WO2022105070A1 - 利用机器学习模型预测电池健康状况的方法和系统 - Google Patents

利用机器学习模型预测电池健康状况的方法和系统 Download PDF

Info

Publication number
WO2022105070A1
WO2022105070A1 PCT/CN2021/077166 CN2021077166W WO2022105070A1 WO 2022105070 A1 WO2022105070 A1 WO 2022105070A1 CN 2021077166 W CN2021077166 W CN 2021077166W WO 2022105070 A1 WO2022105070 A1 WO 2022105070A1
Authority
WO
WIPO (PCT)
Prior art keywords
vehicle
battery
distance
model
health
Prior art date
Application number
PCT/CN2021/077166
Other languages
English (en)
French (fr)
Inventor
瑞吉尔·托纳蒂乌
张轶珍
胡道
傅颖
许永刚
Original Assignee
广州汽车集团股份有限公司
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 广州汽车集团股份有限公司 filed Critical 广州汽车集团股份有限公司
Priority to CN202180005305.XA priority Critical patent/CN114829961A/zh
Publication of WO2022105070A1 publication Critical patent/WO2022105070A1/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
    • B60L58/00—Methods or circuit arrangements for monitoring or controlling batteries or fuel cells, specially adapted for electric vehicles
    • B60L58/10—Methods or circuit arrangements for monitoring or controlling batteries or fuel cells, specially adapted for electric vehicles for monitoring or controlling batteries
    • B60L58/16—Methods or circuit arrangements for monitoring or controlling batteries or fuel cells, specially adapted for electric vehicles for monitoring or controlling batteries responding to battery ageing, e.g. to the number of charging cycles or the state of health [SoH]
    • 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
    • B60L3/12—Recording operating variables ; Monitoring of operating variables
    • 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
    • B60L58/00—Methods or circuit arrangements for monitoring or controlling batteries or fuel cells, specially adapted for electric vehicles
    • B60L58/10—Methods or circuit arrangements for monitoring or controlling batteries or fuel cells, specially adapted for electric vehicles for monitoring or controlling batteries
    • B60L58/12—Methods or circuit arrangements for monitoring or controlling batteries or fuel cells, specially adapted for electric vehicles for monitoring or controlling batteries responding to state of charge [SoC]
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
    • G01R31/367—Software therefor, e.g. for battery testing using modelling or look-up tables
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
    • G01R31/382—Arrangements for monitoring battery or accumulator variables, e.g. SoC
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
    • G01R31/392—Determining battery ageing or deterioration, e.g. state of health
    • 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
    • B60L2240/00—Control parameters of input or output; Target parameters
    • B60L2240/40—Drive Train control parameters
    • B60L2240/54—Drive Train control parameters related to batteries
    • B60L2240/545—Temperature
    • 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
    • B60L2240/00—Control parameters of input or output; Target parameters
    • B60L2240/40—Drive Train control parameters
    • B60L2240/54—Drive Train control parameters related to batteries
    • B60L2240/547—Voltage
    • 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
    • B60L2260/00—Operating Modes
    • B60L2260/40—Control modes
    • B60L2260/46—Control modes by self learning
    • 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
    • B60L2260/00—Operating Modes
    • B60L2260/40—Control modes
    • B60L2260/50—Control modes by future state prediction
    • B60L2260/52—Control modes by future state prediction drive range estimation, e.g. of estimation of available travel distance
    • 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
    • B60L2260/00—Operating Modes
    • B60L2260/40—Control modes
    • B60L2260/50—Control modes by future state prediction
    • B60L2260/54—Energy consumption estimation

Definitions

  • the present application relates to the technical field of battery modeling and battery health monitoring, and in particular, to a method and system for predicting battery health status based on driving distance under full battery load by using a machine learning model, a storage medium, and an electric vehicle.
  • Telematics data for electric vehicles makes it possible to remotely monitor the performance of power battery packs for battery maintenance, prevent major battery damage issues, and avoid costly battery replacements. There is a need to develop methods and systems that enable this functionality.
  • existing operations for battery health monitoring include the use of onboard vehicle monitoring systems to diagnose the condition of vehicle components and transmit vehicle status information via telematics. Transfer engine state data to an external database, and build a normal state model based on historical data from multiple vehicles. The latter enables monitoring by comparing individual vehicle data with a database.
  • a vehicle health monitoring system which transmits vehicle open circuit voltage and usage data to a device to calculate the vehicle health value and compares it with a list value, and when a signal outside a predetermined range occurs, Send out a warning.
  • Embodiments of the present application provide a method and system for predicting battery health status based on a driving distance under full battery load by using a machine learning model, a storage medium, and an electric vehicle, so as to improve the accuracy of vehicle battery health prediction.
  • the embodiments of the present application propose a modeling and monitoring method based on the distance traveled by the battery with full load. Compared with other indicators (such as battery capacity in amp-hours Ah), "driving distance" is more practical and easy to understand for ordinary users. However, driving distance can be affected by many factors other than battery capacity, such as the environment, vehicle mechanics and driver behavior. The methods and systems of the present application eliminate these effects through machine learning methods. As a result, the model-compensated driving distance metric can reliably measure battery health.
  • a method of predicting battery health using a machine learning model includes: remotely obtaining vehicle history information, wherein the vehicle history information includes at least one of the following: odometer reading, battery state of charge, vehicle speed, battery module temperature, and battery voltage; According to the relationship between the historical information of the vehicle, the driving distance model is created; by using the real-time remote information of the vehicle as the model input, the distance traveled by the vehicle under full battery load is obtained based on the driving distance model; the obtained distance is compared with the distance reference value, Predict vehicle battery health.
  • creating a travel distance model based on the relationship between the distance traveled by the vehicle with full battery load and vehicle history information includes: identifying battery travel from the vehicle history information; extracting the following model from the vehicle history information At least one of the characteristics: positive/negative acceleration count, average vehicle speed, average battery temperature, unbalance of battery temperature, regenerative energy, average battery voltage difference, and cumulative distance; based on distance traveled by vehicle with full battery load and model features to model the distance traveled.
  • the method further includes: training the driving distance model based on the historical data of the vehicle to obtain a predictive general model; Vehicle trains individual vehicle models.
  • the method prior to modeling the distance traveled by the vehicle with a full battery load, the method further includes: based on changes in the battery state of charge, calculating the distance traveled, positive/negative acceleration counts, and Regenerative energy quantification.
  • the method before obtaining the distance traveled by the vehicle with full battery load based on the travel distance model by using the vehicle real-time telematics as the model input, the method further comprises: collecting the vehicle real-time information from the vehicle at a preset frequency. remote information.
  • the distance traveled by the vehicle with full battery load is obtained based on a travel distance model by using real-time vehicle telematics as a model input, comprising at least one of the following: according to the following model features: positive/negative acceleration counts, Average vehicle speed and regenerative energy to obtain changes in travel distance caused by travel characteristics; obtain changes in travel distance caused by seasonal temperature fluctuations according to the following model features: average battery temperature and battery temperature imbalance; according to the following model features: Average Battery voltage difference and accumulated distance, get the change in driving distance caused by long-term battery damage.
  • the distance traveled over time is proportional to the accumulated usage and the average battery voltage difference.
  • the relationship between the distance traveled by the vehicle with full battery load and the model features can be determined by the following linear regression formula:
  • y is the distance traveled in the trip
  • X is at least one of the model features
  • ⁇ are the corresponding model coefficients
  • ⁇ t is the matrix transpose of ⁇ .
  • the method further includes sending an alert when the battery health of the vehicle is below a preset threshold.
  • a system for predicting battery health using a machine learning model includes: a first acquisition module for acquiring vehicle history information, wherein the vehicle history information includes at least one of the following: odometer reading, battery state of charge, vehicle speed, battery module temperature, battery voltage; creating a module for According to the relationship between the distance traveled by the vehicle when the battery is fully loaded and the historical information of the vehicle, a driving distance model is created; a second acquisition module is used to obtain the driving distance model by using the real-time remote information of the vehicle as a model input. The distance traveled by the vehicle when the battery is fully loaded; the prediction module is used to compare the obtained distance with a distance reference value to predict the battery health of the vehicle.
  • system further includes a sending module for sending an alert when the vehicle's battery health is below a preset threshold.
  • a non-volatile computer-readable storage medium stores a program, and when the program is executed by a computer, implements the method steps in the above embodiments.
  • an electric vehicle is further provided.
  • the electric vehicle includes the system for predicting battery health using a machine learning model in the above-described embodiments.
  • an interpretable machine learning model is used to monitor the battery health status of the vehicle, and the internal equipment of the vehicle is not required to measure the battery health status, but the battery health status is derived through the machine learning model.
  • an analysis of the factors that lead to the deterioration of battery performance is provided.
  • FIG. 1 shows a flowchart of a method for predicting battery health according to an embodiment of the present application
  • FIG. 2 shows a flowchart of a method for predicting battery health according to another embodiment of the present application
  • FIG. 3 shows a schematic diagram of a time series of vehicle telematics according to an embodiment of the present application
  • FIG. 4 shows a data pipeline according to an embodiment of the present application
  • FIG. 5 shows a block diagram of a system structure for predicting battery health according to another embodiment of the present application
  • FIG. 6 shows a structural block diagram of an electric vehicle according to an embodiment of the present application.
  • This embodiment provides a method for predicting battery health by using a machine learning model.
  • the data-driven approach in this embodiment utilizes an interpretable machine learning model to monitor the vehicle's battery health.
  • Data from the battery management system (BMS) including voltage, current, temperature, state of charge, vehicle speed and odometer readings, is collected into a central database.
  • BMS battery management system
  • the method does not require in-vehicle devices to measure battery health, but derives battery health through machine learning models.
  • the method provides an analysis of factors that cause damage to battery performance, such as temperature, battery voltage imbalance, and usage patterns.
  • the data-driven approach allows manufacturers to monitor the battery health of production vehicles in real time.
  • the method includes the following steps.
  • Step S101 obtaining vehicle history information, wherein the vehicle history information includes at least one of the following: odometer reading, battery state of charge, vehicle speed, battery module temperature, battery voltage;
  • Step S102 creating a travel distance model according to the relationship between the distance traveled by the vehicle when the battery is fully loaded and the historical information of the vehicle;
  • Step S103 by using the real-time remote information of the vehicle as the model input, obtain the distance traveled by the vehicle when the battery is fully loaded based on the travel distance model;
  • step S104 the obtained distance is compared with the distance reference value to predict the battery health of the vehicle.
  • the method may further include: identifying the battery travel from the vehicle history information; extracting at least one of the following model features from the vehicle history information: positive/negative acceleration count, average vehicle speed, average battery temperature , battery temperature imbalance, regenerative energy, average battery voltage difference, and cumulative distance; the driving distance is modeled according to the relationship between the distance the vehicle travels under full battery load and the model features.
  • the method may further include: training the travel distance model based on the historical data of the vehicle to obtain a predictive general model; taking the coefficient of the predictive general model as a starting point, training for different vehicles Individual vehicle models.
  • the method may further include: based on changes in the battery state of charge, calculating the distance traveled, positive/negative acceleration counts, and regenerative energy for each trip quantify.
  • the method may further include: collecting the current remote information of the vehicle at a preset frequency.
  • step S103 may include at least one of the following: according to the following features: positive/negative acceleration count, average vehicle speed and regenerative energy, predict the change in travel distance caused by travel characteristics; according to the following features: battery average temperature and battery temperature The imbalance degree of , predicts the change in driving distance caused by seasonal temperature fluctuations; according to the following characteristics: average battery voltage difference and cumulative distance, predicts the change in driving distance caused by long-term battery damage.
  • the travel distance lost over time is proportional to the accumulated usage and the average battery voltage difference.
  • the driving distance can be determined by the following linear regression formula:
  • ⁇ is the corresponding model coefficient
  • ⁇ t is the matrix transpose of ⁇ .
  • the method may further include: when the battery health of the vehicle is lower than a preset threshold, sending an alarm.
  • This embodiment provides a machine learning model for real-time monitoring of vehicle battery health based on vehicle telematics to enable personalized vehicle care and preventive maintenance.
  • FIG. 2 shows a flowchart of a method for predicting battery health by using a machine learning model according to an embodiment of the present application. As shown in Figure 2, the method includes the following steps.
  • Step S201 receiving the remote information of the vehicle (eg pure electric vehicle or hybrid electric vehicle) in real time at a specific frequency (for example, 10s).
  • the received remote information may include odometer readings, battery state of charge, vehicle speed, battery module temperature, and battery voltage.
  • Figure 3 shows general telematics of a day's SOC signal for a pure electric vehicle. As shown in the upper part of Figure 3, based on the vehicle speed and battery related signals, it can be determined whether the vehicle is running on battery or charging. As shown in the lower part of Figure 3, trips are extracted, defined as trip segments for battery powered vehicles.
  • Step S202 modeling the driving distance when the battery is fully loaded.
  • the driving distance when the battery is fully loaded can be modeled, which is a general measure of battery performance.
  • the key features used for modeling are shown in Table 1.
  • the distance traveled, positive/negative acceleration counts, and regenerated energy per trip are then quantified to 100 SOC, multiplied by a normalization constant 100/ ⁇ SOC, where ⁇ SOC is the net change in SOC over the trip.
  • Step S203 using the transfer learning technology to improve the prediction accuracy of the vehicle.
  • the fleet model is first trained with the historical data of all vehicles in the fleet to obtain a predictive general model of the fleet.
  • the coefficients of this fleet model are then used as a starting point for training individual-vehicle (VIN) models, building a different model for each vehicle to improve model performance individually.
  • VIN vehicle-vehicle
  • Table 2 shows the root mean square error (RMSE) and R 2 for approximately 120,000 trips sampled from approximately 5,000 vehicles.
  • RMSE root mean square error
  • RVIN personalized vehicle model
  • machine learning model can very accurately depict the distance traveled under full load, with variance values as high as 0.90.
  • machine learning models also provide an analysis of the physical factors that cause travel distance to vary over time.
  • model characteristics may depend on vehicle and battery type. For example, different vehicle types may exhibit different correlations between battery temperature and outside temperature. For most vehicle types, this application uses the seven features in the table above, adding second-order terms for average battery temperature and average vehicle speed to calculate the quadratic dependence of distance traveled on temperature and speed.
  • this example aims to account for the variation in traveled distance D caused by trip characteristics, seasons, and battery health.
  • ⁇ is the corresponding model coefficient to be determined
  • ⁇ t is the matrix transpose of ⁇
  • the present embodiment constructs a linear model with interpretable features, which provides analysis of factors that affect changes in driving distance, including travel characteristics, environmental factors, and factors related to battery degradation. The latter can be used to predict battery health, as described below.
  • the traveled distance (loss value D(t)) lost over time is proportional to the accumulated usage (accumulated distance driven by the battery (U(t)) and the average battery voltage difference Vdiff(t). value is linear with U(t) and Vdiff(t),
  • D(t) is the loss value for a given trip t
  • V diff expected value of the vehicle under optimum conditions
  • ci the coefficient to be determined. Note that the minus sign "-" here is used according to the symbol rules.
  • the loss value D of the trip t depends only on the model coefficient of the usage, the battery voltage difference and the reference value rather than depending on any constant (intercept).
  • the first term -c 1 U(t) represents the loss due to cumulative battery usage for trip t
  • the second term for losses due to battery voltage differences.
  • the SOH predicted by the model is compared to a reference value, as shown below
  • the above can be obtained by integrating the battery current I(t) over the entire battery charge or discharge period and dividing by the nominal capacity Ah ref. .
  • the study found that for most vehicles, the SOH value estimated by the model was in good agreement with the SOH reference value, and the root mean square error of the SOH unit was 0.07 (the value of SOH was 0-1). Individual inconsistencies can be explained by noise in the SOH reference values and lack of sufficient data for some vehicles to train the model.
  • the model provides decomposition factors to account for battery health loss over time, measured in SOH or km. This breakdown can explain what causes battery degradation and can be used to design specific maintenance strategies. For example, whenever it is found that the loss of SOH is largely due to an increase in the battery voltage difference over time, updating the software and/or hardware to reduce the battery voltage difference is likely to have a positive impact on battery health.
  • FIG. 4 shows the data pipeline of the method.
  • Raw data is collected from the vehicle at a predetermined frequency through the vehicle telematics system.
  • Models are trained to predict battery health, providing analytical explanations for performance losses due to battery degradation or seasonal effects. As new data is collected, the model is updated in real-time, as are the model predictions. Model results are displayed on a visualization platform for the manufacturer's quality assurance department to monitor vehicle performance in real time.
  • This embodiment further provides a system for predicting battery health using a machine learning model.
  • the system can be applied to a cloud-based server or an onboard computing device for implementing the above-described embodiments in a preferred mode of implementation. It will not be repeated here.
  • the term "module” below may be a combination of software and/or hardware that implements preset functions.
  • the devices described in the following embodiments can preferably be implemented by software, and can also be implemented by hardware or a combination of software and hardware.
  • FIG. 5 shows a block diagram of a system structure for predicting battery health by using a machine learning model according to an embodiment of the present application.
  • the system 100 includes a first acquisition module 10 , a creation module 20 , a second acquisition module 30 and a prediction module 40 .
  • the first acquiring module 10 is configured to acquire vehicle history information, wherein the vehicle history information includes at least one of the following: odometer reading, battery state of charge, vehicle speed, battery module temperature, and battery voltage.
  • the creation module 20 is configured to create a travel distance model according to the relationship between the distance traveled by the vehicle when the battery is fully loaded and the historical information of the vehicle;
  • the second obtaining module 30 is configured to obtain the distance traveled by the vehicle when the battery is fully loaded based on the travel distance model by using the real-time remote information of the vehicle as the model input;
  • the prediction module 40 is used to predict the battery health of the vehicle by comparing the obtained distance with the distance reference value.
  • This embodiment provides a non-volatile computer-readable storage medium, where the non-volatile computer-readable storage medium can store a program, and when the program is executed by a computer, the following steps are performed.
  • Step S1 obtaining vehicle history information, wherein the vehicle history information includes at least one of the following: odometer reading, battery state of charge, vehicle speed, battery module temperature, battery voltage;
  • Step S2 creating a travel distance model according to the relationship between the distance traveled by the vehicle when the battery is fully loaded and the historical information of the vehicle;
  • Step S3 by using the real-time remote information of the vehicle as the model input, obtain the distance traveled by the vehicle when the battery is fully loaded based on the travel distance model;
  • step S4 the obtained distance is compared with the distance reference value to predict the battery health of the vehicle.
  • the storage medium may include, but is not limited to, various media capable of storing program codes, such as a USB flash drive, a ROM, a RAM, a removable hard disk, a magnetic disk, or an optical disk.
  • program codes such as a USB flash drive, a ROM, a RAM, a removable hard disk, a magnetic disk, or an optical disk.
  • the electric vehicle 200 includes the system for predicting the battery health condition using the machine learning model in the above-mentioned embodiment.
  • the electric vehicle in this embodiment may be any type of New Energy Vehicle (NEV) such as EV (Electric Vehicle), HEV (Hybrid Electric Vehicle), and PHEV (Plug-in Hybrid Electric Vehicle).
  • NEV New Energy Vehicle
  • EV Electric Vehicle
  • HEV Hybrid Electric Vehicle
  • PHEV Plug-in Hybrid Electric Vehicle
  • each module or step of the present application can be implemented by a general-purpose computer device, and these modules or steps can be implemented on a single computer device or distributed in a network formed by multiple computer devices.
  • implemented on the computer or in one embodiment by program code executable by a computer device, so that modules or steps can be stored in a storage device for execution by the computer device.
  • steps shown or described may be performed in an order different from that described herein, may separately form a single integrated circuit module, or multiple modules or steps may form a single integrated circuit module for execution. Accordingly, the present application is not limited to any particular combination of hardware and software.

Landscapes

  • Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Sustainable Development (AREA)
  • Sustainable Energy (AREA)
  • Power Engineering (AREA)
  • Transportation (AREA)
  • Mechanical Engineering (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Electric Propulsion And Braking For Vehicles (AREA)
  • Secondary Cells (AREA)
  • Tests Of Electric Status Of Batteries (AREA)

Abstract

本发明提供了一种利用机器学习模型预测基于电池满负载下的行驶距离的电池健康状况的方法和系统。该方法包括:远程获取车辆历史信息，其中车辆历史信息包括以下至少一个:里程表读数、电池充电状态、车速、电池模块温度和电池电压；根据车辆在电池满负载下行驶的距离和车辆历史信息之间的关系，创建行驶距离模型；通过使用车辆实时远程信息作为模型输入，基于行驶距离模型获取车辆在电池满负载下行驶的距离；将获取的距离与距离参考值进行比较，预测车辆的电池健康状况。

Description

利用机器学习模型预测电池健康状况的方法和系统 技术领域
本申请涉及电池建模和电池健康监测技术领域，具体地，涉及一种利用机器学习模型基于电池满负载下的行驶距离预测电池健康状况的方法和系统、存储介质以及电动车辆。
背景技术
电动车辆的远程信息处理数据使得远程监控动力电池组的性能成为可能，以便进行电池维护、防止重大电池损坏问题、避免高成本的电池更换。需要开发能够实现这种功能的方法和系统。
相关技术中，电池健康监控的现有操作包括使用车载车辆监控系统诊断车辆部件状况，并通过远程信息处理传输车辆状态信息。将发动机状态数据传输到外部数据库，根据多辆车的历史数据建立正常状态模型。后者可以通过将个别车辆的数据与数据库进行比较，来实现监控。另一相关技术公开了一种车辆健康监测系统，将车辆开路电压和使用数据发送给一设备，以计算车辆的健康值，并将其与列表值进行比较，当出现超出预定范围的信号时，发出警报。
应当注意，本申请的背景技术中公开的内容仅用于增强对本申请的背景的理解，不是且不应被理解为是对本领域普通技术人员已知的现有技术的认可或任何形式的暗示。
发明内容
本申请实施例提供了一种利用机器学习模型基于电池满负载下的行驶距离预测电池健康状况的方法和系统、存储介质以及电动车辆，提高车辆电池健康预测的准确性。
本申请实施例提出了基于电池满负载行驶的距离的建模和监控方法。与其他指标(比如，以安培小时Ah为单位的电池容量)相比，“行驶距离”更具实际意义，是普通用户容易理解的指标。然而，行驶距离会受到除电池容量之外的许多因素的影响，例如环境、车辆 机械部件和驾驶员行为。本申请的方法和系统通过机器学习方法来排除这些影响。从而，模型补偿的行驶距离指标可以可靠地测量电池健康状况。
根据本申请的一方面，提供了一种利用机器学习模型预测电池健康状况的方法。该方法包括:远程获取车辆历史信息，其中所述车辆历史信息包括以下至少一个:里程表读数、电池充电状态、车速、电池模块温度和电池电压；根据车辆在电池满负载下行驶的距离和所述车辆历史信息之间的关系，创建行驶距离模型；通过使用车辆实时远程信息作为模型输入，基于行驶距离模型获取车辆在电池满负载下行驶的距离；将获取的距离与距离参考值进行比较，预测车辆的电池健康状况。
在一示例性实施例中，根据车辆在电池满负载下行驶的距离和车辆历史信息之间的关系创建行驶距离模型，包括:从车辆历史信息中识别电池行程；从车辆历史信息中提取以下模型特征中的至少一个:正/负加速度计数、平均车速、平均电池温度、电池温度的不平衡度、再生能量、平均电池电压差和累计距离；根据车辆在电池满负载下行驶的距离和模型特征之间的关系，对行驶距离进行建模。
在一示例性实施例中，创建行驶距离模型之后，该方法还包括:基于车辆的历史数据训练该行驶距离模型，得到预测性通用模型；将该预测性通用模型的系数作为开始点，为不同车辆训练个别车辆模型。
在一示例性实施例中，在对车辆在电池满负载下行驶的距离进行建模之前，该方法还包括:基于电池充电状态的变化，将每次行程的行驶距离、正/负加速度计数和再生能量度量化。
在一示例性实施例中，在通过使用车辆实时远程信息作为模型输入，基于行驶距离模型获取车辆在电池满负载下行驶的距离之前，该方法还包括:按预设频率从车辆中收集车辆实时远程信息。
在一示例性实施例中，通过使用实时车辆远程信息作为模型输入，基于行驶距离模型获取车辆在电池满负载下行驶的距离，包括以下至少一项:根据以下模型特征:正/负加速度计数、平均车速和再生能量，获取由行程特性引起的行驶距离变化；根据以下模型特征:电池平均温度和电池温度的不平衡度，获取由季节性温度波动引起的行驶距离变化；根据以下模型特征:平均电池电压差和累计距离，获取由电池长期损坏引起的行驶距离变化。
在一示例性实施例中，随时间损失的行驶距离与累计使用量以及平均电池电压差成比例。
在一示例性实施例中，所述车辆在电池满负载下行驶的距离和所述模型特征之间的关系可由以下线性回归公式确定:
y＝θ tX
其中y是行程中行驶的距离，X是模型特征中的至少一个，θ是相应的模型系数，θ t是θ的矩阵转置。
在一示例性实施例中，在将获取的距离与距离参考值进行比较，预测车辆的电池健康状况之后，该方法还包括:当车辆的电池健康状况低于预设阈值时，发送警报。
根据本申请的另一方面，进一步提供了一种利用机器学习模型预测电池健康状况的系统。该系统包括：第一获取模块，用于获取车辆历史信息，其中所述车辆历史信息包括以下至少一项:里程表读数、电池充电状态、车速、电池模块温度、电池电压；创建模块，用于根据车辆在电池满负载下行驶的距离和所述车辆历史信息之间的关系，创建行驶距离模型；第二获取模块，用于通过使用车辆实时远程信息作为模型输入，基于所述行驶距离模型获取车辆在电池满负载下行驶的距离；预测模块，用于将获取的距离与距离参考值进行比较，预测车辆的电池健康状况。
在一示例性实施例中，该系统还包括发送模块，用于为当车辆的电池健康状况低于预设阈值时发送警报。
根据本申请的另一方面，进一步提供了一种非易失性计算机可读存储介质。该非易失性计算机可读存储介质存储有程序，该程序被计算机执行时实现上述实施例中的方法步骤。
根据本申请的另一方面，进一步提供了一种电动车辆。该电动车辆包括上述实施例中的利用机器学习模型预测电池健康状况的系统。
本申请的上述实施例中，利用了可解释性的机器学习模型来监控车辆的电池健康状况，不需要车辆的内部设备来测量电池健康状况，而是通过机器学习模型导出电池健康状况。此外，还提供了对导致电池性能损坏的因素的分析。
附图简述
这里描述的附图为本申请提供了进一步理解，并且形成本申请的一部分。本申请的示意性实施例及其描述仅用于解释本申请，而非旨在限定本申请。
图1示出了本申请一实施例的用于预测电池健康状况的方法流程图；
图2示出了本申请另一实施例的用于预测电池健康状况的方法流程图；
图3示出了本申请一实施例的车辆远程信息的时间序列的示意图；
图4示出了本申请一实施例的数据流水线；
图5示出了本申请另一实施例的用于预测电池健康状况的系统结构框图；
图6示出了本申请一实施例的电动车辆的结构框图。
具体实施方式
下面将参考附图并结合实施例详细描述本申请。应当理解，以下所说明的优选实施例仅用于说明和解释本发明，并不用于限定本发明。
实施例1
本实施例提供了一种利用机器学习模型预测电池健康状况的方法。本实施例中的数据驱动方案利用了可解释性的机器学习模型来监控车辆的电池健康状况。将电池管理系统(BMS)中的数据(包括电压、电流、温度、充电状态、车速和里程表读数)收集到中央数据库中。该方法不需要车辆内部设备来测量电池健康状况，而是通过机器学习模型来导出电池健康状况。此外，该方法还提供了对导致电池性能损坏的因素的分析，例如温度、电池电压不平衡和使用方式。数据驱动方案使得制造商可以实时监控生产车辆的电池健康状况。
如图1所示，该方法包括以下步骤。
步骤S101，获取车辆历史信息，其中车辆历史信息包括以下至少一项:里程表读数、电池充电状态、车速、电池模块温度、电池电压；
步骤S102，根据车辆在电池满负载下行驶的距离和车辆历史信息之间的关系，创建行驶距离模型；
步骤S103，通过使用车辆实时远程信息作为模型输入，基于行驶距离模型获取车辆在电池满负载下行驶的距离；
步骤S104，将获取的距离与距离参考值进行比较，预测车辆的电池健康状况。
本实施例中，在步骤S102，该方法可以进一步包括:从车辆历史信息中识别电池行程；从车辆历史信息中提取以下模型特征中的至少一个:正/负加速度计数、平均车速、平均电池温度、电池温度的不平衡度、再生能量、平均电池电压差和累计距离；根据车辆在电池满负载下行驶的距离和模型特征之间的关系，对行驶距离进行建模。
本实施例中，在步骤S102之后，该方法还可以进一步包括:基于车辆的历史数据训练该行驶距离模型，得到预测性通用模型；将该预测性通用模型的系数作为开始点，为不同车辆训练个别车辆模型。
本实施例中，在对车辆在电池满负载下行驶的距离进行建模之前，该方法还可以包括:基于电池充电状态的变化，将每次行程的行驶距离、正/负加速度计数和再生能量度量化。
本实施例中，在步骤S103之前，该方法可以进一步包括:按预设频率收集车辆当前远程信息。
本实施例中，步骤S103可包括以下至少一项：根据以下特征:正/负加速度计数、平均车速和再生能量，预测由行程特性引起的行驶距离变化；根据以下特征:电池平均温度和电池温度的不平衡度，预测由季节性温度波动引起的行驶距离变化；根据以下特征:平均电池电压差和累计距离，预测由电池长期损坏引起的行驶距离变化。
本实施例中，随时间损失的行驶距离与累计使用量以及平均电池电压差成比例。
本实施例中，行驶距离可由以下线性回归公式确定：
y＝θ tX
其中目标向量y是行程中的行驶驱动，X是模型特征或多个模型特征的组合，θ是对应的模型系数，θ t是θ的矩阵转置。
本实施例中，在步骤S104之后，该方法还可以包括:当车辆的电池健康状况低于预设阈值时，发送警报。
本实施例中，当电池健康状况低于预定阈值时，发出警报，制造商可以根据需要召回车辆进行预测性维修或通过无线方式进行车辆软件升级。这样，用户方面可收到及时维护，避免重大电池安全问题或更换电池，可延长车辆电池寿命，提高用户满意度，降低制造商保修成本。
实施例2
本实施例提供了用于基于车辆远程信息处理实时监控车辆电池健康状况的机器学习模型，以实现个性化车辆护理和预防性维护。
图2示出了本申请一实施例的利用机器学习模型预测电池健康状况的方法流程图。如图2所示，该方法包括以下步骤。
步骤S201，以特定频率(例如10s)实时接收车辆(如纯电动车辆或混合电动车辆)的远程信息。本实施例中，接收的远程信息可以包括里程表读数、电池充电状态、车辆速度、电池模块温度和电池电压。
图3示出了一纯电动车辆一天的SOC信号的一般远程信息处理。如图3的上部分所示，根据车速和电池相关信号，可确定车辆是在靠电池驱动还是在充电。如图3的下部分所示，提取行程，定义为电池驱动车辆的行程分段。
根据上述电池行程的定义，使用主要指标对电池满载时每次行程的电池性能进行了描述，如下。
步骤S202，对电池满负载时的驱动距离进行建模。
本实施例中，从车辆远程信息中识别电池行程后，可对电池满负载时的驱动距离进行建模，这是电池性能的一般度量。用于建模的关键特性如表1所示。
表1
特征 描述
正/负加速度计数 行程中车辆正/负加速度计数
平均车速 行程中车速的平均值
平均电池温度 行程中电池温度的平均值
电池模块的温度不平衡度 行程中电池模块间的最大温度不平衡度
再生能量 行程中(例如，制动期间)再生的能量
平均电池电压差 两个电池单元间的最大电压差的平均值
累计距离 电池驱动的累计距离
然后将每次行程的行驶距离、正/负加速度计数和再生能量度量化至100SOC，乘以归一化常数100/ΔSOC，其中ΔSOC是行程中SOC的净变化。
步骤S203，使用迁移学习技术提高车辆的预测精度。
本实施例中，首先用车队中所有车辆的历史数据来训练车队模型，得到车队的预测性通用模型。然后将该车队模型的系数用作训练单个车辆(individual-vehicle,VIN)模型的开始点，为每个车辆构建不同的模型，以个性化地提高模型性能。
此时，采用了线性回归模型，其具有可解释性和良好的性能。表2示出了从约5000辆车中抽样的约120000次行程的均方根误差(RMSE)和R 2。用所有车辆训练的车队模型的R 2值为0.83，个性化车辆模型(VIN)进一步提高得到集合R 2值0.90。采用更精细的算法得到的数值类似，例如，使用具有2个隐层和tanh激活函数的全连接神经网络得到R 2值为0.92。
表2
  RMSE(km) R 2
车队模型 23.81 0.83
个性化车辆模型(VIN) 19.83 0.90
神经网络 19.0 0.92
如上表所示，机器学习模型可以非常准确地描绘满负荷行驶的距离，方差值高达0.90。除了预测行驶距离，机器学习模型还提供了对导致行驶距离随时间变化的物理因素的分析。
本实施例中，如上所述，为获得关于影响行程中行驶距离的因素的分析，采用了具有可解释特征的线性模型。模型中只包含具有特定目的的特征，如下表3所示。
表3
Figure PCTCN2021077166-appb-000001
请注意，型号特征可取决于车辆和电池类型。例如，不同的车辆类型表现出的电池温度和外部温度之间的相关性可能不同。对于大多数车辆类型，本申请使用了上表中的七个特征，添加平均电池温度和平均车速的二阶项，以计算行驶距离与温度和速度的二次相关性。
利用下面的线性回归，该实施例旨在解释由行程特性、季节和电池健康状况引起的行驶距离D的变化。
y＝θ tX
其中目标向量y是行程中的行驶驱动，X是模型特征或多个模型特征的组合，θ是待确定的对应模型系数，θ t是θ的矩阵转置。
简单来说，本实施例构建了具有可解释特征的线性模型，其提供了对影响行驶距离变化的因素的分析，包括行程特性、环境因素和电池退化相关因素。后者可用于预测电池健康状况，如下所述。
本实施例发现电池温度和研究的一类车辆满负载下行驶的距离之间具有近似二次相关性。因此，模型中包括线性项和二次温度项，以描述由于温度引起的行驶距离的季节性波动。
S(t)＝k 1+k 2T avg(t)+k 2T avg(t) 2+k 4T diff(t)
其中，k i是常数，T avg(t)，T diff(t)分别是行程t的平均电池温度和电池模块温度的不平衡度。由于季节性影响是和偏移值k 1相关的，所以在实际操作时，我们将其设置为与该值相关，使S(t)最大化，得到S′(t)＝0。
本实施例的模型中，随时间损失的行驶距离(损失值D(t)与累计使用量(电池驱动的累计距离(U(t))以及平均电池电压差Vdiff(t)成比例。假设损失值与U(t)和Vdiff(t)成线性关系，
Figure PCTCN2021077166-appb-000002
其中，D(t)是给定行程t的损失值，
Figure PCTCN2021077166-appb-000003
是电池电压差的标称值(车辆在最佳条件下的预期值V diff)，c i是待确定的系数。注意这里的减号“-”按符号规则使用.
开始时的损失值为0(如，对于新电池，D(t＝0)＝0)。同样，新电池的使用量和V diff为零U(t＝0)＝0，
Figure PCTCN2021077166-appb-000004
Figure PCTCN2021077166-appb-000005
从而c 0＝0，因此
Figure PCTCN2021077166-appb-000006
这表明行程t的损失值D仅取决于使用量的模型系数、电池电压差和参考值
Figure PCTCN2021077166-appb-000007
而不取决于任何常数(截距)。上面的等式中，第一项-c 1U(t)表示由于行程t的电池累计使用量而引起的损失，第二项
Figure PCTCN2021077166-appb-000008
为由电池电压差引起的损失。
电池健康状况(state-of-health，SOH)是根据车辆在时间t行驶距离的损失y(t)，相对于在t＝0行驶的距离y 0来估算的，如下
Figure PCTCN2021077166-appb-000009
假设行程因素和季节条件固定，随时间行驶的距离损失只是由于电池退化。则，
Figure PCTCN2021077166-appb-000010
其中
Figure PCTCN2021077166-appb-000011
是模型预测的行驶距离。
根据这个定义，我们可以很轻易地用模型预测来估计SOH，如下所示
Figure PCTCN2021077166-appb-000012
为了验证本实施例中的模型，将该模型预测的SOH与参考值进行比较，如下所示
Figure PCTCN2021077166-appb-000013
以上可以通过在整个电池充电或放电期间对电池电流I(t)进行积分并除以标称容量来获得Ah ref.。研究发现，对于大多数车辆，该模型估算得到的SOH值与SOH参考值有很好的一致性，其SOH单位的均方根误差为0.07(SOH的取值为0-1)。个别不一致可由SOH参考值中的噪声以及某些车辆缺少足够的数据训练模型来解释。
此外，该模型提供了分解因素，来解释随时间变化电池的健康损失，以SOH或km为单位。这种分解可解释导致电池退化的原因，并且可用于设计特定的维护策略。例如，每当发现SOH的损失大部分是由于电池电压差随时间的增加引起时，更新软件和/或硬件以减少电池电压差很大可能会对电池健康产生积极影响。
本实施例中，图4示出了该方法的数据流水线。通过车辆远程信息处理系统，以预定频率从车辆中收集原始数据。将数据存储在一个集中的平台上。清理和处理数据，以建立模型输入。使用交叉验证方自动调整模型超参数。模型经过训练后，可预测电池的健康状况，为电池退化或季节影响造成的性能损失提供分析解释。收集新数据后，模型会实时更新，模型预测也会更新。模型结果会显示在可视化平台上，供制造商质保部门实时监控车辆性能。
实施例3
本实施例进一步提供了一种利用机器学习模型预测电池健康状况的系统。该系统可应用于基于云的服务器或机载计算设备，用于在优选实现模式下执行上述实施例。此处不再赘 述。例如，下面的用语“模块”可以是实现预设功能的软件和/或硬件的组合。以下实施例中描述的设备优选地可由软件实现，也可以由硬件或软件与硬件的组合实现。
图5示出了本申请一实施例的利用机器学习模型预测电池健康状况的系统结构框图。如图5所示，系统100包括第一获取模块10、创建模块20、第二获取模块30和预测模块40。
第一获取模块10用于获取车辆历史信息，其中车辆历史信息包括以下至少一项:里程表读数、电池充电状态、车速、电池模块温度、电池电压。
创建模块20用于根据车辆在电池满负载下行驶的距离和车辆历史信息之间的关系，来创建行驶距离模型；
第二获取模块30用于通过使用车辆实时远程信息作为模型输入，基于行驶距离模型获取车辆在电池满负载下行驶的距离；
预测模块40用于将获取的距离与距离参考值进行比较，来预测车辆的电池健康状况。
实施例4
本实施例提供了非易失性计算机可读存储介质，该非易失性计算机可读存储介质可存储程序，该程序被计算机执行时执行以下步骤。
步骤S1，获取车辆历史信息，其中车辆历史信息包括以下至少一项:里程表读数、电池充电状态、车速、电池模块温度、电池电压；
步骤S2，根据车辆在电池满负载下行驶的距离和车辆历史信息之间的关系，创建行驶距离模型；
步骤S3，通过使用车辆实时远程信息作为模型输入，基于行驶距离模型获取车辆在电池满负载下行驶的距离；
步骤S4，将获取的距离与距离参考值进行比较，预测车辆的电池健康状况。
示例性实施例中，该存储介质可以包括但不限于能够存储程序代码的各种介质，例如U盘、ROM、RAM、移动硬盘、磁盘或光盘。
实施例5
本实施例提供了一种电动车辆。如图6所示，该电动车辆200包括上述实施例中的利用机器学习模型预测电池健康状况的系统。应注意，本实施例中的电动车辆可以是任何类型的新能源汽车(NEV)，例如EV(电动车辆)、HEV(混合电动车辆)和PHEV(插入式混合电动车辆)。
显然，本领域技术人员应理解，本申请的各个模块或步骤可以由通用计算机设备来实现，并且这些模块或步骤可以集中在单个计算机设备上实现，也可以分布在由多个计算机设备形成的网络上实现，或者在一实施例中由计算机设备可执行的程序代码来实现，因此模块或步骤可存储在存储设备中，以供计算机设备执行。某些情况下，所示出或描述的步骤可以以不同于本申请描述的顺序来执行，也可以分别形成单个集成电路模块，或者多个模块或步骤形成单个集成电路模块，以供执行。因此，本申请不限于任何特定的硬件和软件组合。
以上仅是本申请的示例性实施例，并不旨在限制本申请。对本领域技术人员来说，本申请可有多种改动和变化。凡按照本申请的精神和原理所做的修改、等同替换、改进等都应当理解为落入本申请的保护范围。

Claims (13)

  1. 一种利用机器学习模型预测电池健康状况的方法，其特征在于，包括:
    远程获取车辆历史信息，其中所述车辆历史信息包括以下至少一项:里程表读数、电池充电状态、车速、电池模块温度、电池电压；
    根据车辆在电池满负载下行驶的距离和所述车辆历史信息之间的关系，创建行驶距离模型；
    通过使用车辆实时远程信息作为模型输入，基于所述行驶距离模型获取车辆在电池满负载下行驶的距离；
    将获取的距离与距离参考值进行比较，预测车辆的电池健康状况。
  2. 根据权利要求1所述的方法，所述根据车辆在电池满负载下行驶的距离和所述车辆历史信息之间的关系，创建行驶距离模型包括：
    从所述车辆历史信息中识别电池行程；
    从所述车辆历史信息中提取以下模型特征中的至少一个:正/负加速度计数、平均车速、平均电池温度、电池温度的不平衡度、再生能量、平均电池电压差和累计距离；
    根据所述车辆在电池满负载下行驶的距离和所述模型特征之间的关系，对行驶距离进行建模。
  3. 根据权利要求2所述的方法，在创建行驶距离模型之后，该方法还包括:
    基于车辆的历史数据训练所述行驶距离模型，以提供预测性通用模型；
    使用所述预测性通用模型的系数作为开始点，为不同车辆训练单个车辆模型。
  4. 根据权利要求2所述的方法，对车辆在电池满负载下行驶的距离进行建模之前，该方法还包括:
    根据电池充电状态的变化，将每次行程的行驶距离、正/负加速度计数和再生能量度量化。
  5. 根据权利要求1所述的方法，在通过使用车辆实时远程信息作为模型输入，基于所述行驶距离模型获取车辆在电池满负载下行驶的距离之前，该方法还包括:
    按预设频率收集所述车辆实时远程信息。
  6. 根据权利要求2所述的方法，所述通过使用车辆实时远程信息作为模型输入，基于所述行驶距离模型获取车辆在电池满负载下行驶的距离包括以下至少一项:
    根据以下模型特征:正/负加速度计数、平均车速和再生能量，获取由行程特性引起的行驶距离变化；
    根据以下模型特征:电池平均温度和电池温度的不平衡度，获取由季节性温度波动引起的行驶距离变化；
    根据以下模型特征:平均电池电压差和累计距离，获取由电池长期损坏引起的行驶距离变化。
  7. 根据权利要求6所述的方法，其特征在于，随时间损失的行驶距离与累计使用量以及平均电池电压差成比例。
  8. 根据权利要求2所述的方法，其特征在于，所述车辆在电池满负载下行驶的距离和所述模型特征之间的关系可由以下线性回归公式确定:
    y＝θ tX
    其中y是行程中行驶的距离，X是模型特征中的至少一个，θ是相应的模型系数，θ t是θ的矩阵转置。
  9. 根据权利要求1所述的方法，在将获取的距离与距离参考值进行比较，预测车辆的电池健康状况之后，该方法还包括:
    当车辆的电池健康状况低于预设阈值时，发送警报。
  10. 一种利用机器学习模型预测电池健康状况的系统，其特征在于，包括:
    第一获取模块，用于获取车辆历史信息，其中所述车辆历史信息包括以下至少一项:里程表读数、电池充电状态、车速、电池模块温度、电池电压；
    创建模块，用于根据车辆在电池满负载下行驶的距离和所述车辆历史信息之间的关系，创建行驶距离模型；
    第二获取模块，用于通过使用车辆实时远程信息作为模型输入，基于所述行驶距离模型获取车辆在电池满负载下行驶的距离；
    预测模块，用于将获取的距离与距离参考值进行比较，预测车辆的电池健康状况。
  11. 根据权利要求10所述的系统，进一步包括:
    发送模块，用于当车辆的电池健康状况低于预设阈值时，发送警报。
  12. 一种非易失性计算机可读存储介质，所述存储介质存储有程序，其特征在于，所述程序由计算机执行时实现如权利要求1-9任一项所述的方法。
  13. 一种电动车辆，其特征在于，包括如权利要求10或11所述的系统。
PCT/CN2021/077166 2020-11-19 2021-02-22 利用机器学习模型预测电池健康状况的方法和系统 WO2022105070A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202180005305.XA CN114829961A (zh) 2020-11-19 2021-02-22 利用机器学习模型预测电池健康状况的方法和系统

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US16/952,097 2020-11-19
US16/952,097 US20220153166A1 (en) 2020-11-19 2020-11-19 Method and System for Predicting Battery Health with Machine Learning Model

Publications (1)

Publication Number Publication Date
WO2022105070A1 true WO2022105070A1 (zh) 2022-05-27

Family

ID=81587290

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/077166 WO2022105070A1 (zh) 2020-11-19 2021-02-22 利用机器学习模型预测电池健康状况的方法和系统

Country Status (3)

Country Link
US (1) US20220153166A1 (zh)
CN (1) CN114829961A (zh)
WO (1) WO2022105070A1 (zh)

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20210248442A1 (en) * 2020-02-11 2021-08-12 Distech Controls Inc. Computing device and method using a neural network to predict values of an input variable of a software
CN113554200B (zh) * 2020-04-23 2024-02-27 广州汽车集团股份有限公司 动力电池电压不一致性预测方法、系统及设备
US11527786B1 (en) * 2022-03-28 2022-12-13 Eatron Technologies Ltd. Systems and methods for predicting remaining useful life in batteries and assets
CN115257461A (zh) * 2022-07-28 2022-11-01 上汽通用五菱汽车股份有限公司 一种剩余续驶里程的确定方法和装置
CN115544900B (zh) * 2022-11-28 2023-05-02 深圳联友科技有限公司 一种基于Shap算法分析电动车续航里程影响因素的方法
CN117755150A (zh) * 2023-11-28 2024-03-26 北京潞电电力建设有限公司 一种基于机器学习的快速充电方法

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103969585A (zh) * 2013-01-31 2014-08-06 国际商业机器公司 评估电池的使用状况的方法和装置、相关系统和车辆
CN105044617A (zh) * 2015-08-25 2015-11-11 国家电网公司 一种插电式混合动力汽车用动力电池评价系统及评价方法
CN109204063A (zh) * 2018-08-16 2019-01-15 北京新能源汽车股份有限公司 一种动力电池健康状态soh的获取方法、装置及车辆
CN110018427A (zh) * 2019-05-24 2019-07-16 福建工程学院 一种基于行车数据的电动汽车电池健康状态评估方法
CN110873841A (zh) * 2019-11-05 2020-03-10 北京匠芯电池科技有限公司 一种基于数据驱动与电池特性相结合的电池寿命预测方法

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10197631B2 (en) * 2015-06-01 2019-02-05 Verizon Patent And Licensing Inc. Systems and methods for determining vehicle battery health

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103969585A (zh) * 2013-01-31 2014-08-06 国际商业机器公司 评估电池的使用状况的方法和装置、相关系统和车辆
CN105044617A (zh) * 2015-08-25 2015-11-11 国家电网公司 一种插电式混合动力汽车用动力电池评价系统及评价方法
CN109204063A (zh) * 2018-08-16 2019-01-15 北京新能源汽车股份有限公司 一种动力电池健康状态soh的获取方法、装置及车辆
CN110018427A (zh) * 2019-05-24 2019-07-16 福建工程学院 一种基于行车数据的电动汽车电池健康状态评估方法
CN110873841A (zh) * 2019-11-05 2020-03-10 北京匠芯电池科技有限公司 一种基于数据驱动与电池特性相结合的电池寿命预测方法

Also Published As

Publication number Publication date
US20220153166A1 (en) 2022-05-19
CN114829961A (zh) 2022-07-29

Similar Documents

Publication Publication Date Title
WO2022105070A1 (zh) 利用机器学习模型预测电池健康状况的方法和系统
US11065978B2 (en) Systems, methods, and storage media for adapting machine learning models for optimizing performance of a battery pack
US11813962B2 (en) Supplying power to an electric vehicle
US11555858B2 (en) Systems, methods, and storage media for predicting a discharge profile of a battery pack
US11325494B2 (en) Systems, methods, and storage media for determining a target battery charging level for a drive route
JP5343168B2 (ja) 電池の劣化度を取得するための方法及びそのシステム
JP7269437B2 (ja) 電気車充電ステーションを用いたバッテリー性能管理システム及び方法
US11125822B2 (en) Method for evaluating an electric battery state of health
KR102297343B1 (ko) 하이브리드 varma 및 lstm을 이용한 배터리의 출력 전압 응답 및 충전 상태 예측 방법
CN114609524A (zh) 运行提供电储能器的预测老化状态的系统的方法和装置
US11084387B2 (en) Systems, methods, and storage media for arranging a plurality of cells in a vehicle battery pack
JP2014013245A (ja) 電池劣化度モデルの生成及び更新方法
CN114578248A (zh) 用于预测电池运行的设备中设备电池的老化状态的方法和装置
KR102246633B1 (ko) 하이브리드 varma 및 lstm을 이용한 배터리의 출력 전압 응답 및 충전 상태 예측 최적화 방법
US20230068963A1 (en) Battery fault detection
CN114631032A (zh) 用于监控电池组健康状况的方法和系统
US20230324463A1 (en) Method and Apparatus for Operating a System for Detecting an Anomaly of an Electrical Energy Store for a Device by Means of Machine Learning Methods
CN115219932A (zh) 用于评价设备电池组的相对老化状态的方法和装置
CN115219931A (zh) 运行用于提供电蓄能器的预测老化状态的系统的方法和装置
CN114043875A (zh) 基于大数据的剩余里程预估偏差分析方法和系统
US20230306803A1 (en) Method and apparatus for the user-dependent selection of a battery operated technical device depending on a user usage profile
CN118051852A (zh) 创建识别设备电池组的异常的异常识别模型的方法和装置
CN116256640A (zh) 用于提供老化状态模型以确定电蓄能器的当前和预测老化状态的计算机实现的方法
JP2023125246A (ja) バッテリ劣化度予測装置
CN115469224A (zh) 用于借助机器学习方法运行提供设备的电储能器的预测的老化状态的系统的方法和装置

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: 21893229

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: 21893229

Country of ref document: EP

Kind code of ref document: A1