0163] (2) First trip time: The first trip time in all data is modeled using the GMM formula, and then random sampling is performed to obtain the required first trip time data;

[0164] (3) End point of the journey: Calculation of the end point value:

[0165] Based on the starting point of the trip s The value of H, the end point of the journey l e Value W, first trip time data t s The value T s And the value N of the total number of trips n:

[0166]

[0167] In the denominator:

[0168] P(l s =H,t s =T s ,n=N)=P(l s =H,n=N)*P(t s =T s |l s =H,n=N);

[0169] Substituting, we get:

[0170]

[0171] In the molecule, due to P(l) s =H,n=N,l e Since the variables in =W) are all discrete values, we can directly count the number of data points corresponding to the conditions and divide them by the total number of data points. We can then use the statistical frequency value as the probability value to obtain the result p1.

[0172] In the molecule, calculate P(t) s =T s |l s =H,n=N,l e When =W), it can be assumed that it conforms to GMM. Since the probability value is the integral of the probability density value, the instantaneous probability density value can be regarded as the probability value at that moment, that is, the probability density function is used to replace the probability in the calculation, and the final result is normalized to obtain the calculated probability value;

[0173] Then we can obtain:

[0174]

[0175] Wherein, P(l s The result p2 is obtained by calculating the statistical frequencies (H, n = N).

[0176] Calculate P(t) s =T s |l s When H = N, n = N, we can also assume that it conforms to the GMM and use the probability density function instead of probability calculation, then the calculation formula is:

[0177]

[0178] Substituting the above two equations with p1 and p2, we can calculate the approximate probability that the destination of this trip is W:

[0179]

[0180] The approximate probability values ​​for different trip endpoints are calculated sequentially and then normalized to obtain the probability values ​​for each trip endpoint.

[0181] The trip endpoint value is obtained based on the trip endpoint probability value;

[0182] (4) Travel distance: Calculation of travel distance:

[0183] Based on the trip start point and trip end point, historical driving distance data is filtered from the historical trip data, and the filtered historical driving distance data is used as a training set to construct a probability density model, as shown in the following formula:

[0184]

[0185] The travel distance is determined based on the probability density model;

[0186] (5) Travel time: The travel time is obtained by dividing the travel distance by the average travel speed.

[0187] (6) Parking time: Calculation of parking time:

[0188] Ending point l e With the end time of the trip t e Using the initial variable as an initial variable, the GMM method is used for fitting, where the termination location is l. e The time t is a discrete variable representing the end of the journey. e and parking time t p It is a continuous variable.

[0189] Taking the termination point W and the trip termination time T1 as an example, the parking time is calculated. The process of fitting the parking time is quite different from the previous travel data processing process, mainly because it involves two continuous variables, which need to be fitted using a two-dimensional GMM. However, the main idea is still to separate the continuous variables from the discrete variables using Bayesian principles.

[0190] Similar to the aforementioned treatment of continuous variables, if the instantaneous probability density is used as the probability value in the calculation, then:

[0191]

[0192] For p(t) p =T,t e =T1|l e =W), since there are two discrete variables, a two-dimensional Gaussian mixture model is used for fitting, and the formula is expressed as:

[0193]

[0194] Wherein, the mean vector and the covariance matrix are respectively:

[0195]

[0196] p(t) in the denominator e =T1|l e =W) is a one-dimensional Gaussian mixture model, and satisfies:

[0197]

[0198] Due to mathematical properties, the conditional distribution of a Gaussian mixture model remains a Gaussian mixture model. It is a one-dimensional Gaussian mixture distribution:

[0199]

[0200] Through mathematical derivation, the weights, mean, variance, etc., corresponding to the above formula can all be derived from the variables in the aforementioned formula, where:

[0201]

[0202]

[0203]

[0204] The parking time was obtained based on sampling using a Gaussian mixture model and a roulette wheel algorithm:

[0205] The mean, covariance, and initial variable t obtained from the aforementioned formula are used to... e Substituting the values ​​yields The corresponding Gaussian mixture model's weights, mean, and variance are used to construct a new one-dimensional GMM. Finally, the stopping time t for this trip can be obtained by sampling using the roulette wheel algorithm. p .

[0206] The data obtained from the above fitting process includes: the number of trips, start point, end point, travel time, travel distance, start time, end time, and parking time for this trip. The start time of the next trip can be obtained by adding the end time and parking time. The number of trips is incremented by 1. This process continues until the number of trips reaches the total number of trips predicted at the beginning, at which point the trip chain generation ends.

[0207] S3. Based on the predicted travel chain data and the energy consumption model, predict the travel energy consumption data and generate energy consumption chain data.

[0208] In this embodiment, after obtaining the travel chain information, it is necessary to obtain the electric vehicle parameter information to obtain the energy consumption information. The present invention calculates the energy consumption information using several typical models as examples.

[0209] Table 1 Common Electric Vehicle Parameters

[0210]

[0211]

[0212] Meanwhile, the energy consumption chain model used needs to consider the relationship between electric vehicle energy consumption and travel information. The following is an introduction to the selected physical model (energy consumption model), which mainly considers two variables: vehicle speed and temperature during driving.

[0213] The energy consumption model includes a power consumption model related to speed and a power consumption model related to temperature.

[0214] The power consumption-speed related model is expressed as follows:

[0215] P=(kv 3 +f rl mgv)*K a ;

[0216] Among them, kv 3 f represents the power loss caused by wind resistance. rl mgv represents the power loss caused by rolling resistance, K a This represents the correction factor generated by neglecting the acceleration and deceleration process.

[0217] The power consumption and temperature-related model is expressed as follows:

[0218]

[0219] Among them, P EV T represents power consumption. amb This indicates the current temperature.

[0220] After obtaining common electric vehicle parameters, it is generally assumed that the State of Charge (SOC) of the electric vehicle is evenly distributed between 60% and 90% when the trip begins. Then, the charging behavior is simulated and the charging time is calculated according to the disordered charging and peak avoidance charging strategies. Finally, the charging load change is calculated according to the charging power of common charging piles.

[0221] S4. Based on the energy consumption chain data and combined with the anxiety level model, predict and generate charging chain data under different charging strategies.

[0222] The anxiety level model formula is as follows, which simulates the probability of a user charging based on battery status:

[0223]

[0224] Where k is the gradient at the midpoint of the medium probability model curve. The higher the value, the faster the probability of charging increases as the remaining power decreases. %SOC is the SOC value corresponding to the midpoint of the model probability curve. When SOC = %SOC, the probability of charging is 50%.

[0225] For reference Figure 5 and Figure 6 Below are the implementation procedures for two charging strategies.

[0226] The charging strategy includes a disordered charging strategy:

[0227] S411. Obtain trip chain information, and calculate energy consumption information for each stage and trip start SOC information based on the energy consumption model.

[0228] S412. At the end of each trip, calculate the current SOC and compare it with the SOC consumed in the next trip plus a preset percentage. If the current SOC is less than the SOC consumed in the next trip plus a preset percentage, then charging is required. Otherwise, determine whether the parking time is less than a preset time threshold. If the parking time is less than the time threshold, then charging is not required. Otherwise, determine whether charging is required according to the anxiety level model formula.

[0229] In this embodiment, the preset percentage is 20%, which is used to prevent the battery from being damaged due to excessively low power, thus affecting its lifespan. The time threshold is 30 minutes; if the parking time is less than 30 minutes, people often choose not to let the car charge, based on real-life experience.

[0230] The charging strategy includes a peak-avoidance charging strategy:

[0231] S421. Calculate the energy consumption information of each stroke based on the stroke chain and energy consumption chain, and calculate the total SOC required for the entire stroke chain;

[0232] S422. For each parking time, determine whether the current parking time is within a high electricity price period. If it is, determine whether the current SOC meets the needs of the next trip. If it meets the needs of the next trip, charging is not required; otherwise, charging is required.

[0233] S423. If the current parking time is during a low electricity price period or is entirely during a low electricity price period, determine whether the current trip is the last trip in this trip chain. If it is not the last trip in this trip chain, determine whether charging is needed based on whether the current SOC meets the SOC required for the next trip. If it is the last trip, determine whether the current SOC is greater than the SOC required for the next trip chain plus a preset percentage. If so, determine whether charging is needed based on the anxiety level model formula; otherwise, charging is required.

[0234] If charging is required, calculate the available charging time:

[0235] t c =min(T) H ,t s0 )+24-max(T L ,f overnight *24+t e );

[0236]

[0237] Among them, T H With T L These represent the start times of high electricity prices and low electricity prices, respectively; t s0 This indicates the start time of the next day's itinerary chain, replaced by the start time of the current itinerary chain; f overnight Does this indicate whether it exceeds midnight? e This indicates the end time of this trip.

[0238] The above calculations can determine the duration of low electricity prices at night, allowing for charging at these lower prices and maximizing cost savings.

[0239] S5. Based on the charging chain data, generate charging load change curves under different charging strategies.

[0240] In this embodiment, the corresponding electric vehicle charging load curve can be obtained by continuously substituting the values ​​into the calculations under the two charging strategies described above. The charging load is obtained by calculating the charging duration. The main purpose of the charging strategy is to determine the charging behavior and to statistically analyze the charging duration.

[0241] Please refer to Figure 2 Embodiment two of the present invention is as follows:

[0242] An electric vehicle charging load prediction terminal 1 includes a processor 2, a memory 3, and a computer program stored in the memory 3 and executable on the processor 2. When the processor 2 executes the computer program, it implements the steps in the electric vehicle charging load prediction method described in Embodiment 1 above.

[0243] In summary, the electric vehicle charging load prediction method and terminal provided by this invention consider the uncertainty and spatiotemporal coupling characteristics of electric vehicle users' travel behavior. It derives and predicts users' travel behavior through historical travel data to deduce the energy consumption of electric vehicles. Based on the electric vehicle's power data and subsequent journey, it combines an anxiety model to obtain charging chain information that conforms to the charging pattern. According to different charging strategies, it obtains the charging load curve changes under different charging strategies, thereby realizing the simulation of electric vehicle charging load.

[0244] This invention conducts a detailed simulation and discussion of the travel and charging behavior of electric vehicle clusters. By using a data-driven and model-based approach, it studies the behavior of electric vehicle clusters to gain a deeper understanding of their behavior and patterns, providing a scientific basis for optimizing the management and control of electric vehicle clusters.

[0245] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent modifications made based on the content of the present invention specification and drawings, or direct or indirect applications in related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for predicting the charging load of electric vehicles, characterized in that, Including the following steps: S1. Obtain user's historical travel data, classify and filter the historical travel data to obtain historical trip chain data; S2. Based on the historical travel chain data, predict user travel and generate predicted travel chain data; S3. Based on the predicted travel chain data and energy consumption model, predict travel energy consumption data and generate energy consumption chain data; S4. Based on the energy consumption chain data and combined with the anxiety level model, predict and generate charging chain data under different charging strategies. S5. Based on the charging chain data, generate charging load change curves under different charging strategies; Generating charging chain data using an anxiety level model includes the following steps: The anxiety level model formula is as follows, which simulates the probability of a user charging based on battery status: ； in, The gradient at the midpoint of the medium probability model curve represents the rate of change. The higher the value, the faster the probability of charging increases as the remaining battery power decreases. Let SOC be the value corresponding to the midpoint of the model probability curve, when At that time, the probability of charging behavior occurring is 50%; The charging strategy includes a peak-avoidance charging strategy: S421. Calculate the energy consumption information of each stroke based on the stroke chain and energy consumption chain, and calculate the total SOC required for the entire stroke chain; S422. For each parking time, determine whether the current parking time is within a high electricity price period. If it is, determine whether the current SOC meets the needs of the next trip. If it meets the needs of the next trip, charging is not required; otherwise, charging is required. S423. If the current parking time is during a low electricity price period or is entirely during a low electricity price period, determine whether the current trip is the last trip in this trip chain. If it is not the last trip in this trip chain, determine whether charging is needed based on whether the current SOC meets the SOC required for the next trip. If it is the last trip, determine whether the current SOC is greater than the SOC required for the next trip chain plus a preset percentage. If so, determine whether charging is needed based on the anxiety level model formula; otherwise, charging is required. If charging is required, calculate the available charging time: ； ； in, and These represent the start time of high electricity prices and the start time of low electricity prices, respectively. This indicates the start time of the next day's itinerary chain, replaced by the start time of the current itinerary chain; Does this indicate whether it is past midnight? This indicates the end time of this trip.

2. The electric vehicle charging load prediction method according to claim 1, characterized in that, Step S1 includes the following steps: S11. Filter data based on the number of trips and mode of transportation for each trip chain in the user's historical travel data, and remove abnormal data at the same time; S12. Classify the trip chain data according to the pre-classified travel locations; The abnormal data includes trip chain data where any of the following conditions exist: trip time is 0, trip distance is 0, parking time is 0, the ratio of trip distance to trip time exceeds a preset threshold, trip start point is uncertain, trip end point is uncertain, trip time is uncertain, and parking time is uncertain.

3. The electric vehicle charging load prediction method according to claim 1, characterized in that, Step S2 is as follows: Based on the user's trip chain data, the predicted trip chain data is obtained by iteratively processing the data using a Gaussian mixture model. The predicted trip chain data includes total number of trips, first trip time, trip destination, trip distance, trip time, and parking time.

4. The electric vehicle charging load prediction method according to claim 3, characterized in that, Step S2 includes the following steps: S21. Use the frequency of occurrence of trips to represent the probability of occurrence, and use the roulette wheel algorithm to obtain the total number of trips; S22. Use the GMM formula to fit the first travel time in all data to the model, and then perform random sampling to obtain the required first travel time data; S23. Calculation of the end point value of the journey: Based on the starting point of the trip The value of H, the end point of the journey Value W, first trip time data value and total number of trips value : ； In the denominator: ； Substituting, we get: ； In molecules, due to Since all variables in the dataset are discrete values, we can directly count the number of data points corresponding to each condition, divide the total number of data points by the total number of data points, and use the statistical frequency as the probability value to obtain the result. ; In molecules, calculations Since the probability value is the integral of the probability density value, the probability density function is used to replace the probability in the calculation, and the final result is normalized to obtain the calculated probability value. Then we can obtain: ； in, The result was obtained from the statistical frequency calculation. ; calculate When the probability density function is used instead of probability calculation, the calculation formula is: ； The approximate probability that the destination of the journey is W is: ； The approximate probability values ​​for different trip endpoints are calculated sequentially and then normalized to obtain the probability values ​​for each trip endpoint. The trip endpoint value is obtained based on the trip endpoint probability value; S24. Calculation of travel distance: Based on the trip start point and the trip end point, historical driving distance data is filtered from the historical trip chain data, and the filtered historical driving distance data is used as a training set to construct a probability density model, as shown in the following formula: ； The travel distance is determined based on the probability density model; S25. Divide the travel distance by the average travel speed to obtain the travel time; S26. Calculation of parking time: By the place of termination End time of trip Using GMM as the initial variable, a fitting method is performed, where the termination location is... For discrete variables, the trip termination time and parking time For continuous variables; If the instantaneous probability density is used as the probability value in the calculation, then: ； Among them A two-dimensional Gaussian mixture model is used for fitting, and the formula is expressed as: ； Wherein, the mean vector and the covariance matrix are respectively: ； in the denominator It is a one-dimensional Gaussian mixture model, and satisfies: ； Due to mathematical properties, the conditional distribution of a Gaussian mixture model remains a Gaussian mixture model. It is a one-dimensional Gaussian mixture distribution: ； in: ； ； ； The parking time was obtained by sampling based on a Gaussian mixture model and a roulette wheel algorithm.

5. The electric vehicle charging load prediction method according to claim 1, characterized in that, The energy consumption model includes a power consumption model related to speed and a power consumption model related to temperature. The power consumption-speed related model is expressed as follows: ； in, This indicates the power loss caused by wind resistance. This represents the power loss caused by rolling resistance. This represents the correction factor generated by neglecting the acceleration and deceleration process. The power consumption and temperature-related model is expressed as follows: ； in, Indicates power consumption. This indicates the current temperature.

6. The electric vehicle charging load prediction method according to claim 1, characterized in that, The charging strategy includes a disordered charging strategy: S411. Obtain trip chain information, and calculate energy consumption information for each stage and trip start SOC information based on the energy consumption model. S412. At the end of each trip, calculate the current SOC and compare it with the SOC consumed in the next trip plus a preset percentage. If the current SOC is less than the SOC consumed in the next trip plus a preset percentage, then charging is required. Otherwise, determine whether the parking time is less than a preset time threshold. If the parking time is less than the time threshold, then charging is not required. Otherwise, determine whether charging is required according to the anxiety level model formula.

7. An electric vehicle charging load prediction terminal, comprising a processor, a memory, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps in the electric vehicle charging load prediction method according to any one of claims 1-6.

Images