New energy vehicle driving condition construction and evaluation method
Through the Internet of Vehicles technology, the driving data of new energy vehicles is collected and processed, and the driving working condition mimicry model is constructed, which solves the problem of complexity in the driving working condition data modeling of new energy vehicles, and realizes comprehensive analysis of driving working conditions and energy consumption optimization.
Patent Information
- Application Number
- CN202410944284.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-15
- Publication Date
- 2024-11-15
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The complexity and variability of power systems of new energy vehicles make driving working condition data complex and simulation. The existing technology lacks sufficient data for effective construction and evaluation of driving working condition models.
The Internet of Vehicles technology is used to collect driving data of new energy vehicles, and the high-quality driving working condition division data set is obtained through outlier detection and integration, and multi-modal driving habit state construction and mimicry feature vector correction are carried out to establish a mimicry model of driving working conditions. , conduct four-dimensional indicator evaluation and energy consumption characteristic judgment to obtain a visual model of driving conditions.
A comprehensive analysis and quantitative evaluation of the driving conditions of new energy vehicles has been achieved, and guidance on driving performance and energy consumption optimization has been provided to help users understand and optimize the energy utilization efficiency of vehicles.
Smart Images

Figure CN118964925A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of new energy vehicles, and in particular to a method for constructing and evaluating a driving condition of a new energy vehicle. Background Art
[0002] With the rapid development and popularization of new energy vehicles, the demand for the construction and evaluation standards of new energy vehicle driving conditions is increasing. The driving conditions of traditional fuel vehicles cannot be directly applied to new energy vehicles, and the power system of new energy vehicles is highly complex and changeable, including batteries, motors and electronic control systems. The interaction and dynamics between these components make the data modeling and simulation of driving conditions complex and difficult. At the same time, the existing technology does not obtain enough new energy vehicle driving-related data to complete the simulation construction of the driving condition model. Summary of the invention
[0003] In order to solve the above technical problems, the present invention proposes a new energy vehicle driving condition construction and evaluation method to solve at least one of the above technical problems.
[0004] To achieve the above object, the present invention provides a method for constructing and evaluating a driving condition of a new energy vehicle, comprising the following steps:
[0005] Step S1: using the Internet of Vehicles technology to collect driving data of new energy vehicles, thereby obtaining a driving original data set; performing outlier detection, filtering and integration of abnormal time series fluctuations on the driving original data set, thereby obtaining a driving condition division data set;
[0006] Step S2: constructing a multimodal driving habit state for the driving condition division data set, thereby obtaining a driving habit mimicry model set; performing mimicry feature vector correction on the driving habit mimicry model set, and performing model fusion, thereby obtaining a driving condition mimicry model;
[0007] Step S3: performing a four-dimensional index evaluation based on the driving condition simulation model to obtain four-dimensional comprehensive index data; performing a comprehensive evaluation on the driving condition division data set according to the four-dimensional comprehensive index data to obtain four-dimensional comprehensive evaluation data;
[0008] Step S4: dividing the driving condition division data set into actual driving data based on the driving condition simulation model, thereby obtaining long-term actual driving data; performing feature discrimination of the long-term actual driving data that is strongly correlated with energy consumption, thereby obtaining long-term driving energy consumption data; performing feature set division of the long-term driving energy consumption data into different time series windows, thereby obtaining a driving condition energy consumption data set;
[0009] Step S5: performing a visualization model migration of the driving condition of the new energy vehicle to the driving condition simulation model according to the driving condition energy consumption data set, thereby obtaining a driving condition visualization model.
[0010] The present invention can obtain the driving data of new energy vehicles in real time by using the Internet of Vehicles technology to collect driving data, including key information such as vehicle speed, acceleration, and battery status. By collecting a large amount of raw driving data, the driving characteristics and behavior patterns of new energy vehicles can be fully understood. Outlier detection and filtering of abnormal time series fluctuations can eliminate abnormal conditions in the data and ensure the accuracy and reliability of the data. Such data integration and processing process can obtain a driving condition division data set, providing a high-quality data basis for subsequent steps. By constructing a multimodal driving habit state for the driving condition division data set, the driving condition can be described from different angles, such as driving behavior, vehicle status, etc. Such a driving habit mimic model set can more accurately characterize the driving characteristics and habits of new energy vehicles. Mimic feature vector correction and model fusion can further improve the accuracy and reliability of the driving condition mimic model, so that it can better reflect the actual driving situation. Four-dimensional index evaluation based on the driving condition mimic model can comprehensively consider the four dimensions of the driving condition, such as energy consumption, safety, comfort and environmental protection. Such a comprehensive evaluation can provide a comprehensive driving condition analysis to help evaluate the performance and effect of new energy vehicles under different driving conditions. The four-dimensional comprehensive evaluation data can provide a quantitative evaluation of driving conditions, and provide guidance and reference for subsequent analysis and decision-making. Based on the driving condition simulation model, the driving condition division data set is divided into actual driving data, which can obtain more accurate data according to the actual driving conditions. By distinguishing the features that are strongly related to energy consumption for long-term actual driving data, the features that are strongly related to energy consumption can be identified, thereby obtaining long-term driving energy consumption data. Such energy consumption data can provide the energy consumption of new energy vehicles in the actual driving process, which is helpful for analyzing and evaluating the energy efficiency of vehicles. The feature set division of long-term driving energy consumption data into different time series windows can divide the data into different time windows, such as by hour, day or week. In this way, the driving condition energy consumption data set at different time scales can be obtained, and the energy consumption characteristics of new energy vehicles can be further refined and analyzed. The feature set division of different time series windows helps to discover the relationship between energy consumption and time scale, and provides a more targeted reference for energy consumption optimization and planning. By migrating the driving condition simulation model through the driving condition energy consumption dataset, the abstract driving condition model can be transformed into a visual form, so as to more intuitively display the energy consumption of new energy vehicles under different driving conditions. Such a visual model can help users and decision makers better understand and analyze the energy consumption characteristics of new energy vehicles, and provide visual support and guidance for formulating energy consumption optimization strategies and driving planning. BRIEF DESCRIPTION OF THE DRAWINGS
[0011] Figure 1 A schematic diagram of the steps of the method for constructing and evaluating the driving condition of a new energy vehicle according to the present invention;
[0012] Figure 2 Detailed implementation flow chart of step S1;
[0013] Figure 3 It is a schematic diagram of the detailed implementation steps of step S15. DETAILED DESCRIPTION
[0014] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.
[0015] This application example provides a method for constructing and evaluating the driving condition of a new energy vehicle. The execution subject of the method for constructing and evaluating the driving condition of a new energy vehicle includes but is not limited to the following: mechanical equipment, data processing platform, cloud server node, network upload device, etc. equipped with the system can be regarded as the general computing node of this application, and the data processing platform includes but is not limited to: at least one of an audio and image management system, an information management system, and a cloud data management system.
[0016] See also Figures 1 to 3 The present invention provides a method for constructing and evaluating a driving condition of a new energy vehicle, comprising the following steps:
[0017] Step S1: using the Internet of Vehicles technology to collect driving data of new energy vehicles, thereby obtaining a driving original data set; performing outlier detection, filtering and integration of abnormal time series fluctuations on the driving original data set, thereby obtaining a driving condition division data set;
[0018] Step S2: constructing a multimodal driving habit state for the driving condition division data set, thereby obtaining a driving habit mimicry model set; performing mimicry feature vector correction on the driving habit mimicry model set, and performing model fusion, thereby obtaining a driving condition mimicry model;
[0019] Step S3: performing a four-dimensional index evaluation based on the driving condition simulation model to obtain four-dimensional comprehensive index data; performing a comprehensive evaluation on the driving condition division data set according to the four-dimensional comprehensive index data to obtain four-dimensional comprehensive evaluation data;
[0020] Step S4: dividing the driving condition division data set into actual driving data based on the driving condition simulation model, thereby obtaining long-term actual driving data; performing feature discrimination of the long-term actual driving data that is strongly correlated with energy consumption, thereby obtaining long-term driving energy consumption data; performing feature set division of the long-term driving energy consumption data into different time series windows, thereby obtaining a driving condition energy consumption data set;
[0021] Step S5: performing a visualization model migration of the driving condition of the new energy vehicle to the driving condition simulation model according to the driving condition energy consumption data set, thereby obtaining a driving condition visualization model.
[0022] The present invention can obtain the driving data of new energy vehicles in real time by using the vehicle networking technology to collect driving data, including key information such as vehicle speed, acceleration, battery status, etc. By collecting a large amount of original driving data, the driving characteristics and behavior patterns of new energy vehicles can be fully understood. Outlier detection and filtering of abnormal time series fluctuations can eliminate abnormal conditions in the data and ensure the accuracy and reliability of the data. Such data integration and processing process can obtain a driving condition division data set, providing a high-quality data basis for subsequent steps. By constructing a multimodal driving habit state for the driving condition division data set, the driving condition can be described from different angles, such as driving behavior, vehicle status, etc. Such a driving habit mimic model set can more accurately characterize the driving characteristics and habits of new energy vehicles. Mimic feature vector correction and model fusion can further improve the accuracy and reliability of the driving condition mimic model, so that it can better reflect the actual driving situation. Four-dimensional index evaluation based on the driving condition mimic model can comprehensively consider the four dimensions of driving conditions, such as energy consumption, safety, comfort and environmental protection. Such a comprehensive evaluation can provide a comprehensive driving condition analysis and help evaluate the performance and effect of new energy vehicles under different driving conditions. The four-dimensional comprehensive evaluation data can provide a quantitative evaluation of driving conditions, providing guidance and reference for subsequent analysis and decision-making. The actual driving data division of the driving condition division data set based on the driving condition mimicry model can obtain more accurate data according to the actual driving conditions. By distinguishing the features that are strongly related to energy consumption for long-term actual driving data, the features that are strongly related to energy consumption can be identified, thereby obtaining long-term driving energy consumption data. Such energy consumption data can provide the energy consumption of new energy vehicles during actual driving, which is helpful for analyzing and evaluating the energy efficiency of vehicles. The feature set division of long-term driving energy consumption data in different time series windows can divide the data according to different time windows, such as by hour, day or week. In this way, the driving condition energy consumption data set under different time scales can be obtained, and the energy consumption characteristics of new energy vehicles can be further refined and analyzed. The feature set division of different time series windows helps to discover the relationship between energy consumption and time scale, and provides a more targeted reference for energy consumption optimization and planning. By migrating the driving condition simulation model through the driving condition energy consumption dataset, the abstract driving condition model can be transformed into a visual form, so as to more intuitively display the energy consumption of new energy vehicles under different driving conditions. Such a visual model can help users and decision makers better understand and analyze the energy consumption characteristics of new energy vehicles, and provide visual support and guidance for formulating energy consumption optimization strategies and driving planning.
[0023] In this embodiment, reference Figure 1 The above is a schematic flow chart of the steps of the method for constructing and evaluating the driving condition of a new energy vehicle according to the present invention. The method for constructing and evaluating the driving condition of a new energy vehicle comprises the following steps:
[0024] Step S1: using the Internet of Vehicles technology to collect driving data of new energy vehicles, thereby obtaining a driving original data set; performing outlier detection, filtering and integration of abnormal time series fluctuations on the driving original data set, thereby obtaining a driving condition division data set;
[0025] In this embodiment, data of new energy vehicles is collected by using Internet of Vehicles technology, including vehicle-mounted sensors, GPS, etc. These data include vehicle speed, acceleration, location information, vehicle environment information, road information in electronic map data, weather information, battery information, vehicle parameters, etc. The collected driving data is preprocessed and cleaned, abnormal data and outliers are removed, and abnormal detection of time series fluctuations is performed to identify and filter out abnormal fluctuations that occur during driving, such as abnormal behaviors such as sudden acceleration or deceleration.
[0026] Step S2: constructing a multimodal driving habit state for the driving condition division data set, thereby obtaining a driving habit mimicry model set; performing mimicry feature vector correction on the driving habit mimicry model set, and performing model fusion, thereby obtaining a driving condition mimicry model;
[0027] In this embodiment, the driving habit state of new energy vehicles is constructed by dividing the data set by driving conditions. This includes the extraction and analysis of the vehicle's acceleration, braking, steering and other behavioral characteristics, converting the driving habit state into a multimodal representation, considering the association and interaction between different behavioral characteristics, and constructing a driving habit mimicry model set. The driving habit mimicry model set is subjected to feature vector correction, and is adjusted and optimized according to the feedback of actual driving data to improve the accuracy and reliability of the model. The corrected mimicry model is subjected to model fusion, and the information of multiple modes is comprehensively considered to obtain a driving condition mimicry model for analysis and evaluation in subsequent steps.
[0028] Step S3: performing a four-dimensional index evaluation based on the driving condition simulation model to obtain four-dimensional comprehensive index data; performing a comprehensive evaluation on the driving condition division data set according to the four-dimensional comprehensive index data to obtain four-dimensional comprehensive evaluation data;
[0029] In this embodiment, a four-dimensional index evaluation is performed on the driving condition of new energy vehicles based on the driving condition simulation model. The four-dimensional index can include considerations of energy consumption, working failures, driving speed and safety; it is used to describe and measure the comprehensive performance and characteristics of the driving condition, and the four-dimensional comprehensive index data is used to conduct a comprehensive evaluation of the driving condition division data set, evaluate the advantages and disadvantages and applicable scope of different driving conditions, and provide quantitative information on the comprehensive performance of the driving condition.
[0030] Step S4: dividing the driving condition division data set into actual driving data based on the driving condition simulation model, thereby obtaining long-term actual driving data; performing feature discrimination of the long-term actual driving data that is strongly correlated with energy consumption, thereby obtaining long-term driving energy consumption data; performing feature set division of the long-term driving energy consumption data into different time series windows, thereby obtaining a driving condition energy consumption data set;
[0031] In this embodiment, based on the driving condition simulation model, the long-term driving data in the driving condition division data set are matched and matched to obtain the long-term actual driving data. For the long-term actual driving data, the feature discrimination strongly related to energy consumption is performed to identify the driving features closely related to energy consumption, including factors such as vehicle speed, acceleration, road and environment; the long-term driving energy consumption data is divided into feature sets of different time series windows, and the continuous driving data is divided and counted according to the time window to form a driving condition energy consumption data set for the construction of the energy consumption model and the visual analysis of the driving condition.
[0032] Step S5: performing a visualization model migration of the driving condition of the new energy vehicle to the driving condition simulation model according to the driving condition energy consumption data set, thereby obtaining a driving condition visualization model.
[0033] In this embodiment, by utilizing the driving condition energy consumption data set, the driving characteristics and energy consumption information therein are associated and migrated with the driving condition simulation model. The migrated driving condition simulation model can more accurately describe and reflect the energy consumption of new energy vehicles under long-term use under different driving conditions.
[0034] In this embodiment, reference Figure 2 The above is a schematic flow chart of the detailed implementation steps of step S1. In this embodiment, the detailed implementation steps of step S1 include:
[0035] Step S11: using the Internet of Vehicles technology to collect driving data of new energy vehicles, thereby obtaining a driving original data set, wherein the driving original data set includes driving condition original data, driving vehicle basic data and driving road environment data, wherein the driving condition original data includes speed data, acceleration data and braking data, the driving vehicle basic data includes vehicle performance parameters, energy consumption data, vehicle diagnostic data and battery status data, and the driving road environment data includes road type data, traffic condition data and driving environment data;
[0036] Step S12: dividing the velocity data and the acceleration data into behavioral features, thereby obtaining basic behavioral data;
[0037] Step S13: performing braking effect behavior feature classification on the braking data and the basic behavior data, thereby obtaining driving behavior classification data;
[0038] Step S14: performing regular filtering on the driving behavior classification data according to the preset regular rule data, thereby obtaining driving behavior filtering data;
[0039] Step S15: performing a correlation statistical clustering on the driving behavior filtering data and the driving vehicle basic data, thereby obtaining first driving clustering data;
[0040] Step S16: performing secondary statistical clustering of the driving behavior filtering data and the driving road environment data based on correlation, thereby obtaining second driving clustering data;
[0041] Step S17: performing time series characteristic change analysis on the first driving clustering data and the second driving clustering data, thereby obtaining time series change clustering data;
[0042] Step S18: performing outlier detection, filtering and integration of abnormal time series fluctuations on the first driving clustering data and the second driving clustering data according to the time series variation clustering data, so as to obtain a driving condition division data set.
[0043] In this embodiment, data of new energy vehicles is collected by using vehicle networking technology. The collected data includes original data of driving conditions, basic data of driving vehicles, and environmental data of driving roads. The original data of driving conditions include speed data, acceleration data, and braking data, which can be obtained through devices such as vehicle-mounted sensors. The basic data of driving vehicles include vehicle performance parameters, energy consumption data, vehicle diagnostic data, and battery status data, which can be obtained through the vehicle's electronic control unit (ECU) or vehicle-mounted monitoring system. The environmental data of driving roads include road type data, traffic condition data, and driving environment data, which can be obtained through the vehicle's GPS positioning system, map data, and environmental sensors. The collected speed data and acceleration data are processed and analyzed, and different behavior characteristics are divided according to the changing rules of speed data and acceleration data, such as uniform speed driving, acceleration, deceleration, etc. The basic behavior data is extracted and sorted, and different behavior characteristics are associated with corresponding time periods and values to form a basic behavior data set. Combined with the basic behavior data, the braking data is associated with the behavior characteristics, and different braking effect behavior characteristics are divided, such as light braking, emergency braking, etc. The driving behavior division data is extracted and sorted, and different braking effect behavior characteristics are associated with corresponding time periods and values to form a driving behavior division data set. Set the preset regular rule data, including the reasonable range of driving behavior, abnormal threshold, etc., compare and filter the driving behavior division data with the preset regular rule data, and filter out the data that does not conform to the rules. The driving behavior filtering data after regular filtering is obtained for analysis and processing in subsequent steps, including but not limited to setting a reasonable speed range for new energy vehicles, and data beyond this range can be filtered out. For example, limit the maximum speed and minimum speed to exclude abnormal or unreasonable speed data; set a reasonable range of acceleration of new energy vehicles, and data beyond this range can be filtered out. This helps to exclude abnormal acceleration or deceleration behavior data; define the effect range of braking, such as light braking, moderate braking and emergency braking levels. By setting different braking effect thresholds, abnormal or unreasonable braking behavior data can be filtered out; set the energy consumption range of new energy vehicles, and data beyond this range can be filtered out. This can prevent abnormal or unreasonable data from interfering with the analysis results. Set the battery status range, such as charging status, discharging status, battery capacity, etc. By setting a reasonable battery status range, you can filter out data that does not meet the conditions. Define reasonable road types, such as urban roads, highways, rural roads, etc. Filtering data according to preset road types can make the analysis more accurate and reliable; set reasonable traffic conditions, such as traffic flow, congestion level, etc. By setting different traffic condition thresholds, you can filter out abnormal or unreasonable traffic condition data; define reasonable environmental conditions, such as temperature, humidity, air pressure, etc. Filtering data according to preset environmental conditions can exclude abnormal data that is greatly affected by the environment.The relevant data can be filtered according to the acquisition time period. For example, if the parking time exceeds the preset threshold, the data of the vehicle within a certain period will not be acquired, but the data will be divided into two sections to acquire the driving data of the minimum time unit. The correlation between the driving behavior filtering data and the driving vehicle basic data is calculated using statistical methods (such as correlation coefficient). According to the correlation result, the data is clustered, and the data with similar correlation characteristics are classified into the same category to obtain the first driving cluster data, that is, the data with similar driving behavior and vehicle basic characteristics are clustered together to form the first driving cluster data and the second driving cluster data. The first driving cluster data and the second driving cluster data are subjected to time series feature analysis, including the time series changes of driving behavior and the time series changes of road environment. Fourier transform is used to analyze the changes in time domain and frequency domain in the time series. According to the changes in time series features, the data are clustered, and the data with similar time series change patterns are classified into the same category. To detect outliers of abnormal time series fluctuations in time series change clustering data, statistical methods or machine learning algorithms can be used to filter out abnormal time series fluctuation data, retain data that conform to normal driving conditions, and integrate the first driving clustering data and the second driving clustering data that conform to normal driving conditions to form a driving condition division data set.
[0044] The present invention collects driving data of new energy vehicles through the Internet of Vehicles technology, and can obtain data including driving conditions, basic vehicle information and road environment. The original data of driving conditions provides detailed records of driving behaviors such as speed, acceleration and braking, which is helpful for analyzing and understanding the dynamic behavior of the vehicle. The basic data of driving vehicles provides important information about vehicle performance, energy consumption, diagnosis and battery status, and provides basic data support for subsequent analysis and evaluation. The driving road environment data provides background information about road types, traffic conditions and driving environments, which is helpful for analyzing the driving performance of vehicles in different environments. Behavior feature division of speed data and acceleration data can extract features related to vehicle behavior, such as acceleration peak value, speed change rate, etc. These basic behavior data can reflect the driving style and driving characteristics of new energy vehicles, which is helpful for subsequent driving behavior analysis and model construction. The braking effect behavior feature division of braking data and basic behavior data can extract features related to braking effect, such as braking distance, braking time, etc. These driving behavior division data can help evaluate the braking performance and driving behavior of new energy vehicles, and provide a basis for the optimization and safety evaluation of the braking system. Regular filtering of driving behavior segmentation data by preset regular rule data can exclude data points that do not conform to the rules or are abnormal, ensuring the accuracy and reliability of the data. After regular filtering, the driving behavior filtering data can better reflect the normal driving behavior of new energy vehicles and improve the quality of subsequent analysis and modeling results. By performing a statistical clustering of the correlation of the driving behavior filtering data and the basic data of the driving vehicle, data points with similar behavior patterns and characteristics can be clustered and grouped. In this way, similar driving behaviors and vehicle characteristics can be aggregated to form the first driving clustering data. This step helps to reduce the dimension and organize large-scale data, and extract the main patterns and trends of driving behaviors and vehicle characteristics. By performing a secondary statistical clustering of the correlation of the driving behavior filtering data and the driving road environment data, the correlation between the driving behavior and the road environment can be analyzed and aggregated. In this way, the second driving clustering data can be obtained, which contains the comprehensive characteristics and patterns of driving behavior and road environment. This step helps to more comprehensively understand the mutual influence relationship between driving behavior and road environment. By performing time series feature change analysis on the first driving clustering data and the second driving clustering data, the changing laws and trends of driving behavior and road environment over time can be revealed. In this way, time series change clustering data can be obtained, which contains the change patterns and characteristics of driving behavior and road environment in different time periods. This step helps to discover the periodicity, seasonality and other time series change laws of driving behavior and road environment. By performing outlier detection and filtering of abnormal time series fluctuations on the time series change clustering data, abnormal data points that do not conform to normal driving behavior and road environment can be identified.In this way, a driving condition classification data set can be obtained, which contains normal driving behavior and road environment data after anomaly detection and filtering. This step helps to extract representative and stable driving condition data, providing a reliable data basis for subsequent analysis and modeling.
[0045] In this embodiment, reference Figure 3 The above is a schematic flow chart of detailed implementation steps of step S15. In this embodiment, the detailed implementation steps of step S15 include:
[0046] Step S151: extracting abnormal conditions from vehicle diagnostic data to obtain vehicle abnormal data;
[0047] Step S152: performing time series fluctuation statistics on the battery status data and the energy consumption data, thereby obtaining battery energy consumption fluctuation data;
[0048] Step S153: performing phase correlation analysis on the battery energy consumption fluctuation data to obtain phase fluctuation data;
[0049] Step S154: performing vehicle performance evaluation on the vehicle performance parameters in the same time sequence according to the staged fluctuation data, thereby obtaining staged vehicle performance data;
[0050] Step S155: performing vehicle abnormality trend analysis on the staged vehicle performance data and vehicle abnormality data, thereby obtaining performance abnormality trend data;
[0051] Step S156: marking the abnormal trend of the staged vehicle performance data according to the abnormal performance trend data, thereby obtaining the abnormal performance trend data;
[0052] Step S157: Performing a statistical clustering of different driving behavior correlations on the performance abnormality trend data and the driving behavior filtering data, thereby obtaining first driving clustering data.
[0053] In this embodiment, by collecting vehicle diagnostic data, including sensor data, fault codes, etc., for each sensor or fault code, a threshold or rule is set to determine whether it is an abnormal situation, the diagnostic data is processed and screened, and the data that meets the abnormal situation is extracted to form a vehicle abnormal data set. Collect battery status data, such as battery voltage, current, etc., as well as energy consumption data, and perform time series fluctuation statistics on the battery status data and energy consumption data, such as calculating the mean, standard deviation, fluctuation amplitude and other indicators. The battery energy consumption fluctuation data is divided into different time periods or stages, and can be divided according to the characteristics of the fluctuation, such as the battery energy consumption fluctuation characteristics of driving in different time periods such as morning, morning, noon, afternoon, evening, night, and early morning, and the energy consumption changes sharply, the battery power off changes sharply, etc., and the data is divided into stages, and the fluctuation data of each time period is analyzed for correlation, and the fluctuation conditions between different time periods are compared. Collect vehicle performance parameter data, such as acceleration, braking effect, energy consumption, etc., match the vehicle performance parameter data with the staged fluctuation data to form data pairs of the same time series, and evaluate the vehicle performance based on the data pairs of the same time series, calculate the performance indicators or obtain the evaluation results, and form staged vehicle performance data. Comprehensively consider the staged vehicle performance data and vehicle abnormal data, observe their changing trends in different stages, analyze the time series characteristics of the staged vehicle performance data and vehicle abnormal data, such as trends, periodicity, abnormal points, etc., and extract the performance abnormality trend data for subsequent analysis and marking. Use the marking information of the performance abnormality trend data to mark the staged vehicle performance data, apply the abnormal trend mark to the staged vehicle performance data, and identify the abnormal data points or time periods. The performance abnormality trend data and the driving behavior filtering data are integrated to form a complete data set. The data set is clustered and analyzed using a clustering algorithm (such as K-means, hierarchical clustering, etc.). Data points with similar characteristics are divided into different clusters. According to the correlation of driving behaviors, the clusters are associated with different driving behaviors (such as urban driving, highway driving, rural driving, etc.) to obtain the first driving cluster data. Each cluster represents a specific driving behavior and contains the performance abnormality trend data related to the driving behavior.
[0054] The present invention can detect and identify abnormal problems existing in the vehicle, such as fault codes, warning information, etc., by extracting abnormal conditions from vehicle diagnostic data. In this way, vehicle abnormal data can be obtained for subsequent fault diagnosis and maintenance work. By timely discovering and handling vehicle abnormal conditions, the reliability and safety of the vehicle can be improved, and the occurrence of faults and accidents can be reduced. By performing time series fluctuation statistics on battery status data and energy consumption data, the performance and health of the battery can be analyzed and evaluated. In this way, battery energy consumption fluctuation data can be obtained, which includes the energy consumption changes of the battery in different time periods. These data are helpful in judging the use status, life and performance attenuation of the battery, and provide guidance for battery management and optimization. By performing stage correlation analysis on battery energy consumption fluctuation data, the energy consumption data can be divided and analyzed in different stages. In this way, stage fluctuation data can be obtained, which includes the energy consumption characteristics and patterns of the battery in different stages. These data are helpful in revealing the energy consumption change law of the battery under different working conditions and usage conditions, and provide a reference basis for battery management and optimization. By evaluating the vehicle performance parameters in the same time series according to the stage fluctuation data, the performance of the vehicle in different stages can be analyzed and evaluated. In this way, staged vehicle performance data can be obtained, which includes performance indicators and parameters of the vehicle at different stages. These data are helpful to understand the performance characteristics and change trends of the vehicle under different working conditions and usage conditions, and provide a reference for vehicle design and optimization. By analyzing the staged vehicle performance data and vehicle abnormality data, the trend and change law of vehicle performance abnormality can be identified and analyzed. In this way, performance abnormality trend data can be obtained, which includes the trend characteristics and change trend of vehicle performance abnormality. These data are helpful to timely discover and deal with vehicle performance abnormalities, and provide guidance for preventing failures and improving vehicle reliability. By marking the abnormal trend of the staged vehicle performance data according to the performance abnormality trend data, the trend of performance abnormality can be marked and identified. In this way, performance abnormality trend data can be obtained, in which the stage and trend of vehicle performance abnormality are marked. These marked data are helpful to quickly locate and identify the causes and influencing factors of vehicle performance abnormality, and provide guidance for fault diagnosis and maintenance. By performing a correlation statistical clustering on the performance abnormality trend data and the driving behavior filtering data, data with similar performance abnormality trends and driving behaviors can be grouped and clustered. In this way, the first driving clustering data can be obtained, which includes a set of vehicle data with similar performance abnormality trends and driving behaviors. These clustering data help to identify and analyze the impact of different driving behaviors on vehicle performance abnormalities, and provide guidance for driving behavior optimization and vehicle management.
[0055] Preferably, the specific steps of step S16 are:
[0056] Step S161: integrating the road type data and traffic condition data for all traffic condition occurrence periods of the same road type, thereby obtaining road traffic collation data;
[0057] Step S162: feature mapping is performed on the road traffic collated data to obtain road traffic mapping data, wherein the feature mapping includes feature mapping of continuous traffic condition data of the same road type and road type data, feature mapping of different traffic condition data of the same road type and road type data, and feature mapping of traffic condition data of different road types and road type data;
[0058] Step S163: performing autoregressive correlation analysis on the driving environment data and the driving behavior filtering data, thereby obtaining driving behavior environment correlation data;
[0059] Step S164: performing multi-dimensional factor modeling on the driving environment data according to the driving behavior environment association data, thereby obtaining a multi-dimensional environmental factor model;
[0060] Step S165: constructing a long-short term memory network model for the road traffic mapping data based on the multidimensional environmental factor model, thereby obtaining a driving memory network model;
[0061] Step S166: dividing the driving behavior filtering data into driving behavior patterns under different driving conditions based on the driving memory network model, thereby obtaining driving behavior pattern data, wherein the driving behavior pattern data includes behavior state data, behavior type data, and traffic environment weight data;
[0062] Step S167: constructing a hidden Markov model for the behavior type data and the behavior state data according to the traffic environment weight data, thereby obtaining a driving behavior state model;
[0063] Step S168: performing secondary statistical clustering of the driving behavior filtering data and the driving road environment data based on the driving behavior state model, thereby obtaining second driving clustering data.
[0064] In this embodiment, by collecting road type data and traffic condition data, including traffic flow, congestion, accident data, etc., according to the road type, the same type of road traffic condition data is integrated by time period, such as by hour, by day, by working day, etc., and the integrated data forms road traffic collation data, which contains road traffic condition information in different time periods. Feature extraction is performed on the road traffic collation data, including features such as traffic flow, congestion degree, average vehicle speed, etc., feature mapping is performed on the continuous same traffic condition data and road type data in the same road type, and the association law between them is found, feature mapping is performed on different traffic condition data and road type data in the same road type, and the differences and connections between them are found, feature mapping is performed on traffic condition data and road type data in different road types, and traffic condition characteristics under different road types are explored, and road traffic mapping data is obtained, including the results of different feature mappings. Driving environment data, such as weather conditions, road conditions, traffic signals, etc., driving behavior filtering data, including driving behavior data such as vehicle speed, acceleration, and braking, are collected, and the autoregressive association analysis method is used to analyze the association between driving environment data and driving behavior filtering data, and their temporal relationship and influence are explored. Using the driving behavior environment-related data, multi-dimensional factor analysis is performed to summarize and extract the relevant driving environment factors, establish a multi-dimensional environmental factor model, and map the driving environment data to various environmental factors for subsequent analysis and modeling. Using the multi-dimensional environmental factor model, the road traffic mapping data is used as input to construct a long short-term memory network (LSTM) model. The LSTM model can capture the long-term dependencies and sequence patterns of road traffic data and is used to predict the driving data characteristics under different traffic conditions in the future. Using the driving memory network model, the driving behavior filtering data is divided into driving behavior patterns. According to the output of the model, the driving behavior filtering data is divided into different driving behavior patterns, including behavior state data (such as normal driving, emergency braking, speeding, etc.), behavior type data (such as urban roads, highways, rural roads, etc.) and traffic environment weight data (corresponding to the importance of each driving behavior pattern). Using the traffic environment weight data, the behavior type data and behavior state data are constructed into a Hidden Markov Model (HMM). The HMM model can be used to describe the conversion relationship between driving behavior states, and predict the next possible behavior state based on the current behavior type data and behavior state data, so as to optimize the relationship between the obtained driving behavior data and the driving environment, and predict the complete driving behavior characteristic state. Using the driving behavior state model, the driving behavior filtering data and the driving road environment data are subjected to secondary statistical clustering of correlation. Through statistical analysis and clustering algorithms, the data of driving behavior states and driving road environments with similarities are clustered together to obtain the second driving clustering data.
[0065] The present invention can summarize and organize the data of all traffic conditions in the same road type by integrating the road type data and the traffic condition data. In this way, road traffic organized data can be obtained, which includes a data set of all traffic conditions in the same road type. These organized data are helpful for analyzing and evaluating the traffic conditions of different road types, and provide a reference basis for traffic planning and optimization. By performing feature mapping on the road traffic organized data, the road traffic condition data can be associated and mapped with the road type data. In this way, road traffic mapping data can be obtained, which includes a data set under different feature mapping conditions. These mapping data are helpful for understanding and describing the characteristics and changing trends of traffic conditions in different road types. By performing autoregressive association analysis on the driving environment data and the driving behavior filtering data, the correlation relationship between the driving environment and the driving behavior can be analyzed and evaluated. In this way, driving behavior environment associated data can be obtained, which includes the correlation information between the driving environment and the driving behavior. These associated data are helpful for understanding the influence of the driving environment on the driving behavior. By performing multi-dimensional factor modeling on the driving environment data according to the driving behavior environment associated data, multiple influencing factors in the driving environment can be extracted and described. In this way, a multi-dimensional environmental factor model can be obtained, which includes environmental factors of different dimensions. These factor models help to understand and analyze the complexity of the driving environment and provide a comprehensive representation of environmental characteristics. By constructing a long-term and short-term memory network model for road traffic mapping data based on a multidimensional environmental factor model, a model with memory capacity can be established to predict road traffic conditions. In this way, a driving memory network model can be obtained, which can use historical data for long-term and short-term traffic predictions. This model helps to improve the accuracy and precision of traffic predictions. By dividing the driving behavior patterns of driving behavior filtering data based on the driving memory network model, driving behaviors under different driving conditions can be classified and identified. In this way, driving behavior pattern data can be obtained, which includes information such as behavior state, behavior type, and traffic environment weight. These pattern data help to understand and describe the characteristics and changes of different driving behavior patterns. By constructing a hidden Markov model for behavior type data and behavior state data based on traffic environment weight data, a driving behavior state model can be established. In this way, the state transfer and changes of driving behavior under different traffic environments can be described and predicted. The driving behavior state model helps to understand the dynamic characteristics of driving behavior. By performing secondary statistical clustering of the correlation of driving behavior filtering data and driving road environment data based on the driving behavior state model, data with high correlation can be clustered and classified. In this way, the second driving cluster data can be obtained, which includes a data set with similar characteristics and behavior patterns. These cluster data are helpful to discover and analyze the correlation between driving behavior and road environment.
[0066] Preferably, the specific steps of step S2 are:
[0067] Step S21: randomly dividing the driving condition divided data set into training data, thereby obtaining a first training data set and a second training data set;
[0068] Step S22: performing relevant weight marking of four-dimensional relevant features on the first training data set, thereby obtaining a four-dimensional weight marking data set, wherein the relevant weight marking of the four-dimensional relevant features includes a safety feature marking, a battery energy consumption feature marking, a working failure feature marking, and a driving speed feature marking;
[0069] Step S23: constructing a multimodal driving habit state for the four-dimensional weighted labeled data set, thereby obtaining a driving habit mimicry model set;
[0070] Step S24: performing mutual transformation mimicry among four dimensions based on the driving habit mimicry model set, thereby obtaining driving behavior-operating condition mimicry data;
[0071] Step S25: performing driving condition prediction of different driving behaviors on the driving behavior-condition simulation data, thereby obtaining condition simulation prediction data;
[0072] Step S26: performing generative adversarial training optimization on the operating condition simulation prediction data according to the second training data set, thereby obtaining operating condition simulation optimization data;
[0073] Step S27: Correct the mimetic feature vector of the driving habit mimetic model set according to the working condition simulation optimization data, and perform model fusion to obtain a driving condition mimetic model.
[0074] In this embodiment, the driving condition data set is randomly divided into two parts, namely, a first training data set and a second training data set. It is ensured that the samples between the two data sets are independent and randomly distributed. The first training data set is used to mark the four-dimensional related features therein with relevant weights. The importance of each relevant feature is evaluated by statistical analysis or machine learning methods, and converted into a weight mark. The relevant weight mark of the relevant feature is applied to the first training data set to obtain a four-dimensional weighted marked data set, which reflects the importance of different features in the driving condition. Based on the four-dimensional weighted marked data set, a multimodal driving habit state is constructed. For each relevant feature, the samples in the data set are classified or clustered to form different driving habit states. Combining the driving habit states of different features, a driving habit mimicry model set is obtained to describe the driving habits under different driving conditions. The driving habit mimicry model set is used to perform mutual transformation mimicry on the four-dimensional weighted marked data set. According to the model in the driving habit mimicry model set, the driving condition in the four-dimensional weighted marked data set is mimicked to obtain driving behavior-condition mimicry data, which reflects the mimicry of the driving condition under different driving habits. Based on the driving behavior-operating condition simulation data, the driving behavior and operating condition prediction are carried out. The model between driving behavior and driving condition is established based on the existing driving behavior-operating condition simulation data using machine learning or statistical methods. The driving condition of unknown driving behavior is predicted using this model to obtain operating condition simulation prediction data, which reflects the driving conditions under different driving behaviors. Using the second training data set, the operating condition simulation prediction data is compared with the actual driving condition data. Based on the generative adversarial network (GAN) or other optimization methods, the operating condition simulation prediction data is adjusted and optimized to make it closer to the actual driving condition data. Through generative adversarial training optimization, the operating condition simulation optimization data is obtained, which more accurately reflects the driving condition and the characteristic data related thereto. The operating condition simulation optimization data is used to correct the mimicry feature vector of the driving habit mimicry model set. According to the difference between the operating condition simulation optimization data and the original driving habit mimicry model set, the mimicry feature vector in the driving habit mimicry model set is adjusted. The corrected driving habit mimicry model set is model fused, and the mimicry feature vectors of different models are combined to obtain a driving condition mimicry model, which can more accurately describe the mimicry of the driving condition.
[0075] The present invention can divide the data set into two parts, namely, a first training data set and a second training data set, by randomly dividing the training data set for the driving condition division data set. In this way, the first training data set can be used to train the model, and the second training data set can be used to verify and evaluate the performance of the model. This division helps to accurately evaluate the generalization ability and prediction performance of the model. By marking the relevant weights of the four-dimensional related features of the first training data set, different features in the data set can be weighted. In this way, a four-dimensional weighted labeled data set can be obtained, which contains the labeled features and corresponding weights. These weight labels help to distinguish and optimize the degree of influence of different features on the driving conditions, and provide more accurate feature representation for subsequent modeling and prediction. By constructing a multimodal driving habit state for the four-dimensional weighted labeled data set, the driving habit state in the data set can be modeled and described. In this way, a driving habit mimicry model set can be obtained, which contains models of different driving habit states. These mimicry models help to understand and represent the driving characteristics and behavior patterns under different driving habit states, and provide a basis for subsequent prediction and optimization. By performing mutual transformation and mimicry between four dimensions based on the driving habit mimicry model set, the correlation between driving behavior and working conditions can be modeled and mimicked. In this way, driving behavior-working condition mimicry data can be obtained, which includes mimicry representations between driving behavior and working conditions. These mimicry data are helpful in analyzing and predicting the changes and adaptability of driving behavior under different working conditions. By predicting the driving working conditions of different driving behaviors on the driving behavior-working condition mimicry data, the driving working condition state under a given driving behavior can be inferred and predicted. In this way, working condition simulation prediction data can be obtained, which includes working condition prediction results under different driving behaviors. These prediction data are helpful in understanding and analyzing the impact of driving behavior on working conditions. By optimizing the working condition simulation prediction data through generative adversarial training based on the second training data set, the accuracy and fidelity of the working condition simulation can be improved. Generative adversarial training is a training method based on game theory. By allowing the generative model and the discriminative model to compete and optimize with each other, the data generated by the generative model is closer to the real data distribution. In this way, working condition simulation optimization data can be obtained, which includes working condition simulation results optimized by generative adversarial training. These optimized data help improve the accuracy and reliability of the working condition simulation. By correcting the mimetic feature vectors of the driving habit mimetic model set according to the working condition simulation optimization data and performing model fusion, the performance of the driving habit mimetic model can be further optimized. By combining the working condition simulation optimization data with the driving habit mimetic model, the feature vector representation in the mimetic model can be corrected to more accurately reflect the characteristics of the driving condition. At the same time, through model fusion, the prediction results of multiple driving habit mimetic models can be integrated to improve the prediction and generalization capabilities of the mimetic model. In this way, a driving condition mimetic model can be obtained, providing an effective modeling tool for driving condition analysis and optimization.
[0076] Preferably, the specific steps of step S26 are:
[0077] Step S261: extracting four-dimensional working condition related features from the working condition simulation prediction data, thereby obtaining four-dimensional working condition related data, wherein the four-dimensional working condition related data includes safety related data, battery energy consumption related data, working failure related data, and driving speed related data;
[0078] Step S262: performing speed characteristic evaluation of braking safety on the safety-related data and the driving speed-related data, thereby obtaining safety braking evaluation data;
[0079] Step S263: evaluating the speed fluctuation characteristics of energy consumption change and battery state aging on the battery energy consumption related data and the driving speed related data, thereby obtaining speed fluctuation evaluation data;
[0080] Step S264: evaluating the driving behavior characteristics of the fault trend on the working fault related data and the driving speed related data, thereby obtaining fault trend evaluation data;
[0081] Step S265: extracting correlation features from the second training data set according to the safety braking assessment data, the speed fluctuation assessment data, and the fault trend assessment data, thereby obtaining training-related data;
[0082] Step S266: constructing a generation model for the operating condition simulation prediction data, thereby obtaining a generation model;
[0083] Step S267: constructing an adversarial model for the training-related data and the second training data set, thereby obtaining an adversarial model;
[0084] Step S268: performing generative adversarial training on the generative model and the adversarial model, thereby obtaining a four-dimensional generative adversarial network model;
[0085] Step S269: Generate adversarial training optimization on the operating condition simulation prediction data based on the four-dimensional generative adversarial network model to obtain operating condition simulation optimization data.
[0086] In this embodiment, by extracting safety-related features such as braking distance, safe speed, etc. from the working condition simulation prediction data, extracting battery energy consumption-related features such as energy consumption change rate, battery capacity consumption, etc. from the working condition simulation prediction data, extracting working fault-related features such as fault occurrence frequency, fault type, etc. from the working condition simulation prediction data, extracting driving speed-related features such as average speed, maximum speed, etc. from the working condition simulation prediction data, the extracted features are combined to obtain four-dimensional working condition-related data, which includes safety-related data, battery energy consumption-related data, working fault-related data, and driving speed-related data. The safety-related data and driving speed-related data are used to evaluate the braking safety, and the safety is evaluated according to indicators such as braking distance, braking time, speed change rate, etc., to obtain safe braking evaluation data, which reflects the braking safety under driving conditions. Using the battery energy consumption related data and the driving speed related data, the speed fluctuation characteristics of energy consumption change and battery state aging are evaluated. According to the energy consumption change rate, battery capacity consumption, speed change rate and other indicators, the speed fluctuation is evaluated to obtain speed fluctuation evaluation data, which reflects the energy consumption change and battery state aging speed fluctuation under driving conditions. Using the working fault related data and the driving speed related data, the driving behavior characteristics of the fault trend are evaluated. According to the fault occurrence frequency, fault type, driving speed change rate and other indicators, the fault trend is evaluated to obtain fault trend evaluation data, which reflects the fault trend under driving conditions. Combined with the safety braking evaluation data, the speed fluctuation evaluation data and the fault trend evaluation data, the second training data set is subjected to correlation feature extraction, and the features related to safety braking, speed fluctuation and fault trend are extracted, such as braking distance, speed change rate, fault occurrence frequency, etc., to obtain training related data, which contains features related to safety braking, speed fluctuation and fault trend. Using the working condition simulation prediction data, a generative model is constructed, and the generative model is trained using appropriate machine learning or deep learning algorithms so that it can generate samples with similar working condition simulation prediction data. Using the training-related data and the second training data set, an adversarial model is constructed, and an appropriate adversarial generative network (GAN) algorithm is used to train the adversarial model so that it can generate samples with similar training-related data and the second training data set. The generative model and the adversarial model are subjected to generative adversarial training. By alternately training the generative model and the adversarial model, they compete and improve with each other to provide more accurate generated samples, and finally a four-dimensional generative adversarial network model is obtained, which can generate samples similar to the working condition simulation prediction data. The working condition simulation prediction data is optimized by generative adversarial training using the four-dimensional generative adversarial network model. By inputting the working condition simulation prediction data into the generative adversarial network model, samples more similar to the actual working condition simulation data are generated, and working condition simulation optimization data is obtained, which can be used as a substitute or supplement for more accurate working condition simulation prediction data.
[0087] The present invention can extract characteristic information related to the working condition by performing four-dimensional working condition related feature extraction on the working condition simulation prediction data. In this way, four-dimensional working condition related data can be obtained, which contains relevant information on safety, battery energy consumption, working failure and driving speed. These working condition related data are helpful to deeply understand and describe the characteristics of the working condition, and provide a basis for subsequent evaluation and optimization. By evaluating the speed characteristics of braking safety on the safety-related data and the driving speed-related data, the braking safety during driving can be evaluated. In this way, safe braking evaluation data can be obtained to reflect the braking performance and safety during driving. These evaluation data are helpful to understand the advantages and disadvantages of braking performance. By evaluating the speed fluctuation characteristics of energy consumption changes and battery state aging on the battery energy consumption-related data and the driving speed-related data, the energy consumption changes during driving and the stability of the battery state can be evaluated. In this way, speed fluctuation evaluation data can be obtained to reflect the fluctuation of energy consumption during driving and the state change of the battery. These evaluation data are helpful to understand the use of the battery and the degree of performance attenuation. By evaluating the driving behavior characteristics of the fault trend on the working failure-related data and the driving speed-related data, the fault trend and working state during driving can be evaluated. In this way, fault trend evaluation data can be obtained for analyzing fault behaviors and working condition changes during driving. These evaluation data are helpful in predicting and identifying potential fault conditions. By extracting correlation features from the second training data set based on the safety braking evaluation data, speed fluctuation evaluation data, and fault trend evaluation data, feature information related to these evaluation data can be extracted. In this way, training-related data can be obtained, which contains features related to safety braking, speed fluctuation, and fault trend. These training-related data can be used to establish models and conduct further analysis to improve the safety, performance, and reliability of the system. By constructing a generative model for the working condition simulation prediction data, a generative model can be established to simulate and generate data related to the working condition. In this way, more working condition data samples can be generated using the generative model to expand the training data set and perform data enhancement. The generative model can provide more sample diversity and coverage, which helps to improve the generalization ability and performance of the model. By constructing an adversarial model for the training-related data and the second training data set, an adversarial model can be established for model training and optimization. The adversarial model can improve the robustness and robustness of the model through adversarial training and reduce the sensitivity of the model to noise and interference. In this way, the adaptability and reliability of the model in actual scenarios can be improved. By performing generative adversarial training on the generative model and the adversarial model, the two models can interact and optimize with each other. Generative adversarial training can improve the quality and diversity of data generated by the generative model, and improve the performance of the generative model through feedback from the adversarial model. In this way, a four-dimensional generative adversarial network model can be obtained, which comprehensively utilizes the advantages of the generative model and the adversarial model and improves the generation and discrimination capabilities of the model.Generative adversarial training optimization of the working condition simulation prediction data based on the four-dimensional generative adversarial network model can further optimize the quality and accuracy of the working condition simulation data. Through the optimization of generative adversarial training, the generated working condition simulation data can be made more realistic and accurate, and more consistent with the actual working conditions. This can improve the reliability and simulation effect of the working condition simulation.
[0088] Preferably, the specific steps of step S3 are:
[0089] Step S31: performing a four-dimensional comprehensive importance ranking based on the driving condition simulation model, thereby obtaining importance sequence data;
[0090] Step S32: performing four-dimensional index evaluation on the importance sequence data, thereby obtaining four-dimensional comprehensive index data;
[0091] Step S33: Comprehensively evaluate the driving condition division data set according to the four-dimensional comprehensive index data, so as to obtain four-dimensional comprehensive evaluation data.
[0092] In this embodiment, the driving condition is modeled and simulated by using the driving condition simulation model. By simulating the driving condition, data related to four dimensions (including safety, energy consumption, fault trend, and driving speed) are obtained. The obtained data is comprehensively analyzed, including statistical analysis, machine learning, and other methods to determine the importance of different features in each dimension. According to the analysis results, the different features in the four dimensions are ranked by importance to obtain importance sequence data. According to the importance sequence data, the weights of different features in each dimension are determined, and the data of each dimension are normalized to ensure that the data are comparable. According to the weights and the normalized data, the comprehensive index of each dimension is calculated to obtain four-dimensional comprehensive index data, which reflects the importance and comprehensive performance of each dimension. The driving condition is divided into different data sets, such as a training set, a validation set, and a test set. According to the four-dimensional comprehensive index data, the driving condition in each data set is comprehensively evaluated. The comprehensive evaluation can use different methods, such as rule-based judgment, machine learning models, etc. According to the comprehensive evaluation results, four-dimensional comprehensive evaluation data is obtained, which reflects the comprehensive performance of each driving condition.
[0093] The present invention can determine the importance of different operating conditions in the whole by ranking the importance of four-dimensional comprehensiveness based on the driving condition simulation model. In this way, importance sequence data can be obtained to guide subsequent analysis and decision-making. The importance sequence data reflects the contribution of each operating condition factor in the whole, which helps to optimize and adjust the parameters and weights of the operating condition simulation, and improve the accuracy and reliability of the operating condition simulation. By performing four-dimensional index evaluation on the importance sequence data, the importance and performance of different operating conditions can be comprehensively considered. In this way, four-dimensional comprehensive index data can be obtained for comprehensive evaluation of the characteristics and performance of driving conditions. The four-dimensional comprehensive index data combines the weights and performances of various operating conditions factors, and can provide more comprehensive and accurate operating condition evaluation results, which helps to deeply understand the characteristics and influencing factors of driving conditions. By performing a comprehensive evaluation on the driving condition division data set according to the four-dimensional comprehensive index data, different driving conditions can be classified and evaluated. In this way, four-dimensional comprehensive evaluation data can be obtained to describe and compare the characteristics and performance of different driving conditions. The four-dimensional comprehensive evaluation data can help users understand the advantages and disadvantages and applicable scope of driving conditions, and provide guidance and support for selecting appropriate conditions and formulating corresponding strategies.
[0094] Preferably, the specific steps of step S4 are:
[0095] Step S41: extracting state features when the simulated driving time is extended according to a preset time extension threshold and based on a driving condition simulation model, thereby obtaining long-time simulated driving data;
[0096] Step S42: dividing the driving condition division data set into actual driving data of corresponding extended time according to a preset time extension threshold, so as to obtain long-term actual driving data;
[0097] Step S43: performing feature discrimination of the long-term simulated driving data and the long-term actual driving data with strong correlation to energy consumption according to the four-dimensional comprehensive evaluation data, thereby obtaining the long-term driving energy consumption data;
[0098] Step S44: Divide the long-term driving energy consumption data into feature sets of different time series windows, so as to obtain a driving condition energy consumption data set.
[0099] In this embodiment, according to a preset time extension threshold, the driving time in the driving condition simulation model is extended to the threshold, wherein the time extension threshold refers to multiple different thresholds. In the simulated driving process, data related to the state characteristics are extracted based on the simulation model. These characteristics may include vehicle speed, acceleration, steering angle, fuel consumption, etc. By simulating the process of extending the driving time, long-term simulated driving data is obtained, which includes the complete driving process data from the starting point of driving to the end of the extension time. According to the preset time extension threshold, the driving time in the actual driving data is extended to the threshold, and for each driving condition in the driving condition division data set, the actual driving time is extended to the preset time extension threshold, so that the actual driving data of the corresponding extended time for each driving condition is obtained, which includes the complete driving process data from the starting point of driving to the end of the extension time. The time extension threshold can be adjusted according to the actual long-term driving data set obtained. If the actual long-term driving data set obtained is relatively small, and according to the existing data set training and statistical judgment in related fields, it is concluded that the energy consumption and driving behavior caused by such long-term driving are affected, then the threshold needs to be adjusted to the threshold point with strong influence as much as possible. The energy consumption-related features in the four-dimensional comprehensive evaluation data are used to distinguish the features of the long-term simulated driving data and the long-term actual driving data. According to the distinguishing results, the features strongly related to energy consumption, such as driving speed, acceleration, engine load, etc., are extracted from the long-term simulated driving data and the long-term actual driving data to obtain the long-term driving energy consumption data. The long-term driving energy consumption data is divided into different time series windows, such as every hour, every minute or every second. According to the correlation of the features in different time periods, that is, the correlation with energy consumption and driving behavior, the time window with strong correlation is selected, and the genetic forest algorithm may be used to make the time window as large as possible while keeping the data correlation high. For the data in each time window, the feature set is extracted, including the average energy consumption, the maximum energy consumption, the energy consumption fluctuation, etc., so that the driving condition energy consumption data set is obtained, in which the feature set corresponding to different time windows reflects the driving condition energy consumption in the window, ensuring the accuracy of the condition data during long-term driving.
[0100] The present invention can simulate long-term driving conditions by extracting state features when simulating driving time extension according to a preset time extension threshold and based on a driving condition simulation model. In this way, long-term simulated driving data can be obtained, including state features of driving conditions within the extended time range. Long-term simulated driving data can be used to analyze the changing rules of driving conditions on different time scales, and help predict and evaluate the performance and characteristics of long-term driving. By dividing the driving condition division data set into actual driving data of corresponding extended time according to the preset time extension threshold, long-term actual driving data can be obtained. In this way, long-term actual driving data can be obtained, including driving records of actual driving conditions within the extended time range. Long-term actual driving data can be used to verify the accuracy and reliability of simulation results, and to evaluate the performance and characteristics of actual driving conditions on a long-term scale. By distinguishing features strongly related to energy consumption for long-term simulated driving data and long-term actual driving data according to four-dimensional comprehensive evaluation data, features strongly related to energy consumption can be extracted. In this way, long-term driving energy consumption data can be obtained, which is used to describe and compare the energy consumption of long-term driving conditions. Long-term driving energy consumption data can help analyze and optimize energy consumption, as well as evaluate the impact of different driving conditions on energy consumption, and provide guidance and basis for energy management and energy-saving optimization. By dividing the long-term driving energy consumption data into feature sets of different time series windows, driving condition energy consumption data sets on different time scales can be obtained. In this way, a driving condition energy consumption data set can be obtained, which is used to analyze and compare energy consumption characteristics and trends under different time windows. The driving condition energy consumption data set can help understand and grasp the energy consumption laws of driving conditions, and provide more detailed and accurate data support for energy management and optimization.
[0101] Preferably, the specific steps of step S43 are:
[0102] Step S431: Comprehensively evaluate the long-term actual driving data according to the four-dimensional comprehensive index data, thereby obtaining long-term actual evaluation data;
[0103] Step S432: sorting the importance of the long-time simulated driving data with strong correlation to energy consumption, thereby obtaining long-time energy consumption series data;
[0104] Step S433: Calculating the evaluation difference of the long-term actual evaluation data and the four-dimensional comprehensive evaluation data, thereby obtaining evaluation difference data;
[0105] Step S434: feature identification is performed on the long-term energy consumption series data according to the evaluation difference data, so as to obtain the long-term driving energy consumption data.
[0106] In this embodiment, the long-term actual driving data is evaluated by using four-dimensional comprehensive index data. The evaluation method may include rule-based judgment, machine learning model, etc., and according to the evaluation results, the long-term actual evaluation data is obtained, which reflects the comprehensive performance of the long-term actual driving data. The features related to energy consumption in the long-term simulated driving data are used to sort the importance. Statistical analysis, machine learning and other methods can be used to obtain long-term energy consumption sequence data according to the sorting results, which reflects the order of importance of energy consumption of long-term simulated driving data. The long-term actual evaluation data is compared with the four-dimensional comprehensive evaluation data to calculate the evaluation difference. The difference or percentage can be used for calculation, and the evaluation difference data reflects the difference between the long-term actual driving data and the comprehensive evaluation data, which can be used for further analysis and judgment. The evaluation difference data is used to perform feature discrimination to determine the features in the long-term energy consumption sequence data that are strongly correlated with the evaluation difference. Statistical analysis, machine learning and other methods can be used to extract features related to energy consumption according to the discrimination results to obtain long-term driving energy consumption data.
[0107] The present invention can evaluate the characteristics and performance of long-term actual driving by comprehensively evaluating the long-term actual driving data according to the four-dimensional comprehensive index data. In this way, long-term actual evaluation data can be obtained to describe and compare the characteristics and performance of long-term actual driving conditions. The long-term actual evaluation data can help analyze the advantages and disadvantages and applicable scope of the actual driving conditions, and provide guidance and support for driving strategies and performance improvements. By sorting the importance of long-term simulated driving data with strong correlation to energy consumption, the importance order of energy consumption in different time periods can be determined. In this way, long-term energy consumption sequence data can be obtained to guide energy consumption optimization and analysis. The long-term energy consumption sequence data reflects the relative contribution of energy consumption in different time periods, which helps to identify and improve the time periods or factors with a high proportion of energy consumption in driving conditions, and improve energy utilization efficiency. By calculating the evaluation difference of the long-term actual evaluation data and the four-dimensional comprehensive evaluation data, the difference between the actual driving data and the comprehensive evaluation data can be compared. In this way, evaluation difference data can be obtained to analyze the gap and error between actual driving and simulated evaluation. Evaluation difference data can help discover the limitations and improvement directions of the evaluation model, and improve the accuracy and reliability of the evaluation. By distinguishing the characteristics of the long-term energy consumption series data based on the evaluation difference data, the deviations and errors in the long-term driving energy consumption data can be identified and adjusted. In this way, the long-term driving energy consumption data can be obtained to more accurately describe and analyze the energy consumption of driving conditions. The long-term driving energy consumption data can help understand and grasp the energy consumption patterns and trends of long-term driving.
[0108] Preferably, the specific steps of step S5 are:
[0109] Step S51: performing a visualization model migration of the driving condition of the new energy vehicle according to the driving condition energy consumption data set to the driving condition mimic model, thereby obtaining a driving condition visualization model;
[0110] Step S52: Performing operating condition standard evaluation based on the driving condition visualization model to obtain comprehensive operating condition evaluation data.
[0111] In this embodiment, by collecting a driving condition energy consumption data set, which includes energy consumption information under different driving conditions, the driving condition energy consumption data set is used to train and transfer learning the driving condition mimicry model. A machine learning algorithm, such as a deep learning model, can be used to build a visualization model. During the training process, the driving condition energy consumption data set is matched with the driving condition mimicry model to achieve the visualization model migration of the driving condition, and a driving condition visualization model is obtained, which can convert the input driving condition into a visualization display, such as a chart, a curve, etc. Using the driving condition visualization model, the data of different driving conditions are input into the model for visualization display, and the driving condition is evaluated according to the results of the visualization display. The evaluation can be based on preset condition standards, such as energy consumption, stability, comfort and other indicators. After evaluating each driving condition, comprehensive condition evaluation data is obtained, which reflects the overall performance and quality of different driving conditions.
[0112] The present invention can convert the actual driving condition into a visualization model by migrating the driving condition simulation model into a visualization model according to the driving condition energy consumption data set. In this way, a driving condition visualization model can be obtained, which is used to intuitively display the characteristics and energy consumption of the driving condition. The driving condition visualization model can help analyze and compare the differences and similarities between different driving conditions, and provide a visual reference and basis for driving condition evaluation and optimization. By evaluating the working condition standard based on the driving condition visualization model, the driving condition can be comprehensively evaluated. In this way, comprehensive evaluation data of the working condition can be obtained, which is used to describe and compare the comprehensive performance and characteristics of different driving conditions. The comprehensive evaluation data of the working condition can evaluate the driving condition based on multiple indicators and standards, help determine the advantages and disadvantages and applicable scope of the driving condition, and provide reference and guidance for driving strategy and performance improvement. At the same time, the use of the visualization model can make the evaluation results more intuitive and easy to understand, and improve the effect and credibility of the evaluation.
[0113] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.
[0114] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.
[0115] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.
Claims
1. A method for constructing and evaluating driving conditions of new energy vehicles. It is characterized in that The following steps are involved: Step S1: using the Internet of Vehicles technology to collect driving data of new energy vehicles, thereby obtaining a driving original data set; performing outlier detection, filtering and integration of abnormal time series fluctuations on the driving original data set, thereby obtaining a driving condition division data set; Step S2: constructing a multimodal driving habit state for the driving condition division data set, thereby obtaining a driving habit mimicry model set; performing mimicry feature vector correction on the driving habit mimicry model set, and performing model fusion, thereby obtaining a driving condition mimicry model; Step S3: performing a four-dimensional index evaluation based on the driving condition simulation model to obtain four-dimensional comprehensive index data; performing a comprehensive evaluation on the driving condition division data set according to the four-dimensional comprehensive index data to obtain four-dimensional comprehensive evaluation data; Step S4: dividing the driving condition division data set into actual driving data based on the driving condition simulation model, thereby obtaining long-term actual driving data; The energy consumption-related features of the actual long-term driving data are identified to obtain the energy consumption data of the long-term driving; The long-term driving energy consumption data is divided into feature sets of different time series windows to obtain the driving condition energy consumption data set; Step S5: performing a visualization model migration of the driving condition of the new energy vehicle to the driving condition simulation model according to the driving condition energy consumption data set, thereby obtaining a driving condition visualization model.
2. The method for constructing and evaluating driving conditions of new energy vehicles according to claim 1, It is characterized in that The specific steps of step S1 are: Step S11: using the Internet of Vehicles technology to collect driving data of new energy vehicles, thereby obtaining a driving original data set, wherein the driving original data set includes driving condition original data, driving vehicle basic data and driving road environment data, wherein the driving condition original data includes speed data, acceleration data and braking data, the driving vehicle basic data includes vehicle performance parameters, energy consumption data, vehicle diagnostic data and battery status data, and the driving road environment data includes road type data, traffic condition data and driving environment data; Step S12: dividing the velocity data and the acceleration data into behavioral features, thereby obtaining basic behavioral data; Step S13: performing braking effect behavior feature classification on the braking data and the basic behavior data, thereby obtaining driving behavior classification data; Step S14: performing regular filtering on the driving behavior classification data according to the preset regular rule data, thereby obtaining driving behavior filtering data; Step S15: performing a correlation statistical clustering on the driving behavior filtering data and the driving vehicle basic data, thereby obtaining first driving clustering data; Step S16: performing secondary statistical clustering of the driving behavior filtering data and the driving road environment data based on correlation, thereby obtaining second driving clustering data; Step S17: performing time series characteristic change analysis on the first driving clustering data and the second driving clustering data, thereby obtaining time series change clustering data; Step S18: performing outlier detection, filtering and integration of abnormal time series fluctuations on the first driving clustering data and the second driving clustering data according to the time series variation clustering data, so as to obtain a driving condition division data set.
3. The method for constructing and evaluating driving conditions of new energy vehicles according to claim 2, It is characterized in that The specific steps of step S15 are: Step S151: extracting abnormal conditions from vehicle diagnostic data to obtain vehicle abnormal data; Step S152: performing time series fluctuation statistics on the battery status data and the energy consumption data, thereby obtaining battery energy consumption fluctuation data; Step S153: performing stage correlation analysis on the battery energy consumption fluctuation data to obtain stage fluctuation data; Step S154: performing vehicle performance evaluation on the vehicle performance parameters in the same time sequence according to the staged fluctuation data, thereby obtaining staged vehicle performance data; Step S155: performing vehicle abnormality trend analysis on the staged vehicle performance data and the vehicle abnormality data, thereby obtaining performance abnormality trend data; Step S156: marking the abnormal trend of the staged vehicle performance data according to the abnormal performance trend data, thereby obtaining the abnormal performance trend data; Step S157: Performing statistical clustering of different driving behavior correlations on the performance abnormality trend data and the driving behavior filtering data, thereby obtaining first driving clustering data.
4. The method for constructing and evaluating driving conditions of new energy vehicles according to claim 2, It is characterized in that The specific steps of step S16 are: Step S161: integrating the road type data and traffic condition data for all traffic condition occurrence periods of the same road type, thereby obtaining road traffic collation data; Step S162: feature mapping is performed on the road traffic collated data to obtain road traffic mapping data, wherein the feature mapping includes feature mapping of continuous traffic condition data of the same road type and road type data, feature mapping of different traffic condition data of the same road type and road type data, and feature mapping of traffic condition data of different road types and road type data; Step S163: performing autoregressive correlation analysis on the driving environment data and the driving behavior filtering data, thereby obtaining driving behavior environment correlation data; Step S164: performing multi-dimensional factor modeling on the driving environment data according to the driving behavior environment association data, thereby obtaining a multi-dimensional environmental factor model; Step S165: constructing a long-short term memory network model for the road traffic mapping data based on the multidimensional environmental factor model, thereby obtaining a driving memory network model; Step S166: dividing the driving behavior filtering data into driving behavior patterns under different driving conditions based on the driving memory network model, thereby obtaining driving behavior pattern data, wherein the driving behavior pattern data includes behavior state data, behavior type data, and traffic environment weight data; Step S167: constructing a hidden Markov model for the behavior type data and the behavior state data according to the traffic environment weight data, thereby obtaining a driving behavior state model; Step S168: performing secondary statistical clustering of the driving behavior filtering data and the driving road environment data based on the driving behavior state model, thereby obtaining second driving clustering data.
5. The method for constructing and evaluating driving conditions of new energy vehicles according to claim 1, It is characterized in that The specific steps of step S2 are: Step S21: randomly dividing the driving condition divided data set into training data, thereby obtaining a first training data set and a second training data set; Step S22: performing relevant weight marking of four-dimensional relevant features on the first training data set, thereby obtaining a four-dimensional weight marking data set, wherein the relevant weight marking of the four-dimensional relevant features includes a safety feature marking, a battery energy consumption feature marking, a working failure feature marking, and a driving speed feature marking; Step S23: constructing a multimodal driving habit state for the four-dimensional weighted labeled data set, thereby obtaining a driving habit mimicry model set; Step S24: performing mutual transformation mimicry among four dimensions based on the driving habit mimicry model set, thereby obtaining driving behavior-operating condition mimicry data; Step S25: performing driving condition prediction of different driving behaviors on the driving behavior-operating condition simulation data, thereby obtaining operating condition simulation prediction data; Step S26: performing generative adversarial training optimization on the operating condition simulation prediction data according to the second training data set, so as to obtain operating condition simulation optimization data; Step S27: Correct the mimetic feature vector of the driving habit mimetic model set according to the working condition simulation optimization data, and perform model fusion to obtain a driving condition mimetic model.
6. The method for constructing and evaluating driving conditions of new energy vehicles according to claim 5, It is characterized in that The specific steps of step S26 are: Step S261: extracting four-dimensional working condition related features from the working condition simulation prediction data, thereby obtaining four-dimensional working condition related data, wherein the four-dimensional working condition related data includes safety related data, battery energy consumption related data, working failure related data, and driving speed related data; Step S262: performing speed characteristic evaluation of braking safety on the safety-related data and the driving speed-related data, thereby obtaining safety braking evaluation data; Step S263: evaluating the speed fluctuation characteristics of energy consumption change and battery state aging on the battery energy consumption related data and the driving speed related data, thereby obtaining speed fluctuation evaluation data; Step S264: evaluating the driving behavior characteristics of the fault trend on the working fault related data and the driving speed related data, thereby obtaining fault trend evaluation data; Step S265: extracting correlation features from the second training data set according to the safety braking assessment data, the speed fluctuation assessment data, and the fault trend assessment data, thereby obtaining training-related data; Step S266: constructing a generation model for the operating condition simulation prediction data, thereby obtaining a generation model; Step S267: constructing an adversarial model for the training-related data and the second training data set, thereby obtaining an adversarial model; Step S268: performing generative adversarial training on the generative model and the adversarial model, thereby obtaining a four-dimensional generative adversarial network model; Step S269: Generate adversarial training optimization on the operating condition simulation prediction data based on the four-dimensional generative adversarial network model to obtain operating condition simulation optimization data.
7. The method for constructing and evaluating driving conditions of new energy vehicles according to claim 1, It is characterized in that The specific steps of step S3 are: Step S31: performing a four-dimensional comprehensive importance ranking based on the driving condition simulation model, thereby obtaining importance sequence data; Step S32: performing four-dimensional index evaluation on the importance sequence data, thereby obtaining four-dimensional comprehensive index data; Step S33: Comprehensively evaluate the driving condition division data set according to the four-dimensional comprehensive index data, so as to obtain four-dimensional comprehensive evaluation data.
8. The method for constructing and evaluating driving conditions of new energy vehicles according to claim 1, It is characterized in that The specific steps of step S4 are: Step S41: extracting state features when the simulated driving time is extended according to a preset time extension threshold and based on a driving condition simulation model, thereby obtaining long-time simulated driving data; Step S42: dividing the driving condition division data set into actual driving data of corresponding extended time according to a preset time extension threshold, so as to obtain long-term actual driving data; Step S43: performing feature discrimination of the long-term simulated driving data and the long-term actual driving data with strong correlation to energy consumption according to the four-dimensional comprehensive evaluation data, thereby obtaining the long-term driving energy consumption data; Step S44: Divide the long-term driving energy consumption data into feature sets of different time series windows, so as to obtain a driving condition energy consumption data set.
9. The method for constructing and evaluating driving conditions of new energy vehicles according to claim 8, It is characterized in that The specific steps of step S43 are: Step S431: Comprehensively evaluate the long-term actual driving data according to the four-dimensional comprehensive index data, thereby obtaining long-term actual evaluation data; Step S432: sorting the importance of the long-time simulated driving data with strong correlation to energy consumption, thereby obtaining long-time energy consumption series data; Step S433: Calculating the evaluation difference of the long-term actual evaluation data and the four-dimensional comprehensive evaluation data, thereby obtaining evaluation difference data; Step S434: feature identification is performed on the long-term energy consumption series data according to the evaluation difference data, so as to obtain the long-term driving energy consumption data.
10. The method for constructing and evaluating driving conditions of new energy vehicles according to claim 1, It is characterized in that The specific steps of step S5 are: Step S51: performing a visualization model migration of the driving condition of the new energy vehicle according to the driving condition energy consumption data set to the driving condition mimic model, thereby obtaining a driving condition visualization model; Step S52: Performing operating condition standard evaluation based on the driving condition visualization model to obtain comprehensive operating condition evaluation data.
Citation Information
Patent Citations
Driving condition intelligent forecast and estimation method based on driving habits and traffic road condition
CN107862864A
Vehicle driving cost evaluation method based on data driving scene
CN113034210A
Electric vehicle driving condition construction and evaluation method
CN113222385A
Cited By
New energy automobile energy loss evaluation and optimization method
CN120552618A
Simulated working condition test method for remote vehicle control
CN120891816A
Vehicle performance data acquisition method and device, vehicle and program product
CN121009292A
