WO2019018974A1 - 用于对电池容量进行建模和估计的方法及系统 - Google Patents

用于对电池容量进行建模和估计的方法及系统 Download PDF

Info

Publication number
WO2019018974A1
WO2019018974A1 PCT/CN2017/094036 CN2017094036W WO2019018974A1 WO 2019018974 A1 WO2019018974 A1 WO 2019018974A1 CN 2017094036 W CN2017094036 W CN 2017094036W WO 2019018974 A1 WO2019018974 A1 WO 2019018974A1
Authority
WO
WIPO (PCT)
Prior art keywords
capacity
regression model
battery
data set
curve
Prior art date
Application number
PCT/CN2017/094036
Other languages
English (en)
French (fr)
Inventor
张彩萍
郭琦沛
姜久春
张维戈
高洋
姜研
肖鹏飞
Original Assignee
罗伯特·博世有限公司
北京交通大学
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 罗伯特·博世有限公司, 北京交通大学 filed Critical 罗伯特·博世有限公司
Priority to PCT/CN2017/094036 priority Critical patent/WO2019018974A1/zh
Priority to CN201780093207.XA priority patent/CN111448467B/zh
Publication of WO2019018974A1 publication Critical patent/WO2019018974A1/zh

Links

Images

Classifications

    • 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]
    • 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
    • Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
    • Y02E60/00—Enabling technologies; Technologies with a potential or indirect contribution to GHG emissions mitigation
    • Y02E60/10—Energy storage using batteries
    • 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
    • Y02T10/10—Internal combustion engine [ICE] based vehicles
    • Y02T10/40—Engine management systems

Definitions

  • the present invention relates to battery measurement detection techniques and, more particularly, to methods, apparatus and systems for modeling and estimating battery capacity.
  • a battery pack in an EV/HEV usually includes tens to hundreds of single cells. Battery life varies with time and with different electrical loads.
  • the battery packs included in the battery pack on the EV/HEV will age at different speeds and operate at different states of charge (SOC). Therefore, it is critical to provide a battery management system (BMS) to determine or predict the status of the battery pack.
  • BMS battery management system
  • An important challenge in real-time application of a battery management system is how to determine the aging state of the battery, ie, the battery available capacity (maximum available capacity).
  • the conventional method of determining the usable capacity of a battery is to fully charge and discharge with a small current.
  • the conventional battery usable capacity determining method has a problem in that it is unrealistic to directly fill the individual cells in the battery pack without disassembling the battery pack on the EV/HEV into a single battery. of.
  • the power battery is often unable to be discharged to the air-power state, and therefore the actual usable capacity of the battery cannot be directly measured by the measuring device.
  • the present invention is to solve the above problems in the conventional method for determining the battery capacity in the prior art, and provides a method for modeling and estimating the battery capacity. Set the system.
  • a method for modeling battery capacity comprising: acquiring charging data to generate a capacity increment curve; and selecting a parameter related to a peak in the capacity increment curve as The independent variable is selected as the dependent variable to form a data set; a regression model is established for the battery capacity, and the regression model is trained based on the data set to obtain an optimized regression model.
  • a system for modeling battery capacity comprising: a memory storing executable instructions; a processor coupled to the memory, the The instructions, when executed by the processor, cause the processor to perform the above method.
  • an apparatus for modeling battery capacity comprising: a curve generation module for acquiring charging data to generate a capacity increment curve; and an argument selection module for Selecting a parameter related to a peak in the capacity increment curve as an independent variable and selecting a battery available capacity as a dependent variable to form a data set; and a regression model establishing an optimization module for establishing a regression model for the battery capacity, and The regression model is trained and optimized based on the data set to obtain an optimized regression model.
  • a system for estimating a battery capacity comprising: an acquisition device that collects charging data of a battery; a storage device that stores a regression model obtained by the above method; a battery management device configured to: generate a capacity increment curve using the collected charging data, calculate a value of an independent variable in the regression model according to the generated capacity increment curve; utilize the regression model and calculate The value of the argument determines the available capacity of the battery.
  • a method for estimating a battery capacity comprising: acquiring charging data of a battery to generate a capacity increment curve; and calculating the obtained method according to the generated capacity increment curve The value of the independent variable in the regression model; and using the regression model and the value of the calculated independent variable to determine the available capacity of the battery.
  • a computer readable storage medium having executable instructions thereon that, when executed, cause a processor to perform the above method.
  • the above-described scheme provided by the present invention can collect data collected according to a normal daily charging process (which requires only a partial charging process) by a mathematical model established for battery capacity as compared with the prior art. To accurately estimate the actual available capacity of the battery, especially the actual available capacity of each individual battery in a battery pack such as in an EV/HEV.
  • the above described methods and apparatus provided in accordance with the present invention can be readily applied to online BMS or cloud-based monitoring systems to perform battery capacity determination with very low computing power without additional testing effort.
  • FIG. 1 shows a flow chart of a method for modeling battery capacity in accordance with an embodiment of the present invention
  • FIG. 2 shows a capacity increment (IC) graph in accordance with an embodiment of the present invention
  • FIG. 3 shows a flow chart of a method of generating an IC graph in accordance with an embodiment of the present invention
  • FIG. 4 illustrates a data table of a data set including arguments and dependent variables, in accordance with an embodiment of the present invention
  • FIG. 5 illustrates a flow chart of a method for modeling battery capacity in accordance with another embodiment of the present invention
  • Figure 6 shows a test chart for verifying the trained regression model
  • FIG. 7 shows a schematic diagram of a system for modeling battery capacity in accordance with an embodiment of the present invention.
  • FIG. 8 shows a schematic diagram of an apparatus for modeling battery capacity in accordance with an embodiment of the present invention
  • FIG. 9 shows a schematic diagram of an apparatus for modeling battery capacity in accordance with another embodiment of the present invention.
  • Figure 10 shows a schematic diagram of a system for estimating battery capacity in accordance with an embodiment of the present invention.
  • FIG. 11 illustrates a method for estimating battery capacity in accordance with an embodiment of the present invention. flow chart.
  • a method 100 for modeling battery capacity provided in an embodiment of the present invention will be specifically described in conjunction with FIGS. 1-4.
  • the battery capacity modeling method 100 and the modeling method 500 described below or its corresponding operations can be loaded or run in the processor 702 of the modeling system 700 shown in FIG. The following method or operation can thus be performed by the processor 702, which will be described in detail later.
  • the method 100 first obtains the charging data of the experimental battery based on the cycle experiment at block 101, and calculates a capacity increment (IC) curve (charge amount increment curve) for each cycle experiment based on the charging data. ).
  • IC capacity increment
  • a battery test station and a test battery may be first connected in a laboratory, and then subjected to a full-charge aging (life) cycle test of the test battery.
  • the measuring instrument such as voltage, current, charging capacity (charging amount) known in the art is used as a collecting device to collect the charging and discharging data of the experimental battery in real time, and extract the required charging data.
  • the extracted or derived charging data may include, for example, voltage, current, charging capacity, and battery available capacity for each cycle.
  • the IC curve for each cycle that is, the charge capacity change rate (dQ/dV) versus voltage (V) can be generated by calculation using the charge data of each cycle, as shown in FIG. 2.
  • the IC curve of the battery is an important analytical tool for the capacity increment analysis method.
  • the advantage of the capacity increment analysis method is that the voltage platform involving the first-order phase transition of the battery on the conventional charge-discharge voltage curve can be clearly identified by the voltage increment curve. dQ/dV peak. Therefore, small changes that are not easily found on the charging voltage curve can be reflected on the capacity increment curve.
  • the following method 300 can be employed to generate an IC curve, as shown in FIG.
  • the charging capacity (Q) and the corresponding battery voltage (V) are obtained based on the measured charging data, and the voltage V is derived by the charging capacity Q, thereby determining the charging capacity change rate (dQ/dV).
  • dQ/dV can be obtained by differential calculation as shown in the following equation (1).
  • the IC curve is plotted with the battery voltage V as the horizontal axis and dQ/dV as the vertical axis, as shown in FIG.
  • the IC curve varies with different electrode materials of the battery, and in FIG. 2, only the IC curve of the lithium ion battery is exemplarily shown.
  • peaks 1 and 2 peaks 2 and 3, peaks 3, and peaks corresponding to the corresponding peaks in FIG. 2 (peak height, ie, peak)
  • the ordinate the peak position (that is, the abscissa (voltage) corresponding to the peak), the area (for example, the peak area of No. 4, the peak area of No. 2 and No. 2, and the peak area of No. 3), and the slope of the left and right sides (for example, No.
  • the left boundary of the peak No. 1 is the point where the slope change rate of the curve on the left side of the No. 1 peak (that is, the second derivative of the curve) is equal to 0, and the boundary between the peaks of No. 1 and No. 2 is 1, 2
  • the minimum value between the peaks ie, the curve derivative is equal to 0
  • the boundary of the peaks 2 and 3 is the minimum value of the curve to the right of the 2 peaks (the curve derivative is equal to 0).
  • the right boundary of the 3rd peak is the charging cutoff.
  • the slopes of the left and right sides corresponding to the corresponding peaks can be determined based on the peaks and the left and right boundary points.
  • the area of the corresponding peak is the area wrapped between each peak curve and the left and right boundaries, and the curve can be integrated after the respective peaks define the left and right side boundaries.
  • the peaks, peak positions, areas, left and right side slopes, and the like corresponding to the corresponding peaks in FIG. 2 can be correspondingly obtained by those skilled in the art, and the details of the present invention will not be repeated here.
  • the battery material and the IC curve for different usage times may also have other numbers of peaks (such as 4, 5, etc.) and corresponding parameter values.
  • other parameters related to the peak can be defined and calculated as needed.
  • the parameters associated with the peaks in the IC curve are selected as the independent variables of the regression model described below, and the available battery capacity in each cycle is selected as the cause of the regression model described below. Variables to form a data set containing independent and dependent variables.
  • the parameter related to the peak in the IC curve may be, for example, a parameter such as the height, position, area, or left/right slope of the above peak.
  • a parameter such as the height, position, area, or left/right slope of the above peak.
  • the loss of battery capacity can be divided into the loss of lithium ions and the loss of active materials, and the changes in the two correspond to different changes in the curve of the capacity increase curve: the loss of lithium ions usually only leads to one The decrease in the peak of the capacity increment peak, which in turn causes the capacity corresponding to the peak to decrease, while the other peaks are unaffected; the loss of the active material causes the height of each peak on the capacity increment curve to decrease. .
  • the loss of battery capacity is mainly reflected in the increase of the impedance of the lithium-ion battery.
  • the displacement on the capacity increase curve is the overall offset of the curve. For the capacity increase curve during the charging process, if the impedance increases, the curve is to the right. (high voltage direction) offset.
  • Figure 4 shows a data table containing a data set such as the peak value, the area, the position, and the left and right side slopes of the peak, wherein the first column of the data table is the number of cycles for the full charge of the experimental battery, and the last column For the battery available capacity of the cycle, that is, the maximum available capacity, the middle column is the above-mentioned selected independent variable.
  • a regression model is established for the battery capacity, and the regression model is trained optimized based on the formed data set to obtain an optimized regression model.
  • the regression model established for battery capacity The capacity formula can be expressed as:
  • h is the target value (ie, estimated battery available capacity)
  • ⁇ is the regression model parameter to be optimized or calculated
  • X is the input value from the data set (ie, the selected independent variable).
  • the regression model is a multiple linear regression model.
  • the above-described regression model (i.e., equation (2)) established is trained based on the data set formed at block 102 to obtain an optimized regression model.
  • the established regression model can be trained and optimized based on the formed data set and using a loss function (cost function), thereby calculating an optimized regression model parameter ⁇ , thereby obtaining an optimized regression.
  • the model ie, achieves the best estimate state.
  • the loss function is the optimization goal of parameter optimization, that is, when the loss function reaches a minimum, the optimal estimation state is achieved.
  • the loss function can be expressed as:
  • m is the total number of training data (i.e., for the total number of test cell cycle)
  • X-i is the i-th set of input data (i.e., the i-th cycle corresponding argument)
  • h (Xi) is the group i
  • the output data of the regression model ie, the estimated battery available capacity in the ith cycle
  • y i is the i-th set of true dependent variable data (ie, the battery available capacity measured in the ith cycle).
  • the loss function represents the sum of the least squared differences between the expected output and the true output on all input data.
  • equation (2) can be substituted into equation (3) to evolve equation (3) to the following equation (4):
  • the above loss function is a convex function, and therefore, the convex function must have a minimum point in the numerical space, whereby the purpose of the regression model parameter tuning is to find the minimum point of the loss function.
  • the standard method for finding the spatial minimum of the convex function can be obtained by using the gradient descent method known in the art to obtain the minimum value of the loss function, and the obtained regression model parameter ⁇ value is the parameter optimized by the regression model, that is, The coefficients of the respective variables.
  • the loss function is used to calculate the correlation between the independent variable and the dependent variable
  • the present invention is not intended to be limited thereto, and can be expected by those skilled in the art.
  • other functions known in the art can be used to tune the regression model parameters ⁇ to obtain an optimized regression model.
  • a method 500 for modeling battery capacity provided by another embodiment of the present invention is described in conjunction with FIG.
  • the operations in blocks 501, 502 in the modeling method 500 are the same as those in the above-described blocks 101, 102, and are not described herein again. Only the differences between the modeling methods 500 and 100 will be specifically described below.
  • the method 500 proceeds to block 503 at block 502, after forming a data set containing independent variables (which may also be referred to as alternative independent variables) and dependent variables, as described above.
  • independent variables which may also be referred to as alternative independent variables
  • dependent variables dependent variables
  • the correlation of each independent variable (the alternative independent variable) with the dependent variable is calculated, and the highly correlated independent variable is selected as the final independent variable of the regression model (ie, the final independent variable).
  • the high correlation independent variable is preferably the highest correlation one or the first several independent variables, that is, the highest correlation one or the high correlation Independent variables, for example, the first two, the first three, the first four, or the first five independent variables.
  • a Pearson correlation coefficient can be employed to evaluate the correlation between each independent variable and the dependent variable.
  • the Pearson correlation coefficient can be expressed as follows:
  • COV(X, Y) is the covariance of the two variables
  • ⁇ X and ⁇ Y are the standard deviations of X and Y.
  • the Pearson correlation coefficient of each independent variable and dependent variable calculates the Pearson correlation coefficient of each independent variable and dependent variable, and select one or the first few independent variables with the highest correlation (the absolute value of the Pearson correlation coefficient is the largest) from the independent variables.
  • the final independent variable of the regression model For example, in the example of a 25Ah ternary battery, the peak height and maximum available capacity of the highest peak in the IC curve can be determined. It is highly positively correlated.
  • the Pearson correlation coefficient is used to calculate the correlation between the independent variable and the dependent variable
  • the present invention is not intended to be limited thereto, and is a person skilled in the art. It will be appreciated that other correlation coefficient calculation methods known in the art can also be used to calculate the correlation between the independent variable and the dependent variable.
  • the data set formed at block 502 is updated based on the selected final argument.
  • the data set is updated in the following manner: the unselected independent variables and corresponding data in the data set are deleted, leaving the selected independent variable (ie, the final independent variable), the dependent variable And the corresponding data.
  • the data set may be updated by extracting the selected independent variable (ie, the final independent variable), the dependent variable, and the corresponding data to form a new data set.
  • method 500 proceeds to block 505, where the operations at block 505 are substantially the same as those at block 103 above, except that the established regression model is trained and optimized based on the updated data set to obtain an optimization.
  • Regression model It will be understood by those skilled in the art that in method 100, the data set formed at block 102 is used at block 103 for correlation operations, and in the present embodiment, the updated data set is employed at block 505. Related operations are performed, and the present invention will not be described again in order to simplify the present invention.
  • the updated data set can be divided into a training data set and a verification data set, wherein the training data set can be used to train the established regression model, and the test data set can be used for verification training.
  • the accuracy of the optimized regression model can be further preferably, the data set can be divided into a training data set and a verification data set proportionally according to the number of cycles. The ratio may preferably be 8:2, 9:1 or the like.
  • the assigned verification data set can be selected as an input value to test the accuracy of the optimized regression model. As shown in Fig. 6, the absolute average error of the test results is within 0.2 Ah, which indicates the high accuracy of the training-optimized regression model in the capacitance capacity estimation.
  • the training optimized regression model may be stored for estimation of battery capacity, wherein the stored regression model includes calculated or optimized regression model parameters ⁇ and selected in the regression model The argument X.
  • the above modeling method 500 reduces the regression on the one hand with respect to the modeling method 100.
  • the number of independent variables X selected in the model further saves the computing power of the processor, and on the other hand, the correlation calculation is used to further improve the accuracy of estimating the battery capacity.
  • Figure 7 shows a schematic diagram of a system for modeling battery capacity in accordance with an embodiment of the present invention.
  • the modeling system 700 can include a processor 701 and a memory 702 coupled to the processor 701.
  • the memory 702 stores executable instructions that, when executed, cause the processor 701 to perform the operations included in the foregoing methods 100, 300, and 500.
  • Figure 8 shows a schematic diagram of an apparatus for modeling battery capacity in accordance with an embodiment of the present invention.
  • the modeling apparatus 800 shown in FIG. 8 can be implemented in a combination of software, hardware, or a combination of software and hardware.
  • the modeling apparatus 800 includes a curve generation module 801, an independent variable selection module 802, and a regression model establishment optimization module 803.
  • the curve generation module 801 acquires the charging data of the experimental battery based on the cycle experiment, and generates a corresponding IC curve for each cycle experiment.
  • the independent variable selection module 802 selects a parameter related to a peak in the IC curve as an independent variable of the regression model according to the IC curve generated by the curve generation module 801 and selects a battery available capacity as a dependent variable of the regression model, thereby forming an inclusion variable and The data set of the dependent variable.
  • the regression model establishment optimization module 803 establishes a regression model for the battery capacity, and performs training optimization on the regression model based on the formed data set.
  • FIG. 9 shows a schematic diagram of an apparatus for modeling battery capacity in accordance with another embodiment of the present invention.
  • the modeling apparatus 900 shown in FIG. 9 includes a curve generation module 901, an argument selection module 902, a correlation calculation module 903, a data set update module 904, and a regression model establishment optimization module 905.
  • the curve generation module 901 and the argument selection module 902 have the same functions as the curve generation module 801 and the argument selection module 802 in the modeling apparatus 800, and are not described herein again.
  • the correlation calculation module 903 calculates the correlation between each independent variable and the dependent variable according to the independent variable selected by the independent variable selection module 902, and selects a high correlation from each of the independent variables.
  • the independent variable is used as the final independent variable of the regression model (ie, the final independent variable).
  • the data set update module 904 updates the formed data set based on the final argument selected by the correlation calculation module 903.
  • the unselected independent variables and corresponding data in the data set are deleted, leaving the selected final independent variable, dependent variable and corresponding data.
  • the regression model establishment optimization module 905 has substantially the same function as the regression model establishment optimization module 803, except that it trains the regression model based on the data set updated by the data set update module 904 to obtain an optimal regression model.
  • the battery capacity estimation system 1000 includes an acquisition device 1001, a storage device 1002, and a battery management device 1003.
  • the collection device 1001 and the storage device 1002 can be communicably connected to the battery management device 1003 by wire or wirelessly, respectively.
  • the collecting device 1001 is configured to collect charging data of the battery when the battery to be estimated enters a charging state.
  • the collecting device can employ measuring instruments such as voltage, current, and charging capacity known in the art.
  • the charging data may include a voltage measured or collected during a full charge or a partial charge, a charge capacity corresponding thereto, and the like.
  • the storage device 1002 is configured to store and train an optimized regression model for battery capacity according to the above method 100 or 500, wherein the regression model has an optimized parameter vector ⁇ and a selected independent variable X.
  • the storage device may be a local computer readable storage medium such as a hard drive, random access memory (RAM), and/or optical read only memory, or may be a network storage medium such as a cloud drive, a memory, a server, or the like.
  • the battery management device 1003 generates a capacity increment curve by using the collected charging data of the battery to be estimated, and calculates a value (ie, an input value) of the selected independent variable in the stored regression model according to the generated capacity increment curve.
  • a value ie, an input value
  • the charging data is measured during the process (full charge or partial charge), and the capacity increase curve is drawn according to the measured charge data to calculate the stored regression The argument selected by the model.
  • the battery management device 1003 calls the regression model stored in the storage device 1002 and uses the regression model and the value of the calculated argument to estimate the actual available capacity of the battery.
  • the method 1100 collects the charging data of the battery to be estimated by the above-mentioned collecting device 1001 at block 1101. Specifically, when the battery enters the charging state, the above-described collecting device 1001 collects and records the voltage of the battery during charging and the charging capacity corresponding to the voltage, and the like. In an example, when estimating the available capacity of the battery pack in the EV/HEV, the voltage of each of the individual cells in the battery pack and the charging capacity corresponding to the voltage may be separately collected. In another example, when estimating the available capacity of the battery pack in the EV/HEV, the voltage of the entire battery pack and the charging capacity corresponding to the voltage may also be collected.
  • the battery management device 1003 can utilize the acquired charging capacity and voltage to generate an IC curve and calculate the selected one of the regression models stored in the storage device 1002 based on the generated IC curve. The value of the argument.
  • the battery management device 1003 utilizes the stored regression model and the calculated values of the independent variables to determine or estimate the available capacity of the battery to be estimated. For example, in a battery pack for an EV/HEV, the actual usable capacity of each of the battery cells or the battery pack can be estimated.
  • embodiments of the present invention also provide a computer readable storage medium having executable instructions thereon that, when executed, cause a processor to perform the methods 100, 300, 500, and 1100 as described above.

Landscapes

  • Engineering & Computer Science (AREA)
  • Manufacturing & Machinery (AREA)
  • Chemical & Material Sciences (AREA)
  • Chemical Kinetics & Catalysis (AREA)
  • Electrochemistry (AREA)
  • General Chemical & Material Sciences (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Secondary Cells (AREA)
  • Charge And Discharge Circuits For Batteries Or The Like (AREA)

Abstract

一种用于对电池容量进行建模的方法，其包括：获取充电数据，以生成容量增量曲线；选择与所述容量增量曲线中的峰相关的参数作为自变量并选择电池可用容量作为因变量，以形成数据集；针对电池容量建立回归模型，并基于所述数据集对所述回归模型进行训练优化以获得优化的回归模型。所提供的上述方法与现有技术相比，能够基于对电池容量所建立的数学模型而根据正常的日常充电过程(这仅需要部分充电过程)所采集的数据来确定电池容量，尤其是诸如EV/HEV中的电池组中各个单体电池的容量。

Description

用于对电池容量进行建模和估计的方法及系统 技术领域
本发明涉及电池测量检测技术，更具体地，涉及用于对电池容量进行建模和估计的方法、装置及系统。
背景技术
诸如锂离子电池(LiB)之类的电池已广泛应用于电动或混合动力电动汽车(EV/HEV)中。EV/HEV中的电池组通常包括数十到数百个单体电池。电池寿命随时间以及不同的电气负载而变化。EV/HEV上的电池组所包含的多个电池会以不同的速度老化，并且在不同的充电状态(SOC)下工作。因此，提供电池管理系统(BMS)来确定或预测电池组的状态是至关重要的。实时应用电池管理系统的一个重要的挑战是如何确定电池的老化状态，即，电池可用容量(最大可用容量)。
要获得电池准确的实际可用容量，需要将电池连接特定的电池测试设备，做完整区间的充电或放电(即，满充满放)测试。在现有技术中，确定电池可用容量的常规方法是以小电流完全充电和放电。然而，这种常规的电池可用容量确定方法的问题在于，在不将EV/HEV上的电池组拆卸成单体电池的情况下，直接对电池组中的各个单体电池满充满放是不现实的。此外，在电动汽车的实际使用中，考虑到电池的寿命问题与实际使用情况，动力电池往往也不能被放电至空电状态，因此也无法通过测量设备直接测量电池的实际可用容量。
发明内容
本发明正是为了解决现有技术中确定电池容量的常规方法所存在的上述问题，而提供了用于对电池容量进行建模和估计的方法、装 置及系统。
根据本发明的一方面，提供了一种用于对电池容量进行建模的方法，包括：获取充电数据，以生成容量增量曲线；选择与所述容量增量曲线中的峰相关的参数作为自变量并选择电池可用容量作为因变量，以形成数据集；针对电池容量建立回归模型，并基于所述数据集对所述回归模型进行训练优化以获得优化的回归模型。
根据本发明的另一方面，还提供了一种用于对电池容量进行建模的系统，所述系统包括：存储器，其存储有可执行指令；处理器，其耦合到所述存储器，所述指令在由所述处理器执行时使得所述处理器执行上述方法。
根据本发明的另一方面，还提供了一种用于对电池容量进行建模的装置，包括：曲线生成模块，其用于获取充电数据以生成容量增量曲线；自变量选择模块，其用于选择与所述容量增量曲线中的峰相关的参数作为自变量并选择电池可用容量作为因变量，以形成数据集；以及回归模型建立优化模块，其用于针对电池容量建立回归模型，并基于所述数据集对所述回归模型进行训练优化，以获得优化的回归模型。
根据本发明的另一方面，还提供了一种用于估计电池容量的系统，所述系统包括：采集设备，其采集电池的充电数据；存储设备，其存储上述方法所获得的回归模型；以及电池管理设备，其被配置为：利用所采集的充电数据来生成容量增量曲线，根据所生成的容量增量曲线计算所述回归模型中的自变量的值；利用所述回归模型和所计算的自变量的值来确定电池的可用容量。
根据本发明的另一方面，还提供了一种用于估计电池容量的方法，包括：获取电池的充电数据，以生成容量增量曲线；根据所生成的容量增量曲线计算上述方法所获得的回归模型中的自变量的值；以及利用所述回归模型和所计算的自变量的值来确定电池的可用容量。
根据本发明的另一方面，还提供了一种计算机可读存储介质，其上具有可执行指令，当所述可执行指令被执行时，使得处理器执行上述方法。
从以上描述可以看出，本发明所提供的上述方案与现有技术相比，能够通过对电池容量所建立的数学模型而根据正常的日常充电过程(这仅需要部分充电过程)所采集的数据来精确估计电池实际的可用容量，尤其是诸如EV/HEV中的电池组中各个单体电池的实际可用容量。根据本发明所提供的上述方法及装置可以很容易地应用于在线BMS或基于云的监控系统，以非常低的计算能力来执行电池容量确定，而无需额外的测试工作。
附图说明
本发明的其它特征、特点、益处和优点通过以下结合附图的详细描述将变得更加显而易见。其中：
图1示出了根据本发明一实施例的用于对电池容量进行建模的方法的流程图；
图2示出了根据本发明一实施例的容量增量(IC)曲线图；
图3示出了根据本发明一实施例的生成IC曲线图的方法的流程图；
图4示出了根据本发明一实施例的包含自变量和因变量的数据集的数据表；
图5示出了根据本发明另一实施例的用于对电池容量进行建模的方法的流程图；
图6示出了对训练优化后的回归模型进行验证的测试图；
图7示出了根据本发明一实施例的用于对电池容量进行建模的系统的示意图；
图8示出了根据本发明一实施例的用于对电池容量进行建模的装置的示意图；
图9示出了根据本发明另一实施例的用于对电池容量进行建模的装置的示意图；
图10示出了根据本发明一实施例的用于估计电池容量的系统的示意图。
图11示出了根据本发明一实施例的用于估计电池容量的方法的 流程图。
具体实施方式
下面结合附图和具体实施例对本发明提供的用于对电池容量进行建模和估计的方法、装置以及系统进行详细描述。下述各个实施例及其特征在没有明确表示相互矛盾的情况下，可相互组合、替换等。
首先，结合图1-图4来具体描述本发明在一实施例中所提供的用于对电池容量进行建模的方法100。此外，可以理解的是，该电池容量建模方法100和下述的建模方法500或其对应的下述操作均可加载或运行在图7所示的建模系统700的处理器702中，从而可由处理器702来执行以下方法或操作，稍后将对此进行详细描述。
如图1所示，方法100首先在块101处，基于循环实验来获取实验电池的充电数据，并基于充电数据计算出针对每次循环实验的容量增量(IC)曲线(充电量增量曲线)。
在本发明的一优选实施例中，可首先在实验室中将电池测试站和实验电池(例如，锂离子电池)相连接，随后通过对实验电池进行满充满放衰退老化(寿命)循环实验，并利用本领域公知的电压、电流、充电容量(充电量)等测量仪器作为采集设备来实时采集实验电池的充放电数据，并提取出所需的充电数据。所提取或导出的充电数据可包括例如针对每次循环的电压、电流、充电容量、以及电池可用容量等。
可以利用每次循环的充电数据通过计算来生成针对每次循环的IC曲线，即，充电容量变化率(dQ/dV)与电压(V)的关系曲线，如图2所示。电池的IC曲线为容量增量分析法的重要分析工具，容量增量分析法的优点是将传统充放电电压曲线上涉及电池一阶相变的电压平台转化成容量增量曲线上能明确识别的dQ/dV峰。因此，在充电电压曲线上不易发现的微小的变化都可以在容量增量曲线上反应出来。
在本发明的一优选实施例中，可采用以下方法300来生成IC曲线，如图3所示。
在块301处，根据所测得的充电数据来获取充电容量(Q)和对应的电池电压(V)，通过充电容量Q对电压V求导，从而求得充电容量变化率(dQ/dV)。在一优选实施例中，可以通过差分计算来求得dQ/dV，如下式(1)所示。
Figure PCTCN2017094036-appb-000001
在块302处，根据计算出的dQ/dV，在直角坐标系中，以电池电压V作为水平轴而以dQ/dV作为纵轴来绘制IC曲线，如图2所示。
IC曲线随电池的不同电极材料而变化，在图2中，仅示例性示出了锂离子电池的IC曲线。可以根据图2中的IC曲线识别出数个峰，例如在图2中的1号峰1、2号峰2、3号峰3、以及与相应峰所对应的峰值(峰高度，即峰的纵坐标)、峰位置(即，峰对应的横坐标(电压))、面积(例如，1号峰面积4、2号峰面积7以及3号峰面积10)、左右侧斜率(例如，1号峰左侧斜率5和右侧斜率6、2号峰左侧斜率8和右侧斜率9)等。另外，如图2所示，1号峰的左边界为1号峰左侧曲线斜率变化率(即，曲线的二次导数)等于0的点，1、2号峰的分界为1、2号峰之间的极小值处(即，曲线导数等于0处)，2、3号峰的分界为2峰右侧曲线极小值处(曲线导数等于0处)。3号峰右边界为充电截止处。相应峰所对应的左右侧斜率可基于峰值与左右边界点连直线后求斜率。相应峰的面积为各峰曲线与左右边界之间包裹的面积，可在各个峰确定左右侧边界之后，对曲线进行积分而得出。
本领域技术人员基于以上限定可相应求得图2中与相应峰所对应的峰值、峰位置、面积、左右侧斜率等，为了不模糊本发明的主题，这里不再一一赘述。同时，在这里需要说明的是，尽管在图2中示出了锂离子电池容量增量曲线中3个峰以及与相应峰相关的参数，然而，对于本领域技术人员可以预期的是，对于不同电池材料、不同使用时间的IC曲线也可具有其它数量的峰(如4个、5个等)以及对应的参数值。此外，也可以根据需要来定义与峰相关的其它参数，并进行计算。
再次回到图1，接下来，在块102处，选择与IC曲线中的峰相关的参数作为下述回归模型的自变量，并选择每个循环中的电池可用容量作为下述回归模型的因变量，从而形成包含自变量和因变量的数据集。
如上所述，与IC曲线中的峰相关的参数可以例如为上述峰的高度、位置、面积或左/右侧斜率等参数。在针对锂离子电池的优选示例中，优选将容量增量曲线中峰的峰值、面积、位置、和左右侧斜率选取为自变量。这是因为，本发明人认识到锂离子电池的老化一般分为热力学特性的改变和动力学特性的改变。在热力学特性方面，电池容量的损失可以分为锂离子的损失和活性材料的损失，而两者的改变在容量增量曲线上对应着曲线不同的变化：锂离子的损失通常只会导致某一个容量增量峰在峰值上的减小，进而导致该峰所对应的容量减小，而其他的峰不受影响；活性材料的损失会导致容量增量曲线上的每一个峰的高度都减小。在动力学特性方面，主要体现为锂离子电池阻抗的增加，在容量增量曲线上的体现为曲线整体的偏移，对于充电过程中的容量增量曲线，若阻抗增加，则曲线向右侧(高电压方向)偏移。因此所选参数与电池老化衰退具有对应关系。图4示出了包含峰的峰值、面积、位置、和左右侧斜率这样的数据集的一个数据表，其中，数据表的第1列为针对实验电池进行满充满放的循环次数，最后1列为该次循环的电池可用容量，即最大可用容量，中间列为上述所选的自变量。
在这里需要说明的是，尽管在图4示出的数据表中示出了峰值、面积、位置等11个参数作为自变量，然而，本发明并不旨在限制于此，对于本领域技术人员可以理解的是，针对不同的电池及其特性，也可以选择与峰相关的其它参数，或仅选择该数据表中的若干参数作为自变量。
接下来，在块103处，针对电池容量建立回归模型，并基于上述所形成的数据集对该回归模型进行训练优化，从而获得优化的回归模型。
在本发明的一优选实施例中，针对电池容量所建立的回归模型的 容量公式可表示为：
h(X)＝θTX           等式(2)
其中h是目标值(即，估计出的电池可用容量)，θ是待优化或计算的回归模型参数，X是来自数据集的输入值(即，所选的自变量)。其中，在所选的自变量X为多个时，该回归模型即为多元线性回归模型。基于在块102处所形成的数据集对所建立的上述回归模型(即等式(2))进行训练优化，从而获得优化的回归模型。
在本发明的一优选实施例中，可基于所形成的数据集并利用损失函数(cost函数)对所建立的回归模型进行训练优化，从而计算出优化的回归模型参数θ，从而获得优化的回归模型，即，实现最佳估计状态。损失函数是参数优化的优化目标，即，当损失函数达到最小时，实现最佳估计状态。损失函数可表示为：
Figure PCTCN2017094036-appb-000002
其中m是训练数据的总数量(即，针对实验电池的总循环次数)，Xi是第i组输入数据(即，第i次循环所对应的自变量)，h(Xi)是第i组回归模型的输出数据(即，第i次循环所估计的电池可用容量)，yi是第i组真实因变量数据(即，第i次循环实测的电池可用容量)。损失函数表示在所有输入数据上预期输出与真实输出的最小平方差之和。利用损失函数对回归模型进行训练优化的具体过程如下：
首先，可将等式(2)代入等式(3)从而将等式(3)演变为以下等式(4)：
Figure PCTCN2017094036-appb-000003
对于本领域技术人员而言已知上述损失函数为凸函数，因此，在数值空间中该凸函数必有一个最小值点，由此回归模型参数调优的目的为寻找此损失函数的最小值点。对该凸函数求空间最小值的标准方法可采用本领域已知的梯度下降法，来求得该损失函数的最小值，此时得到的回归模型参数θ值即为回归模型优化的参数，即各自变量的系数。
在这里需要说明的是，尽管在上述优选实施例中，采用损失函数来计算自变量与因变量之间的相关性，然而，本发明并不旨在限制于此，对于本领域技术人员可以预期的是，也可以采用本领域已知的其它函数来对回归模型参数θ进行调优，来获得优化的回归模型。
下面，结合图5来具体描述本发明在另一实施例中所提供的用于对电池容量进行建模的方法500。该建模方法500中的块501、502中的操作与上述块101、102中的操作相同，这里不再赘述。下面仅针对建模方法500与100的不同之处进行具体描述。
方法500在块502处，如上所述，形成包含自变量(也可称为备选自变量)和因变量的数据集之后，行进到块503。在块503处，计算每个自变量(备选自变量)与因变量的相关性，并选择相关性高的自变量作为回归模型最终的自变量(即，最终自变量)。
在本发明的一优选实施例中，该相关性高的自变量优选为相关性最高的一个自变量或前几个自变量，即，相关性最高的一个自变量或相关性高的前几个自变量，例如，前两个、前三个、前四个、或前五个自变量等。
在本发明的一优选实施例中，可采用皮尔逊(Pearson)相关系数来评估每个自变量与因变量之间的相关性。皮尔逊相关系数越高，从一个变量去预测另一个变量的精确度就越高。皮尔逊相关系数可如下表示为：
Figure PCTCN2017094036-appb-000004
其中X和Y为两个变量，COV(X,Y)是两个变量的协方差，σX，σY是X和Y的标准偏差。皮尔逊相关系数ρX，Y在-1和1之间，其中ρX，Y>0表示正相关，ρX，Y<0表示负相关，ρX，Y＝0表示不相关。
根据上述等式(5)，计算每个自变量和因变量的皮尔逊相关系数，并从自变量中选择相关性最高(皮尔逊相关系数的绝对值最大)的一个或前几个自变量作为回归模型最终的自变量。例如，在25Ah三元电池的示例中，可确定出IC曲线中最高峰的峰高度与最大可用容量 呈高度正相关。
在这里需要说明的是，尽管在上述优选实施例中，采用皮尔逊相关系数来计算自变量与因变量之间的相关性，然而，本发明并不旨在限制于此，对于本领域技术人员可以理解的是，也可以采用本领域已知的其它相关系数计算方法来计算自变量与因变量之间的相关性。
接下来，在块504处，基于所选的最终自变量来对在块502处形成的数据集进行更新。
在本发明的一优选实施例中，采用以下方式对数据集进行更新：将数据集中未选的自变量及对应的数据删除，留下所选的自变量(即，最终自变量)、因变量及对应的数据。在本发明的另一优选实施例中，对数据集进行更新的方式也可以是提取出所选的自变量(即，最终自变量)、因变量及对应的数据而构成新的数据集。
接下来，方法500行进到块505处，块505处的操作与上述块103处的操作基本相同，不同之处在于基于所更新的数据集对所建立的回归模型进行训练和优化，从而获得优化的回归模型。对于本领域技术人员可以理解的是，在方法100中，在块103处采用在块102处形成的数据集进行相关操作，而在本实施例中，在块505处均采用更新后的数据集进行相关操作，为了使本发明简明，这里不再赘述。
在本发明的一优选实施例中，可按比例将更新后的数据集分为训练数据集和验证数据集，其中，训练数据集可用于训练所建立的回归模型，测试数据集可用于验证训练优化后的回归模型的准确性。进一步优选地，可根据循环次数按比例将数据集分为训练数据集和验证数据集。该比例可优选为8:2、9:1等。
可选择所划分的验证数据集作为输入值来测试优化后的回归模型的准确性。如图6所示，测试结果的绝对平均误差在0.2Ah以内，从而表明该训练优化后的回归模型在电容容量估计中的高准确度。
在本发明的一优选实施例中，可将训练优化后的回归模型进行存储以用于电池容量的估计，其中所存储的回归模型包括计算出或优化的回归模型参数θ以及回归模型中所选的自变量X。
上述建模方法500相对于建模方法100而言，一方面减少了回归 模型中所选的自变量X个数，从而进一步节省了处理器的计算能力，另一方面利用相关性计算而进一步提高了对电池容量估计的准确度。
图7示出了按照本发明一实施例的用于对电池容量进行建模的系统的示意图。如图7所示，该建模系统700可以包括处理器701和与处理器701耦合的存储器702。其中，存储器702存储有可执行指令，该可执行指令当被执行时使得处理器701可执行前述方法100、300以及500所包括的操作。
图8示出了按照本发明一实施例的用于对电池容量进行建模的装置的示意图。图8所示的建模装置800可以利用软件、硬件或软硬件结合的方式来执行。该建模装置800包括曲线生成模块801、自变量选择模块802、以及回归模型建立优化模块803。
曲线生成模块801基于循环实验获取实验电池的充电数据，并针对每次循环实验生成对应的IC曲线。
自变量选择模块802根据曲线生成模块801所生成的IC曲线来选择与IC曲线中的峰相关的参数作为回归模型的自变量并选择电池可用容量作为回归模型的因变量，从而形成包含自变量和因变量的数据集。
回归模型建立优化模块803针对电池容量建立回归模型，并基于所形成的数据集对回归模型进行训练优化。
图9示出了根据本发明另一实施例的用于对电池容量进行建模的装置的示意图。图9所示的建模装置900包括曲线生成模块901、自变量选择模块902、相关性计算模块903、数据集更新模块904以及回归模型建立优化模块905。其中，曲线生成模块901和自变量选择模块902与建模装置800中的曲线生成模块801和自变量选择模块802的功能相同，这里不再赘述。
相关性计算模块903根据自变量选择模块902所选择的自变量来计算每个自变量与因变量的相关性，并从每个自变量中选择相关性高 的自变量作为回归模型最终的自变量(即，最终自变量)。
数据集更新模块904基于由相关性计算模块903选择的最终自变量对所形成的数据集进行更新。优选地，将数据集中未选的自变量及对应的数据删除，留下所选的最终自变量、因变量及对应的数据。
回归模型建立优化模块905与回归模型建立优化模块803的功能基本相同，不同之处在于其基于由数据集更新模块904更新的数据集对回归模型进行训练，从而获得最优回归模型。
下面具体描述基于所建立并优化后的回归模型来估计电池容量的系统和方法。
首先，结合图10来描述本发明一实施例的用于估计电池容量的系统1000。该电池容量估计系统1000包括采集设备1001、存储设备1002以及电池管理设备1003。采集设备1001和存储设备1002可分别通过有线或无线方式与电池管理设备1003通信连接。
采集设备1001用于在待估计电池进入充电状态时采集电池的充电数据。采集设备可采用本领域公知的电压、电流、充电容量等测量仪器。该充电数据可包括在满充或部分充电过程中测量或采集的电压和与其对应的充电容量等。
存储设备1002用于存储根据上述方法100或500针对电池容量所建立并训练优化后的回归模型，其中该回归模型具有优化的参数矢量θ以及所选的自变量X。存储设备可以是诸如硬盘驱动器、随机存取存储器(RAM)和/或光只读存储器等本地计算机可读存储介质，也可以是诸如云端硬盘、存储器、服务器等网络存储介质。
电池管理设备1003利用所采集的待估计电池的充电数据来生成容量增量曲线，并根据所生成的容量增量曲线计算出所存储的回归模型中所选的自变量的值(即，输入值)。对于本领域技术人员可理解的是，在实际电池可用容量估计过程中，由于并不需要实际测量电池的可用容量，因此也无需进行满充满放整个完整区间的充电数据测量，而仅需要在充电过程(满充或部分充电)期间测得充电数据，并根据所测的充电数据来绘制容量增量曲线从而计算出所存储的回归 模型所选的自变量。随后，电池管理设备1003调用存储设备1002中存储的回归模型，并利用该回归模型和所计算的自变量的值来估计电池的实际可用容量。
接下来，结合图11来进一步描述本发明一实施例的用于估计电池容量的方法1100。
在实际应用中，如图11所示，方法1100在块1101处，利用上述采集设备1001采集待估计电池的充电数据。具体来说，当电池进入充电状态时，利用上述采集设备1001采集并记录充电过程中电池的电压和与电压相对应的充电容量等。在一示例中，当对EV/HEV中的电池组可用容量进行估计时，可分别采集电池组中各个单体电池的电压和与电压相对应的充电容量。在另一示例中，当对EV/HEV中的电池组可用容量进行估计时，也可采集整个电池组的电压和与电压相对应的充电容量。
在充电过程完成后，在块1102处，电池管理设备1003可利用所采集的充电容量和电压来生成IC曲线，并根据所生成的IC曲线计算存储在存储设备1002中的回归模型中所选的自变量的值。
最后，在块1103处，电池管理设备1003利用所存储的回归模型和计算出的自变量的值来确定或估计出待估计电池的可用容量。例如，在针对EV/HEV中的电池组中，可估计出电池组或电池组中每个单体电池的实际可用容量。
利用本发明的上述方法和系统，能够基于对电池容量预先建立好的回归模型而根据正常的日常充电过程(这仅需要部分充电过程)所测量或采集的充电数据来精确估计电池实际的最大可用容量，而无需对电池组或电池完整充放电或放电至空电状态，提高了电池可用容量估计的安全性和精确性，同时还节省了时间，并易于电池管理系统对电池的监控和管理。
此外，本发明实施例还提供一种计算机可读存储介质，其上具有可执行指令，当所述可执行指令被执行时，使得处理器执行如上所述的方法100、300、500以及1100。
以上结合具体实施例对本发明进行了详细描述。显然，以上描述以及在附图中示出的实施例均应被理解为是示例性的，而不构成对本发明的限制。对于本领域技术人员而言，可以在不脱离本发明的精神的情况下对其进行各种变型或修改，这些变型或修改均不脱离本发明的范围。因此，本发明的保护范围由所附的权利要求书来限定。

Claims (13)

  1. 一种用于对电池容量进行建模的方法，包括：
    获取充电数据，以生成容量增量曲线；
    选择与所述容量增量曲线中的峰相关的参数作为自变量并选择电池可用容量作为因变量，以形成数据集；
    针对电池容量建立回归模型，并基于所述数据集对所述回归模型进行训练优化以获得优化的回归模型。
  2. 如权利要求1所述的方法，其中，所述生成容量增量曲线包括：
    根据所述充电数据来获取充电容量和电压，通过所述充电容量对所述电压求导以得到充电容量变化率；
    以所述电压作为水平轴而以所述充电容量变化率作为纵轴来生成所述容量增量曲线。
  3. 如权利要求1所述的方法，其中，所述基于所述数据集对所述回归模型进行训练优化以获得优化的回归模型包括基于所述数据集并利用损失函数对回归模型进行训练优化，并对优化的回归模型进行存储。
  4. 如权利要求1-3中的任一项所述的方法，其中，所述参数为与相应峰所对应的峰值、位置、面积以及左右侧斜率。
  5. 如权利要求1-3中的任一项所述的方法，在所述形成数据集之后且在所述建立回归模型之前，所述方法还包括：
    计算每个所述自变量与所述因变量的相关性，并选择相关性最高的一个自变量或前几个自变量作为最终自变量；以及
    基于所选的最终自变量对所形成的数据集进行更新，
    其中，所述基于数据集对所述回归模型进行训练优化是基于更新 后的数据集对所述回归模型进行训练优化。
  6. 如权利要求5所述的方法，其中，采用皮尔逊相关系数来计算每个所述自变量与所述因变量之间的相关性，并选择相关性最高的一个自变量或前几个自变量作为最终自变量。
  7. 如权利要求5所述的方法，其中，按比例将更新后的数据集分为训练数据集和验证数据集，并利用按比例划分的所述训练数据集对所述回归模型进行训练，并利用按比例划分的验证数据集来验证所述回归模型的准确性。
  8. 一种用于对电池容量进行建模的系统，所述系统包括：
    存储器，其存储有可执行指令；
    处理器，其耦合到所述存储器，所述指令在由所述处理器执行时使得所述处理器执行如权利要求1-7中的任一项所述的方法。
  9. 一种用于对电池容量进行建模的装置，包括：
    曲线生成模块，其用于获取充电数据以生成容量增量曲线；
    自变量选择模块，其用于选择与所述容量增量曲线中的峰相关的参数作为自变量并选择电池可用容量作为因变量，以形成数据集；以及
    回归模型建立优化模块，其用于针对电池容量建立回归模型，并基于所述数据集对所述回归模型进行训练优化，以获得优化的回归模型。
  10. 如权利要求9所述的装置，还包括：
    相关性计算模块，其用于计算所述自变量与所述因变量的相关性，并选择相关性最高的一个自变量或前几个自变量作为最终自变量；以及
    数据集更新模块，其用于基于所选定的最终自变量对所形成的数 据集进行更新，
    其中，所述回归模型建立优化模块基于更新后的数据集对所述回归模型进行训练优化。
  11. 一种用于估计电池容量的系统，所述系统包括：
    采集设备，其采集电池的充电数据；
    存储设备，其存储如权利要求1-7中的任一项所述的方法所获得的回归模型；以及
    电池管理设备，其被配置为：
    利用所采集的充电数据来生成容量增量曲线，根据所生成的容量增量曲线计算所述回归模型中的自变量的值；
    利用所述回归模型和所计算的自变量的值来确定电池的可用容量。
  12. 一种用于估计电池容量的方法，包括：
    获取电池的充电数据，以生成容量增量曲线；
    根据所生成的容量增量曲线计算如权利要求1-7中任一项所述的方法所获得的回归模型中的自变量的值；以及
    利用所述回归模型和所计算的自变量的值来确定电池的可用容量。
  13. 一种计算机可读存储介质，其上具有可执行指令，当所述可执行指令被执行时，使得处理器执行如权利要求1-7、12中的任一项所述的方法。
PCT/CN2017/094036 2017-07-24 2017-07-24 用于对电池容量进行建模和估计的方法及系统 WO2019018974A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
PCT/CN2017/094036 WO2019018974A1 (zh) 2017-07-24 2017-07-24 用于对电池容量进行建模和估计的方法及系统
CN201780093207.XA CN111448467B (zh) 2017-07-24 2017-07-24 用于对电池容量进行建模和估计的方法及系统

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2017/094036 WO2019018974A1 (zh) 2017-07-24 2017-07-24 用于对电池容量进行建模和估计的方法及系统

Publications (1)

Publication Number Publication Date
WO2019018974A1 true WO2019018974A1 (zh) 2019-01-31

Family

ID=65039450

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2017/094036 WO2019018974A1 (zh) 2017-07-24 2017-07-24 用于对电池容量进行建模和估计的方法及系统

Country Status (2)

Country Link
CN (1) CN111448467B (zh)
WO (1) WO2019018974A1 (zh)

Cited By (7)

* 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
CN110927591A (zh) * 2019-12-11 2020-03-27 北京理工大学 一种电池容量估计方法、计算机可读介质及车辆
CN111414709A (zh) * 2020-04-14 2020-07-14 上海交通大学 一种基于神经网络的超级电容器循环寿命和容量衰减的早期预测方法
CN112014745A (zh) * 2020-08-27 2020-12-01 洛阳理工学院 一种针对动力电池的三级故障诊断方法
CN113158345A (zh) * 2021-04-29 2021-07-23 浙江吉利控股集团有限公司 一种新能源车辆动力电池容量预测方法及系统
CN113640674A (zh) * 2021-06-30 2021-11-12 昆明理工大学 基于优化高斯过程回归的锂离子电池可用容量估计方法
CN116205379A (zh) * 2023-05-04 2023-06-02 上海叁零肆零科技有限公司 锂亚电池存活天数预测模型建立方法、系统、设备及介质

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112415400B (zh) * 2020-10-21 2023-09-12 欣旺达电动汽车电池有限公司 电池容量估计方法及系统
CN112834946B (zh) * 2020-12-31 2024-06-25 东软睿驰汽车技术（沈阳）有限公司 一种评估电池电芯衰减程度的方法及装置
CN113884912A (zh) * 2021-10-18 2022-01-04 苏州元启动力科技有限公司 基于模拟退火算法的插值拟合锂电池dQ/dV曲线方法及装置
CN114210591B (zh) * 2021-12-02 2023-12-22 格林美股份有限公司 一种基于ic曲线的锂电池梯次利用分选方法及装置
CN115184805A (zh) * 2022-06-21 2022-10-14 东莞新能安科技有限公司 电池健康状态获取方法、装置、设备及计算机程序产品

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103035961A (zh) * 2011-10-10 2013-04-10 波音公司 电池组适应性学习管理系统
CN103278777A (zh) * 2013-05-24 2013-09-04 杭州电子科技大学 一种基于动态贝叶斯网络的锂电池健康状况估计方法
CN105789716A (zh) * 2016-03-03 2016-07-20 北京交通大学 一种广义电池管理系统

Family Cites Families (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101067644B (zh) * 2007-04-20 2010-05-26 杭州高特电子设备有限公司 蓄电池性能分析专家诊断方法
CN103336913B (zh) * 2013-07-25 2016-03-16 哈尔滨工业大学 一种适合于空间锂离子电池状态监测和截止电压预测的方法
US10393813B2 (en) * 2013-08-27 2019-08-27 The Regents Of The University Of Michigan On-board state of health monitoring of batteries using incremental capacity analysis
CN104914723B (zh) * 2015-05-22 2018-10-16 浙江大学 基于协同训练偏最小二乘模型的工业过程软测量建模方法
US9800073B2 (en) * 2015-08-20 2017-10-24 International Business Machines Corporation Modeling a change in battery degradation
CN105320114B (zh) * 2015-11-09 2018-07-06 长春合成兴业能源技术有限公司 基于数据挖掘的火力发电站锅炉燃烧调整模型获取方法
CN106055775B (zh) * 2016-05-27 2018-12-11 哈尔滨工业大学 一种粒子滤波与机理模型相结合的二次电池寿命预测方法
CN106125007A (zh) * 2016-08-31 2016-11-16 北京新能源汽车股份有限公司 一种电池剩余电量的确定方法、装置及汽车
CN106443467A (zh) * 2016-09-18 2017-02-22 北京交通大学 基于充电过程的锂离子电池充入电量建模方法与应用
CN106772064B (zh) * 2016-11-25 2019-05-28 广东电网有限责任公司电力科学研究院 一种锂离子电池健康状态预测方法及装置

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103035961A (zh) * 2011-10-10 2013-04-10 波音公司 电池组适应性学习管理系统
CN103278777A (zh) * 2013-05-24 2013-09-04 杭州电子科技大学 一种基于动态贝叶斯网络的锂电池健康状况估计方法
CN105789716A (zh) * 2016-03-03 2016-07-20 北京交通大学 一种广义电池管理系统

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
SUN PEIKUN: "Research of State of Health Estimation Method for Electric Vehicle Lithium-ion Power Battery", BEIJING INSTITUTE OF TECHNOLOGY-SCHOOL OF MECHANICAL ENGINEERING - MASTER'S THESIS, 2016-01, XP055676847, ISSN: 1674-0246 *

Cited By (10)

* 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
CN110927591A (zh) * 2019-12-11 2020-03-27 北京理工大学 一种电池容量估计方法、计算机可读介质及车辆
CN110927591B (zh) * 2019-12-11 2020-10-02 北京理工大学 一种电池容量估计方法、计算机可读介质及车辆
CN111414709A (zh) * 2020-04-14 2020-07-14 上海交通大学 一种基于神经网络的超级电容器循环寿命和容量衰减的早期预测方法
CN112014745A (zh) * 2020-08-27 2020-12-01 洛阳理工学院 一种针对动力电池的三级故障诊断方法
CN113158345A (zh) * 2021-04-29 2021-07-23 浙江吉利控股集团有限公司 一种新能源车辆动力电池容量预测方法及系统
CN113640674A (zh) * 2021-06-30 2021-11-12 昆明理工大学 基于优化高斯过程回归的锂离子电池可用容量估计方法
CN113640674B (zh) * 2021-06-30 2022-09-06 昆明理工大学 基于优化高斯过程回归的锂离子电池可用容量估计方法
CN116205379A (zh) * 2023-05-04 2023-06-02 上海叁零肆零科技有限公司 锂亚电池存活天数预测模型建立方法、系统、设备及介质
CN116205379B (zh) * 2023-05-04 2023-07-07 上海叁零肆零科技有限公司 锂亚电池存活天数预测模型建立方法、系统、设备及介质

Also Published As

Publication number Publication date
CN111448467B (zh) 2023-04-14
CN111448467A (zh) 2020-07-24

Similar Documents

Publication Publication Date Title
WO2019018974A1 (zh) 用于对电池容量进行建模和估计的方法及系统
US20190285704A1 (en) Method and apparatus for correcting error occurring in estimation of battery life
JP5683175B2 (ja) 電気化学システムの計測不能な特性を推定する改良された方法
CN102565710B (zh) 用于估计蓄电池健康状态的方法和装置
CN102468521B (zh) 用于估计蓄电池健康状态的方法和装置
CN104577242B (zh) 一种电池组管理系统和方法
WO2019033391A1 (zh) 用于检测电池组一致性的方法、装置及系统
US20150369875A1 (en) Battery state estimating device
US10310019B2 (en) System for estimating the state of health of a battery using liquid-phase diffusivity of lithium-ion parameters
Yang et al. A voltage reconstruction model based on partial charging curve for state-of-health estimation of lithium-ion batteries
CN104237795A (zh) 通过相同电压传感器测量多个电池单元的失衡探测
CN109100655B (zh) 一种动力电池的数据处理方法和装置
KR20060116724A (ko) 하이브리드 차량용 배터리의 충전상태 추정 방법
JP2015060761A (ja) 二次電池の劣化診断システム及び劣化診断方法
KR102572652B1 (ko) 배터리의 충전상태를 추정하는 방법
US20140184233A1 (en) Method for checking deterioration of battery
CN104833917B (zh) 用于锂蓄电池中荷电状态实时估算的标称电池电阻的确定
CN106696712A (zh) 动力电池故障检测方法、系统及电动车辆
Usman et al. Universal adaptive stabilizer based optimization for Li-ion battery model parameters estimation: An experimental study
Sheikhan et al. State of charge neural computational models for high energy density batteries in electric vehicles
KR20210000206A (ko) 이상 배터리 셀 검출 방법
EP4270033A1 (en) Method and apparatus for estimating state of health of battery
KR20210000207A (ko) 내부 단락 셀 검출 방법
CN114609523A (zh) 一种电池容量的在线检测方法、电子设备及存储介质
CN112731162B (zh) 一种基于v2g使用场景下的电池健康度检测方法

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: 17918960

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: 17918960

Country of ref document: EP

Kind code of ref document: A1