[go: up one dir, main page]
More Web Proxy on the site http://driver.im/

WO2023124220A1 - 车辆检测方法及装置 - Google Patents

车辆检测方法及装置 Download PDF

Info

Publication number
WO2023124220A1
WO2023124220A1 PCT/CN2022/117872 CN2022117872W WO2023124220A1 WO 2023124220 A1 WO2023124220 A1 WO 2023124220A1 CN 2022117872 W CN2022117872 W CN 2022117872W WO 2023124220 A1 WO2023124220 A1 WO 2023124220A1
Authority
WO
WIPO (PCT)
Prior art keywords
features
vehicle
model
temperature
data
Prior art date
Application number
PCT/CN2022/117872
Other languages
English (en)
French (fr)
Inventor
付振
梁小明
孙建蕾
邵天东
彭凯
刘相超
Original Assignee
中国第一汽车股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 中国第一汽车股份有限公司 filed Critical 中国第一汽车股份有限公司
Publication of WO2023124220A1 publication Critical patent/WO2023124220A1/zh

Links

Images

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60L—PROPULSION OF ELECTRICALLY-PROPELLED VEHICLES; SUPPLYING ELECTRIC POWER FOR AUXILIARY EQUIPMENT OF ELECTRICALLY-PROPELLED VEHICLES; ELECTRODYNAMIC BRAKE SYSTEMS FOR VEHICLES IN GENERAL; MAGNETIC SUSPENSION OR LEVITATION FOR VEHICLES; MONITORING OPERATING VARIABLES OF ELECTRICALLY-PROPELLED VEHICLES; ELECTRIC SAFETY DEVICES FOR ELECTRICALLY-PROPELLED VEHICLES
    • B60L3/00—Electric devices on electrically-propelled vehicles for safety purposes; Monitoring operating variables, e.g. speed, deceleration or energy consumption
    • H—ELECTRICITY
    • H01—ELECTRIC ELEMENTS
    • H01M—PROCESSES OR MEANS, e.g. BATTERIES, FOR THE DIRECT CONVERSION OF CHEMICAL ENERGY INTO ELECTRICAL ENERGY
    • H01M10/00—Secondary cells; Manufacture thereof
    • H01M10/42—Methods or arrangements for servicing or maintenance of secondary cells or secondary half-cells
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T10/00—Road transport of goods or passengers
    • Y02T10/60—Other road transportation technologies with climate change mitigation effect
    • Y02T10/70—Energy storage systems for electromobility, e.g. batteries

Definitions

  • the present application relates to the field of vehicles, and in particular, to a vehicle detection method and device.
  • Battery thermal runaway refers to the phenomenon of spontaneous combustion caused by the common electrothermal positive feedback in bipolar transistors and excessive junction temperature. If the temperature of the battery reaches a certain threshold, the working state of the battery will be completely out of control, causing the internal temperature of the battery to rise sharply, or even burn and explode. Therefore, how to monitor the data indicators related to battery thermal runaway in real time and to give early warning of battery thermal runaway is very important.
  • the embodiment of the present application provides a vehicle detection method and device to at least solve the technical problem in the related art that battery thermal runaway accidents frequently occur due to poor accuracy, insufficient timeliness, and poor real-time performance of early warning of battery thermal runaway phenomena. .
  • a vehicle detection method including: acquiring driving data of the vehicle; extracting features of multiple dimensions in the driving data, wherein the features are used to characterize the thermal runaway state of the power battery of the vehicle ; Using the state evaluation model to process the features of multiple dimensions to obtain the detection result of the vehicle, wherein the detection result is used to represent whether the vehicle has thermal runaway phenomenon, and the state evaluation model is obtained through machine learning.
  • using the state evaluation model to process the features of multiple dimensions to obtain the detection result of the vehicle includes: using multiple sub-models to process the features of multiple dimensions respectively to obtain the processing results of multiple sub-models, wherein the multiple sub-models The model has a one-to-one correspondence with multiple dimensions; the processing results of multiple sub-models are weighted and summed to obtain the detection result.
  • the multiple dimensions include: voltage dimension, temperature dimension, current dimension, power dimension and entropy dimension
  • the multiple sub-models include: voltage logistic regression model, temperature logistic regression model, current identification model, entropy determination model and power evaluation model .
  • the method also includes: obtaining multiple training features, wherein the multiple training features are features of the voltage dimension or temperature dimension; respectively dividing each training feature to determine the probability density of each training feature; The probability density of each training feature is used to determine the target training feature; the logistic regression model is trained using the probability density of the target training feature to obtain a voltage logistic regression model or a temperature logistic regression model.
  • determining the target training features based on the probability densities of the multiple training features includes: sorting the multiple training features according to the probability densities of the multiple training features; acquiring the top-ranked multiple training features to obtain the target training features.
  • the method further includes: acquiring multiple charging and discharging currents; acquiring the difference between the charging and discharging currents at two adjacent moments in the multiple charging and discharging currents to obtain multiple current differences; based on the multiple charging and discharging currents and the multiple The distribution fitting of each current difference is carried out to obtain the fitting result; a current identification model is constructed based on the fitting result.
  • the method further includes: acquiring the cell voltage and the temperature of the temperature measuring point; determining the first information entropy of the cell voltage and the second information entropy of the temperature measuring point temperature; Combined, the target coordinate points in the two-dimensional coordinate system are obtained; based on the target coordinate points, the single classification support vector machine model is trained to obtain the entropy judgment model.
  • the method further includes: obtaining the current remaining power and the number of historical overshoots; determining the first coefficient based on the current remaining power; determining the second coefficient based on the historical number of overshoots, wherein the second coefficient is used to represent the loss of the power battery degree; based on the first coefficient, the second coefficient, the current remaining power and the number of historical overshoots, a power evaluation model is constructed.
  • the method further includes: determining the target value based on the composition structure and materials of the power battery; The processing result and the target value are weighted and summed to obtain the detection result.
  • the method further includes: performing data cleaning on the driving data to obtain cleaned data; and extracting features of multiple dimensions in the cleaned data.
  • performing data cleaning on the driving data includes at least one of the following: performing deduplication processing on the driving data; deleting data whose time stamp exceeds a preset time range in the driving data; The temperature, charge and discharge current and voltage in the data are converted; the data in the driving data whose value is less than the preset value is deleted.
  • a vehicle detection device including:
  • the acquisition module is configured to acquire the driving data of the vehicle
  • the extraction module is configured to extract features of multiple dimensions in the driving data, wherein the features are used to characterize the thermal runaway state of the power battery of the vehicle;
  • the processing module is configured to use the state evaluation model to process the features of multiple dimensions to obtain the detection result of the vehicle, wherein the detection result is used to represent whether the vehicle has a thermal runaway phenomenon, and the state evaluation model is obtained through machine learning.
  • the processing module includes: a processing unit configured to use multiple sub-models to process the features of multiple dimensions respectively to obtain processing results of multiple sub-models, wherein the multiple sub-models have a one-to-one correspondence with multiple dimensions;
  • the weighted sum unit is configured to perform a weighted sum of the processing results of multiple sub-models to obtain a detection result.
  • multiple dimensions include: voltage dimension, temperature dimension, current dimension, electric quantity dimension and entropy dimension
  • multiple sub-models include: voltage logistic regression model, temperature logistic regression model, current identification model, entropy determination model and electric quantity evaluation model .
  • the device further includes: a first acquisition unit configured to acquire multiple training features, wherein the multiple training features are features of voltage dimension or temperature dimension; a division unit configured to divide each training feature respectively , to determine the probability density of each training feature; the first determination unit is set to determine the target training feature based on the probability density of multiple training features; the first training unit is set to use the probability density of the target training feature to carry out the logistic regression model Train to get a voltage logistic regression model or a temperature logistic regression model.
  • a first acquisition unit configured to acquire multiple training features, wherein the multiple training features are features of voltage dimension or temperature dimension
  • a division unit configured to divide each training feature respectively , to determine the probability density of each training feature
  • the first determination unit is set to determine the target training feature based on the probability density of multiple training features
  • the first training unit is set to use the probability density of the target training feature to carry out the logistic regression model Train to get a voltage logistic regression model or a temperature logistic regression model.
  • the first determining unit is further configured to sort the multiple training features according to the probability density of the multiple training features, and acquire the top-ranked multiple training features to obtain the target training feature.
  • the device further includes: a second acquisition unit configured to acquire a plurality of charging and discharging currents; a third acquiring unit configured to acquire a difference between charging and discharging currents at two adjacent moments in the plurality of charging and discharging currents, A plurality of current differences are obtained; the fitting unit is configured to perform distribution fitting based on the plurality of charging and discharging currents and the plurality of current differences to obtain a fitting result; the first construction unit is configured to construct a current identification model based on the fitting results.
  • the device further includes: a fourth acquisition unit configured to acquire the cell voltage and the temperature of the temperature measurement point; a second determination unit configured to determine the first information entropy of the cell voltage and the second information entropy of the temperature measurement point Information entropy; combination unit, is set to combine based on the first information entropy and the second information entropy, obtains the target coordinate point in the two-dimensional coordinate system; The second training unit, is set to single classification support vector machine model based on the target coordinate point Carry out training to obtain the entropy judgment model.
  • the device further includes: a fifth acquiring unit, configured to acquire the current remaining power and the number of historical overshoots; a third determining unit, configured to determine the first coefficient based on the current remaining power; a fourth determining unit, configured to The number of historical overshoots determines the second coefficient, where the second coefficient is used to characterize the degree of loss of the power battery; the second construction unit is set to construct the power based on the first coefficient, the second coefficient, the current remaining power and the number of historical overshoots Evaluate the model.
  • a fifth acquiring unit configured to acquire the current remaining power and the number of historical overshoots
  • a third determining unit configured to determine the first coefficient based on the current remaining power
  • a fourth determining unit configured to The number of historical overshoots determines the second coefficient, where the second coefficient is used to characterize the degree of loss of the power battery
  • the second construction unit is set to construct the power based on the first coefficient, the second coefficient, the current remaining power and the number of historical over
  • the device further includes: a determination module, configured to determine the target value based on the composition structure and materials of the power battery; a weighted sum module, configured to weight the processing results of multiple sub-models and the target value, and obtain the detection result .
  • a determination module configured to determine the target value based on the composition structure and materials of the power battery
  • a weighted sum module configured to weight the processing results of multiple sub-models and the target value, and obtain the detection result .
  • the device further includes: a data cleaning module configured to perform data cleaning on the driving data to obtain cleaned data; the extraction module is also configured to extract features of multiple dimensions in the cleaned data.
  • a data cleaning module configured to perform data cleaning on the driving data to obtain cleaned data
  • the extraction module is also configured to extract features of multiple dimensions in the cleaned data.
  • the data cleaning module is configured to perform at least one of the following: deduplicating the driving data; deleting data whose time stamp exceeds a preset time range in the driving data; extracting the voltage of the monomer and the temperature of the temperature measurement point; Convert the temperature, charge and discharge current and voltage in the data; delete the data whose value is less than the preset value in the driving data.
  • a computer-readable storage medium includes a stored program, wherein, when the program is running, the device where the computer-readable storage medium is located is controlled to perform the above-mentioned vehicle detection method.
  • a processor is also provided, and the processor is used to run a program, wherein the above-mentioned vehicle detection method is executed when the program is running.
  • the driving data of the vehicle is obtained; the features of multiple dimensions in the driving data are extracted, wherein the features are used to characterize the thermal runaway state of the power battery of the vehicle; the state evaluation model is used to analyze the features of multiple dimensions
  • the detection results of the vehicle are obtained.
  • the detection results are used to represent whether the vehicle has thermal runaway.
  • the state evaluation model is obtained through machine learning, and the multi-dimensional real-time data and characteristic data of the vehicle battery are processed through machine learning. Analysis and comparison have achieved the purpose of accurately evaluating the state of the battery, thereby realizing the technical effect of detecting and early warning of the battery thermal runaway phenomenon of the vehicle, and then solving the problem of early warning of the battery thermal runaway phenomenon in related technologies. Poor accuracy, insufficient timeliness and poor real-time performance lead to technical problems of frequent battery thermal runaway accidents.
  • Fig. 1 is a schematic flow chart of a vehicle detection method according to an embodiment of the present application
  • Fig. 2 is a schematic diagram of an optional fitting result according to an embodiment of the present application.
  • Fig. 3 is an optional model rendering according to an embodiment of the present application.
  • Fig. 4 is a schematic structural diagram of a vehicle detection device according to an embodiment of the present application.
  • an embodiment of a vehicle detection method is provided. It should be noted that the steps shown in the flow charts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and, although In the flowcharts, a logical order is shown, but in some cases the steps shown or described may be performed in an order different from that shown or described herein.
  • Fig. 1 is a flow chart of a vehicle detection method according to an embodiment of the present application. As shown in Fig. 1, the method includes the following steps:
  • Step S102 acquiring driving data of the vehicle.
  • the vehicle in the above steps refers to a vehicle equipped with a power battery, such as a new energy vehicle, etc.
  • the driving data in the above steps is the real-time vehicle data collected by the car in different states, including but not limited to current, voltage , temperature, pressure difference, pressure rise, temperature difference, temperature rise, entropy and other parameter information that can characterize the thermal runaway state of the power battery.
  • the driving data of the car after obtaining the driving data of the car, perform cleaning and preprocessing on the driving data, including but not limited to, deduplication of data, processing of time stamp outliers, extraction of cell voltage and temperature of temperature measurement points, temperature correction, charging/ Correction of discharge current, correction of total battery pack voltage, targeted data processing of accumulated mileage and other data, ignoring the impact of data whose monomer voltage and temperature at the temperature measurement point in the data are all 0 on the score of the prediction model, ignoring the temperature measurement in the data The lowest value of point temperature is 0, the impact of the data with the highest value less than 5 on the prediction model score, ignore the impact of data with the highest value of cell voltage equal to the lowest value of cell voltage equal to 3.650 in the data on the prediction model score, ignore the impact of cell voltage less than The impact of data of 0 or greater than 4.4 on the score of the prediction model, the influence of data with a temperature of less than 0 or greater than 200 on the score of the prediction model, etc. are preprocessed on the data.
  • Step S104 extracting features of multiple dimensions in the driving data, wherein the features are used to characterize the thermal runaway state of the power battery of the vehicle.
  • the characteristics in the above steps include voltage, temperature, current, soc (state of charge, state of charge), entropy and other multi-dimensional indicators as the characteristics of effectively reflecting the state of thermal runaway, for example, the voltage of each single cell, power The total internal voltage of the battery, the temperature of each single cell, the lowest single temperature of the power battery, etc.
  • multi-dimensional expansion of each indicator can be carried out.
  • the change of voltage and temperature is also included in the measurement content of the model when considering the value itself.
  • the monomer measures its consistency through information entropy, and fits a suitable distribution for data with high precision to judge outliers.
  • the following features can be extracted but not limited to:
  • CellVoltageDiff the difference between the highest cell voltage and the lowest cell voltage
  • CellVoltageDelta the difference between the next moment and the previous moment of the voltage of each single cell
  • InnerVoltage total internal voltage of the power battery
  • HIghVoltage the highest single voltage of the power battery
  • CellTemp the temperature of each single cell
  • CellTempDiff the difference between the temperature of the highest temperature measurement point and the temperature of the lowest temperature measurement point
  • CellTempDelta the difference between the next moment and the previous moment of the temperature of each single cell
  • HIghTemp the highest single temperature of the power battery
  • LowTemp the lowest single temperature of the power battery
  • VoltEntropy the information entropy obtained by calculating the voltage of the single cell
  • TempEntropy The information entropy calculated by the temperature of the temperature measurement point
  • Point the point in the two-dimensional coordinates composed of Entropy_colt and Entropy_temp;
  • Soc0 SOC at the beginning of charging
  • Soc1 SOC at the end of charging
  • Soc_delta the end SOC minus the start SOC indicates the change in SOC during the charging process
  • OverChargeTimes historical cumulative overshoot times
  • TimeDiff The difference between the timestamp of the currently received data and the timestamp of the previous data.
  • the driving data can be processed through existing feature extraction methods to obtain features of different dimensions, but it is not limited thereto, and other methods can also be used, which are not specifically limited in this application.
  • Step S106 using the state evaluation model to process the features of multiple dimensions to obtain the detection result of the vehicle, wherein the detection result is used to represent whether the vehicle has thermal runaway phenomenon, and the state evaluation model is obtained through machine learning.
  • the state assessment model can be an additive model with 6 parts.
  • the 6 parts are voltage model based on probability density, temperature model based on probability density, abnormal current identification model based on distribution, and unitary model based on information entropy.
  • ⁇ 1 , ⁇ 2 , ⁇ 3 , ⁇ 4 , and ⁇ 5 are the weight values of each model, that is, the influence of each model on the thermal runaway of the power battery. In actual use, it can be used for different models of new energy vehicles and power batteries Different values are used to identify more abnormal vehicles as the standard; V d is the score of the voltage probability density model, T d is the score of the temperature probability density model, C d is the score of the abnormal current identification model, and E a is the information The score of the entropy monomer consistency judgment model, S OC is the score based on the SOC overcharge risk assessment model.
  • the driving data of the vehicle is obtained; the features of multiple dimensions in the driving data are extracted, wherein the features are used to characterize the thermal runaway state of the power battery of the vehicle; the state evaluation model is used to analyze the features of multiple dimensions
  • the detection results of the vehicle are obtained.
  • the detection results are used to represent whether the vehicle has thermal runaway.
  • the state evaluation model is obtained through machine learning, and the multi-dimensional real-time data and characteristic data of the vehicle battery are processed through machine learning. Analysis and comparison have achieved the purpose of accurately evaluating the state of the battery, thereby realizing the technical effect of detecting and early warning of the battery thermal runaway phenomenon of the vehicle, and further solving the problem of early warning of the battery thermal runaway phenomenon in related technologies. Poor accuracy, insufficient timeliness and poor real-time performance lead to technical problems of frequent battery thermal runaway accidents.
  • using the state evaluation model to process the features of multiple dimensions to obtain the detection result of the vehicle includes: using multiple sub-models to process the features of multiple dimensions respectively to obtain the processing results of multiple sub-models, wherein the multiple sub-models The model has a one-to-one correspondence with multiple dimensions; the processing results of multiple sub-models are weighted and summed to obtain the detection result.
  • the above-mentioned multiple dimensions include: voltage dimension, temperature dimension, current dimension, electric quantity dimension and entropy dimension
  • the above-mentioned multiple sub-models include: voltage logistic regression model, temperature logistic regression model, current identification model, entropy determination model and power assessment models.
  • a voltage logistic regression model is established by machine learning for the voltage dimension
  • a temperature logistic regression model is established for the temperature dimension by machine learning
  • a current identification model is established for the current dimension by machine learning
  • establish a power evaluation model for the power dimension by machine learning establish an entropy judgment model for the entropy dimension by machine learning.
  • the method also includes: obtaining multiple training features, wherein the multiple training features are features of the voltage dimension or temperature dimension; respectively dividing each training feature to determine the probability density of each training feature; The probability density of each training feature is used to determine the target training feature; the logistic regression model is trained using the probability density of the target training feature to obtain a voltage logistic regression model or a temperature logistic regression model.
  • the above-mentioned training features may include: features of the voltage dimension or temperature dimension collected during the driving of different vehicles, and label information of whether the vehicle has thermal runaway.
  • voltage and temperature indicators the following specific features can be selected to be included in the model data range: CellVoltage, CellVoltageDiff, CellVoltageDelta, InnerVoltage, HIghVoltage, LowVoltage, CellTemp, CellTempDiff, CellTempDelta, HIghTemp, and LowTemp.
  • features are extracted from all normal vehicle data as positive samples; for thermal runaway vehicles, data during a period of time before thermal runaway occurs are extracted as negative samples. Separately divide the corresponding feature interval options for each feature, and calculate the probability function of each feature.
  • the formula for calculating the probability density is as follows:
  • outliers are such a small fraction of the overall data, outliers can be ignored and all probability densities calculated for all features. Further, all features can be selected according to the probability of occurrence of the feature, and features with a large number of occurrences and high probability are selected as the target training features for training, and the obtained model is more accurate. Therefore, multiple features with low probability density can be screened out as the target training features.
  • Target training features The probability density of the target training feature can be used as the training data, and the logistic model can be used to train these mathematical mechanical energies to obtain the logistic model of the voltage and temperature with respect to the probability density.
  • determining the target training features based on the probability densities of the multiple training features includes: sorting the multiple training features according to the probability densities of the multiple training features; acquiring the top-ranked multiple training features to obtain the target training features.
  • the probability density of the training features is sorted, and the top ten training features are selected as the target training features, wherein the training features with lower probability densities are ranked higher , select the training feature with the higher ranking as the target training feature, so that the processing result of the trained model is more accurate.
  • feature screening can be performed on positive samples and negative samples separately.
  • the method further includes: acquiring multiple charging and discharging currents; acquiring the difference between the charging and discharging currents at two adjacent moments in the multiple charging and discharging currents to obtain multiple current differences; based on the multiple charging and discharging currents and the multiple The distribution fitting of each current difference is carried out to obtain the fitting result; a current identification model is constructed based on the fitting result.
  • the accuracy of the charging/discharging current of the power battery is relatively high, so the actual value of the charging/discharging current and the calculated rise at two moments are extracted as data sources.
  • Use the EM algorithm to fit the distribution select the lomax distribution to fit, visualize the results and evaluate the fitting results with mean square error, absolute error, goodness of fit, and explainable variance. Among them, the fitting results are shown in Figure 2, the Fitted curve in the figure represents the fitting result, and the Actual curve represents the actual result.
  • the evaluation indicators are as follows: the indicator of Mean_squared_error: 2.3965469674281517e-07, the indicator of Mean_absolute_error: 0.00015565805321614162, the indicator of Mean_squared_log_error: 2.3375983167095477e-07, the indicator of r2_score Metric: 0.9914617992918154, Metric for Explained_variance_score: 0.9914908945389258.
  • the threshold value for judging abnormal current is obtained by setting the quantile.
  • the method further includes: acquiring the cell voltage and the temperature of the temperature measuring point; determining the first information entropy of the cell voltage and the second information entropy of the temperature measuring point temperature; Combined, the target coordinate points in the two-dimensional coordinate system are obtained; based on the target coordinate points, the single classification support vector machine model is trained to obtain the entropy judgment model.
  • information entropy is selected as the basis for judging monomer consistency.
  • each piece of data will get a point composed of voltage and temperature information entropy, and then input these points obtained from the normal car data into the OneClassSvm algorithm model , train a model that includes all normal data points.
  • the model will judge whether the point is a normal point (>0) or an abnormal point ( ⁇ 0).
  • the effect of the model is shown in Figure 3.
  • the concentric circles in the figure represent the segmentation hyperplane, and the hollow circles represent the training samples.
  • the formula for calculating entropy is: Inside the split hyperplane are all normal points. Every time a piece of data is generated, the model will calculate whether the points composed of voltage and temperature information entropy are in the split hyperplane. If it is, it will return a number greater than 0, indicating a normal point; otherwise, it will return A number less than 0 indicates an outlier point. In addition, if the value less than 0 is farther away from 0, it means the farther away from the segmentation hyperplane, that is, the greater the degree of anomaly.
  • the method further includes: obtaining the current remaining power and the number of historical overshoots; determining the first coefficient based on the current remaining power; determining the second coefficient based on the historical number of overshoots, wherein the second coefficient is used to represent the loss of the power battery degree; based on the first coefficient, the second coefficient, the current remaining power and the number of historical overshoots, a power evaluation model is constructed.
  • the first one is the current SOC (that is, the current remaining power), and the first coefficient is determined based on the current remaining power, only for the current SOC greater than or equal to 90
  • the data is processed, and if it is not, it returns 0 directly.
  • the second is the historical number of overcharges.
  • the second coefficient is determined based on the historical overcharge times.
  • the second coefficient is used to characterize the degree of loss of the power battery. According to different values of the current SOC from 90 to 100, different first coefficients are selected. Then give different battery loss levels according to different historical overshoot times.
  • these two indicators and the corresponding first coefficient are combined, and the second coefficient gives the final coefficient of thermal runaway, and then the power evaluation model is constructed.
  • the method further includes: determining the target value based on the composition structure and materials of the power battery; The processing result and the target value are weighted and summed to obtain the detection result.
  • the above-mentioned target value can be a constant determined based on the composition and structure of the power battery and the influence of materials (ternary lithium, lithium iron phosphate, etc.). If it is a different type of battery, the influence of the entire constant term needs to be considered.
  • a model coefficient is given to each sub-model, and the specific coefficient is obtained by cross-validation (that is, each coefficient is given a certain range, here, 100 coefficients are taken with an equal difference between 0-1) , to get the final coefficient composition.
  • cross-validation that is, each coefficient is given a certain range, here, 100 coefficients are taken with an equal difference between 0-1) , to get the final coefficient composition.
  • a constant term is added to make the final result more accurate.
  • the method further includes: performing data cleaning on the driving data to obtain cleaned data; and extracting features of multiple dimensions in the cleaned data.
  • performing data cleaning on the driving data includes at least one of the following: performing deduplication processing on the driving data; deleting data whose time stamp exceeds a preset time range in the driving data; The temperature, charge and discharge current and voltage in the data are converted; the data in the driving data whose value is less than the preset value is deleted.
  • data cleaning preprocessing includes, but is not limited to, several or more of the following:
  • Timestamp outlier processing Delete data whose year, month, day, hour, minute, and second exceed the normal value range.
  • the temperature value in the original data is not the real temperature, and needs to be converted accordingly to get the real temperature value
  • the charging/discharging current in the original data is not the real charging/discharging current, and needs to be converted accordingly to obtain the real charging/discharging current.
  • the total voltage of the battery pack is corrected.
  • the total voltage data in the original data is not the real total voltage, and needs to be converted accordingly to obtain the real total voltage data.
  • a vehicle detection device which can implement the vehicle detection method provided in the above-mentioned embodiment 1, the specific implementation and preferred application scenarios are the same as the above-mentioned embodiment 1, here I won't go into details.
  • Fig. 4 is a schematic structural diagram of a vehicle detection device according to an embodiment of the present application. As shown in Fig. 4, the device includes:
  • An acquisition module 42 configured to acquire the driving data of the vehicle
  • the extraction module 44 is configured to extract features of multiple dimensions in the driving data, wherein the features are used to characterize the thermal runaway state of the power battery of the vehicle;
  • the processing module 46 is configured to use the state evaluation model to process the features of multiple dimensions to obtain the detection result of the vehicle, wherein the detection result is used to indicate whether the vehicle has thermal runaway phenomenon, and the state evaluation model is obtained through machine learning.
  • the above acquisition module 42, extraction module 44, and processing module 46 can be run in the computer terminal as part of the device, and the functions realized by the above modules can be executed by the processor in the computer terminal, and the computer terminal can also be Smartphones (such as Android phones, IOS phones, etc.), tablet computers, handheld computers, and mobile Internet devices (Mobile Internet Devices, MID), PAD and other terminal equipment.
  • Smartphones such as Android phones, IOS phones, etc.
  • tablet computers tablet computers
  • handheld computers handheld computers
  • mobile Internet devices Mobile Internet Devices, MID
  • PAD PAD and other terminal equipment.
  • the processing module 46 includes: a processing unit configured to use multiple sub-models to process the features of multiple dimensions respectively to obtain processing results of multiple sub-models, wherein the multiple sub-models have a one-to-one correspondence with multiple dimensions ;
  • the weighted sum unit is set to weight the processing results of multiple sub-models to obtain the detection result.
  • multiple dimensions include: voltage dimension, temperature dimension, current dimension, electric quantity dimension and entropy dimension
  • multiple sub-models include: voltage logistic regression model, temperature logistic regression model, current identification model, entropy determination model and electric quantity evaluation model .
  • the above-mentioned processing unit and the weighted sum unit can be run in a computer terminal as a part of the device, and the functions realized by the above-mentioned modules can be executed by the processor in the computer terminal, and the computer terminal can also be a smart phone (such as Android mobile phones, IOS mobile phones, etc.), tablet computers, handheld computers, and mobile Internet devices (Mobile Internet Devices, MID), PAD and other terminal equipment.
  • the device further includes: a first acquisition unit configured to acquire multiple training features, wherein the multiple training features are features of voltage dimension or temperature dimension; a division unit configured to divide each training feature respectively , to determine the probability density of each training feature; the first determination unit is set to determine the target training feature based on the probability density of multiple training features; the first training unit is set to use the probability density of the target training feature to carry out the logistic regression model Train to get a voltage logistic regression model or a temperature logistic regression model.
  • a first acquisition unit configured to acquire multiple training features, wherein the multiple training features are features of voltage dimension or temperature dimension
  • a division unit configured to divide each training feature respectively , to determine the probability density of each training feature
  • the first determination unit is set to determine the target training feature based on the probability density of multiple training features
  • the first training unit is set to use the probability density of the target training feature to carry out the logistic regression model Train to get a voltage logistic regression model or a temperature logistic regression model.
  • the first determining unit is further configured to sort the multiple training features according to the probability density of the multiple training features, and acquire the top-ranked multiple training features to obtain the target training feature.
  • the above-mentioned first acquisition unit, division unit, first determination unit, and first training unit can be run in a computer terminal as a part of the device, and the above-mentioned modules can be executed by the processor in the computer terminal.
  • the computer terminal can also be a smart phone (such as an Android phone, an IOS phone, etc.), a tablet computer, a handheld computer, and a mobile Internet device (Mobile Internet Devices, MID), PAD and other terminal devices.
  • the device further includes: a second acquisition unit configured to acquire a plurality of charging and discharging currents; a third acquiring unit configured to acquire a difference between charging and discharging currents at two adjacent moments in the plurality of charging and discharging currents, A plurality of current differences are obtained; the fitting unit is configured to perform distribution fitting based on the plurality of charging and discharging currents and the plurality of current differences to obtain a fitting result; the first construction unit is configured to construct a current identification model based on the fitting results.
  • the device further includes: a fourth acquisition unit configured to acquire the cell voltage and the temperature of the temperature measurement point; a second determination unit configured to determine the first information entropy of the cell voltage and the second information entropy of the temperature measurement point Information entropy; combination unit, is set to combine based on the first information entropy and the second information entropy, obtains the target coordinate point in the two-dimensional coordinate system; The second training unit, is set to single classification support vector machine model based on the target coordinate point Carry out training to obtain the entropy judgment model.
  • the device further includes: a fifth acquiring unit, configured to acquire the current remaining power and the number of historical overshoots; a third determining unit, configured to determine the first coefficient based on the current remaining power; a fourth determining unit, configured to The number of historical overshoots determines the second coefficient, where the second coefficient is used to characterize the degree of loss of the power battery; the second construction unit is set to construct the power based on the first coefficient, the second coefficient, the current remaining power and the number of historical overshoots Evaluate the model.
  • a fifth acquiring unit configured to acquire the current remaining power and the number of historical overshoots
  • a third determining unit configured to determine the first coefficient based on the current remaining power
  • a fourth determining unit configured to The number of historical overshoots determines the second coefficient, where the second coefficient is used to characterize the degree of loss of the power battery
  • the second construction unit is set to construct the power based on the first coefficient, the second coefficient, the current remaining power and the number of historical over
  • the third determination unit, the fourth determination unit, and the second construction unit can run in the computer terminal as a part of the device, and the functions realized by the above-mentioned modules can be executed by the processor in the computer terminal, and the computer terminal can also be a smart phone (such as Android mobile phones, IOS mobile phones, etc.), tablet computers, handheld computers, and mobile Internet devices (Mobile Internet Devices, MID), PAD and other terminal equipment.
  • the device further includes: a determination module, configured to determine the target value based on the composition structure and materials of the power battery; a weighted sum module, configured to weight the processing results of multiple sub-models and the target value, and obtain the detection result .
  • a determination module configured to determine the target value based on the composition structure and materials of the power battery
  • a weighted sum module configured to weight the processing results of multiple sub-models and the target value, and obtain the detection result .
  • the device further includes: a data cleaning module configured to perform data cleaning on the driving data to obtain cleaned data; the extraction module 44 is also configured to extract features of multiple dimensions in the cleaned data.
  • a data cleaning module configured to perform data cleaning on the driving data to obtain cleaned data
  • the extraction module 44 is also configured to extract features of multiple dimensions in the cleaned data.
  • the data cleaning module is used to perform at least one of the following: deduplicating the driving data; deleting data whose time stamp exceeds a preset time range in the driving data; extracting the voltage of the single cell and the temperature of the temperature measurement point; Convert the temperature, charge and discharge current and voltage in the data; delete the data whose value is less than the preset value in the driving data.
  • the above determination module, weighted sum module, data cleaning module and extraction module can be run in the computer terminal as part of the device, and the functions realized by the above modules can be executed by the processor in the computer terminal. It can also be smart phones (such as Android mobile phones, IOS mobile phones, etc.), tablet computers, handheld computers, mobile Internet devices (Mobile Internet Devices, MID), PAD and other terminal devices.
  • a computer-readable storage medium includes a stored program, wherein, when the program is running, the device where the computer-readable storage medium is located is controlled to execute the first embodiment above vehicle detection method.
  • a processor is also provided, and the processor is used to run a program, wherein the vehicle detection method in the above-mentioned embodiment 1 is executed when the program is running.
  • the disclosed technical content can be realized in other ways.
  • the device embodiments described above are only illustrative.
  • the division of the units may be a logical function division.
  • multiple units or components may be combined or may be Integrate into another system, or some features may be ignored, or not implemented.
  • the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces, and the indirect coupling or communication connection of units or modules may be in electrical or other forms.
  • the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple units. Part or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.
  • each functional unit in each embodiment of the present application may be integrated into one processing unit, each unit may exist separately physically, or two or more units may be integrated into one unit.
  • the above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.
  • the integrated unit is realized in the form of a software function unit and sold or used as an independent product, it can be stored in a computer-readable storage medium.
  • the technical solution of the present application is essentially or part of the contribution to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium , including several instructions for enabling a computer device (which may be a personal computer, server or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application.
  • the aforementioned storage media include: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disc, etc., which can store program codes. .
  • the solutions provided in the embodiments of the present application can be applied in the field of vehicles.
  • the state assessment model is obtained through machine learning, through Machine learning analyzes and compares the multi-dimensional real-time data and characteristic data of the vehicle battery to achieve the purpose of accurately evaluating the state of the battery, thereby realizing the technical effect of detecting and early warning of the thermal runaway phenomenon of the vehicle battery.
  • it solves the technical problem in related technologies that the battery thermal runaway accidents frequently occur due to the poor accuracy, insufficient timeliness and poor real-time performance of early warning of the battery thermal runaway phenomenon.

Landscapes

  • Engineering & Computer Science (AREA)
  • Mechanical Engineering (AREA)
  • Sustainable Development (AREA)
  • Sustainable Energy (AREA)
  • Power Engineering (AREA)
  • Transportation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Manufacturing & Machinery (AREA)
  • Chemical & Material Sciences (AREA)
  • Chemical Kinetics & Catalysis (AREA)
  • Electrochemistry (AREA)
  • General Chemical & Material Sciences (AREA)
  • Secondary Cells (AREA)
  • Traffic Control Systems (AREA)

Abstract

本申请公开了一种车辆检测方法及装置。其中，该方法包括：获取车辆的行驶数据；提取行驶数据中的多个维度的特征，其中，特征用于表征车辆的动力电池的热失控状态；利用状态评估模型对多个维度的特征进行处理，得到车辆的检测结果，其中，检测结果用于表征车辆是否出现热失控现象，状态评估模型是通过机器学习得到的。本申请解决了相关技术中由于对电池热失控现象进行预警的准确性差，不够及时且实时性较差，导致电池热失控事故频发的技术问题。

Description

车辆检测方法及装置
本申请要求于2021年12月31日提交中国专利局、优先权号为202111676319.2、发明名称为“车辆检测方法及装置”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及车辆领域，具体而言，涉及一种车辆检测方法及装置。
背景技术
随着新能源汽车的推广和普及，其安全性也越来越受到重视。新能源电动汽车的动力电池易发生热失控的事故，电池热失控是指双极晶体管中常见的电热正反馈，结温过高而导致自燃的现象。如果电池升温达到一定阈值，电池的工作状态就会完全失控，导致电池内部温度直线上升，甚至燃烧爆炸。因此如何实时监测电池热失控相关的数据指标，并对电池热失控现象进行预警至关重要。
针对上述的问题，目前尚未提出有效的解决方案。
发明内容
本申请实施例提供了一种车辆检测方法及装置，以至少解决相关技术中由于对电池热失控现象进行预警的准确性差，不够及时且实时性较差，导致电池热失控事故频发的技术问题。
根据本申请实施例的一个方面，提供了一种车辆检测方法，包括：获取车辆的行驶数据；提取行驶数据中的多个维度的特征，其中，特征用于表征车辆的动力电池的热失控状态；利用状态评估模型对多个维度的特征进行处理，得到车辆的检测结果，其中，检测结果用于表征车辆是否出现热失控现象，状态评估模型是通过机器学习得到的。
可选地，利用状态评估模型对多个维度的特征进行处理，得到车辆的检测结果包括：利用多个子模型分别对多个维度的特征进行处理，得到多个子模型的处理结果，其中，多个子模型与多个维度具有一一对应关系；将多个子模型的处理结果进行加权和，得到检测结果。
可选地，多个维度包括：电压维度、温度维度、电流维度、电量维度和熵维度， 多个子模型包括：电压逻辑回归模型、温度逻辑回归模型、电流识别模型、熵判定模型和电量评估模型。
可选地，该方法还包括：获取多个训练特征，其中，多个训练特征为电压维度或温度维度的特征；分别对每个训练特征进行划分，确定每个训练特征的概率密度；基于多个训练特征的概率密度，确定目标训练特征；利用目标训练特征的概率密度对逻辑回归模型进行训练，得到电压逻辑回归模型或温度逻辑回归模型。
可选地，基于多个训练特征的概率密度，确定目标训练特征包括：按照多个训练特征的概率密度，对多个训练特征进行排序；获取排序最前的多个训练特征，得到目标训练特征。
可选地，该方法还包括：获取多个充放电电流；获取多个充放电电流中相邻两个时刻的充放电电流的差值，得到多个电流差；基于多个充放电电流和多个电流差进行分布拟合，得到拟合结果；基于拟合结果构建电流识别模型。
可选地，该方法还包括：获取单体电压和测温点温度；确定单体电压的第一信息熵和测温点温度的第二信息熵；基于第一信息熵和第二信息熵进行组合，得到二维坐标系中的目标坐标点；基于目标坐标点对单分类支持向量机模型进行训练，得到熵判定模型。
可选地，该方法还包括：获取当前剩余电量和历史过冲次数；基于当前剩余电量确定第一系数；基于历史过冲次数确定第二系数，其中，第二系数用于表征动力电池的损耗程度；基于第一系数、第二系数、当前剩余电量和历史过冲次数，构建电量评估模型。
可选地，在利用多个子模型分别对多个维度的特征进行处理，得到多个子模型的处理结果之后，该方法还包括：基于动力电池的组成结构和材料，确定目标值；将多个子模型的处理结果和目标值进行加权和，得到检测结果。
可选地，在获取车辆的行驶数据之后，该方法还包括：对行驶数据进行数据清洗，得到清洗后的数据；提取清洗后的数据中的多个维度的特征。
可选地，对行驶数据进行数据清洗包括如下至少之一：对行驶数据进行去重处理；删除行驶数据中时间戳超过预设时间范围的数据；提取单体电压和测温点温度；对行驶数据中的温度、充放电电流和电压进行换算；删除行驶数据中取值小于预设值的数据。
根据本申请实施例的另一方面，还提供了一种车辆检测装置，包括：
获取模块，设置为获取车辆的行驶数据；
提取模块，设置为用于提取行驶数据中的多个维度的特征，其中，特征用于表征车辆的动力电池的热失控状态；
处理模块，设置为利用状态评估模型对多个维度的特征进行处理，得到车辆的检测结果，其中，检测结果用于表征车辆是否出现热失控现象，状态评估模型是通过机器学习得到的。
可选地，处理模块包括：处理单元，设置为利用多个子模型分别对多个维度的特征进行处理，得到多个子模型的处理结果，其中，多个子模型与多个维度具有一一对应关系；加权和单元，设置为将多个子模型的处理结果进行加权和，得到检测结果。
可选地，多个维度包括：电压维度、温度维度、电流维度、电量维度和熵维度，多个子模型包括：电压逻辑回归模型、温度逻辑回归模型、电流识别模型、熵判定模型和电量评估模型。
可选地，该装置还包括：第一获取单元，设置为获取多个训练特征，其中，多个训练特征为电压维度或温度维度的特征；划分单元，设置为分别对每个训练特征进行划分，确定每个训练特征的概率密度；第一确定单元，设置为基于多个训练特征的概率密度，确定目标训练特征；第一训练单元，设置为利用目标训练特征的概率密度对逻辑回归模型进行训练，得到电压逻辑回归模型或温度逻辑回归模型。
可选地，第一确定单元还设置为按照多个训练特征的概率密度，对多个训练特征进行排序，并获取排序最前的多个训练特征，得到目标训练特征。
可选地，该装置还包括：第二获取单元，设置为获取多个充放电电流；第三获取单元，设置为获取多个充放电电流中相邻两个时刻的充放电电流的差值，得到多个电流差；拟合单元，设置为基于多个充放电电流和多个电流差进行分布拟合，得到拟合结果；第一构建单元，设置为基于拟合结果构建电流识别模型。
可选地，该装置还包括：第四获取单元，设置为获取单体电压和测温点温度；第二确定单元，设置为确定单体电压的第一信息熵和测温点温度的第二信息熵；组合单元，设置为基于第一信息熵和第二信息熵进行组合，得到二维坐标系中的目标坐标点；第二训练单元，设置为基于目标坐标点对单分类支持向量机模型进行训练，得到熵判定模型。
可选地，该装置还包括：第五获取单元，设置为获取当前剩余电量和历史过冲次数；第三确定单元，设置为基于当前剩余电量确定第一系数；第四确定单元，设置为基于历史过冲次数确定第二系数，其中，第二系数用于表征动力电池的损耗程度；第二构建单元，设置为基于第一系数、第二系数、当前剩余电量和历史过冲次数，构建电量评估模型。
可选地，该装置还包括：确定模块，设置为基于动力电池的组成结构和材料，确定目标值；加权和模块，设置为将多个子模型的处理结果和目标值进行加权和，得到检测结果。
可选地，该装置还包括：数据清洗模块，设置为对行驶数据进行数据清洗，得到清洗后的数据；提取模块还设置为提取清洗后的数据中的多个维度的特征。
可选地，数据清洗模块设置为执行如下至少之一：对行驶数据进行去重处理；删除行驶数据中时间戳超过预设时间范围的数据；提取单体电压和测温点温度；对行驶数据中的温度、充放电电流和电压进行换算；删除行驶数据中取值小于预设值的数据。
根据本申请实施例的另一方面，还提供了一种计算机可读存储介质，计算机可读存储介质包括存储的程序，其中，在程序运行时控制计算机可读存储介质所在设备执行上述的车辆检测方法。
根据本申请实施例的另一方面，还提供了一种处理器，处理器用于运行程序，其中，程序运行时执行上述的车辆检测方法。
在本申请实施例中，采用获取车辆的行驶数据；提取行驶数据中的多个维度的特征，其中，特征用于表征车辆的动力电池的热失控状态；利用状态评估模型对多个维度的特征进行处理，得到车辆的检测结果，其中，检测结果用于表征车辆是否出现热失控现象，状态评估模型是通过机器学习得到的方式，通过机器学习对车辆电池多维度的实时的数据与特征数据进行分析对比，达到了能够准确地对电池的状态进行评估的目的，从而实现了对车辆的电池热失控现象进行检测，预警的技术效果，进而解决了相关技术中由于对电池热失控现象进行预警的准确性差，不够及时且实时性较差，导致电池热失控事故频发的技术问题。
附图说明
此处所说明的附图用来提供对本申请的进一步理解，构成本申请的一部分，本申请的示意性实施例及其说明用于解释本申请，并不构成对本申请的不当限定。在附图中：
图1是根据本申请实施例的一种车辆检测方法的流程示意图；
图2是根据本申请实施例的一种可选的拟合结果的示意图；
图3是根据本申请实施例的一种可选的模型效果图；
图4是根据本申请实施例的一种车辆检测装置的结构示意图。
具体实施方式
为了使本技术领域的人员更好地理解本申请方案，下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本申请一部分的实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都应当属于本申请保护的范围。
需要说明的是，本申请的说明书和权利要求书及上述附图中的术语“第一”、“第二”等是用于区别类似的对象，而不必用于描述特定的顺序或先后次序。应该理解这样使用的数据在适当情况下可以互换，以便这里描述的本申请的实施例能够以除了在这里图示或描述的那些以外的顺序实施。此外，术语“包括”和“具有”以及他们的任何变形，意图在于覆盖不排他的包含，例如，包含了一系列步骤或单元的过程、方法、系统、产品或设备不必限于清楚地列出的那些步骤或单元，而是可包括没有清楚地列出的或对于这些过程、方法、产品或设备固有的其它步骤或单元。
实施例1
根据本申请实施例，提供了一种车辆检测的方法实施例，需要说明的是，在附图的流程图示出的步骤可以在诸如一组计算机可执行指令的计算机系统中执行，并且，虽然在流程图中示出了逻辑顺序，但是在某些情况下，可以以不同于此处的顺序执行所示出或描述的步骤。
图1是根据本申请实施例的一种车辆检测方法的流程图，如图1所示，该方法包括如下步骤：
步骤S102，获取车辆的行驶数据。
上述步骤中的车辆指的是装有动力电池的车辆，例如，新能源汽车等，上述步骤中的行驶数据是汽车在不同的状态下的实时采集到的车辆数据，包括但不限于电流、电压、温度、压差、压升量、温差、温升量、熵等一切能够表征动力电池热失控状态的参数信息。
可选地，获取汽车的行驶数据之后，对行驶数据进行清洗预处理，包括但不限于，对数据去重、时间戳异常值处理、提取单体电压和测温点温度、温度矫正、充/放电电流矫正、电池包总电压矫正、对累计里程等数据进行针对性的数据处理、忽略数据中单体电压和测温点温度全部为0的数据对预测模型得分的影响、忽略数据中测温点温度最低值为0，最高值小于5的数据对预测模型得分的影响、忽略数据中单体电压最高值等于单体电压最低值等于3.650的数据对预测模型得分的影响、忽略单体电压小 于0或大于4.4的数据对预测模型得分的影响、忽略温点温度小于0或大于200的数据对预测模型得分的影响等对数据的预处理。通过对数据进行清洗预处理可以减少一些错误数据，无关数据等对后续判断数据的误差，提高准确度，提升效率。
步骤S104，提取行驶数据中的多个维度的特征，其中，特征用于表征车辆的动力电池的热失控状态。
上述步骤中的特征包括电压、温度、电流、soc(state of charge即，荷电状态)、熵等多个维度的指标作为有效反应热失控状态的特征，例如，各个单体电芯电压、动力电池内部总电压、各个单体电芯温度、动力电池最低单体温度等。并且，可以基于丰富的数据源，对每个指标进行多维度拓展，例如对于电压和温度，在考虑其数值本身的情况下，将电压和温度的变化量也纳入模型的衡量内容，对于多个单体通过信息熵衡量其一致性，对于精度较高的数据拟合合适的分布来判断异常值。例如，在本申请实施例中，可以提取但不限于如下特征：
CellVoltage：各个单体电芯电压；
CellVoltageDiff：最高单体电压和最低单体电压之间的差值；
CellVoltageDelta：各个单体电芯电压下一个时刻与上一个时刻的差值；
InnerVoltage：动力电池内部总电压；
HIghVoltage：动力电池最高单体电压；
LowVoltage：动力电池最低单体电压；
CellTemp：各个单体电芯温度；
CellTempDiff：没高测温点温度和最低测温点温度之间的差值；
CellTempDelta：各个单体电芯温度下一个时刻和上一个时刻的差值；
HIghTemp：动力电池最高单体温度；
LowTemp：动力电池最低单体温度；
Current：动力电池充/放电电流；
CurrentDelta：动力电池充/放电电流下一个时刻和上一个时刻的差值；
VoltEntropy：所以单体电压计算得到的信息熵；
TempEntropy：所以测温点温度计算得到的信息熵；
Point：Entropy_colt和Entropy_temp组成的二维坐标中的点；
Soc0：充电开始时的SOC；
Soc1：充电结束时的SOC；
Soc_delta：结束SOC减去起始SOC表示充电过程SOC的变化量；
OverChargeTimes：历史累计过冲次数；
TimeDiff：当前接收数据的时间戳和上一条数据时间戳的差值。
在一种可选的实施例中，可以通过现有的特征提取方式对行驶数据进行处理，得到不同维度的特征，但不仅限于此，也可以采用其他方式实现，本申请对此不作具体限定。
步骤S106，利用状态评估模型对多个维度的特征进行处理，得到车辆的检测结果，其中，检测结果用于表征车辆是否出现热失控现象，状态评估模型是通过机器学习得到的。
在一种可选的实施例中，可以针对电压、温度、电流、soc、熵等指标采用不同的机器学习方法分别建立状态评估模型。同时，考虑到其余因素对动力电池热失控的影响，可以根据经验引入合适的常数项修正模型误差。根据以上指标，状态评估模型可以是具有6个部分的可加模型，6个部分分别为基于概率密度的电压模型、基于概率密度的温度模型、基于分布的异常电流识别模型、基于信息熵的单体一致性判定模型、基于soc的过充风险评估模型以及常数项。
在利用状态评估模型对上述多个维度的特征进行处理的过程中，可以针对不同指标通过不同模型进行处理，得到不同模型的得分值，进而通过如下公式得到最终的得分，基于该得分值可以确定车辆是否会出现热失控现象：
Score＝β 1*V d+β 2*T d+β 3*C d+β 4*E a+β 5*S OC+C，
其中，β 1，β 2，β 3，β 4，β 5为各模型的权重值，即各模型对动力电池热失控的影响，在实际使用过程中，可以针对不同型号新能源汽车和动力电池采用不同的取值，以识别更多的异常车为标准；V d为电压概率密度模型的得分，T d为温度概率密度模型的得分，C d为异常电流识别模型的得分，E a为信息熵单体一致性判定模型的得分，S OC为基于SOC过充风险评估模型的得分。
在本申请实施例中，采用获取车辆的行驶数据；提取行驶数据中的多个维度的特征，其中，特征用于表征车辆的动力电池的热失控状态；利用状态评估模型对多个维度的特征进行处理，得到车辆的检测结果，其中，检测结果用于表征车辆是否出现热 失控现象，状态评估模型是通过机器学习得到的方式，通过机器学习对车辆电池多维度的实时的数据与特征数据进行分析对比，达到了能够准确地对电池的状态进行评估的目的，从而实现了对车辆的电池热失控现象进行检测及预警的技术效果，进而解决了相关技术中由于对电池热失控现象进行预警的准确性差，不够及时且实时性较差，导致电池热失控事故频发的技术问题。
可选地，利用状态评估模型对多个维度的特征进行处理，得到车辆的检测结果包括：利用多个子模型分别对多个维度的特征进行处理，得到多个子模型的处理结果，其中，多个子模型与多个维度具有一一对应关系；将多个子模型的处理结果进行加权和，得到检测结果。
通过利用多个子模型分别对多个维度的数据特征进行处理，分别建立相应的子模型，再把子模型的处理结果进行加权和得到检测结果，达到了每个子模型的检测结果都具有针对性，进而把子模型的结果加权和得到最终的检测结果使得检测结果的准确性更高。
可选地，上述的多个维度包括：电压维度、温度维度、电流维度、电量维度和熵维度，上述的多个子模型包括：电压逻辑回归模型、温度逻辑回归模型、电流识别模型、熵判定模型和电量评估模型。
在一种可选地实施例中，对电压维度通过机器学习的方法建立电压逻辑回归模型，对温度维度通过机器学习的方法建立温度逻辑回归模型，对电流维度通过机器学习的方法建立电流识别模型，对电量维度通过机器学习的方法建立电量评估模型，对熵维度通过机器学习的方法建立熵判定模型。
可选地，该方法还包括：获取多个训练特征，其中，多个训练特征为电压维度或温度维度的特征；分别对每个训练特征进行划分，确定每个训练特征的概率密度；基于多个训练特征的概率密度，确定目标训练特征；利用目标训练特征的概率密度对逻辑回归模型进行训练，得到电压逻辑回归模型或温度逻辑回归模型。
上述的训练特征可以包括：不同车辆行驶过程中采集到的电压维度或温度维度的特征，以及车辆是否发生热失控的标注信息。对于电压和温度指标，可以选取以下具体特征纳入模型数据范围：CellVoltage、CellVoltageDiff、CellVoltageDelta、InnerVoltage、HIghVoltage、LowVoltage、CellTemp、CellTempDiff、CellTempDelta、HIghTemp和LowTemp。
针对上述特征，在所有正常车数据中提取特征作为正样本；对于热失控车辆，提取在发生热失控前的一段时间内的数据作为负样本。分别对每个特征划分对应的特征区间选项，计算每个特征的概率函数。概率密度计算公式如下：
Figure PCTCN2022117872-appb-000001
由于异常值是整体数据中的很小一部分，因此可以忽略异常值，对所有特征的所有概率密度进行计算。进一步可以根据征出现的概率对所有特征进行选择，选择出现次数多，概率大特征作为目标训练特征进行训练，进而得到的模型更加的准确，因此，可以筛选出概率密度较低的多个特征作为目标训练特征。可以将目标训练特征的概率密度作为训练数据，用logistic模型对这些数机械能训练，得到电压温度关于概率密度的logistic模型。
可选地，基于多个训练特征的概率密度，确定目标训练特征包括：按照多个训练特征的概率密度，对多个训练特征进行排序；获取排序最前的多个训练特征，得到目标训练特征。
在一种可选地实施例中，在得到训练特征后，对训练特征的概率密度进行排序，选取排名前十的训练特征作为目标训练特征，其中，概率密度越低的训练特征的排序越高，选择排名越靠前的训练特征作为目标训练特征，使得训练得到的模型的处理结果更加准确。
需要说明的是，可以分别对正样本和负样本进行特征筛选。
可选地，该方法还包括：获取多个充放电电流；获取多个充放电电流中相邻两个时刻的充放电电流的差值，得到多个电流差；基于多个充放电电流和多个电流差进行分布拟合，得到拟合结果；基于拟合结果构建电流识别模型。
在一种可选地实施例中，动力电池充/放电电流精度较高，因此提取充/放电电流真实值以及计算得出的两个时刻的升量作为数据源。利用EM算法拟合分布，选取lomax分布拟合，进行结果可视化和均方误差、绝对误差、拟合优度、可解释方差评估拟合结果。其中，拟合结果见图2，图中Fitted曲线表示拟合结果，Actual曲线表示实际结果。
评价指标如下：Mean_squared_error的指标：2.3965469674281517e-07，Mean_absolute_error的指标：0.00015565805321614162，Mean_squared_log_error的指标：2.3375983167095477e-07，r2_score的指标：0.9914617992918154，Explained_variance_score的指标：0.9914908945389258。
利用分布拟合的结果，通过设置分位数，获得判别电流异常的阈值。
可选地，该方法还包括：获取单体电压和测温点温度；确定单体电压的第一信息熵和测温点温度的第二信息熵；基于第一信息熵和第二信息熵进行组合，得到二维坐标系中的目标坐标点；基于目标坐标点对单分类支持向量机模型进行训练，得到熵判 定模型。
在一种可选地实施例中，选取信息熵作为单体一致性判断的依据。将单体电压和温度测温点的信息熵组合成二维坐标中的单个点，每条数据都会得到一个电压温度信息熵组成的点，然后将正常车的数据得到的这些点输入OneClassSvm算法模型中，训练出一个包含所有正常数据点的模型。当传入一条新的数据点时，模型会判断该点是正常点(>0)还是异常点(<0)。其中，模型效果见图3，图中同心圆环表示分割超平面，空心圆圈表示训练样本。
熵的计算公式：
Figure PCTCN2022117872-appb-000002
分割超平面内部的是所有正常的点，每产生一条数据，模型会计算电压温度信息熵组成的点是否在分割超平面内，如果在会返回一个大于0的数，表示正常点；否则会返回一个小于0的数，表示异常点，此外如果小于0的值离0越远，表示离分割超平面越远，也就是异常程度比较大。
可选地，该方法还包括：获取当前剩余电量和历史过冲次数；基于当前剩余电量确定第一系数；基于历史过冲次数确定第二系数，其中，第二系数用于表征动力电池的损耗程度；基于第一系数、第二系数、当前剩余电量和历史过冲次数，构建电量评估模型。
在一种可选地实施例中，通过两个指标来衡量，第一个是当前的SOC(即，当前的剩余电量)，基于当前剩余电量确定第一系数，只对当前SOC大于等于90的数据进行处理，否的直接返回0，第二个是历史的过充次数，基于历史过冲次数确定第二系数，第二系数用于表征动力电池的损耗程度。根据当前SOC从90-100之间的取值不同，选取不同的第一系数。然后根据不同的历史过冲次数给到不同的电池损耗程度。最后结合这两个指标以及相应的第一系数，第二系数给出最终的发生热失控的系数，进而构建电量评估模型。
可选地，在利用多个子模型分别对多个维度的特征进行处理，得到多个子模型的处理结果之后，该方法还包括：基于动力电池的组成结构和材料，确定目标值；将多个子模型的处理结果和目标值进行加权和，得到检测结果。
上述的目标值可以是基于动力电池的组成结构和材料(三元锂、磷酸铁锂等)的影响所确定的常数，如果是不同类型的电池，就需要考虑整个常数项的影响。
在一种可选地实施例中，赋予每个子模型一个模型系数，具体系数由交叉验证获取(就是每个系数给定一定的范围，此处是0-1之间等差取100个系数)，得到最终系数组成。考虑不同电池包组成结构和材料，加入了一个常数项，通过加入常数项，使得最终的结果更加准确。
可选地，在获取车辆的行驶数据之后，该方法还包括：对行驶数据进行数据清洗，得到清洗后的数据；提取清洗后的数据中的多个维度的特征。
可选地，对行驶数据进行数据清洗包括如下至少之一：对行驶数据进行去重处理；删除行驶数据中时间戳超过预设时间范围的数据；提取单体电压和测温点温度；对行驶数据中的温度、充放电电流和电压进行换算；删除行驶数据中取值小于预设值的数据。
在一种可选地实施例中，数据清洗预处理包括但不限于以下的几种或多种：
(1)数据去重。对完全重复的数据，只保留一条。
(2)时间戳异常值处理。删除年、月、日、时、分、秒超过正常取值范围的数据。
(3)提取单体电压和测温点温度。将每个单体电压和测温点温度都提取出来，各自作为单独的一列数据，并将所有单体电压和测温点温度数据格式转化成int16数据类型，以减小占用空间。
(4)温度矫正。原始数据中的温度数值不是真实温度，需要进行相应换算得到真实温度值
(5)充/放电电流矫正。原始数据中的充/放电电流不是真实的充/放电电流，需要进行相应换算得到真实充/放电电流。
(6)电池包总电压矫正。原始数据中总电压的数据不是真实的总电压，需要进行相应换算得到真实的总电压数据。
(7)对累计里程等数据进行针对性的数据处理。
(8)忽略数据中单体电压和测温点温度全部为0的数据对预测模型得分的影响。
(9)忽略数据中测温点温度最低值为0，最高值小于5的数据对预测模型得分的影响。
(10)忽略数据中单体电压最高值等于单体电压最低值等于3.650的数据对预测模型得分的影响。
(11)忽略单体电压小于0或大于4.4的数据对预测模型得分的影响。
(12)忽略温点温度小于0或大于200的数据对预测模型得分的影响。
通过进行上述的数据清洗，可以去除一些不必要的数据，减少不必要的数据处理时间，提升工作效率。
实施例2
根据本申请实施例的另一方面，还提供了一种车辆检测装置，该装置可以执行上述实施例1中提供的车辆检测方法，具体实现方式和优选应用场景与上述实施例1相同，在此不做赘述。
图4是根据本申请实施例的一种车辆检测装置的结构示意图，如图4所示，该装置包括：
获取模块42，设置为获取车辆的行驶数据；
提取模块44，设置为提取行驶数据中的多个维度的特征，其中，特征用于表征车辆的动力电池的热失控状态；
处理模块46，设置为利用状态评估模型对多个维度的特征进行处理，得到车辆的检测结果，其中，检测结果用于表征车辆是否出现热失控现象，状态评估模型是通过机器学习得到的。
此处需要说明的是，上述获取模块42、提取模块44、处理模块46可以作为装置的一部分运行在计算机终端中，可以通过计算机终端中的处理器来执行上述模块实现的功能，计算机终端也可以是智能手机(如Android手机、IOS手机等)、平板电脑、掌上电脑以及移动互联网设备(Mobile Internet Devices，MID)、PAD等终端设备。
可选地，处理模块46包括：处理单元，设置为利用多个子模型分别对多个维度的特征进行处理，得到多个子模型的处理结果，其中，多个子模型与多个维度具有一一对应关系；加权和单元，设置为将多个子模型的处理结果进行加权和，得到检测结果。
可选地，多个维度包括：电压维度、温度维度、电流维度、电量维度和熵维度，多个子模型包括：电压逻辑回归模型、温度逻辑回归模型、电流识别模型、熵判定模型和电量评估模型。
此处需要说明的是，上述处理单元、加权和单元可以作为装置的一部分运行在计算机终端中，可以通过计算机终端中的处理器来执行上述模块实现的功能，计算机终端也可以是智能手机(如Android手机、IOS手机等)、平板电脑、掌上电脑以及移动互联网设备(Mobile Internet Devices，MID)、PAD等终端设备。
可选地，该装置还包括：第一获取单元，设置为获取多个训练特征，其中，多个训练特征为电压维度或温度维度的特征；划分单元，设置为分别对每个训练特征进行划分，确定每个训练特征的概率密度；第一确定单元，设置为基于多个训练特征的概率密度，确定目标训练特征；第一训练单元，设置为利用目标训练特征的概率密度对 逻辑回归模型进行训练，得到电压逻辑回归模型或温度逻辑回归模型。
可选地，第一确定单元还设置为按照多个训练特征的概率密度，对多个训练特征进行排序，并获取排序最前的多个训练特征，得到目标训练特征。
此处需要说明的是，上述第一获取单元、划分单元、第一确定单元、第一训练单元可以作为装置的一部分运行在计算机终端中，可以通过计算机终端中的处理器来执行上述模块实现的功能，计算机终端也可以是智能手机(如Android手机、IOS手机等)、平板电脑、掌上电脑以及移动互联网设备(Mobile Internet Devices，MID)、PAD等终端设备。
可选地，该装置还包括：第二获取单元，设置为获取多个充放电电流；第三获取单元，设置为获取多个充放电电流中相邻两个时刻的充放电电流的差值，得到多个电流差；拟合单元，设置为基于多个充放电电流和多个电流差进行分布拟合，得到拟合结果；第一构建单元，设置为基于拟合结果构建电流识别模型。
可选地，该装置还包括：第四获取单元，设置为获取单体电压和测温点温度；第二确定单元，设置为确定单体电压的第一信息熵和测温点温度的第二信息熵；组合单元，设置为基于第一信息熵和第二信息熵进行组合，得到二维坐标系中的目标坐标点；第二训练单元，设置为基于目标坐标点对单分类支持向量机模型进行训练，得到熵判定模型。
可选地，该装置还包括：第五获取单元，设置为获取当前剩余电量和历史过冲次数；第三确定单元，设置为基于当前剩余电量确定第一系数；第四确定单元，设置为基于历史过冲次数确定第二系数，其中，第二系数用于表征动力电池的损耗程度；第二构建单元，设置为基于第一系数、第二系数、当前剩余电量和历史过冲次数，构建电量评估模型。
此处需要说明的是，上述第二获取单元、第三获取单元、拟合单元、第一构建单元、第四获取单元、第二确定单元、组合单元、第二训练单元、第五获取单元、第三确定单元、第四确定单元、第二构建单元可以作为装置的一部分运行在计算机终端中，可以通过计算机终端中的处理器来执行上述模块实现的功能，计算机终端也可以是智能手机(如Android手机、IOS手机等)、平板电脑、掌上电脑以及移动互联网设备(Mobile Internet Devices，MID)、PAD等终端设备。
可选地，该装置还包括：确定模块，设置为基于动力电池的组成结构和材料，确定目标值；加权和模块，设置为将多个子模型的处理结果和目标值进行加权和，得到检测结果。
可选地，该装置还包括：数据清洗模块，设置为对行驶数据进行数据清洗，得到清洗后的数据；提取模块44还设置为提取清洗后的数据中的多个维度的特征。
可选地，数据清洗模块用于执行如下至少之一：对行驶数据进行去重处理；删除行驶数据中时间戳超过预设时间范围的数据；提取单体电压和测温点温度；对行驶数据中的温度、充放电电流和电压进行换算；删除行驶数据中取值小于预设值的数据。
此处需要说明的是，上述确定模块、加权和模块、数据清洗模块和提取模块可以作为装置的一部分运行在计算机终端中，可以通过计算机终端中的处理器来执行上述模块实现的功能，计算机终端也可以是智能手机(如Android手机、IOS手机等)、平板电脑、掌上电脑以及移动互联网设备(Mobile Internet Devices，MID)、PAD等终端设备。
实施例3
根据本申请实施例的另一方面，还提供了一种计算机可读存储介质，计算机可读存储介质包括存储的程序，其中，在程序运行时控制计算机可读存储介质所在设备执行上述实施例1的车辆检测方法。
实施例4
根据本申请实施例的另一方面，还提供了一种处理器，处理器用于运行程序，其中，程序运行时执行上述实施例1的车辆检测方法。
上述本申请实施例序号仅仅为了描述，不代表实施例的优劣。
在本申请的上述实施例中，对各个实施例的描述都各有侧重，某个实施例中没有详述的部分，可以参见其他实施例的相关描述。
在本申请所提供的几个实施例中，应该理解到，所揭露的技术内容，可通过其它的方式实现。其中，以上所描述的装置实施例仅仅是示意性的，例如所述单元的划分，可以为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个单元或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口，单元或模块的间接耦合或通信连接，可以是电性或其它的形式。
所述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。
另外，在本申请各个实施例中的各功能单元可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。上述集成的单元既可以采用硬件的形式实现，也可以采用软件功能单元的形式实现。
所述集成的单元如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个计算机可读取存储介质中。基于这样的理解，本申请的技术方案本质上或者说对现有技术做出贡献的部分或者该技术方案的全部或部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个存储介质中，包括若干指令用以使得一台计算机设备(可为个人计算机、服务器或者网络设备等)执行本申请各个实施例所述方法的全部或部分步骤。而前述的存储介质包括：U盘、只读存储器(ROM，Read-Only Memory)、随机存取存储器(RAM，Random Access Memory)、移动硬盘、磁碟或者光盘等各种可以存储程序代码的介质。
以上所述仅是本申请的优选实施方式，应当指出，对于本技术领域的普通技术人员来说，在不脱离本申请原理的前提下，还可以做出若干改进和润饰，这些改进和润饰也应视为本申请的保护范围。
工业实用性
本申请实施例中提供的方案，可应用于车辆领域中。通过采用获取车辆的行驶数据；提取行驶数据中的多个维度的特征；利用状态评估模型对多个维度的特征进行处理，得到车辆的检测结果，状态评估模型是通过机器学习得到的方式，通过机器学习对车辆电池多维度的实时的数据与特征数据进行分析对比，达到了能够准确地对电池的状态进行评估的目的，从而实现了对车辆的电池热失控现象进行检测，预警的技术效果，进而解决了相关技术中由于对电池热失控现象进行预警的准确性差，不够及时且实时性较差，导致电池热失控事故频发的技术问题。

Claims (20)

  1. 一种车辆检测方法，包括：
    获取车辆的行驶数据；
    提取所述行驶数据中的多个维度的特征，其中，所述特征用于表征所述车辆的动力电池的热失控状态；
    利用状态评估模型对所述多个维度的特征进行处理，得到所述车辆的检测结果，其中，所述检测结果用于表征所述车辆是否出现热失控现象，所述状态评估模型是通过机器学习得到的。
  2. 根据权利要求1所述的方法，其中，利用状态评估模型对所述多个维度的特征进行处理，得到所述车辆的检测结果包括：
    利用多个子模型分别对所述多个维度的特征进行处理，得到所述多个子模型的处理结果，其中，所述多个子模型与所述多个维度具有一一对应关系；
    将所述多个子模型的处理结果进行加权和，得到所述检测结果。
  3. 根据权利要求2所述的方法，其中，所述多个维度包括：电压维度、温度维度、电流维度、电量维度和熵维度，所述多个子模型包括：电压逻辑回归模型、温度逻辑回归模型、电流识别模型、熵判定模型和电量评估模型。
  4. 根据权利要求3所述的方法，其中，所述方法还包括：
    获取多个训练特征，其中，所述多个训练特征为电压维度或温度维度的特征；
    分别对每个训练特征进行划分，确定所述每个训练特征的概率密度；
    基于所述多个训练特征的概率密度，确定目标训练特征；
    利用所述目标训练特征的概率密度对逻辑回归模型进行训练，得到所述电压逻辑回归模型或所述温度逻辑回归模型。
  5. 根据权利要求4所述的方法，其中，基于所述多个训练特征的概率密度，确定目标训练特征包括：
    按照所述多个训练特征的概率密度，对所述多个训练特征进行排序；
    获取排序最前的多个训练特征，得到所述目标训练特征。
  6. 根据权利要求3所述的方法，其中，所述方法还包括：
    获取多个充放电电流；
    获取所述多个充放电电流中相邻两个时刻的充放电电流的差值，得到多个电流差；
    基于所述多个充放电电流和所述多个电流差进行分布拟合，得到拟合结果；
    基于所述拟合结果构建所述电流识别模型。
  7. 根据权利要求3所述的方法，其中，所述方法还包括：
    获取单体电压和测温点温度；
    确定所述单体电压的第一信息熵和所述测温点温度的第二信息熵；
    基于所述第一信息熵和所述第二信息熵进行组合，得到二维坐标系中的目标坐标点；
    基于所述目标坐标点对单分类支持向量机模型进行训练，得到所述熵判定模型。
  8. 根据权利要求3所述的方法，其中，所述方法还包括：
    获取当前剩余电量和历史过冲次数；
    基于所述当前剩余电量确定第一系数；
    基于所述历史过冲次数确定第二系数，其中，所述第二系数用于表征所述动力电池的损耗程度；
    基于所述第一系数、所述第二系数、所述当前剩余电量和所述历史过冲次数，构建所述电量评估模型。
  9. 根据权利要求2至8中任意一项所述的方法，其中，在利用多个子模型分别对所述多个维度的特征进行处理，得到所述多个子模型的处理结果之后，所述方法还包括：
    基于所述动力电池的组成结构和材料，确定目标值；
    将所述多个子模型的处理结果和所述目标值进行加权和，得到所述检测结果。
  10. 根据权利要求1所述的方法，其中，在获取车辆的行驶数据之后，包括：
    对所述行驶数据进行数据清洗，得到清洗后数据；
    对所述清洗后数据进行多维度特征提取，得到多维度特征。
  11. 根据权利要求10所述的方法，其中，对所述行驶数据进行数据清洗，包括：
    对所述行驶数据进行去重处理；
    删除所述行驶数据中时间戳超过预设时间范围的数据；
    提取单体电压和测温点温度；
    对所述行驶数据中的温度、充放电电流和电压进行换算；
    删除所述行驶数据中取值小于预设值的数据。
  12. 一种车辆检测装置，包括：
    获取模块，设置为获取车辆的行驶数据；
    提取模块，设置为提取所述行驶数据中的多个维度的特征，其中，所述特征用于表征所述车辆的动力电池的热失控状态；
    处理模块，设置为利用状态评估模型对所述多个维度的特征进行处理，得到所述车辆的检测结果，其中，所述检测结果用于表征所述车辆是否出现热失控现象，所述状态评估模型是通过机器学习得到的。
  13. 一种计算机可读存储介质，所述计算机可读存储介质包括存储的程序，其中，在所述程序运行时控制所述计算机可读存储介质所在设备执行如下方法：
    获取车辆的行驶数据；
    提取所述行驶数据中的多个维度的特征，其中，所述特征用于表征所述车辆的动力电池的热失控状态；
    利用状态评估模型对所述多个维度的特征进行处理，得到所述车辆的检测结果，其中，所述检测结果用于表征所述车辆是否出现热失控现象，所述状态评估模型是通过机器学习得到的。
  14. 如权利要求13所述的存储介质，在所述程序运行时控制所述计算机可读存储介质所在设备还执行如下方法：
    利用多个子模型分别对所述多个维度的特征进行处理，得到所述多个子模型的处理结果，其中，所述多个子模型与所述多个维度具有一一对应关系；
    将所述多个子模型的处理结果进行加权和，得到所述检测结果。
  15. 如权利要求14所述的存储介质，在所述程序运行时控制所述计算机可读存储介质所在设备还执行如下方法：
    获取多个训练特征，其中，所述多个训练特征为电压维度或温度维度的特征；
    分别对每个训练特征进行划分，确定所述每个训练特征的概率密度；
    基于所述多个训练特征的概率密度，确定目标训练特征；
    利用所述目标训练特征的概率密度对逻辑回归模型进行训练，得到所述电压逻辑回归模型或所述温度逻辑回归模型。
  16. 如权利要求14所述的存储介质，在所述程序运行时控制所述计算机可读存储介质所在设备还执行如下方法：
    按照所述多个训练特征的概率密度，对所述多个训练特征进行排序；
    获取排序最前的多个训练特征，得到所述目标训练特征。
  17. 如权利要求14所述的存储介质，在所述程序运行时控制所述计算机可读存储介质所在设备还执行如下方法：
    获取多个充放电电流；
    获取所述多个充放电电流中相邻两个时刻的充放电电流的差值，得到多个电流差；
    基于所述多个充放电电流和所述多个电流差进行分布拟合，得到拟合结果；
    基于所述拟合结果构建所述电流识别模型。
  18. 如权利要求14所述的存储介质，在所述程序运行时控制所述计算机可读存储介质所在设备还执行如下方法：
    获取单体电压和测温点温度；
    确定所述单体电压的第一信息熵和所述测温点温度的第二信息熵；
    基于所述第一信息熵和所述第二信息熵进行组合，得到二维坐标系中的目标坐标点；
    基于所述目标坐标点对单分类支持向量机模型进行训练，得到所述熵判定模型。
  19. 如权利要求14所述的存储介质，在所述程序运行时控制所述计算机可读存储介质所在设备还执行如下方法：
    获取当前剩余电量和历史过冲次数；
    基于所述当前剩余电量确定第一系数；
    基于所述历史过冲次数确定第二系数，其中，所述第二系数用于表征所述动力电池的损耗程度；
    基于所述第一系数、所述第二系数、所述当前剩余电量和所述历史过冲次数，构建所述电量评估模型。
  20. 一种处理器，所述处理器用于运行程序，其中，所述程序运行时执行如下方法：
    获取车辆的行驶数据；
    提取所述行驶数据中的多个维度的特征，其中，所述特征用于表征所述车辆的动力电池的热失控状态；
    利用状态评估模型对所述多个维度的特征进行处理，得到所述车辆的检测结果，其中，所述检测结果用于表征所述车辆是否出现热失控现象，所述状态评估模型是通过机器学习得到的。
PCT/CN2022/117872 2021-12-31 2022-09-08 车辆检测方法及装置 WO2023124220A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202111676319.2A CN114312322B (zh) 2021-12-31 2021-12-31 车辆检测方法及装置
CN202111676319.2 2021-12-31

Publications (1)

Publication Number Publication Date
WO2023124220A1 true WO2023124220A1 (zh) 2023-07-06

Family

ID=81022766

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/117872 WO2023124220A1 (zh) 2021-12-31 2022-09-08 车辆检测方法及装置

Country Status (2)

Country Link
CN (1) CN114312322B (zh)
WO (1) WO2023124220A1 (zh)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN118722237A (zh) * 2024-09-02 2024-10-01 浙江凌骁能源科技有限公司 电池异常预警方法、装置、设备、可读存储介质和程序产品
CN119068572A (zh) * 2024-08-16 2024-12-03 重庆赛力斯凤凰智创科技有限公司 车辆健康状况的评估方法、装置、电子设备及存储介质
WO2025061179A1 (zh) * 2023-09-20 2025-03-27 蜂巢能源科技股份有限公司 电池异常检测模型的确定方法、装置及电子设备

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114312322B (zh) * 2021-12-31 2024-04-12 中国第一汽车股份有限公司 车辆检测方法及装置
CN114757297A (zh) * 2022-04-29 2022-07-15 江苏第三代半导体研究院有限公司 一种多参数化的半导体检测方法、装置及可读存储介质
CN115294745B (zh) * 2022-05-23 2023-09-29 电子科技大学 一种基于神经网络与数据差分的锂电池热失控分层预警方法
CN115320385B (zh) * 2022-07-28 2024-04-30 重庆金康赛力斯新能源汽车设计院有限公司 车辆电池的热失控预警方法、装置、设备和存储介质
CN115372838A (zh) * 2022-08-22 2022-11-22 上海褒科智能科技有限公司 一种新能源汽车bms故障检测方法

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109978229A (zh) * 2019-02-12 2019-07-05 常伟 一种对动力电池组全电芯多点温度及连接点温度进行热失控预测的方法
CN112038716A (zh) * 2020-08-14 2020-12-04 合肥国轩高科动力能源有限公司 一种锂离子电池包热失控预警方法
CN112092675A (zh) * 2020-08-31 2020-12-18 长城汽车股份有限公司 一种电池热失控预警方法、系统及服务器
CN112590613A (zh) * 2020-09-10 2021-04-02 安徽江淮汽车集团股份有限公司 车辆电池热失控故障诊断方法、装置、设备及存储介质
CN114312322A (zh) * 2021-12-31 2022-04-12 中国第一汽车股份有限公司 车辆检测方法及装置

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8719195B2 (en) * 2011-10-10 2014-05-06 The Boeing Company Battery adaptive learning management system
US11691518B2 (en) * 2017-07-21 2023-07-04 Quantumscape Battery, Inc. Predictive model for estimating battery states
CN111381169B (zh) * 2020-03-05 2021-01-19 大连理工大学 一种动力电池热失控预警方法
CN113500916A (zh) * 2020-03-23 2021-10-15 北京新能源汽车股份有限公司 一种电池系统安全状态的预警方法、预警装置及控制设备
CN112622624B (zh) * 2020-12-18 2022-08-16 华人运通(江苏)技术有限公司 动力电池热失控预警方法、装置、存储介质及终端设备
CN113437371A (zh) * 2021-05-19 2021-09-24 湖南大学 一种新能源汽车锂离子电池热失控预警系统及预警方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109978229A (zh) * 2019-02-12 2019-07-05 常伟 一种对动力电池组全电芯多点温度及连接点温度进行热失控预测的方法
CN112038716A (zh) * 2020-08-14 2020-12-04 合肥国轩高科动力能源有限公司 一种锂离子电池包热失控预警方法
CN112092675A (zh) * 2020-08-31 2020-12-18 长城汽车股份有限公司 一种电池热失控预警方法、系统及服务器
CN112590613A (zh) * 2020-09-10 2021-04-02 安徽江淮汽车集团股份有限公司 车辆电池热失控故障诊断方法、装置、设备及存储介质
CN114312322A (zh) * 2021-12-31 2022-04-12 中国第一汽车股份有限公司 车辆检测方法及装置

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2025061179A1 (zh) * 2023-09-20 2025-03-27 蜂巢能源科技股份有限公司 电池异常检测模型的确定方法、装置及电子设备
CN119068572A (zh) * 2024-08-16 2024-12-03 重庆赛力斯凤凰智创科技有限公司 车辆健康状况的评估方法、装置、电子设备及存储介质
CN118722237A (zh) * 2024-09-02 2024-10-01 浙江凌骁能源科技有限公司 电池异常预警方法、装置、设备、可读存储介质和程序产品

Also Published As

Publication number Publication date
CN114312322A (zh) 2022-04-12
CN114312322B (zh) 2024-04-12

Similar Documents

Publication Publication Date Title
WO2023124220A1 (zh) 车辆检测方法及装置
CN111241154B (zh) 一种基于大数据的蓄电池故障预警方法及系统
CN109655754B (zh) 一种基于充电过程多维分阶的电池性能评估方法
CN110161425B (zh) 一种基于锂电池退化阶段划分的剩余使用寿命的预测方法
CN101738585B (zh) 一种判定蓄电池容量与健康度的方法及系统
CN109100655B (zh) 一种动力电池的数据处理方法和装置
CN106556802A (zh) 一种蓄电池组异常单体电池识别方法及装置
CN115219903B (zh) 基于车联网数据分析的电池自放电率异常判断方法及装置
CN117951629B (zh) 一种电池组充放电温度异常行为的检测方法及系统
CN113030761A (zh) 一种超大规模储能电站电池健康状态的评估方法及系统
CN114094218B (zh) 电池模组的电压的一致性评估方法、装置、设备及介质
CN114764600B (zh) 新能源车电池包的寿命预测方法、装置和电子设备
Liu et al. Big field data-driven battery pack health estimation for electric vehicles: A deep-fusion transfer learning approach
CN114646888A (zh) 一种动力电池容量衰减的评估方法及系统
CN116881701A (zh) 一种新能源汽车电源安全检测方法
CN116901707A (zh) 一种动力电池组故障预警方法、系统及车辆
CN118425818A (zh) 一种融合健康衰退和一致性的动力电池风险评估方法
CN119224592A (zh) 基于大数据分析的新能源电池健康状态监测方法及系统
CN113359036A (zh) 一种矿用动力电池组的用电健康状态在线监管系统
Wang et al. State of Health Estimations for Lithium-Ion Batteries Based on MSCNN.
CN117764601A (zh) 基于电力数据的碳排放量预测方法、装置、设备及介质
CN117471326A (zh) 一种锂离子电池组动态不一致性与健康状态评估方法
CN116008815A (zh) 一种电芯内短路检测方法、装置、设备、存储介质及车辆
CN116794544A (zh) 动力电池的寿命检测方法、计算机可读存储介质及车辆
CN116819364A (zh) 电池健康度的确定方法、应用方法、电子设备及存储介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22913493

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 22913493

Country of ref document: EP

Kind code of ref document: A1