• 专利附录
  • 专利说明
  • 权利要求
  • 类似技术
  • 同族专利
  • 引用文献
  • 法律状态
  • 优先权
    • 专利摘要:
    This method of estimating the state of charge of a cell of a battery comprising: calculating (116) a prediction of the state of charge SOCk of the cell using a model of state comprising a state transition matrix Fk, this prediction being marred by a state noise whose covariance is given by a matrix Qk, - the calculation (124) of a prediction ýk of a measured value yk using an observation model comprising an observability matrix Hk, this prediction ýk being marred by a measurement noise whose covariance is given by a matrix Rk, and the setting (102, 120) of values of matrices Qk and Rk using the following relations: Qk = [N0G0, k (N0)] - 1 and Rk = I, where • N0 is a predetermined integer strictly greater than one, • G0, k (N0) is given by the following relation: • I is the identity matrix.
    公开号:FR3029296A1
    申请号:FR1461615
    申请日:2014-11-28
    公开日:2016-06-03
    发明作者:Vincent Heiries；Sylvain Leirens
    申请人:Renault SAS；
    IPC主号:
    专利说明:

    [0001] The invention relates to an automatic method for estimating the state of charge of a cell of a battery as well as to a method for estimating the state of charge of a battery cell. a recording medium and a battery management system for carrying out this method.  The invention also relates to a motor vehicle comprising this battery management system.  [002] Known methods for estimating the state of charge of a cell of a battery comprise the repetition of the following steps at each successive instant k: a) the acquisition of a measured value yk of a voltage between cell terminals and a measured intensity ik of a charging and discharging current of this cell, b) calculating a prediction of the state of charge SOCk, at time k, of the cell using a state model linking the prediction SOCk to a product Fk * SOCk_i and the value yk-1 or the intensity ik-1 measured at time k-1, where SOCk_i is the state of charge of this cell at time k-1 and Fk is a state transition matrix, this SOCk prediction being tainted with a state noise whose covariance is given by a covariance matrix Qk of state noise, c) the prediction of the covariance of the error made on the prediction of the state of charge as a function, inter alia, of the Qk-1 matrix of covarian this of the state noise at time k-1 and a matrix Rk-i of covariance of the measurement noise at time k-1, d) the calculation of a prediction S / k of the value measured yk at time k using an observation model which links the measured value yk to a product Hk * SOCk, where Hk is an observability matrix, this prediction j / k being tainted with a measurement noise whose covariance is given by the covariance matrix Rk of the measurement noise, and e) the correction of the state of charge prediction SOCk as a function of the difference between the measured value yk and its prediction S [003] Such a method is in particular described in Part 3 of the following article: L.  Plett, et al.  : "Extended Kalman filtering for battery management systems of LiPBbased HEV battery packs", Journal of Power Sources, 2004, page 252-292.  Later on, this article is referred to by the abbreviation "Plett 2004".  [004] In practice, it is difficult to adjust the matrices Qk and Rk of these estimation methods.  This is particularly true for the matrix Qk because it is difficult to quantify a priori the modeling error committed.  [005] The invention aims to remedy this drawback by proposing an automatic method for estimating the state of charge of a cell in which the adjustment of the matrices Qk and Rk is simplified.  It therefore relates to a method of estimating the state of charge of a cell of a battery comprising the setting of the values of the matrices Qk and Rk using the following relations: Qk = [NoGo, k (No. )] - 1 and Rk = I, where - No is an integer strictly greater than one whose value is predetermined, - Go, k (No) is given by the following relation: N0-1 - I is the identity matrix.  [6] In the above method, the matrices Qk and Rk are set from the known matrices Fk and Hk of the state and observation models, and an integer No.  In the above method, the user only has to choose a value for the integer N instead of having to set, as in the known methods, each of the coefficients of the matrices Qk and Rk.  In the above method, the adjustment of the matrices Qk and Rk is therefore much simpler.  [7] Embodiments of this automatic estimation method may include one or more of the following features: - the setting of the values of the matrices Qk and Rk is reiterated each time one of the matrices Fk and Hk is modified ; the method comprises the construction of the matrix Hk by linearizing a nonlinear observability function which relates the measured value yk to the SOCk prediction of the state of charge at the instant k of the cell; - the nonlinear function of observability is the following function: yk = OCV (SOCk) + VD, k + R0k2 * ik, where: - OCV (SOCk) is a nonlinear known function which returns the empty voltage of the cell according to its state of charge, the empty voltage of the cell being the voltage between the terminals of this cell when these terminals are electrically isolated from any electric charge for two hours, - VD, k is a voltage across the terminals of a parallel RC circuit having a resistor RD and a capacitor CD, wherein the resistor RD and capacitance of the capacitor CD are parameters of the cell, and R0k2 is the internal resistance of the cell; the state model used in step b) for calculating the prediction of the state of charge SOCk is as follows: [1 0 0 K + 11 = T, k + 1 rock] ± v pr,.  "D, k D where 30 - VD, k + 1 and VD, k are voltages across a parallel RC circuit, - RD and CD are the resistance and capacitance values of the parallel RC circuit, - Te is the period of time separating any two instants k and k-1 immediately successive execution of step b), -Cn, k is the capacity of the cell, the capacity of the cell being its maximum capacity to store of electric energy, - wk is a centered Gaussian white state noise whose matrix of covariance is the matrix Qk.  [8] These embodiments of the estimation method further have the following advantages: - Repeating the setting of the matrices Qk and Rk each time one of the matrices Fk and Hk is modified increases the accuracy of the estimated state of charge.  [9] The invention also relates to an information recording medium including instructions for performing the automatic estimation method above when these instructions are executed by an electronic computer.  The subject of the invention is also a system for managing a battery equipped with at least one cell, this system comprising an electronic calculator programmed to reiterate the following steps at each successive instant k: a) the acquisition a measured value yk of a voltage between terminals of the cell and a measured intensity ik of a charge and discharge current of this cell, b) the calculation of a prediction of the state of charge SOCk, at instant k, of the cell using a state model connecting the prediction SOCk to a product Fk * SOCk_i and the value yk-1 or the intensity ik-1 measured at the instant k-1, where SOCk_i is the state of charge of this cell at time k-1 and Fk is a state transition matrix, this prediction of the state of charge SOCk being tainted with a state noise whose covariance is given by a state noise covariance matrix Qk, c) prediction of the covariance of the comma error ise on the prediction of SOCk charge state as a function, in particular, of the state noise covariance matrix Qk-1 at time k-1 and of a noise measurement covariance matrix Rk-i at time k-1, d) calculating a prediction j / k of the measured value yk at time k using an observation model which relates the measured value yk to a product Hk * SOCk, where Hk is an observability matrix, this j / k prediction being tainted by a measurement noise whose covariance is given by the covariance matrix Rk of the measurement noise, and e) the correction of the prediction the state of charge SOCk as a function of the difference between the measured value yk and its prediction S'ik, in which the computer is also programmed to set values of the matrices Qk and Rk by means of the following relations: Qk = [NoGo, k (No)] - 1 and Rk = I, where - No is an integer strictly greater than one whose value is predetermined, - Go, k (No) is given by the rel following: N0-1, k - I is the identity matrix.  [0011] Finally, the invention also relates to a motor vehicle comprising: at least one drive wheel, an electric motor capable of driving this driving wheel in rotation to move the motor vehicle, a battery comprising at least one cell adapted to store electrical energy and, alternately, to restore electrical energy to power the electric motor, this cell having two terminals through which it is electrically connected to the electric motor, 10 - a connected voltmeter electrically between the terminals of the cell to measure the voltage between these terminals, - an ammeter connected in series with the electric cell for measuring the intensity of the charging or discharging current of that cell, and - the claimed system of management of the battery connected to the voltmeter and to the ammeter, this management system comprising a programmable electronic calculator suitable for estimate the state of charge of the battery cell from voltmeter and ammeter measurements.  The invention will be better understood on reading the description which follows, given solely by way of nonlimiting example and with reference to the drawings in which: FIG. 1 is a partial schematic illustration of FIG. a motor vehicle equipped with an electric battery; FIG. 2 is a schematic illustration of an electric model of a cell of the battery of the vehicle of FIG. 1; FIG. 3 is a schematic illustration of an estimator arrangement used for estimating the state of charge of a cell of the vehicle battery of FIG. 1; FIGS. 4 to 9 represent equations of different models; state and observation used by the estimators of Figure 3; FIG. 10 is a flowchart of a method of estimating the state of charge of a cell using the estimators of FIG. 3; FIG. 11 is a flowchart of a method for determining the state of charge of the battery of the vehicle of FIG. 1; FIG. 12 is a flowchart of a method for scheduling refresh times of charge state estimates of different cells of a battery; FIG. 13 is a timing chart illustrating different refresh times planned using the method of FIG. 12; FIG. 14 is a schematic illustration of another estimator arrangement used to estimate the state of charge of a cell of the vehicle battery of FIG. 40; FIGS. 15 and 16 show, respectively, a state model and an observation model used by the estimators of FIG. 14; FIG. 17 is a flowchart of a method for estimating the state of charge of a cell using the estimators of FIG. 14; FIG. 18 is an illustration of another possible state model for predicting the capacity and internal resistance of a cell of a battery.  In these figures, the same references are used to designate the same elements.  In the remainder of this description, the features and functions well known to those skilled in the art are not described in detail.  In this description, the term "computing power" designates the number of operations to be performed by an electronic computer.  Thus, reducing the computing power means reducing the number of operations to achieve to achieve the same result or a result of the same nature.  [0015] Figure 1 shows a motor vehicle 2 with electrical traction plus known as the "electric vehicle".  Electric vehicles are well known and only the structural elements necessary to understand the rest of this description are presented.  The vehicle 2 comprises: - an electric motor 4, able to drive in rotation the driving wheels 6 to roll the vehicle 2 on a roadway 8, and 20 - a battery 10 which supplies the motor 4 with electrical energy.  The battery 10 comprises two terminals 12, 14 for electrical connection and several electrical cells electrically connected between these terminals 12 and 14.  Terminals 12 and 14 are connected to the electrical loads to be powered.  Here, they are therefore in particular connected to the electric motor 4.  To simplify FIG. 1, only four electric cells 18 to 21 are shown.  Typically, these electric cells are grouped into several stages and these stages are connected in series between the terminals 12 and 14.  Here, only two floors are represented.  The first stage comprises the cells 18 and 19, and the second stage comprises the cells 20 and 21.  Each stage has several branches connected in parallel.  Each branch of a stage comprises an electric cell or several electric cells in series.  Here, the first stage has two branches, and each branch has a single electric cell.  The second stage is structurally identical to the first stage in the example shown in FIG.  Here, all the cells of the battery 10 are structurally identical to the manufacturing tolerances.  Therefore, only the cell 18 is now described in more detail.  The cell 18 comprises two electrical connection terminals 30, 32 which electrically connects it to the other cells and terminals 12 and 14 of the battery 10.  The cell 18 is also mechanically fixed without any degree of freedom to the other cells of the battery 10 to form what is often called a "pack" of cells.  Cell 18 is capable of storing electrical energy when not in use.  This stored electrical energy is then used to power the motor 4, which discharges the cell 18.  In alternation, the cell 18 can also receive electrical energy which charges it.  The complete discharge of a cell followed by its complete recharge constitutes what is called a charge / discharge cycle, which is simply called a "cycle of a cell".  The cell 18 is a known type of cell, for example, it is a LiPB cell (Lithium-ion Polymer Battery) or other.  The cell 18 is characterized by an initial nominal capacity Cn'n ', an initial internal resistance RO' ', a current intensity Imax current, a maximum voltage Umax, a minimum voltage Um, n and an OCV function (SOCk ).  The capacity Cn'n 'is the initial capacity of cell 18.  The capacity of a cell represents the maximum amount of electrical energy that can be stored in that cell.  This capacity is expressed in Ah.  As the cell 18 ages, i.e. as the number of charge and discharge cycles increases, the capacity of the cell decreases.  At the moment k, the nominal capacity of the cell 18 is noted Cn, k thereafter.  The initial internal resistance RO '' is the value of the internal resistance of the cell 18 before it begins to age.  The internal resistance of a cell is a physical quantity found in most electrical models of an electric cell.  As the cell ages, typically, the internal resistance increases.  At time k, the internal resistance of cell 18 is denoted ROk.  Imax is the maximum intensity of the current that can be delivered by the cell 18 without damaging it.  Umax is the maximum voltage that can be present permanently between the terminals 30 and 32 of the cell without damaging it.  The voltage Uni is the minimum voltage between the terminals 30 and 32 when the cell 18 is completely discharged.  Subsequently, it is considered that lm. Umax, Um, n are constant physical quantities that do not change over time.  OCV (SOCk) is a predetermined function which returns the empty voltage of the cell 18 according to its state of charge SOCk.  The no-load voltage is the measurable voltage between terminals 30 and 32 after the cell 18 has been electrically isolated from any electrical load for two hours.  The state of charge at the instant k of the cell 18 is SOCk noted.  The state of charge represents the filling rate of the cell 18.  It is equal to 100% when the amount of electrical energy stored in cell 18 is equal to its capacity Cn, k.  It is equal to 0% when the amount of energy stored in the cell 18 is zero, that is to say that it can no longer be extracted from the electrical energy of the cell 18 to supply an electric charge.  The parameters Cnn, ROn, Imax, U. , Umin and the OCV function (SOCk) are known parameters of the cell.  For example, they are given by the constructor of the cell or are determined experimentally from measurements made on this cell.  The battery 10 also comprises for each cell: a voltmeter which measures the voltage between the terminals of this cell, and an ammeter which measures the intensity of the charging or discharging current of this cell.  To simplify Figure 1, only a voltmeter 34 and an ammeter 36 of the cell 18 have been shown.  Unlike the various parameters of the cell 18 introduced previously, the SOCk charge state of the cell 18 is not measurable.  It must therefore be estimated.  For this purpose, the vehicle 2 comprises a battery management system 40, better known by the acronym BMS ("Battery Management System").  This system 40 has the particular function of determining the state of charge of the battery 10 as well as the state of health of this battery.  To determine this state of charge and this state of health, the system 40 is capable of estimating the state of charge and the state of health of each cell of the battery 10.  The state of health of a cell represents the state of progress of the aging of this cell.  Here, the state of health of a cell, at instant k, is denoted SOHk.  Subsequently, it is measured by the ratio Cn, k / Cn'n '.  To calculate the state of health of a cell, the system 40 is also able to estimate the capacity Cn, k of this cell at the current time k.  To achieve these different estimates, the system 40 is electrically connected to each voltmeter and each ammeter of the battery 10 to acquire the measurements of the voltage and the intensity of the current between the terminals of each cell.  Here, the system 40 includes a memory 42 and a programmable electronic calculator 44, able to execute instructions stored in the memory 42.  For this purpose, the memory 42 comprises the instructions necessary for carrying out the methods of FIGS. 10 to 12 and / or of FIG. 17.  This memory 42 also includes the initial values of the various parameters necessary for the execution of these methods.  The structure of the system 40 is therefore identical or similar to those of known battery management systems and is not described in more detail.  FIG. 2 represents an electric model 50 of the cell 18.  This model is known as the "Thevenin model of the first order" or "Lumped parameter model".  It comprises successively connected in series from terminal 32 to terminal 30: a generator 52 of the empty voltage OCV (SOCk), a parallel circuit 54, and an internal resistor 56 called by the next, at the instant k, "internal resistance ROk".  The circuit 54 includes a capacitor CD capacitor connected in parallel with a resistance RD value.  Subsequently, it is considered that these two parameters CD and RD of the model 50 are known and constant over time.  The voltage at the instant k across the circuit 54 is denoted VD, k.  The value at the instant k of the voltage between the terminals 30 and 32 of the cell 18 is denoted yk and the intensity, at the same instant, of the charging or discharging current of the cell 18 is denoted ik.  FIG. 3 represents a first embodiment of an estimator arrangement implemented in the system 40 for estimating the state of charge and the state of health of the cell 18.  Each estimator is implemented in the form of an estimation algorithm executed by the calculator.  Thus, we will not talk about the sequence of "execution of an estimator" as "execution of an estimation algorithm".  In this first embodiment, the system 40 comprises an estimator 60 of the charge state SOCk and the voltage VD, k from the measured value yk of the voltage and the measured intensity ik.  The estimator 60 is here implemented in the form of a Kalman filter.  It therefore uses a state model 62 (FIG. 4) and an observation model 64 (FIG. 5).  In these figures 4 and 5, the equations of these models are represented using the previously defined notations.  Notations R0k2 and Cn, k3 represent, respectively, the capacity and internal resistance of cell 18, respectively, at times k2 and k3.  These moments k2 and k3 are defined later.  Moreover, in the model 62, xk designates the state vector [SOCk, vp, k] r at time k.  In this description, the symbol "T" designates the transposed mathematical operation.  The multiplication operation is represented by the operator ".  " or " * ".  Subsequently, it is considered that the origin of the time corresponds to the zero value of the instant k.  Under these conditions, the current instant k is equal to kTe, where T, is the sampling period of the measurements of the ammeters and voltmeters of the battery 10.  Thus, Te is the period of time which separates any two successive instants k and k-1 of acquisition of the voltage and the intensity of the current by the system 40.  The period Te is typically a constant between 0.1 s and 10 s.  Here, the period Te is equal to 1 s to plus or minus 20%.  For example, Te is equal to one second.  In the model 62, wk is a state noise vector.  Here, the noise wk is a Gaussian white noise centered.  This noise represents the uncertainty on the model used.  The covariance matrix, at the instant k, of the noise wk is denoted Qk It is defined by the following relation: Qk = E (Wk * WkT), where E (. . .  ) is the expected expectation function.  Model 62 is also written as Xk + 1 = FkXk Bkik Wk, where - Fk is the state transition matrix at time k, - Bk is the control vector at time k.  The model 62 allows in particular to predict the state of charge SOCk + i at time k + 1 from the state of charge preceding SOCk.  The model 64 can be used to predict the value yk of the voltage at instant k from the state of charge SOCk, the voltage VD, k and the measured intensity ik.  In this model, vk is a centered Gaussian white measurement noise.  The covariance matrix of the noise vk at time k is noted Rk thereafter.  In the particular case described here, this matrix Rk is a single-column, single-row matrix.  It is defined by the relation Rk = E (Vk * VkT).  This noise vk is independent of the noise wk and the initial state vector xo.  It will be noted that the model 64 is non-linear, since the OCV function (SOCk) is generally nonlinear.  Because of this, the estimator 60 implements the extended version of the Kalman filter, better known by the acronym EKF (Extended Kalman Filter).  In this extended version, we come back to a linear observation model of the form yk = Hkxk + ROk2. ik + vk by linearizing the model 64 near the vector xk.  Typically, the model 64 is developed in Taylor series in the vicinity of the vector xk.  Then we neglect the contributions of derivatives from the second order.  Here, the matrix H k is therefore equal to the first derivative of the OCV function in the vicinity of the state of charge SOCk.  This linearization of the model 64 is typically performed for each new value of the state of charge SOC k.  The estimator 60 needs to know the capacitance C n, k3 and the internal resistance R0k2 to be able to estimate the SOCk charge state. i.  The capacity and internal resistance of cell 18 vary as it ages.  To account for this aging, the capacity and the internal resistance of the cell 18 are estimated, respectively, at times k3 and k2.  Here, an estimator 66 estimates the internal resistance R0k2 from the measured value yk2, the measured intensity ik2 and the state of charge SOCk2.  Another estimator 68 estimates the capacitance C n, k3 from the intensity i k3 and the state of charge SOCk3.  The internal resistance and the capacity of the cell 18 vary more slowly than its state of charge.  Thus, to limit the computing power required to estimate the state of charge of the cell without degrading the accuracy of this estimate, the estimators 66 and 68 are executed less frequently than the estimator 60.  Subsequently, the execution times of the estimators 66 and 68 are noted, respectively, k2 and k3 to distinguish them from the instants k.  Here, the set of instants k2 and the set of instants k3 are subsets of the set of instants k.  Between two successive instants k2 and k2-1 and between two successive instants k3 and k3-1, several periods Te and several instants k elapse.  These estimators 66 and 68 are also implemented each in the form of a Kalman filter.  The estimator 66 uses a state model 70 (FIG. 6) and an observation model 72 (FIG. 7).  In these models, the noises 2 w -, k2 and v 2, k2 are Gaussian white noise centered.  The covariances of the noise W - 2, k2 and V2, k2 are noted, respectively, 0 -2, k2 and R2, k2 thereafter.  The observation model 72 makes it possible to predict the value of a directly measurable physical quantity uk2.  The physical quantity uk2 is here the sum of the last N measured values yk.  It is defined by the following relation: 3 0 2 9 2 9 6 10 L.  = [0045] N is an integer strictly greater than one which is counted as will be described later.  In the relation above and in the model 72, the instant k is equal to the instant k2.  The model 72 takes into account not only the charge state SOCk, the voltage VD, k and the measured intensity ik at the instant k = k2 but also the previous N estimates of the estimator 60 and N previous measured intensities, between moments k2 and k2-1.  Taking into account the measurements and intermediate estimates between the instants k2 and k2-1, makes it possible to increase the precision of the estimation of the internal resistance R0k2.  The estimator 68 uses a state model 74 (FIG. 8) and an observation model 76.  In models 74 and 76, the noises W3, k3 and V3, k3 are Gaussian white noise centered.  The covariances of the noises W3, k3 and V3, k3 are noted respectively Q3, k3 and R3, k3 thereafter.  It will be noted that the model 76 is a linear model so that a simple Kalman filter can be used for the estimator 68 instead of an extended Kalman filter.  The observation model 76 makes it possible to estimate a directly measurable physical quantity zk3.  The physical quantity zk3 is here the sum of the N last measured intensities ik.  It is defined by the following relation: m = k- [0049] In the relation above and in the model 76, the instant k is equal to the instant k3.  This physical quantity zk3 takes into account not only the measured intensity ik-1 at the instant k-1 preceding the instant k3 but also of N preceding intensities measured between the instants k3 and k3-1.  Here, N is an integer strictly greater than one which is counted as will be described later.  It is not necessarily equal to the N introduced in the model 72.  Taking into account measurements and intermediate estimates between instants k3 and k3-1, makes it possible to increase the precision of the estimation of the capacity Cn, k3.  The operation of the estimators 60, 66 and 68 will now be described using the method of FIG. 10 and in the particular case of estimating the state of charge of the cell 18.  The method begins with a phase 100 of adjusting the different covariance matrices necessary to execute the estimators 60,66 and 68.  More specifically, during an operation 102, the covariance matrices Q k and Rk of the estimator 60 are automatically adjusted using the following relations: Qk = [NoGo, k (No)] - 1 and Rk = I, where - No is a predetermined integer strictly greater than 1, - I is the identity matrix, and - Go, k (No) is defined by the following relation: i = 0 [0052] No is generally chosen during the design of the system 40 then frozen once and for all.  Generally, No is less than 100.  For example, No is between 5 and 15.  Here No is chosen equal to 10.  The fact of using the above relations greatly simplifies the setting of matrices Qo and Ro and the setting of matrices Qk and Rk as we shall see later.  Indeed, the only parameter to choose is the value of the integer No.  In an operation 104, the covariances Q2,0 and R2,0 are also set.  For example, Q2,0 is chosen equal to [(13 * ROini) / (3 * NCeoms)] 2, where - [3 is a chosen constant greater than or equal to 0.3 or 0.5 and, preferably, greater than at 0.8 and generally less than three, - Nced is the expected number of cycles of charge and discharge of cell 18 before it reaches its end of life, and - Ns is the number of times the internal resistance is estimated per charge and discharge cycle of the cell18.  The constant [3 represents, expressed as a percentage divided by 100, the difference between the value of the initial internal resistance RO '' and its end-of-life value.  Typically, [3 is set by the user or measured experimentally.  Nced is a number of cycles that can be measured experimentally or obtained from the data of the constructor of cell 18.  Ns is set by the state of charge estimation method implemented by the computer 44.  In this embodiment, as will be seen below, the internal resistance is estimated only once per cycle.  Therefore, Ns is taken equal to 1.  By way of illustration, the covariance R2.0 is chosen equal to (2cniUmax / 300) 2, where Cm 30 is the maximum error of the voltmeter 34 expressed as a percentage.  Subsequently, the covariances 0 -2, k2 and R2, k2 are considered to be constant and equal, respectively, to Q2, o and R2,0.  In an operation 106, the covariances Q3,0 and R3,0 are set.  For example, the covariance Q3, o is taken equal to [y * Cn '' / (3 * Nced * Ns)] 2, where y is, expressed as a percentage divided by 100, the difference between the capacity Cn'n and the capacity of the cell 18 at the end of its life.  y is a constant selected by the user between 0.05 and 0.8 and preferably between 0.05 and 0.3.  Here, y = 0.2.  The covariance R3.0 is for example chosen to be equal to [2 * E, m * Imax / 300] 2, where En is the maximum error of the ammeter 36 expressed as a percentage.  Subsequently, the covariances Q3, k3 and R3, k3 are considered constant and taken equal, respectively, to Q3,0 and R3,0.  Once the covariance matrices are set, the estimate of the state of charge of the cell 18 can begin.  During a phase 110, at each instant k, the voltmeter 34 and the ammeter 36 measure, respectively, the value yK and the intensity ik and these measurements are immediately acquired by the system 40 and stored in the memory 42.  Phase 110 is repeated every moment k.  In parallel, the estimator 60 executes a phase 114 for estimating the state of charge at the instant k of the cell 18.  For this, during a step 116, the estimator 60 calculates a prediction SOCkik-i and a prediction VD, k / ki of, respectively, the state of charge of the cell 18 and the voltage VD at circuit terminals 54 at time k.  In the notations used here, the index k / k-1 indicates that this prediction is made taking into account only the measurements made between the instants 0 and k-1.  This is called prediction a priori.  The index kik indicates that the prediction at the instant k takes into account all the measurements made between the instants 0 and k.  This is called a posteriori prediction.  The predictions SOCkik_i and VD, k / k-1 are calculated using the model 62, the measured intensity ik-1 and the capacity Cn, k3.  It will be noted that in the model 62, the state transition matrix F k-i is constant regardless of k and therefore does not need to be re-evaluated at each instant k.  In a step 117, the estimator 60 also calculates the prediction P k / k-i of an estimation error covariance matrix on the state vector xk.  Typically, this is done using the following relationship: Pk / k-1 = Fk-1Pk-1 / k-1Fk-1T + Qk-1 [0066] These different matrices Fk-1, Pk-1 / k -1 and Qk-1 have already been defined previously.  Then, in a step 118, the estimator 60 constructs the matrix H k by linearizing the model 64 around the predictions SOCkik_i and VD, k / k-1.  In a step 120, the covariance matrices Q k and Rk are automatically updated.  Here, for this, the step 120 is identical to the operation 102 taking into account this time the matrix H k constructed during the step 118.  After this, in a step 122, the estimator 60 corrects the predictions SOCkik-i and VD, k / k-1 as a function of a difference between the measured value yk and a predicted value j / k. from the model 64.  This gap is known as "innovation".  This step 122 typically comprises: an operation 124 for calculating the prediction j / k, then an operation 126 for correcting the predictions SOCkik-i and VD, k / k-1 and the matrix Pk / k-1 for obtain the corrected predictions SÔCk / k, VD, k / k and Pk / k.  In the operation 124, the prediction S / k is calculated using the model 64 in which the value of the state of charge is taken equal to SOCkik_i and the value of the voltage VD, k is taken equal to VD, k / ki.  The difference between the measured value yk and its prediction j / k is noted Ek thereafter.  Many methods exist to correct the estimates a priori SOCkik-i and VD, k / k-1 from the innovation E k.  For example, in step 126, these estimates are corrected using the Kalman Kk gain.  The gain Kk is given by the following relation Kk = Pkik-1Erk (HkPkik-1Wk + Rk) -1.  Then, the predictions a priori are corrected with the following relation: xkik = KkEk.  The matrix Pk / ki is corrected using the following relation: Pk / k = Pk / k-1 -KkFl kPk / k-1 - Steps 116 to 122 are repeated at each instant k where new estimates of the state of charge of the cell 18 must be made.  During each new iteration, the state vector xk_i is initialized with the values obtained during the previous iteration of the phase 114 for the cell 18.  In parallel, during a step 130, the computer 44 compares each new measurement of the intensity ik to a threshold SH, predetermined current.  As long as the measured intensity does not exceed this threshold SH 'the execution of the estimator 66 and inhibited.  On the other hand, as soon as the measured intensity ik exceeds this threshold SH 'then the estimator 66 is immediately executed.  The threshold SH, is generally greater than I max / 2 and advantageously greater than 0.81.  or 0.91max.  The estimator 66 executes a phase 140 for estimating the internal resistance R0k2 at the instant k2.  Here, the instant k2 is equal to the instant k where the intensity ik crosses the threshold SH ,.  For this, during a step 142, the estimator 66 calculates the a priori prediction RO of the internal resistance from the model 70.  Then, in a step 144, the estimator 66 calculates the prediction P - 2, k2 / k2-1 of the covariance matrix of the estimation error on the internal resistance.  For example, this prediction is calculated using the following relationship: P - 2, k2 / k2-1 = P2, k2 - 1 / k2-1 ± Q2.0.  It will be noted that here, the model 72 is a linear function of the state variable.  It is therefore not necessary to linearize it in the vicinity of the prediction R k2 / k2-1 to obtain the matrix H2, k2.  Here, this matrix H2, k2 is equal to -N.  In a step 148, the estimator 66 corrects the prediction RO k2 / k2-1 as a function of the difference between the measured physical quantity uk2 and a prediction Ûk2 of this same physical quantity.  Here, N is a predetermined constant selected strictly at one and, preferably, greater than 10 or 30.  The size uk2 is acquired by the estimator 66 as the values yk are measured and acquired.  More specifically, during an operation 150, the computer 44 acquires the measured quantity uk2 and calculates the prediction Û k2.  The acquisition of the size uk2 is performed by summing the last N measures of the measured value yk.  The prediction Ûk2 is calculated using the model 72.  In this model 72, the value RO k2 is taken equal to the value ROk2 / k2_1 previously calculated.  Then, in an operation 152, the estimator 66 corrects the prediction R0 k2 / k2-1 as a function of the innovation E k2.  The innovation E k2 is equal to the difference between the measured quantity uk2 and the predicted quantity Ûk2.  For example, during the operation 152, the same method as that implemented during the operation 126 is used.  Thus, this operation 152 is not described here in more detail.  Then, the new estimate R 0 k2 / k2 is used in subsequent executions of the estimator 60 in place of the previous estimate R0k2-1 / k2-1.  Triggering the execution of the estimator 66 only when the measured intensity ik is high makes it possible to increase the accuracy of the estimation of the internal resistance while at the same time decreasing the computing power necessary to implement this method.  Indeed, the accuracy of the ammeter measurement is higher when the intensity ik is higher.  Also in parallel with the phases 110 and 114, the method comprises a step 160 in which at each instant k, the estimate SOCk is compared to a predetermined high threshold SHs'.  If the SOCk estimate falls below this threshold SH0, then the process proceeds immediately by steps 162 and 164.  In the opposite case, step 160 is repeated at the next instant k.  Typically, the threshold S Flsoc is between 90% and 100%.  In step 162, the computer 44 starts by initializing a counter to zero and then increments it by 1 to each new measurement of the intensity ik since the beginning of this step.  Moreover, at each instant k, the measured intensity ik and the estimate SOCk generated at the same time are recorded, associated with this instant k, in a database.  In parallel with step 162, during step 164, the calculator 44 compares each new estimate SOCk with a predetermined threshold SLs'.  The threshold SLs' is for example between 0% and 10%.  As long as the estimate SOCk remains above this threshold SLs', step 162 is repeated at the instant k following.  In the opposite case, as soon as the estimate SOCk for the cell 18 falls below this threshold SLs' then, the computer 44 immediately triggers the execution of the estimator 68 and stops incrementing the counter.  Thus, as long as this threshold SLs' is not crossed, the execution of the estimator 68 is inhibited.  The estimator 68 estimates, during a phase 166, capacity C n, k3 at time k3.  The instant k3 is therefore equal to the instant k when the execution of the estimator 68 is triggered.  As for phase 140, since the estimator 68 is not executed at each instant k, the time k3-1 does not correspond to the instant k-1.  In contrast, instants k3 and k3-1 are separated by a time interval greater than or equal to NT, where N is the number counted in step 162.  The parameters of the Kalman filter of the estimator 68 are initialized with the 40 previous values of these parameters obtained at the end of the previous iteration at the instant k3-1 of the phase 166.  Phase 166 comprises: calculating, during a step 170, the prediction Cn, k3 / k3-1 with the aid of the model 74, the calculation, during a step 172, of the prediction P - 3, k3 / k3-1 of the covariance matrix of the estimation error of the capacity, and - the correction, during a step 174, of the predictions Cn, k3 / k3-1 and P - 3, k3 / k3-1- [0089] In steps 172 and 174, the observability matrix 113, k3 is equal to [(SOCk - SOCk_N)] * 3600 / (NTe).  Here is the number of times k elapsed between the moment when the estimated state of charge has dropped below the threshold SHs 'and the moment when the estimated state of charge has dropped below the threshold SLs'.  The value N is equal to the value counted during step 162.  Step 174 comprises an operation 176 for acquiring the measured physical quantity zk3 and calculating the prediction 2k3 of the quantity zk3.  The acquisition of the magnitude zk3 here consists in calculating the sum of the N last intensities measured between the instants k-1 and k-N.  The 2k3 prediction is obtained from the model 76.  Then, during an operation 178, the estimator 68 corrects the prediction C n, k3 / k3-1 as a function of the difference between the measured quantity zk3 and the predicted quantity k3 to obtain the posterior estimate of the capacity C n, k3 / k3.  This correction is for example carried out as described in step 126.  Then, the capacitance Cn, k3 / k3 is transmitted to the estimator 60 which uses it to estimate the state of charge of the cell 18 at the following instants.  Triggering the execution of the estimator 68 only after the cell 18 has largely discharged increases the accuracy of the estimation while at the same time decreasing the computing power necessary to this process.  At the end of phase 166, during a step 180, the calculator calculates the state of health SOHk3 at time k3 using the following formula: SOHk3 = Cn, k3 / Cn ' not'.  FIG. 11 represents a method for determining the state of charge of the battery 10.  At time k, the state of charge of the battery 10 is determined from the state of charge of each of the cells of this battery.  For example, this is done as follows.  During a step 190, the computer 44 determines the state of charge of each stage of the battery by adding the state of charge of each cell of this stage.  Then, during a step 192, the state of charge of the battery is taken equal to the smallest of the states of charge of a determined stage during step 190.  As illustrated by the method of FIG. 11, the determination of the state of charge of the battery at each instant k necessitates having an estimate of the state of charge for each of the cells. instant k.  A first solution therefore consists in executing in parallel, for each of the cells, the estimation method of FIG. 10 by executing phase 114 at each instant.  However, in order to limit the computing power required without degrading the accuracy of the state of charge determined for the battery, it is also possible to plan the execution of the estimations of the states of charge of the cells as described with reference to the method of Figure 12.  The method of FIG. 12 is described in the simplified case in which only three priority levels are used, respectively called high level, medium level and low priority level.  Moreover, it is assumed that the state of charge of a cell whose priority level is high must be estimated at each instant k and therefore at a frequency fe.  The state of charge of a cell whose priority level is average must only be estimated with a frequency three times lower and therefore at a frequency fe / 3.  Finally, the state of charge of the cells of low priority level must be estimated at a frequency ten times lower is therefore at a frequency fe / 10.  In this example, for the high and medium priority levels, there are a limited number of known places in advance.  In other words, the number of cells assigned to the high priority level is limited to a predetermined maximum number in advance.  The same is true for the number of cells assigned to the medium priority level.  In order to plan the times at which the estimates of the state of charge of each of the cells must be refreshed, the computer first assigns, in a step 198, a priority level to each cell.  Step 198 begins with an operation 200 in which the system 40 acquires the measured value yk of the voltage between the terminals of each of the cells.  Then, during an operation 202, if the measured value yk is above a high threshold SHy or, conversely, below a low threshold SLy, then the computer 44 assigns to this cell the high priority as long as there is enough room in this level.  The threshold SHy is greater than or equal to 0.9 * Umax and, preferably, greater than 0.95 * Umax.  The threshold SLy is greater than or equal to Umni, and less than 1.1 * Um ,, 'or 1.05 * Umni.  It is important to frequently refresh the estimation of the state of charge of cells whose voltage is close to Umax or on the contrary close to Umm.  Indeed, an error in the estimation of the state of charge of a cell in such a situation can lead to a degradation of the electrical and mechanical properties of this cell.  Then, for the other cells, during an operation 204, the computer 44 calculates the voltage difference between the current measured value yk and a previous value yk_x, where X is a predetermined integer greater than or equal to one and generally less than 5 or 10.  Here, X = 1.  In an operation 205, the computer 44 identifies twin cells.  Cells are considered to be "binocular" if, at the same instant k, they have the same voltage difference and the same measured value yk.  For this, during step 205, the computer 44 compares the voltage difference and the measured value yk for a cell at the voltage differences and at the measured values yk for the other cells at the same instant in order to identify, among these other cells, the cell or cells of this cell.  The identifier of this cell and the identifiers of the cell or cells identified as being its binoculars are then grouped together in a set that is stored in the memory 42.  The above comparison is for example made for each of the cells of the battery 10 whose identifier has not already been incorporated into one of the registered sets of twin cells.  Subsequently, a priority level is assigned to only one of the cells in each set of twin cells.
    [0002] Thus, step 206 and subsequent steps 208 and 210 are performed only for cells that do not have a twin cell and for a single cell in each set of twin cells. During an operation 206, the computer classifies the cells in decreasing order of absolute value of the difference calculated during the operation 204. Then, it assigns to the first cells of this classification the remaining places associated with the levels. of high priority. Then, it assigns the remaining places associated with an average priority level to the next cells in that ranking. Finally, it assigns to the last cells of this classification the low priority level. Once a priority level has been assigned to each cell, during a step 208, the computer 44 schedules the refresh times of the estimations of the charge states of the cells according to their priority level. Step 208 is performed so as to respect the refresh frequency of the estimates associated with each of the priority levels. For this purpose, for example, the computer 44 first reserves the times at which the estimates of the cells of high priority level must be refreshed. Then, it reserves the moments at which the estimates of the state of charge of the cells of average priority level must be refreshed taking into account the times of refreshment already reserved. Finally, it does the same with cells assigned a low priority level. [00106] To illustrate this, it is assumed that a high priority level was assigned to the cell 18, that a mean priority level was assigned to the cells 19 and 20 and that a low priority level was assigned. in cell 21. In addition, it is required that during a period T ,, the computer executes at most twice the phase 114 of the method of FIG. 10. The result obtained with these hypotheses is shown in FIG. the instants k to k + 11 have been represented on the abscissa. Above each of these instants k, two boxes symbolize the fact that the computer 44 can, at each instant k, perform twice the phase 114 of the method of FIG. 10. In each of these boxes, the number is indicated. of the cell for which phase 114 is executed. When there is no number in this box, this means that the process of Figure 10 is not executed and therefore the saved computing power can be used for other purposes such as, for example, the execution of the estimators. and 68. [00107] Finally, during a step 210, for each cell assigned a priority level, the computer 44 executes the phase 114 at the scheduled time for this cell. Outside these planned times, the computer inhibits the complete execution of the phase 114 for this cell. Similarly, the execution of phase 114 for twin cells to which no priority level has been assigned is also inhibited. In parallel, during a step 212, for each twin cell to which no priority level has been assigned, the estimation of the state of charge of this cell is taken equal to the last estimate calculated during of step 210 for a twin cell of this cell. Thus, phase 114 is executed for only one of the twin cells. This reduces the computing power required to determine the state of charge of battery without degrading the accuracy of this determination. Optionally, in parallel with step 210, at each instant k, the computer 44 also executes a step 214 of predicting the state of charge for each of the cells that are not processed during step 210 to this moment k. Step 214 consists in executing only the prediction step 116, without performing the correction step 122, for all the cells for which, at the same time, the complete estimation phase 114 is not executed. Indeed, the prediction step 116 is much less computationally intensive than the step 122 and can therefore be executed, for example, at each instant k. Thus, when step 214 is implemented, there is at each moment k of a new estimation of the state of charge for each of the cells of the battery. Steps 198 and 208 are repeated at regular intervals to update the priority level assigned to each of these cells and therefore the refresh rates of the estimate of the state of charge of these cells. This method for scheduling refresh times of the estimates of the charge states of the cells makes it possible to limit the computing power required without degrading the accuracy of the state of charge determined for the battery. Indeed, the method of Figure 12 exploits the fact that the cells whose voltage differences are low are cells that discharge or load little and therefore whose state of charge does not change quickly. It is therefore possible to estimate the state of charge of these cells at a lower frequency without degrading the accuracy of the state of charge determined for the battery. During the execution of the methods of FIGS. 10 and 11, whenever the state of charge SOCk of a cell at a given instant must be used for a calculation, the state of charge SOCk is taken. equal to the last estimated state of charge or predicted for this cell. In other words, it is considered that the state of charge remains constant between two successive instants where it is estimated or predicted. It will also be noted that each time the calculator 44 executes the estimation phase 114 for a cell, it retrieves the information necessary for this execution from the values obtained at the end of the previous execution of this phase. for the same cell. This is the case, for example, for state variables. Note, however, that the previous execution time is not necessarily the time k-1 but may be the time k-3 or k-10 depending on the priority level assigned to this cell. [00113] Many other embodiments of the method for estimating the state of charge of a cell are possible. For example, Fig. 14 shows another arrangement of estimators. This other arrangement is identical to that of FIG. 3 except that the estimators 66 and 68 are replaced by a single estimator 230. The estimator 230 simultaneously estimates the capacity and the internal resistance of the cell 18. The estimator 230 is executed less It is frequently the estimator 60. Here, we write k4 the execution times of the estimator 230 and hence C, -, k4 and ROka the capacity and the estimated internal resistance at time k4. The set of instants k4 is a subset of instants k. The estimator 230 estimates at the same time the capacity Cn, k4 and the internal resistance ROka. This estimator 230 implements a Kalman filter that uses a state model 232 (FIG. 15) and an observation model 234 (FIG. 16). The operation of this estimator 230 will now be described with reference to the method of FIG. 17 and in the particular case of the cell 18. This method of FIG. 17 is identical to the preceding one of FIG. 10 except that the steps 130 to 174 are replaced by steps 240, 242, 244 and a phase 246 for estimating the capacitance and the internal resistance. In step 240, the computer 44 compares at each instant k the measured value yk to a high threshold SHy2. Typically, this threshold SHy2 is greater than or equal to 0.8 * Umax or 0.9 * Umax. Steps 242 and 244 are executed only if measured value yk falls below this threshold SHy2. In step 242, the computer 44 begins by initializing a counter to zero and then increments this counter by 1 at each new time k. Moreover, at each of these instants k, the measured intensity ik, the value yk, the state of charge SOCk and the estimated voltage VD, k are recorded, associated at this instant k, in a database. [00118] In parallel with step 242, during step 244, the computer 44 compares, at each instant k, the new measured value yk with a low voltage threshold SLy2. This threshold SLy2 is less than or equal to 1.2 * Um ,, or 1.1 * Um, n and greater than or equal to Unn. As soon as the measured value yk goes below the threshold SLy2, the incrementation of the counter during step 242 is stopped and the execution of the estimator 230 is triggered. On the other hand, as long as the measured value yk remains above this threshold SLy2, the execution of the estimator 230 is inhibited. The estimator 230 executes the phase 246. As previously, it is noted that the instants ka and k4-1 are separated by a time interval greater than or equal to NTe, where N is the value of the counter incremented during the Step 242. The operation of the estimator 230 is deduced from the operation described above for the estimators 66 and 68. It will therefore not be described here in more detail. Other electrical models and therefore other state models can be used to estimate the state of charge of the cell 18. For example, in a simplified variant 40, the circuit 54 is omitted. Conversely, a more complex electric model may contain several parallel RC circuits electrically connected in series with each other. The state model of the cell 18 must then be modified accordingly to correspond to this new electrical model of the cell. However, everything described above applies without difficulty with such a modified state model. For examples of modified state models, the reader may refer to application WO2006057468. The RD and CD parameters of the model 50 can also be estimated instead of being considered as predetermined constant parameters. For this purpose, these two parameters RD and CD are for example introduced into the state vector xk which then becomes [SOCk, VD, k, RD, k and CD, k] T. For example, the state model is modified to incorporate the two following RD equations, k + 1 = RD, k and CD, k + 1 = CD, k. The state vector xk can also be supplemented by the temperature of the cell so as to estimate this temperature at the same time as the state of charge of this cell. The cell may also be equipped with additional sensors such as a temperature sensor. In this case, the observation model is modified to take into account these additional measured physical quantities. The reader can refer to application WO2006057468 for examples of modified observation models. [00125] Other possible electrical models for modeling the electric cell are also presented in Part 2 of Plett 2004, Chapter 3.3. The continuous automatic adjustment of the covariance matrices Rk and Qk can be performed differently. For example, the so-called "Covariance Matching" method described in the following article can be applied: Mehra, RK: "On the identification of variances and adaptive Kalman Filtering", Automatic Control, IEEE Transaction on, Volume 15, No. 2 , pages 175-184, April 1970. This method applies after an initial setting of the matrices Ro and Qo, for example, as described during the operation 102. [00127] In another variant, the matrices Qo, Ro, Qk and Rk are not set as described with reference to operations 102 and 120. For example, these matrices are set using a conventional method. In a simplified case, they are constant. For example, the matrix Ro is then set from the data provided by the sensor manufacturer or from tests performed on these sensors and the matrix Qo by successive tests. The step 122 for correcting the prediction can be performed differently. For example, in a preferred method, the correction of the prediction of the state of charge and the voltage VD, k, is achieved by minimizing a quadratic cost function J which has two terms: a term related to prediction error of the measured value, and - another term related to the estimation error of the state vector. This method is described in detail in Chapter 10.5.2 of the following book: Y. Bar-40 Shalom, et al .: "Estimation With Applications to Tracking and Navigation, Theory Algorithms and Software", Wiley Inter-science, 2001. [00129 In another variant, the estimator 60 is not implemented in the form of a Kalman filter. For example, the state of charge is estimated by modeling its evolution over time in the form of an Infinite Impulse Response (IRN) filter whose coefficients are estimated by the Recursive Least Square (RLS) method. Other state models can be used to estimate the internal resistance and the capacity of the cell. For example, the model 232 may be replaced by a model 250, shown in FIG. 18. In the model 250, a, 13 and y are constants whose values are obtained from data of the cell constructor or measured experimentally. . Typically: - a is equal to 1 to plus or minus 30% or 10%, -13 is also 1 to plus or minus 30% or 10%, and - y is typically between 0.1 and 0, 5. For example, y is 0.2 to plus or minus 30% or 10%. In the model 250, Nck is equal to the number of charge / discharge cycles of the cell performed before the instant k. This number of cycles is for example measured by counting the number of times the state of charge of the cell falls below the high threshold SHs 'and then below the low threshold SLs'. wad, k is a centered Gaussian white noise. y is the difference, expressed as a percentage divided by 100, between the initial capacity Cn '' of the cell and its end-of-life capacity. This model takes into account the fact that: - the internal resistance increases as the cell ages, and - the capacity of the cell decreases as the cell ages. [00132] Similarly, the state model 70 can be replaced by the following state model: R0k2 + 1 = (a + 13Nck2 / N 1RnW CEOL, - - k2 + - 2, k2, where the different symbols of this model have already been previously defined [00133] The state model 74 can be replaced by the following state model: Cn, k3 + 1 = (1 -YNCk3 / NCEOL) Cn, k3 ± V3, k3 where the different symbols of this model have already been previously defined [00134] According to the observation model used by the estimator 68, the magnitude zk3 can be calculated differently, For example, the quantity zk3 is equal to the sum of the last Ns intensities measured between instants k and k-N + 1. In this case, when N is equal to 1, zk3 = 1k3 [00135] What has been previously described for the initialization of the covariance matrices Qk and Rk can be also be applied for the initialization of the covariance matrices of the estimators 68 and 230. [00136] As a variant, the estimator 68 is not im For example, the capacity is estimated by modeling its evolution over time in the form of an Infinite Impulse Response (IIR) filter whose 40 coefficients are estimated by the RLS method ( Recursive Least Square). The methods of FIGS. 10 and 17 can be simplified by taking N equal to a predetermined constant. In this case, N is not counted and steps 160, 162, 240 and 242 may be omitted. For example, N is chosen equal to one or, conversely, strictly greater than 1 or 5 or 10. In another variant, at each instant k between the instants k3 and k3-1, only the step 170 of calculation of a prediction Cn, k is executed but the correction step 174 of this prediction is not executed. Thus, we obtain a new prediction of the capacity of the cell at each of these instants k while limiting the computing power required. Similarly, at each instant k between the instants k4 and k4-1, only the step of calculating the predictions of the capacitance and the internal resistance is executed without performing the step of correcting these predictions. Thus, in these variants, the capacity of the cell is predicted at each instant k but this prediction is corrected only at times k3 or k4. The algorithm for estimating this capacity is therefore only partially executed between the instants k3 and k3-1 or k4 and k4-1 and completely executed only at the instant k3 or k4. At each instant k between the instants k3 and k3-1 or between the instants k4 and k4-15, the capacity can be estimated by executing a first algorithm, then at the instant k3 or k4, the capacity is estimated by executing a second algorithm different from the first algorithm and requiring a larger computing power. The first and second algorithms do not necessarily correspond, as previously described, respectively to step 170 and phase 166 or 246 of a Kalman filter. It can also be two completely different estimation algorithms. The step 166 or 246 for estimating the capacity of the cell can be triggered in response to the crossing of a threshold on the state of charge, as described with reference to FIG. 10, or in response to the crossing. A threshold on the voltage as described with reference to FIG. 17. These steps 166 and 246 can also be triggered in response to crossing a current amount threshold. For this, from the moment when the voltage or the state of charge of the cell has fallen below a predetermined high threshold, at each instant k, the calculator 44 calculates the quantity QCk of the current delivered with the aid of the following relationship: QCk = QCk_i + kTe. As soon as QCk exceeds a high threshold SHQ, then phase 166 or 246 is executed. On the other hand, as long as the quantity QCk remains greater than the threshold SHQ, the execution of the phases 166 or 246 is inhibited. Alternatively, the quantity QCk can also be calculated on a sliding window containing the last N instants k, where N is a predetermined constant. In another embodiment, the triggering of the estimates of the capacitance and / or the internal resistance in response to crossing a threshold is omitted. For example, these estimates are triggered at regular intervals. This regular interval is equal to Te if the available computing power is sufficient to estimate this capacity and this internal resistance at each instant k. Many other embodiments of the method of FIG. 12 are possible. For example, operation 205 may be omitted. In this case, no twin cell is identified and step 212 is also omitted. [00143] The operation 202 can be performed differently. For example, only one of the high and low thresholds is used. Operation 202 can also be omitted. [00144] The number of priority levels can be any and greater than at least two or three. Other methods for assigning a priority level to the cells are possible. For example, the priority level of a cell can be calculated using a formula that relates its priority level to its voltage difference and voltage. In the latter case, the comparison operations are omitted. The method described for associating refresh times with the cells according to their priority levels is only one example. Any other known method of scheduling tasks according to the priority level of these tasks can be adapted to the case described here of the scheduling of the refresh times of the estimations of the states of charge of the cells. The planning of the refresh instants of the estimation of the state of charge of each of the cells described with reference to FIG. 12 can be omitted.
    [0003] For example, this will be the case if the computing power needed to estimate the state of charge of each of the cells at each instant k is available. [00147] In a variant, the computer 44 comprises a plurality of programmable sub-computers capable of executing each and in parallel the estimation method of FIG. 10 or 17 for respective cells. [00148] The state of health of a cell can also be calculated using the following relation: SOHK = ROK / RO "[00149] The battery 10 can be replaced by any type of battery, as by For example, a lead-acid battery, a super-capacitor, or a fuel cell, in which case the state model and / or the observation model of the estimator 60 may eventually be adapted to take into account the technology. of the battery [00150] What has been described above also applies to the hybrid vehicle, that is to say to the vehicle whose drive wheel drive is at the same time or, alternatively, provided by an engine The vehicle 2 may also be a truck, a motorcycle or a tricycle and generally any vehicle capable of moving by driving driving wheels with the aid of a powered electric motor. for example, it may be a freight elevator [00151] The battery 10 can be recharged via an electrical outlet that allows the electrical connection to a power distribution network. The battery 10 can also be recharged by an internal combustion engine.
    权利要求:
    Claims (8)
    [0001]
    REVENDICATIONS1. A method for estimating the state of charge of a cell of a battery, said method comprising the repetition of the following steps at each successive instant k: a) the acquisition (110) of a measured value yk of a voltage between the terminals of the cell and a measured intensity ik of a charging and discharging current of this cell, b) calculating (116) a prediction of the state of charge SOCk, at the moment k, of the cell by means of a state model connecting the prediction SOCk to a product Fk * SOCk_i and to the value yk-1 or to the intensity ik-i measured at time k-1, where SOCk_i is the state of charge of this cell at time k-1 and Fk is a state transition matrix, this state of charge prediction SOCk being tainted with a state noise whose covariance is given by a state noise covariance matrix Qk, c) the prediction (117) of the covariance of the error made on the prediction of the SOCk charge state in function, nota of the state noise covariance matrix Qk-1 at time k-1 and a noise covariance matrix Rk-i at time k-1, d) the calculation (124 ) of a prediction S / k of the measured value yk at time k using an observation model which relates the measured value yk to a product Hk * SOCk, where Hk is a matrix of observability, this S / k prediction being tainted by a measurement noise whose covariance is given by the covariance matrix Rk of the measurement noise, and e) the correction (126) of the prediction of the state of charge SOCk in a function of the difference between the measured value yk and its prediction j / k, characterized in that this method comprises adjusting (102, 120) the values of the matrices Qk and Rk by means of the following relations: Qk = [ NoGo, k (No)] - 1 and Rk = I, where - No is an integer strictly greater than one whose value is predetermined, - Go, k (No) is given by the following relation: - I is the identity matrix . 30
    [0002]
    2. Method according to claim 1, wherein the setting (120) of the values of the matrices Qk and Rk is reiterated each time one of the matrices F k and Hk is modified.
    [0003]
    3. A method according to any of the preceding claims, wherein the method comprises constructing (118) the Hk matrix by linearizing a non-linear observability function which relates the measured value yk to the SOCk prediction of the state of charge at the instant k of the cell.
    [0004]
    4. The method according to claim 3, wherein the nonlinear function of observability is the following function: yk = OCV (SOCk) + VD, k + R0k2 * ik, where: OCV (SOCk) is a known function not linear, which returns the empty voltage of the cell according to its state of charge, the empty voltage of the cell being the voltage between the terminals of this cell when these terminals are electrically isolated from any electric charge for two hours, 10 - VD, k is a voltage across a parallel RC circuit having a resistor RD and a capacitor CD, where the resistance RD and capacitance of the capacitor CD are parameters of the cell, and - R0k2 is the internal resistance of the cell . 15
    [0005]
    A method according to any one of the preceding claims, wherein the state model used in step b) for computing SOCk state of charge prediction is as follows: SO (k + ii [1 ^ +1 = [v 0 1 D, k + 1 CD where 20 - VD, k + i and VD, k are voltages across a parallel RC circuit, - RD and CD are the values of the resistance and the capacity of the parallel RC circuit, - Te is the period of time which separates any two instants k and k-1 immediately successive execution of step b), - Cn, k is the capacity of the cell, the capacity of the Since cell is its maximum capacity for storing electrical energy, wk is a centered white Gaussian state noise whose covariance matrix is the Qk matrix.
    [0006]
    6. Information recording medium (42), characterized in that it comprises instructions for executing an estimation method according to any one of the preceding claims, when these instructions are executed by an electronic calculator.
    [0007]
    7. A system (40) for managing a battery equipped with at least one cell, this system comprising an electronic calculator (44) programmed to repeat successive steps at each successive instant k: a) the acquisition of a value measured yk of a voltage between terminals of the cell and a measured intensity ik of a charging and discharging current of this cell, b) calculating a prediction of the state of charge SOCk, to the instant k, of the cell by means of a state model connecting the prediction SOCk to a product Fk * SOCk_i and the value yk-1 or the intensity ik-1 measured at the instant k -1, where SOCk_i is the state of charge of this cell at time k-1 and Fk is a state transition matrix, this SOCk state of charge prediction being tainted by a noise of state whose covariance is given by a state noise covariance matrix Qk, c) the prediction of the covariance of the error committed on the prediction of the state of charge SOCk based, in particular, on the QK-1 matrix of covariance of the state noise at time k-1 and a matrix Rk-i of covariance of the measurement noise at time k-1, 15 d) computing a prediction S / k of the measured value yk at time k using an observation model which relates the measured value yk to a product Hk * SOCk, where Hk is a matrix of observability, this prediction j / k being tainted with a measurement noise whose covariance is given by the covariance matrix Rk of the measurement noise, and e) the correction of the prediction of the charge state SOCk according to the difference between the measured value yk and its prediction j / k, characterized in that the computer (44) is also programmed to set values of the matrices Qk and Rk using the following relations: Qk = [NoGo , k (N0)] - 1 and Rk = I, where - No is an integer strictly greater than one whose value is predetermined, 25 - Go, k (No) is given by the following relation: N o-1 - I is the identity matrix.
    [0008]
    8. Motor vehicle comprising: - at least one driving wheel (6), - an electric motor (4) capable of driving this driving wheel in rotation to move the motor vehicle, - a battery (10) comprising at least one cell ( 18-21) adapted to store electrical energy and, alternately, to restore electrical energy to supply the electric motor, said cell having two terminals (30, 32) through which it is electrically connected with an electric motor; - a voltmeter (34) electrically connected between the terminals of the cell for measuring the voltage between these terminals; - an ammeter (36) connected in series with the electric cell for measuring the intensity of the charging or charging current; discharge of this cell, and - a system (40) for managing the battery connected to the voltmeter and the ammeter, this management system comprising a programmable electronic calculator (44) able to estimate the state charging the battery cell from the voltmeter and ammeter measurements, characterized in that the battery management system (40) is in accordance with claim 7.
    类似技术:
    公开号 | 公开日 | 专利标题
    EP3224634B1|2019-05-01|Automatic method of estimating the charge state of a battery cell
    EP3224637B1|2018-10-31|Automatic method for estimating the capacity of a cell of a battery
    EP3224635B1|2019-05-01|Automatic method for estimating the state of charge of a cell of a battery
    EP3224636B1|2019-05-01|Automatic method for estimating the state of charge of a cell of a battery
    EP3224633B1|2019-04-24|Automatic method for determining the state of charge of a battery
    FR3009093A1|2015-01-30|ESTIMATING THE AGING CONDITION OF AN ELECTRIC BATTERY
    FR2884928A1|2006-10-27|Neural network device for detecting e.g. state of charge of lead storage battery, has neural network controller providing, by applying neutral network calculation on input parameters, output signal indicating internal state of battery
    FR2963109A1|2012-01-27|METHOD FOR DETERMINING A PARAMETER OF AT LEAST ONE BATTERY ACCUMULATOR
    EP3079940A1|2016-10-19|Assessing the quantity of energy in a motor vehicle battery
    EP2850445B1|2019-07-10|System and corresponding method for estimating the charge status of a battery
    FR2944358A1|2010-10-15|Battery's e.g. lithium-ion battery, health state estimating device for electrical traction of e.g. hybrid vehicle, has samplers to calculate difference between symptomatic and effective parameters, and comparator to indicate health state
    EP2775313A1|2014-09-10|Method of determining the residual capacity of a battery
    EP3667345B1|2021-04-21|Method for determining the state of health of the cells of a battery
    EP3465240A1|2019-04-10|Method for estimating the state of health of a battery
    EP3221710B1|2018-10-10|Method for characterising an accumulator
    WO2021170345A1|2021-09-02|Method for estimating the state of health of a battery
    WO2013060688A1|2013-05-02|Method and system for determining the state of charge of a basic cell and a battery
    FR3098922A1|2021-01-22|Method for determining the state of charge of the cells of a battery
    FR3090116A1|2020-06-19|Method for determining the state of health of the cells of a battery
    FR3101429A1|2021-04-02|Method for determining the state of health of a lithium-ion battery.
    FR3045218A1|2017-06-16|DETERMINATION OF PARAMETERS OF A DYNAMIC MODEL FOR AN ELECTROCHEMICAL BATTERY CELL
    同族专利:
    公开号 | 公开日
    JP6712594B2|2020-06-24|
    JP2017539060A|2017-12-28|
    US10267861B2|2019-04-23|
    CN107110915A|2017-08-29|
    CN107110915B|2020-04-21|
    US20170356964A1|2017-12-14|
    FR3029296B1|2016-12-30|
    EP3224634A1|2017-10-04|
    EP3224634B1|2019-05-01|
    WO2016083753A1|2016-06-02|
    KR20170088424A|2017-08-01|
    引用文献:
    公开号 | 申请日 | 公开日 | 申请人 | 专利标题
    US6441586B1|2001-03-23|2002-08-27|General Motors Corporation|State of charge prediction method and apparatus for a battery|
    US20050046388A1|2003-08-28|2005-03-03|Tate Edward D.|Simple optimal estimator for PbA state of charge|FR3097482A1|2019-06-24|2020-12-25|Renault S.A.S|Method of exchanging electrical energy between a motor vehicle battery and an electrical network|US5570087A|1994-02-18|1996-10-29|Lemelson; Jerome H.|Motor vehicle performance monitor and method|
    US5714866A|1994-09-08|1998-02-03|National Semiconductor Corporation|Method and apparatus for fast battery charging using neural network fuzzy logic based control|
    US6329793B1|1996-07-29|2001-12-11|Midtronics, Inc.|Method and apparatus for charging a battery|
    BR0308950A|2002-03-29|2007-01-30|Estco Battery Man Inc|fuel cell health management system|
    JP3714321B2|2002-11-25|2005-11-09|日産自動車株式会社|Secondary battery charge rate estimation device|
    US7458433B2|2003-07-31|2008-12-02|Arvinmeritor Technology, Llc|Electric motor and gear drive assembly for driving a vehicle wheel|
    US7321220B2|2003-11-20|2008-01-22|Lg Chem, Ltd.|Method for calculating power capability of battery packs using advanced cell model predictive techniques|
    CA2588856C|2004-11-29|2012-11-13|Lg Chem, Ltd.|Method and system for battery state and parameter estimation|
    KR100804698B1|2006-06-26|2008-02-18|삼성에스디아이 주식회사|The method of assuming the state of charge of the battery, battery management system using the method and the driving method of the battery management system using the method|
    FR2971854B1|2011-02-21|2013-10-25|Renault Sa|DEVICE FOR ESTIMATING THE AGING OF A MOTOR VEHICLE POWER BATTERY AND METHOD THEREFOR.|
    FR2975190B1|2011-05-13|2013-05-17|Valeo Equip Electr Moteur|METHOD FOR ESTIMATING THE OPERATING STATE OF A BATTERY FOR AN AUTOMATIC STOP / RELEASE SYSTEM OF THE THERMAL MOTOR OF A VEHICLE, SENSOR AND BATTERY MANAGEMENT SYSTEM ADAPTED|
    US8880253B2|2011-06-28|2014-11-04|Ford Global Technologies, Llc|Nonlinear adaptive observation approach to battery state of charge estimation|
    FR2990516B1|2012-05-11|2015-10-16|Renault Sas|ESTIMATING THE CHARGE OF A BATTERY|
    JP6111275B2|2013-02-05|2017-04-05|日立オートモティブシステムズ株式会社|Battery control device|
    FR3029299B1|2014-11-28|2016-12-09|Renault Sa|AUTOMATIC METHOD FOR DETERMINING THE CHARGING STATE OF A BATTERY|
    FR3029298B1|2014-11-28|2016-12-30|Renault Sa|AUTOMATIC METHOD OF ESTIMATING THE CHARGING STATE OF A CELL OF A BATTERY|FR3029297B1|2014-11-28|2016-12-30|Renault Sa|AUTOMATIC METHOD OF ESTIMATING THE CHARGING STATE OF A CELL OF A BATTERY|
    FR3084170B1|2018-07-17|2020-07-17|Zodiac Aero Electric|METHOD FOR DETERMINING A STATE NOISE COVARIANCE MATRIX FOR THE ADJUSTMENT OF A BATTERY CHARGE OBSERVER AND DEVICE THEREOF|
    WO2021014482A1|2019-07-19|2021-01-28|三菱電機株式会社|Storage battery diagnosis device and charge/discharge pattern setting method|
    WO2021168836A1|2020-02-28|2021-09-02|华为技术有限公司|Anomaly detection method and device|
    DE102020112566A1|2020-05-08|2021-11-11|Bayerische Motoren Werke Aktiengesellschaft|Method for operating an electrically powered vehicle taking into account component limits for electrochemical storage systems|
    KR102356369B1|2020-09-08|2022-02-07|주식회사 에스티|Battery pack monitoring apparatus for monitoring the charge status of the battery pack which is used as the energy source for an agricultural electrical vehicle and the operating method tehreof|
    法律状态:
    2015-11-19| PLFP| Fee payment|Year of fee payment: 2 |
    2016-06-03| PLSC| Search report ready|Effective date: 20160603 |
    2016-11-18| PLFP| Fee payment|Year of fee payment: 3 |
    2017-11-21| PLFP| Fee payment|Year of fee payment: 4 |
    2019-11-20| PLFP| Fee payment|Year of fee payment: 6 |
    2021-08-06| ST| Notification of lapse|Effective date: 20210705 |
    优先权:
    申请号 | 申请日 | 专利标题
    FR1461615A|FR3029296B1|2014-11-28|2014-11-28|AUTOMATIC METHOD OF ESTIMATING THE CHARGING STATE OF A CELL OF A BATTERY|FR1461615A| FR3029296B1|2014-11-28|2014-11-28|AUTOMATIC METHOD OF ESTIMATING THE CHARGING STATE OF A CELL OF A BATTERY|
    JP2017528441A| JP6712594B2|2014-11-28|2015-11-26|How to automatically estimate the state of charge of a battery cell|
    US15/531,547| US10267861B2|2014-11-28|2015-11-26|Automatic method for estimating the state of charge of a battery cell|
    KR1020177017629A| KR20170088424A|2014-11-28|2015-11-26|Automatic method of estimating the charge state of a battery cell|
    CN201580069145.XA| CN107110915B|2014-11-28|2015-11-26|Method for automatic estimation of the state of charge of the battery cells of a battery pack|
    EP15808755.1A| EP3224634B1|2014-11-28|2015-11-26|Automatic method of estimating the charge state of a battery cell|
    PCT/FR2015/053239| WO2016083753A1|2014-11-28|2015-11-26|Automatic method of estimating the charge state of a battery cell|
    [返回顶部]