Data-driven method for estimating the driving range of new energy vehicles
By using a data-driven approach to screen relevant features and build an XGBoost model, the impact of battery aging on driving range estimation is resolved, achieving accurate driving range estimation, reducing data collection costs, and overcoming users' range anxiety.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHONGQING UNIV
- Filing Date
- 2025-02-21
- Publication Date
- 2026-05-26
Smart Images
Figure CN119928667B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of battery technology and relates to a data-driven method for estimating the driving range of new energy vehicles. Background Technology
[0002] Methods for estimating the driving range of electric vehicles can generally be categorized into three types: energy-based calculation, vehicle parameter and operating environment-based methods, and data-driven methods. Energy-based calculation methods calculate the remaining usable energy of the battery at a given moment and estimate the future energy consumption of the electric vehicle based on historical operating data. Dividing these two values yields the vehicle's remaining driving range. Vehicle parameter and operating environment-based methods consider battery charge, real-time traffic conditions, and driving status, and estimate the remaining driving range using parameters such as speed, vehicle weight, and driving resistance. Data-driven methods, due to their reliance on sufficient data for high-precision estimation, have received widespread attention in recent years. The basic process for estimating the remaining driving range using data-driven methods involves: extracting vehicle operating data, preprocessing the data, and training and optimizing the model to obtain an empirical model. This empirical model is then used to predict the vehicle's remaining driving range. Because data-driven methods learn their predictive models from historical vehicle operating data, the resulting models possess good versatility and high prediction accuracy. However, these methods rely on the vehicle's current SOC or SOE to estimate the available energy of electric vehicles, ignoring the effects of battery aging. This makes the models unsuitable for long-term use in estimating the driving range of electric vehicles.
[0003] To address the aforementioned issues, no effective estimation method based on machine learning and historical operating data that takes into account battery aging has yet been proposed for the efficient and accurate estimation of the driving range of new energy vehicles. Summary of the Invention
[0004] In view of this, the purpose of this invention is to provide a data-driven method for estimating the driving range of new energy vehicles.
[0005] To achieve the above objectives, the present invention provides the following technical solution:
[0006] A data-driven method for estimating the driving range of new energy vehicles, comprising the following steps:
[0007] S1. Collect the operating data of new energy vehicles of corresponding models and establish a new energy vehicle operating database;
[0008] S2. Establish an estimation model based on charging data and data augmentation technology, and calculate the vehicle's current total available capacity based on the estimation results and the Ampere integral method;
[0009] S3. Based on the Pearson correlation coefficient index, extract the charging feature set and driving feature set that are correlated with the driving range from the driving range estimation feature set;
[0010] S4. Based on the selected charging feature set and driving feature set, construct a driving range estimation model based on XGBoost, and estimate the driving range through the driving range estimation model.
[0011] Furthermore, in step S1, the collected operating data of the new energy vehicle includes: charging and discharging current, maximum value of battery cell voltage, minimum value of battery cell voltage, maximum value of battery cell temperature, minimum value of battery cell temperature, time, charging status, vehicle speed, mileage, and SOC; then, based on the collected vehicle operating data, a battery pack operating database for the corresponding new energy vehicle is established.
[0012] Furthermore, step S2 specifically includes the following steps:
[0013] S21. Analyze the charging data of new energy vehicles and extract stable charging segments from it;
[0014] S22. Calculate the charging capacity of the sudden charging SOC range using the ampere-hour integration method, and use it as a tag;
[0015] S23. Arbitrarily extract charging data with a SOC span of 20% from the charging segment where the charging cutoff SOC is greater than 99% and the initial SOC is less than 80% to extract features related to the charging capacity in the sudden charging SOC interval, as a capacity estimation feature set.
[0016] S24. Based on the capacity estimation feature set, construct a capacity estimation model for the SOC interval of sudden charging based on GPR, and expand the training set through data augmentation techniques.
[0017] S25. Calculate the charging capacity in the stable charging SOC range using the ampere-hour integral method, and then estimate the charging capacity in the sudden charging SOC range using the sudden charging SOC range capacity estimation model to obtain the current total available capacity of the vehicle.
[0018] Furthermore, in step S21, a stable charging segment refers to a charging segment where the charging cutoff SOC is greater than 99% and the initial SOC is less than 80%.
[0019] In step S22, the calculation method for the data-driven abrupt charging SOC interval capacity estimation tag is as follows:
[0020]
[0021] In the formula, I(t) is the total current when the vehicle is charging, and [90,100] is the SOC range of sudden charging.
[0022] In step S23, the charging data for a charging segment with a charging SOC span of 20% includes the current mileage driven, average current, average temperature, the highest single-cell voltage at the end of charging, the difference between the highest and lowest single-cell voltage at the end of charging, the change in the highest single-cell voltage in the 20% SOC range, the change in the lowest single-cell voltage in the 20% SOC range, the charging cutoff SOC, and the characteristics of the charged capacity.
[0023] In step S24, the mutation charging SOC interval capacity estimation model constructed based on GPR includes: processing the input features into input feature vectors, constructing a mapping relationship between the input features and the target output based on the kernel function, and optimizing the parameters in the kernel function through learning from training samples, so that the model can accurately capture the complex relationship between the input features and the target output.
[0024] Data augmentation technology refers to dividing a stable charging segment into multiple charging sub-segments with equal span SOC, extracting a sample from each charging sub-segment, and extracting features such as mileage, average charging current, average temperature during charging, the difference between the beginning and end of the highest single-cell voltage, the difference between the beginning and end of the lowest single-cell voltage, and the difference between the highest and lowest single-cell voltages at charging cutoff from all the augmented samples to construct a capacity estimation feature set.
[0025] In step S25, the formula for calculating the current total available capacity of new energy vehicles is:
[0026]
[0027] In the formula, Q represents the current total available capacity; Q 90_100 ′ represents the capacity of the 90%-100% SOC range estimated by the sudden charging SOC range capacity estimation model; x represents the initial SOC value of the charging segment; and y represents the cutoff SOC value of the charging segment.
[0028] Furthermore, step S3 includes the following steps:
[0029] S31. Analyze the vehicle operation data and extract complete charging data as well as some data of the corresponding discharge process.
[0030] S32. Based on the selected charging data and driving data, calculate the statistical characteristics of the charging data and driving data, including the mean and standard deviation, and extract the range estimation feature set.
[0031] S33. Through correlation analysis based on Pearson correlation coefficient, a range estimation feature set that is correlated with the driving range is selected. The range estimation feature set includes a charging feature set and a driving feature set.
[0032] Furthermore, in step S33, the Pearson correlation coefficient (PCC) can provide information on the direction and strength of the linear correlation between the features in the range estimation feature set and the range. Its calculation method is as follows:
[0033]
[0034] In the formula, x i Let represent the i-th feature in the range estimation feature set, and y represent the actual range. and These represent the average values of the corresponding feature sequence and driving range sequence, respectively.
[0035] Furthermore, the charging feature set selected based on the Pearson correlation coefficient includes average current, maximum battery cell voltage, minimum battery cell voltage, average temperature during charging, total available energy, and initial state of charge (SOC); the driving feature set includes energy consumption during partial driving, average temperature during driving, depth of discharge, average vehicle speed, average current during discharge, and current variance during discharge.
[0036] Furthermore, step S4 includes the following steps:
[0037] S41. Using the selected range estimation feature set as input, a vehicle range estimation model is built based on XGBoost. The model is trained iteratively through the gradient boosting framework. The first iteration is based on the training sample set containing feature vectors and real mileage to build the first decision tree: a greedy algorithm is used to divide the feature space, and the optimal split point is determined by maximizing information gain. Subsequent iterations generate new trees and optimize feature combinations and split points.
[0038] S42. Input the test data after extracting the feature set into the trained vehicle range estimation model to estimate the vehicle range.
[0039] The beneficial effects of this invention are as follows:
[0040] This invention calculates the current total available energy of the battery using historical vehicle operating data, effectively taking into account the impact of battery aging on driving range. This invention requires less data, reducing high data collection costs, and accurately estimates the vehicle's driving range, helping users overcome range anxiety. This invention extracts features from both battery status and driving behavior, providing a relatively comprehensive reflection of the vehicle's actual operating condition.
[0041] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description
[0042] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein:
[0043] Figure 1 This is an overall flowchart of the data-driven new energy vehicle range estimation method of the present invention;
[0044] Figure 2 This is a diagram illustrating the overall architecture of the data-driven new energy vehicle range estimation method of the present invention.
[0045] Figure 3 This is a schematic diagram illustrating the correlation between the set of health indicators and the battery health status in an example embodiment. Detailed Implementation
[0046] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.
[0047] The accompanying drawings are for illustrative purposes only and are schematic diagrams, not actual pictures. They should not be construed as limiting the invention. To better illustrate the embodiments of the invention, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual product dimensions. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.
[0048] In the accompanying drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components. In the description of the present invention, it should be understood that if terms such as "upper," "lower," "left," "right," "front," and "rear" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, the terms used to describe positional relationships in the drawings are only for illustrative purposes and should not be construed as limiting the present invention. For those skilled in the art, the specific meaning of the above terms can be understood according to the specific circumstances.
[0049] Please see Figures 1-3 This is a data-driven method for estimating the driving range of new energy vehicles.
[0050] Example
[0051] Figure 1 The present invention provides an overall flowchart of a data-driven method for estimating the driving range of new energy vehicles, which can be divided into the following steps:
[0052] Step S1: Collect operating data of new energy vehicles of the same model, including charging and discharging data, and establish a new energy vehicle operating database;
[0053] Step S2: Based on charging data and data augmentation technology, establish an estimation model, and use the estimation results and the Ampere integral method to calculate the total available capacity of the vehicle;
[0054] Step S3: Extract charging feature sets and driving feature sets that are correlated with driving range based on the Pearson correlation coefficient (PCC) index;
[0055] Step S4: Based on the selected charging feature set and driving feature set, construct a driving range estimation model based on XGBoost;
[0056] As an optional embodiment, the complete technical roadmap of this solution is as follows: Figure 2 As shown.
[0057] As an optional embodiment, step S1 specifically includes steps S11-S12:
[0058] Step S11: Collect the operating data of a certain electric vehicle, including charging and discharging current, maximum value of battery cell voltage, minimum value of battery cell voltage, maximum value of battery cell temperature, minimum value of battery cell temperature, time, charging status, vehicle speed, mileage and SOC;
[0059] Step S12: Based on the collected vehicle operation data, establish a battery pack operation database for a certain car model.
[0060] As an optional embodiment, step S2 specifically includes steps S21-S23:
[0061] Step S21: Analyze the charging data of the electric vehicle. As the battery ages, when fully charged, the vehicle's SOC will suddenly change from 90-100% to 100%. To more accurately calculate the capacity corresponding to this part, the charging segments with a cutoff SOC greater than 99% and a starting SOC less than 80% are selected.
[0062] Step S22: Calculate the filling capacity in the 90%-100% SOC range using the ampere-hour integration method, and use it as a tag;
[0063] Step S23: Extract features related to the charging capacity in the 90%-100% SOC range from the charging data with a SOC span of 20%, and use them as the capacity estimation feature set;
[0064] Step S24: Based on the capacity estimation feature set, construct a sudden charging SOC interval capacity estimation model based on GPR, and expand the training set using data augmentation techniques;
[0065] Step S25: Calculate the capacity below 90% SOC using the ampere-hour integration method, then estimate the capacity in the 90%-100% SOC range using a trained abrupt charging SOC range capacity estimation model. Add the two estimates together to obtain the current total available capacity, reflecting the battery's current health status. As the vehicle is used, its health status will decrease, resulting in a smaller total available battery capacity and a shorter driving range. The current total available capacity is used as a feature and input into the driving range prediction model.
[0066] As an optional implementation, an estimation model is established based on charging data and data augmentation technology. The estimation results and the ampere integral method are used to calculate the total available capacity of the vehicle, specifically including:
[0067] Filter charging segments that meet the SOC usage range spanning 90%-100% and with a span greater than 20%;
[0068] The charging capacity in the 90%-100% SOC range is calculated using the ampere-hour integration method and used as a data-driven capacity estimation label for the SOC range of sudden charging. The formula is shown below:
[0069]
[0070] In the charging data with a SOC span of 20%, the following features are extracted as capacity estimation features for the 90%-100% SOC range: current mileage driven, average current, average temperature, highest single-cell voltage at the end of charging, difference between the highest and lowest single-cell voltage at the end of charging, change in the highest single-cell voltage in the 20% SOC range, change in the lowest single-cell voltage in the 20% SOC range, charging cutoff SOC, and charged capacity.
[0071] The training set was expanded using data augmentation techniques. Within charging segments with a SOC range greater than 20, multiple charging segments with an SOC range equal to 20 were created, and a sample was extracted from each segment. From all the expanded samples, the following features were extracted: mileage, average charging current, average temperature during charging, the difference between the start and end of the highest single-cell voltage, the difference between the start and end of the lowest single-cell voltage, and the difference between the highest and lowest single-cell voltage at charging cutoff. Based on GPR, a capacity estimation model for the SOC range during abrupt charging was constructed to estimate Q. 90_100 ′.
[0072] The capacity below 90% SOC is calculated using the ampere-hour integration method. Then, the capacity in the 90%-100% SOC range is estimated using a trained abrupt charging SOC range capacity estimation model. The two estimates are then added together to obtain the current total available capacity, as shown in the following formula:
[0073]
[0074] In the formula, Q represents the current total available capacity; Q 90_100 ′ represents the capacity of the 90%-100% SOC range estimated by the sudden charging SOC range capacity estimation model; x represents the initial SOC value of the charging segment (removing the percentage sign); y represents the cutoff SOC value of the charging segment, with a maximum value of 90.
[0075] As an optional embodiment, step S3 specifically includes steps S31-S33:
[0076] Step S31: Analyze the vehicle operation data to extract complete charging data and partial data of the corresponding discharge process;
[0077] Step S32: Based on the selected charging data and driving data, calculate the statistical characteristics of the charging data and driving data, including the mean and standard deviation, and extract the range estimation feature set;
[0078] Step S33: Through correlation analysis based on Pearson correlation coefficient (PCC), select a range estimation feature set that is highly correlated with the driving range.
[0079] As an optional embodiment, the range estimation feature set includes two parts: a charging feature set and a driving feature set. The charging feature set includes average current, maximum battery cell voltage, minimum battery cell voltage, average temperature during charging, total available energy, and initial state of charge (SOC). The driving feature set includes energy consumption during a portion of the driving process, average temperature during driving, depth of discharge, average vehicle speed, average current during discharge, and current variance during discharge.
[0080] In order to describe the factors affecting driving range from multiple perspectives, the charging feature set and the driving feature set are used simultaneously to establish a vehicle driving range estimation model to achieve the estimation of vehicle driving range.
[0081] As an optional implementation, the Pearson correlation coefficient (PCC) metric is used to assess the correlation between the range estimation feature set and the range, such as... Figure 3 The diagram shown illustrates the correlation analysis between the range estimation feature set and the range in this embodiment. The Pearson correlation coefficient (PCC) provides information on the direction and strength of the linear correlation, and its calculation formula is shown below:
[0082]
[0083] In the formula, x i Let represent the i-th feature in the range estimation feature set, and y represent the actual range. and These represent the average values of the corresponding feature sequence and driving range sequence, respectively.
[0084] As an optional implementation, to take into account the impact on driving range from multiple perspectives, the relevant threshold is set to 0.1.
[0085] As an optional embodiment, step S4 specifically includes steps S41-S42:
[0086] Step S41: Using the selected range estimation feature set as input, a vehicle range estimation model is built based on XGBoost. The model is trained and tuned using the training set; the model is iteratively trained using a gradient boosting framework. The first iteration constructs the first decision tree based on the training sample set containing feature vectors and actual mileage: a greedy algorithm is used to partition the feature space (e.g., vehicle speed segmentation), and the optimal split point is determined by maximizing information gain. Subsequent iterations generate new trees, optimizing feature combinations and split points to reduce residuals and improve model accuracy.
[0087] Step S42: Input the test data after extracting the feature set into the trained vehicle range estimation model to estimate the vehicle range.
[0088] As an optional implementation, mean absolute error and root mean square error are used to evaluate the driving range estimation performance.
[0089] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.
Claims
1. A data-driven method for estimating the driving range of new energy vehicles, characterized in that: The method includes the following steps: S1. Collect the operating data of new energy vehicles of corresponding models and establish a database of battery pack operation for new energy vehicles; S2. Establish an estimation model based on charging data and data augmentation technology, and calculate the vehicle's current total available capacity based on the estimation results and the ampere-hour integral method. Step S2 specifically includes the following steps: S21. Analyze the charging data of new energy vehicles and extract stable charging segments from it; S22. Calculate the charging capacity of the sudden charging SOC range using the ampere-hour integration method, and use it as a tag; S23. Arbitrarily extract charging data with a SOC span of 20% from the charging segment where the charging cutoff SOC is greater than 99% and the initial SOC is less than 80% to extract features related to the charging capacity in the sudden charging SOC interval, as a capacity estimation feature set. S24. Based on the capacity estimation feature set, construct a capacity estimation model for the SOC interval of sudden charging based on GPR, and expand the training set through data augmentation techniques. S25. Calculate the charging capacity in the stable charging SOC range using the ampere-hour integral method, and then estimate the charging capacity in the sudden charging SOC range using the sudden charging SOC range capacity estimation model to obtain the current vehicle's total available capacity. In step S21, a stable charging segment refers to a charging segment where the charging cutoff SOC is greater than 99% and the initial SOC is less than 80%. In step S22, the calculation method for the data-driven abrupt charging SOC interval capacity estimation tag is as follows: In the formula, The total current used to charge the vehicle. For the sudden charging SOC range; In step S23, the charging data for a charging segment with a charging SOC span of 20% includes the current mileage driven, average current, average temperature, the highest single-cell voltage at the end of charging, the difference between the highest and lowest single-cell voltage at the end of charging, the change in the highest single-cell voltage in the 20% SOC range, the change in the lowest single-cell voltage in the 20% SOC range, the charging cutoff SOC, and the characteristics of the charged amount. In step S24, the mutation charging SOC interval capacity estimation model constructed based on GPR includes: processing the input features into input feature vectors, constructing a mapping relationship between the input features and the target output based on the kernel function, and optimizing the parameters in the kernel function through learning from training samples, so that the model can accurately capture the complex relationship between the input features and the target output. Data augmentation technology refers to dividing a stable charging segment into multiple charging sub-segments with equal span SOC, extracting a sample from each charging sub-segment, and extracting features such as mileage, average charging current, average temperature during charging, the difference between the beginning and end of the highest single-cell voltage, the difference between the beginning and end of the lowest single-cell voltage, and the difference between the highest and lowest single-cell voltages at charging cutoff from all the augmented samples to construct a capacity estimation feature set. In step S25, the formula for calculating the current total available capacity of new energy vehicles is: In the formula, This represents the total available capacity. The capacity of the 90%-100% SOC range estimated by the sudden charging SOC range capacity estimation model; This is the initial SOC value for the charging segment; This is the cutoff SOC value for the charging segment; S3. Based on the Pearson correlation coefficient index, extract the charging feature set and driving feature set that are correlated with the driving range from the driving range estimation feature set; S4. Based on the selected charging feature set and driving feature set, construct a driving range estimation model based on XGBoost, and estimate the driving range through the driving range estimation model.
2. The data-driven method for estimating the driving range of new energy vehicles according to claim 1, characterized in that: In step S1, the collected operating data of the new energy vehicle includes: charging and discharging current, maximum value of battery cell voltage, minimum value of battery cell voltage, maximum value of battery cell temperature, minimum value of battery cell temperature, time, charging status, vehicle speed, mileage, and SOC; then, based on the collected vehicle operating data, a battery pack operating database for the corresponding new energy vehicle is established.
3. The data-driven method for estimating the driving range of new energy vehicles according to claim 1, characterized in that: Step S3 includes the following steps: S31. Analyze the vehicle operation data and extract complete charging data as well as some data of the corresponding discharge process. S32. Based on the selected charging data and driving data, calculate the statistical characteristics of the charging data and driving data, including the mean and standard deviation, and extract the range estimation feature set. S33. Through correlation analysis based on Pearson correlation coefficient, a range estimation feature set that is correlated with the driving range is selected. The range estimation feature set includes a charging feature set and a driving feature set.
4. The data-driven method for estimating the driving range of new energy vehicles according to claim 3, characterized in that: In step S33, the Pearson correlation coefficient (PCC) provides information on the direction and strength of the linear correlation between features in the range estimation feature set and the range. Its calculation method is as follows: In the formula, This represents the i-th feature in the range estimation feature set. This indicates the actual driving range. and These represent the average values of the corresponding feature sequence and driving range sequence, respectively.
5. The data-driven method for estimating the driving range of new energy vehicles according to claim 4, characterized in that: The charging feature set, selected based on the Pearson correlation coefficient, includes average current, maximum cell voltage, minimum cell voltage, average temperature during charging, total available energy, and initial state of charge (SOC). The driving feature set includes energy consumption during part of the driving process, average temperature during the driving process, depth of discharge, average vehicle speed, average current during the discharge process, and current variance during the discharge process.
6. The data-driven method for estimating the driving range of new energy vehicles according to claim 3, characterized in that: Step S4 includes the following steps: S41. Using the selected range estimation feature set as input, a vehicle range estimation model is built based on XGBoost. The model is trained iteratively through the gradient boosting framework. The first iteration is based on the training sample set containing feature vectors and real mileage to build the first decision tree: a greedy algorithm is used to divide the feature space, and the optimal split point is determined by maximizing information gain. Subsequent iterations generate new trees and optimize feature combinations and split points. S42. Input the test data after extracting the feature set into the trained vehicle range estimation model to estimate the vehicle range.