EP2710391A2

EP2710391A2 - Device and method for determining a state parameter of a battery

Info

Publication number: EP2710391A2
Application number: EP12721470.8A
Authority: EP
Inventors: Simon SCHWUNK; Stefan MATTING; Matthias Vetter; Nils ARMBRUSTER
Original assignee: Fraunhofer Gesellschaft zur Forderung der Angewandten Forschung eV
Current assignee: Fraunhofer Gesellschaft zur Forderung der Angewandten Forschung eV
Priority date: 2011-05-18
Filing date: 2012-05-08
Publication date: 2014-03-26
Also published as: WO2012156233A2; WO2012156233A3; DE102011079159A1

Abstract

The invention relates to a device (100) for determining a state parameter (122) of a battery, comprising a voltage detector (110) and a processor (120). The voltage detector (110) measures a terminal voltage (U_mess) of the battery. The processor (120) calculates a plurality of current samples based on a plurality of known samples (102) of an earlier probability distribution of a state indicator of the battery and a plurality of error probability samples of an error probability distribution. Moreover, the processor (120) determines a plurality of weighting factors for the plurality of current samples based on the plurality of current samples and the measured terminal voltage (U_mess, SOC, ΔSOC). The processor (120) further calculates the state parameter (122) of the battery based on the plurality of current samples and the plurality of weighting factors.

Description

Apparatus and method for determining a state parameter of a battery

DESCRIPTION Embodiments according to the invention relate to the state determination of batteries and in particular to an apparatus and a method for determining a state parameter of a battery.

In applications with batteries, for example, it is necessary to know the exact state of charge of a battery for regulating the system and, in part, for determining the aging. Based on the state of charge, decisions are then made, such as whether a battery may continue to be charged, when a convenient time to charge or preferentially discharge the battery, or for a power or energy prediction based on a battery model.

Common applications include cell phones and laptops. Increasingly, however, applications for electric and hybrid vehicles, as well as stationary applications in the grid-connected and grid-independent area for power supply, in which more complex consumption profiles and higher demands on the reliability and quality of the state of charge determination prevail.

The problem of reliable state of charge determination is known and has been somewhat resolved for some types of batteries, such as lead or LiCoO2, against graphite. There are various approaches to this, such as the use of rule-based systems or Kalman filters. However, none of these approaches is able to determine the charge state reliably and with sufficient accuracy for battery cells with a flat characteristic curve and hysteresis.

Rule-based systems are difficult to develop because creating the rules is very time-consuming and often requires a wealth of experience.

The Kalman filter as described in Gregory L. Plett's "Sigma-point Cayman filtering for battery management Systems of LiPB-based HEV battery packs, Part 1: Introduction and State Empiri- tion", Gregory L. Plett "Sigma-point Cayman filtering HEV battery packs Part 2: Simultaneous state and parameter estimation ", Gregory L. Plett" Extended Kaiman filtering for battery management Systems of LiPB- Based HEV battery packs, Part 1. Background ", Gregory L. "Plett" Extended Kaiman filtering for battery management Systems of LiPB-based HEV battery packs, Part 2. Modeling and identification "and Gregory L. Plett" Extended Kaiman filtering for battery management Systems of LiPB-based HEV battery packs, Part 3. State and parameter estimation "is nowadays commonly used for state determination It is based on two estimation methods - usually an ampere-hour counter and a terminal voltage model - which are calculated according to their error.However, since the errors in the Kalman filter method must be normally distributed with a mean of zero, then it is not The object of the present invention is to provide a concept for determining a state parameter of a battery which enables the state parameter with high reliability even for batteries having a flat open circuit voltage characteristic and / or hysteresis of the open circuit voltage characteristic and / or high gena to determine. This object is achieved by a device according to claim 1 or a method according to claim 17.

An embodiment according to the invention provides an apparatus for determining a state parameter of a battery having a voltage detector and a processor. The voltage detector is designed to measure a terminal voltage of the battery. The processor is configured to calculate a plurality of current samples based on a plurality of known samples of a previous probability distribution of a state indicator of the battery and a plurality of error probability samples of an error probability distribution. Further, the processor is configured to determine a plurality of weighting factors for the plurality of current samples based on the plurality of current samples and the measured clamping voltage. Furthermore, the processor is configured to calculate the state parameter of the battery based on the plurality of current samples and the plurality of weighting factors.

Embodiments according to the invention are based on the core idea of using the principle of a particle filter (eg based on the probability theorem of Bayes) for the calculation of a state parameter of a battery. This is done by calculating current samples taking into account an earlier probability distribution of a state indicator of the battery and an error probability distribution and calculating weighting factors to the current samples taking into account a measured clamping voltage. Any distributions for the error can be taken into account, whereby As well as a hysteresis behavior in the filter can be mapped. Therefore, by using the described concept, the state parameter of the battery can also be determined with high reliability and / or high accuracy even in batteries with a flat open circuit voltage characteristic and / or a hysteresis of the open circuit voltage characteristic.

In some embodiments according to the invention, the processor determines a plurality of unweighted current samples based on the plurality of current samples and the plurality of weighting factors, such that the plurality of unweighted present samples correspond to a probability distribution represented by the weighting factor-weighted present samples are distributed. The plurality of unweighted present samples may then be used for a later recalculation of the condition parameter of the battery as a plurality of known samples. Thereby, the condition parameter of the battery can be e.g. be determined at regular intervals in order to be able to continuously check the status parameter.

Some embodiments according to the invention include a current detector configured to measure an amount of charge that has flowed into or out of the battery during a time interval. The measured charge amount can then be considered by the processor in calculating the plurality of current samples. Thereby, the reliability and / or the accuracy of the determination of the state parameter of the battery can be increased. Embodiments according to the invention are explained below with reference to the accompanying figures. 1 is a block diagram of a device for determining a state parameter of a battery;

FIG. 2 is a block diagram of an apparatus for determining a state parameter of a battery; FIG.

Fig. 3 is a schematic representation of samples and weighting factors;

4 shows a schematic illustration of a determination of a state parameter of a battery; 5 shows a schematic representation of a determination of a state parameter of a battery;

Fig. 6 is an impedance model of a battery;

Fig. 7 is a diagram of a possible density function for the probability based on samples (sampling);

Fig. 8 is a schematic illustration of a low variance sampling method;

Fig. 9 is a diagram of the open-circuit voltage characteristic of a

Lithium iron phosphate battery with hysteresis due to charging and discharging; and

10 is a flowchart of a method for determining a state parameter of a battery.

Hereinafter, the same reference numerals are used in part for objects and functional units having the same or similar functional properties. Furthermore, optional features of the various embodiments may be combined with each other or interchangeable.

1 shows a device 100 for determining a state parameter 122 of a battery according to an embodiment of the invention. The device 100 comprises a voltage detector 110, which is connected to a processor 120. The voltage detector 1 10 measures a terminal voltage U _{mess of} the battery. The processor 120 calculates a plurality of current samples based on a plurality of known samples 102 of a previous probability distribution of a state indicator of the battery and a plurality of error probability samples of an error probability distribution. In addition, the processor 120 determines a plurality of weighting factors for the plurality of current samples based on the plurality of current samples and the measured clamping voltage U _meSs . Further, the processor 120 calculates the state parameter 122 of the battery based on the plurality of current samples and the plurality of weighting factors.

By calculating current samples taking into account a previous probability distribution of a state indicator of the battery and an error probability distribution, and a weighting of the calculated, current one Samples by weighting factors that take into account the measured clamping voltage, the state parameter of the battery can be determined very reliable and / or with high accuracy. The computed present samples may then be the basis for the known samples 102 for recalculation of the state parameter of the battery at a later time. By iteratively repeating the state parameter determination based on the results of the previous iteration (or iterations), the accuracy can be further increased and / or continuous monitoring of the state parameter can be enabled. The state parameter 122 of the battery may be, for example, the state of charge (eg, percent of the maximum amount of charge), an age of the battery (eg, the ratio C / C "), an internal resistance of the battery, or an internal operating temperature of the battery. The battery (battery type to which the described concept may be applied) is, for example, a lead battery, nickel-based systems, sodium-based batteries, lithium to ion batteries of any kind (such as LiCo0 _2, LiNi0 _2, LiMn ₂ 0 _4, LiFeP0 _4, LiMnP0 ₄ , LiNiP0 ₄ and combinations of these on the cathode side, carbon and silicon based electrodes, titanates on the anode side, any materials as electrolytes), redox flow batteries, lithium-sulfur batteries, lithium metal batteries, lithium-air batteries, Zinc-air batteries or similar batteries. The battery is thus mainly batteries, but it can also be a disposable battery.

The samples (also called samples) represent individual points of a probability distribution and are distributed according to the represented probability distribution over the corresponding value range (eg from 0 to 1). Thus, the plurality of known samples 102 represent an earlier probability distribution of a state indicator. For example, the status indicator corresponds to the state parameter of the battery or depends deterministically on the state parameter of the battery. An earlier probability distribution is a probability distribution from an earlier (previous) calculation of the state parameter 122 of the battery or an initial probability distribution (initialization) of the state indicator. In other words, the former probability distribution of the state indicator, for example, represents an earlier probability distribution of the state parameter of the battery, or an earlier probability distribution of the accessory parameter of the battery is deterministically computable based on the earlier probability distribution of the state indicator. For example, the state indicator may represent a state of charge of the battery in a value range of 0 to 1, while the state parameter of the battery represents a state of charge of the battery in percent where 100% corresponds to a fully charged battery. However, the state parameter 122 of the battery can also depend on the state indicator in a more complex manner than only by mapping to different value ranges. The plurality of error probability samples represent an error probability distribution. For example, the error probability distribution may be represented by a set of samples from which the error probability samples are selected or the error probability distribution may be represented by a function with which the error probability samples may be calculated. The error probability distribution depends, for example, on the battery type. With the described concept, any error probability distributions can be taken into account, which is not possible with known approaches. As a result, the described concept can be used, for example, with batteries with a flat open-circuit voltage characteristic and / or hysteresis behavior with high reliability and accuracy.

The current samples are calculated based on the known samples and the error probability samples and represent, for example, a transition of the state of the battery at an earlier time (defined by the known samples) to a state of the battery (which may also correspond to the previous state of the battery if the state has not changed) at the present time.

The number of samples may vary depending on the application. By a high number of samples, a probability distribution with high accuracy can be mapped and thereby a high accuracy of the determination of the state parameter of the battery can be achieved and / or the stability of the algorithm can be improved. On the other hand, the calculation time can be shortened significantly with a smaller number of samples.

The weighting factors represent a measure of the probability that a value of a current sample will be obtained taking into account the measured clamping voltage UMess. In doing so, a weighting factor is determined for each current sample of the plurality of current samples.

From the weighting factors and the values of the current samples, the state parameter 122 of the battery can then be calculated. This can be done, for example, by calculating an average, a center of gravity, a clustering or a maintenance values of the current samples taking into account the weighting factors or samples based on the current and weighting factors (eg unweighted, current samples). The determination of the state parameter 122 of the battery may be repeated at random, predefined, or periodic intervals, for example, to allow repeated indication of the battery's state parameter 122 (eg, charge status indicator of a laptop or cell phone). For this purpose, for example, a plurality of new samples can be generated from the current samples and the weighting factors, which can be used in the next iteration of the state parameter determination as a plurality of known samples. To this end, the processor 120 may determine a plurality of unweighted current samples based on the plurality of current samples and the plurality of weighting factors such that the plurality of unweighted present samples are distributed according to a probability distribution represented by the weighting factors-weighted present samples , These unweighted, current samples may then represent the plurality of known samples 102 in the next iteration of state parameter determination. The determination of the plurality of unweighted present samples may be made, for example, based on a low-variance importance sampling method of the weighting factor-weighted present samples.

Optionally, the state parameter 122 of the battery may then also be calculated based on the plurality of unweighted present samples (based on the plurality of current samples and the plurality of weighting factors).

For each re-determination of the state parameter 122 of the battery, the voltage _{detector 110} can again measure the _terminal voltage U _{meSs of} the battery. This can then be used to recalculate weighting factors. In other words, the voltage detector 110 may again measure the clamp voltage U _mess after a predefined (or random) time interval, and the processor 120 may determine the plurality of unweighted samples determined in a previous time interval as a plurality of known samples 102 of the probability distribution of the state indicator Use the battery to recalculate a plurality of current samples and again determine a plurality of weighting factors for the plurality of current samples based on the plurality of current samples and the re-measured clamp voltage U _mess . To calculate the plurality of weighting factors, the processor 120 may first calculate a plurality of unnormalized weighting factors and a sum of the plurality of unnormalized weighting factors, and thereafter divide each unnormalized weighting factor of the plurality of unnormalized weighting factors by the calculated sum to obtain the plurality of (normalized) weighting factors ) To obtain weighting factors. This can ensure that the sum over the plurality of normalized weighting factors 1 results.

For example, a weighting factor for a current sample may be calculated based on a comparison of the measured clamp voltage U _mess and a modeled clamp voltage. The modeled clamping voltage can be based on the current sample and calculated, for example, based on an impedance model of the battery. In some embodiments according to the invention, the reliability and / or accuracy of the state parameter determination of the battery is additionally increased by means of a current measurement. 2 shows a block diagram of a device 200 for determining a state parameter 122 of a battery. The device 200 corresponds to the device shown in FIG. 1, but additionally includes a current detector 230 connected to the processor 120. The current detector 230 determines an amount of charge I that has flowed into or out of the battery during a time interval. For example, the current detector 230 may measure the current in or out of the battery and summate over the time interval or measure the charge current and multiply it over time to obtain the charge amount I. Alternatively, current detector 230 may measure charging current I and multiplied by the duration of the time interval, for example, later by processor 120. Processor 120 then calculates the plurality of current samples based on the plurality of known samples 102, the plurality of error probability samples (calculated, for example, by means of a distribution function) and the measured charge amount I. In other words, the charge amount I which has flowed into or out of the battery since the last determination of the state parameter 122 of the battery can be taken into account for calculate a value of a known sample taking into account the error probability sample the current sample. Some embodiments according to the invention relate to determining a plurality of different state parameters of a battery. For example, the state of charge and battery capacity or any other combination or number of different state parameters of a battery may be calculated. The state parameters can be calculated independently or one state parameter can depend on another state parameter. For example, the processor 120 may have another plurality of current samples (for determining another or second state parameter of the battery) based on another plurality of known samples of another prior probability distribution of another state indicator (corresponding to the other or second state parameter) of the battery calculate another plurality of error probability samples of an error probability distribution (which may correspond to the error probability distribution for the determination of the first state parameter or may be another error probability distribution). Further, the processor 120 may determine another plurality of weighting factors for the other plurality of current samples based on the other plurality of current samples and the calculated state parameter of the battery. As a result, the dependence of the other or second state parameter (eg battery capacity) on the first or already calculated state parameter (eg state of charge) of the battery can be taken into account. Thereafter, the processor 120 may calculate the other (or second) state parameter of the battery based on the other plurality of current samples and the other plurality of weighting factors.

In the following, several embodiments of the invention will be described in detail. However, the aspects described in combination with each other can also be realized independently of each other based on the concept generally described.

The underlying mathematical methodology of the new state determination solution comes from the field of "Pattern recognition and machine learning" and is used in the "Computer Vision" area for other applications.

The family of these algorithms is based on the probability theorem of Bayes, where u denotes the inputs (mostly also measured values), x the states and z the measured values: The functions of the probabilities P are mapped over samples. By means of this it is possible to map arbitrary probabilities with a sufficient number of samples, and thus also to map ambiguities of a hysteresis sufficiently well stochastically. The function P * (x _t . _\ ) Shows the probability that the battery is in a certain state of charge after the previous time step. The function p (x, | x ₍ _, w, ",) represents the probability that the state of charge of a battery changes from one instant to the other at given inputs - eg the current - in another state of charge Integrally over these two probabilities, a first estimate of the probability of the state of charge is given.The function ^ (zx,) indicates the probability that for a given

Measured value z - here the terminal voltage or the calculated open circuit voltage of a battery - a certain state of charge is given. The multiplication of the previously calculated integral with the just mentioned probability function results in a probability for the state of charge taking into account all available measured values and inputs.

An advantage of the described concept is the assumption of arbitrary distributions for the error. This allows the error to be matched with any distributions, e.g. caused by the hysteresis behavior, also be imaged in the filter. It is also possible to dynamically adjust the errors according to the history of the system.

An example of an implementation is shown below. A device for management and monitoring of batteries based on a microcontroller (processor) measures e.g. Current, voltage and optionally temperature of one or more batteries. Alternatively, the measured values can also be measured via external devices (current detector, voltage detector, temperature sensor) and transmitted to the microcontroller via bus systems.

On this microcontroller (processor) then runs the algorithm, which was previously described and determines the state of charge of the battery from the measured values.

The particulate filter is e.g. an implementation of the recursive Bayesian filter using a Monte Carlo method. The prior distribution is approximated by a fixed number of samples in the state space. The more samples fall into a certain area in the state space, the higher its probability.

For example, the algorithm can be divided into three steps. The first step corresponds to the integral p {x, in the filter equation. Starting from each sample from the prior distribution (known samples), a new sample (current sample) is drawn from the state transition distribution. _» (*, _,) Represents the estimate for the state of charge in the previous section. P (x,) on the other hand represents the probability of transferring the state of charge to a new state of charge. In practice, this transition is to be ensured by first Calculates what charge has flowed into the battery in the last time step. The measurement uncertainty (error probabil- ity sample) and its distribution are then used together with the previously determined value to determine the probability distribution of the estimate.

The second step corresponds to the part (Z, | X,) in the filter equation. Each in the first

Level drawn sample is weighted with its probability (likelihood). Subsequently, all weights (unnormalized weighting factors) are normalized so that they add up to the value 1 in the sum. The basis for calculating the probability is, for example, an impedance model. From this, one or more possible solutions can be determined, which correlate the measured value (clamping voltage) with the battery state (state parameter). Shown is such an impedance model and the possible probability distribution determined therefrom in FIGS. 6 and 7. FIG. 6 shows a possible impedance model for a battery. Here, the open circuit voltage U ₀ (open circuit voltage) as a function of the battery temperature TAMB and the state of charge SOC (state of charge) is shown. t / ₀ = / (T _AMB , SOC)

Furthermore, the internal resistance Rj is defined as a function of the battery temperature TAMB, the current Iβat and the state of charge SOC. In the illustrated model, the current I _Ba denotes t <0 for discharging the battery and Iatat> 0 for charging the battery. Correspondingly, the internal resistance of the Rj is once an internal discharge resistor RDIS and once an internal charging resistance RCHA-

I _BAr <ö: R _j = R _dis

I _BAT > 0: R _j = R _cha Depending on whether the battery is charged or discharged, a load or source is included in the impedance model.

Furthermore, FIG. 7 shows an example of a possible probability distribution of the state indicator according to which the plurality of known line samples are distributed.

In the final step, importance sampling produces an unweighted set of N samples from the weighted N samples. The unweighted quantity is generated by random sampling of the weighted quantity. A weighted sample is drawn exactly with the probability that corresponds to its weight (weighting factor). It is therefore possible to pull the same sample several times. One possible implementation of importance sampling is low-variance sampling. Low-variance sampling is an importance sampling method that pulls N unweighted from a set of N weighted samples. The starting value is a random number in the interval [0; 1N]. Starting from this starting value, samples are then selected in 1 / N large steps. Low-variance sampling has the positive property that in the case of N equally-weighted samples, each sample is drawn exactly once. The amount of samples remains the same in this case.

An example of a possible implementation of the described concept is shown below as pseudocode. for i = 1 to N do

Sc / - scan Ρ ^ Χ ^ χ, '^, ιι ^^

End for

make

End for

for i = 1 to N do

W _t τ - η ^{~ λ} - w

End for

for i = 1 to N do , w, ^w ))

End for

Matching the illustrated algorithm, Fig. 3 shows the various steps. First, a plurality of known samples x _t '(initial distribution or samples from an earlier determination of the state parameter) are shown. Based on this, considering the error probability distribution 310, the plurality of current samples 5c, 'are calculated. For each current sample 5c / a weighting factor w \ is then calculated. With the current samples 5c / and the associated weighting factors w, resampling (eg, based on a low-variance importance sampling method) may be performed to obtain a plurality of unweighted present samples x _t 'that may be used for later re-determination of the condition parameter known samples can be used. In general, in FIG. 3 the index t stands for the time t (the state parameter can be determined at several successive times) and the index i for the ith samples.

An example of an implementation of a particulate filter is described below with reference to an implementation of the algorithm.

Above all, it is shown how the sample quantity develops from charge states over time, since it already fully describes the probability distribution of the state of charge. The parameters are for example:

• N number of samples

• ARS time interval (eg in hours) between two resampling steps Since there are no measured values at the beginning (initialization), the state of charge can be assumed to be equally distributed. At the beginning, therefore, N random samples are drawn randomly from the uniform distribution on the charge state space and defined as a sample quantity (known samples). Since the state of charge state is normalized, this can be achieved by generating N random numbers in the range between 0 and 1. Next, the process of the algorithm will be described after taking readings (iteration). The measured values of the following variables are given at the same time:

U terminal voltage of the battery

/ Electricity flowing through the battery

T battery temperature

At time (eg in hours) since last measurement Depending on / and At, each charge state sample is moved based on the process model in the state of charge state. The evolution of the state of charge (SOC) is calculated using a discrete-time ampere-hour counter: soc _{n + 1} = soc _n + _c where C is the battery capacity and ε an error value (error probability sample) randomly drawn randomly from a distribution for each sample, which describes the measuring error of the current - and thus the process noise. SOC _{n +} i is a current sample, SOC _n is a known sample, I is the measured current (charging current) and At is the duration of the last time interval (since the last iteration).

For example, if at least a period of RS hours has elapsed since the last resampling step, a resampling step may be performed again, otherwise this step may be skipped. However, there is no principle minimum, but e.g. a limitation through measurement mimic and computing capacity.

Each load state sample (current sample) is assigned a weight (weighting factor) that describes how well the measurement model for that sample explains the actual measurement.

The measurement model describes the expected clamping voltage Umodell on the battery for a given state of charge SOC. Umodell is calculated e.g. using an impedance model (s.o.) as follows:

Umodell = U ₀ (SOC) + R (T, l, SOC) l where U0 models the quiescent voltage and R the Irmenwiderstand and SOC the state of charge, T the battery temperature and I the charging current. The weight for this state of charge is then f (U - U _mod ) where f is the density function of a distribution that describes the error of the measurement model and was determined heuristically or during processing (online) during the development of the algorithm Clamping voltage and U _mo deii the expected or modeled clamping voltage.

From the resulting N-weighted samples (current samples) an unweighted sample of N samples (unweighted, current samples) is drawn again, the frequency with which a weighted sample is drawn is proportional to its weight (weighting factor). This can be realized with the help of the deterministic low-variance sampling method. As a starting value, a random number X between 0 and (\ / N) is generated and then the normalized sum of the weights at the positions [X + O / N, X + 1 / N, ..., X + (N -) / N] scans. When a weight is "hit", the associated sample is selected and the method is illustrated in FIG.

The N sampled samples define the new sample set (new plurality of known samples for a subsequent iteration).

The distribution, mean and / or variance of the samples can be determined and output. Then wait until new readings are recorded and start a new iteration.

Instead of using an impedance model, it is possible to use any other battery models, including models of a purely stochastic nature. It is also possible to dispense with current integration and e.g. just to use a trend of state of charge evolution to calculate priori distribution.

In principle, any sampling method can be used to model the probability distributions.

For the internal resistance current jumps and the resulting voltage response could be evaluated. This could then be incorporated into the so-called measurement model, ie the correction be taken over. Eventually one takes state of charge and temperature ranges for which the internal resistance is corrected to approximate a function Ri = f (SOC, T). For example, in a variant without current measurement, one of the following formulas could be used for the process step: a) SOC (t) = SOC (tl)

b) SOC (t) = SOC (t-1) + SOC (t-1) -SOC (t-2)

For the correction step one could use the following:

Uklemmfk] = UQ (SP [k]) where Uklemm [k] is the measured terminal voltage and UQ (SP [k]) is the modeled terminal voltage.

The f (U-Umodell) can be used in the implementation, but does not have to. In the measurement model, it can be calculated which terminal voltage causes a state of charge and the current:

Uklemm = U0 (SOC) + Ri * I The weight of the individual samples can be determined, for example, by using a probability density function with an expected value in the height of the measured value. If the value estimated in the process model meets the expected value, then it gets the highest weight. To a state of charge value may e.g. from simple average value formation of the random samples. Other variants are conceivable (focus, clustering, expected value ...).

FIGS. 4 and 5 show further examples for calculating state parameters of a battery in the form of flowcharts for state of charge determination and capacity determination.

4 shows a possible determination of a state of charge SOC of a battery. Initially, an initial uniform distribution of samples or a plurality of unweighted samples SP of a previous state of charge determination may be used as a plurality of known samples SP (k = 0 to N samples) 410. Using previously calculated samples, these may be assigned to time interval t-1 and the current samples then assigned to time interval t. Furthermore, the current I can be measured by a current detector. With the plurality of known samples and the measured current I and a plurality of error probability samples, which are also referred to as noise, the plurality of current samples can be calculated 420 using the process model.

SP (t - HEN) ^* At) / C or too

Where SP (t) [k] is the kth current sample, SP (tl) [k] is the kth known sample, I is the measured current, the noise (also n [k]) is the error probability sample, At is the time period the time interval between time t-1 and time t, C the capacity of the battery and N the number of samples.

The plurality of current samples (SOCs Sampling Quantity) are then used 430 to determine the weighting factors based on the measurement model. For this, the voltage detector measures the clamp voltage Umess and the processor calculates a modeled clamp voltage Ukiemm- A comparison of the modeled clamp voltage and the measured Clamping voltage then gives the weighting factors. For example, a probability density function with expected value U _{mess can be} compared with the samples of the clamping voltage Uki _emm and calculated from it with what probability the estimated value (sample of the clamping voltage) is correct.

Uklemmfk J = U0 (sP [k] + or else

Uklemm [k] = U0 (samples [k]) + R ^* I Where Ukiemm [k] is the modeled clamp voltage, U ₀ (SP [k]) is an open circuit voltage, R is the internal resistance and I is the measured current. The weighting factors (SP weights, weights) may then be normalized 440. Based on the current samples and the calculated weighting factors, resampling 450 may be performed to obtain a plurality of unweighted, current samples. The larger the weighting factors (weights [k]), the more often, for example, an associated random sample (SP [k]) is drawn.

Based on the weighted, current samples obtained, a subsequent further iteration of the state parameter determination may be made.

Additionally, based on the unweighted, current samples, the state of charge of the battery may be determined. In this case, for example, a mean value formation of the sample quantity results in a current state of charge estimate (SOC estimate).

Similar to FIG. 4, FIG. 5 shows the determination of the battery capacity as the condition parameter of the battery. The battery capacity Cb _a t is calculated as a function of the determination of the state of charge by the particular state of charge is taken into account in the measurement model.

Initially, an initial uniform distribution of random samples or a plurality of unweighted random samples SP of a preceding capacity determination can be used as a plurality of known random samples SP (k = 0 to N random samples) 410. If previously calculated samples are used, they can be assigned to the time interval t-1 and the current samples are then assigned to the time interval t. SCHEN or too

sam

Where SP (f) [k] is the kth current sample, SP (tl) [k] is the kth known Sample and the noise (also n [k]) the error probability sample. In other words, only the capacity remains the same and a slight noise is added.

The plurality of current samples (sample set with capacities Cβatt) are then used to determine 530 the weighting factors based on the measurement model. For this, the processor may use the determined state of charge SOC or a difference of the particular state of charge ÄSOC (with respect to an earlier state of charge). A comparison of a modeled difference of the state of charge ÄSOC [k] and the determined state of charge ÄSOC then gives the weighting factors.

ASoC [k] = (/ * (? 2-tl)) / SP [k]

Here, ASOC [k] is the modeled difference of the state of charge, t2-tl the duration since the previous state of charge determination, SP [k] a sample of the battery capacity and I the measured current.

The weighting factors (SP weights, weights) may then be normalized 540. Based on the current samples and the calculated weighting factor, re-sampling 550 may be performed to obtain a plurality of unweighted, current samples. The larger the weighting factors (weights [k]), the more often, for example, an associated random sample (SP [k]) is drawn.

Based on the unweighted, current samples obtained, a subsequent further iteration of the state parameter determination may be made.

Additionally, based on the unweighted, current samples, the capacity of the battery could be determined. For example, averaging of the sample quantity results in a current battery capacity (SOH estimate). One aim of the described concept is, for example, to be able to carry out the state of charge determination for all battery types from the readily available measured values voltage, current and temperature on a microcontroller, the temperature being merely an optional measured value. One advantage of the method is that it can carry out a state of charge determination even with complex battery cell systems, such as LiFePO4 for graphite. A flat open-circuit voltage characteristic paired with a hysteresis of this open-circuit voltage characteristic curve - as occurs in the aforementioned example - produce ambiguities which conventional methods do not adequately resolve or depict. Fig. 9 shows an example of this Open circuit voltage characteristic of a lithium iron phosphate battery with hysteresis due to charging and discharging. The concept proposed here is able to determine both the state of charge of conventional batteries such as LiCoO 2 for graphite and novel batteries such as LiFePO 4 for graphite with their ambiguities on a microcontroller.

The described concept can in principle be used for a large number of different state determinations. For example, an age determination, an internal resistance determination or a determination of the internal battery temperature can also be made.

In general, processor 120, voltage detector 110, optional current detector 230, the optional temperature sensor, and any other optional components may be stand-alone hardware units or part of a computer, microcontroller, or digital signal processor, as well as a computer program or software product for execution on a computer Microcontroller or a digital signal processor.

10 shows a flow chart of a method 1000 for determining a state parameter of a battery according to an exemplary embodiment of the invention. The method 1000 includes measuring 1010 a terminal voltage of a battery and calculating 1020 a plurality of current samples based on a plurality of known samples of a previous probability distribution of a state indicator of the battery and a plurality of error probability samples of an error probability distribution. Further, the method 1000 includes determining 1030 a plurality of weighting factors for the plurality of current samples based on the plurality of current samples and the measured clamping voltage. Further, the method includes calculating 1040 the state parameter of the battery based on the plurality of current samples and the plurality of weighting factors.

The method 1000 may include further steps to implement one or more of the optional aspects of the proposed concept described above.

Although some aspects have been described in the context of a device, it will be understood that these aspects also constitute a description of the corresponding method, so that a block or a component of a device is also to be understood as a corresponding method step or as a feature of a method step. Similarly, aspects related to or as a methodology A description of a corresponding block or detail or feature of a corresponding device.

Depending on particular implementation requirements, embodiments of the invention may be implemented in hardware or in software. The implementation may be performed using a digital storage medium such as a floppy disk, a DVD, a CD, a ROM, a PROM, an EPROM, an EEPROM or FLASH memory, a hard disk, or other magnetic or optical memory. are stored on the electronically readable control signals that can cooperate with a programmable computer system or cooperate such that the respective method is performed. Therefore, the digital storage medium can be computer readable. Thus, some embodiments according to the invention include a data carrier having electronically readable control signals capable of interacting with a programmable computer system to perform one of the methods described herein.

In general, embodiments of the present invention may be implemented as a computer program product having a program code, wherein the program code is operable to perform one of the methods when the computer program product runs on a computer. The program code can also be stored, for example, on a machine-readable carrier.

Other embodiments include the computer program for performing any of the methods described herein, wherein the computer program is stored on a machine-readable medium.

In other words, an exemplary embodiment of the method according to the invention is thus a computer program which has program code for carrying out one of the methods described here when the computer program runs on a computer. A further embodiment of the inventive method is thus a data carrier (or a digital storage medium or a computer-readable medium) on which the computer program is recorded for carrying out one of the methods described herein. A further embodiment of the method according to the invention is thus a data stream or a sequence of signals, which represent the computer program for performing one of the methods described herein. The data stream or the sequence of signals can be configured, for example be to be transferred via a data communication connection, for example via the Internet.

Another embodiment includes a processing device, such as a computer or a programmable logic device, that is configured or adapted to perform one of the methods described herein.

Another embodiment includes a computer on which the computer program is installed to perform one of the methods described herein.

In some embodiments, a programmable logic device (eg, a field programmable gate array, an FPGA) may be used to perform some or all of the functionality of the methods described herein. In some embodiments, a field programmable gate array may cooperate with a microprocessor to perform one of the methods described herein. In general, in some embodiments, the methods are performed by any hardware device. This may be a universal hardware such as a computer processor (CPU) or hardware specific to the process, such as an ASIC.

The embodiments described above are merely illustrative of the principles of the present invention. It will be understood that modifications and variations of the arrangements and details described herein will be apparent to others of ordinary skill in the art. Therefore, it is intended that the invention be limited only by the scope of the appended claims and not by the specific details presented in the description and explanation of the embodiments herein.

Claims

claims

Apparatus (100) for determining a condition parameter (122) of a battery, comprising: a voltage detector (110) adapted to measure a terminal voltage (Umess, z _t ) of the battery; and a processor (120) configured to generate a plurality of current samples (5c /, SP (t) [k]) based on a plurality of known samples (102, SP (tl) [k]) of an earlier probability distribution a condition indicator (SOC, Cbatt) of the battery and a plurality of error probability samples (n [k]) of an error probability distribution, wherein the processor (120) is adapted to generate a plurality of weighting factors (w _t ', w [k] ) for the plurality of current samples (5c /, SP (t) [k]) based on the plurality of current samples (5c /, SP (t) [k]) and the measured clamp voltage (U _mess , z _t ) wherein the processor (120) is configured to acquire the state parameter (122) of the battery based on the plurality of current samples (5c /, SP (t) [k]) and the plurality of weighting factors (w [k]) to calculate.

The apparatus of claim 1, wherein the processor (120) is configured to generate a plurality of unweighted present samples based on the plurality of current samples (5c /, SP (t) [k]) and the plurality of weighting factors (5). , w [k]), such that the plurality of unweighted present samples correspond to a current sample (5c /, SP (t) [k]) weighted by the weighting factors (w, ', w [k]) distributed probability distribution are distributed.

The apparatus of claim 2, wherein the processor (120) is configured to calculate the state parameter (122) of the battery based on the plurality of unweighted, current samples. Apparatus according to claim 2 or 3, wherein the voltage _detector ( ₁₁₀ ) is arranged to again measure the _clamp voltage (U _meS s _> _Zt ) after a predefined time interval, the processor being arranged to determine the one in a previous time interval A plurality of unweighted samples as a plurality of known samples (102, x, '_,, SP (tl) [k]) of the probability distribution of

State indicator (SOC, C _ba tt) of the battery to recalculate a plurality of current samples (5c /, SP (t) [k]) and again one

Plurality of weighting factors (w \, w [k]) for the plurality of current samples (5c /, SP (t) [k]) based on the plurality of current samples (/, SP (t) [k]) and the again measured clamping voltage (U _me ss, z _t ) to determine.

5. The apparatus of claim 2, wherein the plurality of unweighted current samples are determined based on a low-variance importance sampling method of the weighted-mean-time (w, w [k]) weighted present samples.

6. The apparatus of claim 1, wherein the processor is configured to calculate a sum of a plurality of unnormalized weighting factors based on the plurality of current samples and the measured one Clamping voltage (U _mess , z _t ) and dividing each weighting factor (w _t ', w [k]) of the plurality of unnormalized weighting factors by the calculated sum to obtain the plurality of weighting factors (w _t ', w [k]) to obtain.

The apparatus of any one of claims 1 to 6, wherein the processor (120) is adapted to provide a weighting factor {w _t ', w [k] for a current sample

(5c /, SP (t) [k]) based on a comparison of the measured clamping voltage

(Umess, z _t ) and a modeled clamp voltage (Ukiemm, z _t ), where the modeled clamp voltage (Ukiemm, Zt) on the current sample (5c /,

SP (t) [k]).

8. The apparatus of claim 7, wherein the processor is configured to calculate the modeled clamping voltage (Ukiemm) based on

Uk, emm [k] = U ₀ (SP [k]) + R * I where Ukiemm M is the modeled clamp voltage, Uo (SP [k]) is a modeled open circuit voltage, R is the internal resistance and I is the measured current.

Apparatus according to any of claims 1 to 8, comprising a current detector (230) adapted to determine an amount of charge (I) which has flowed into or out of the battery during a time interval, or a charging current (I) wherein the processor (120) is adapted to generate the plurality of current samples (3c /, SP (t) [k]) based on the plurality of known ones

To compute samples (102, x, '_ ₁ , SP (tl) [k]), the plurality of error probability samples and the determined charge amount (I) or the determined charge current (I).

The apparatus of claim 9, wherein the processor (120) is configured to generate the plurality of current samples (SP (t) [k]) based on

SP (t) [k] = SP (tl) [k] + ((I + n [k]) * At) / C where SP (t) [k] is the kth current sample SP ( tl) [k] the kth known sample, I the measured current, the noise (also n [k]) the error probability sample, At the time interval of the time interval between time t-1 and time t and C the capacity of the battery.

Apparatus according to any one of claims 1 to 10, wherein the state parameter (122) represents a state of charge, an aging condition, an internal resistance or an internal battery temperature.

The apparatus of any of claims 1 to 11, wherein the state indicator (SOC, Cbatt) corresponds to the state parameter (122) or the state parameter (122) depends deterministically on the state indicator (SOC, Cbatt).

Apparatus according to any one of claims 1 to 12, wherein the processor (120) is adapted to calculate the condition parameter of the battery based on an average, center of gravity, accumulation or expected value of the current samples (3c /, SP (t) [k] ) and the weighting factors (w, ', w [k]) or samples based on the current samples (3c /, SP (t) [k]). 14. The apparatus of claim 1, wherein the processor is configured to iteratively recalculate the state parameter at predefined or random intervals, wherein for each iteration, a plurality of known samples (102, SP (tl ) [k]) obtained in the previous iteration

A plurality of current samples (5c /, SP (t) [k]) or a plurality of unweighted present samples, and the voltage detector (110) recalculates the terminal voltage (U _mess , Zt) of the battery for each iteration for each iteration Measure a plurality of weighting factors (w _t ', w [k]).

15. The apparatus of claim 1, further comprising a temperature sensor configured to measure a battery temperature, wherein the processor is configured to determine the plurality of weighting factors (ww [k]). ) for the plurality of current samples (5c /, SP (t) [k]) based on the plurality of current samples (5c /, SP (t) [k]), the measured clamping voltage (Umess, z _t ) and measured battery temperature (T).

16. The apparatus of claim 1, wherein the processor is configured to generate a different plurality of current samples based on another plurality of known samples of another prior probability distribution of another condition indicator (SOC, Cbatt) of the battery and calculating a different plurality of error probability samples of an error probability distribution, wherein the processor (120) is configured to generate another plurality of weighting factors (w, ', w [k]) for the other plurality of current samples based on the other plurality of the current sample and the calculated state parameter (122) of the battery, wherein the processor (120) is adapted to another state parameter (122) of the battery based on the other plurality of current samples and the other plurality of weighting factors (w, w [k]).

17. A method (1000) for determining a condition parameter of a battery, comprising the following steps:

Measuring (1010) a terminal voltage of a battery; Calculating (1020) a plurality of current samples based on a plurality of known samples of a previous probability distribution of a state indicator of the battery and a plurality of error probability samples of an error probability distribution;

Determining (1030) a plurality of weighting factors for the plurality of current samples based on the plurality of current samples and the measured clamping voltage; and

Calculating (1040) the state parameter of the battery based on the plurality of current samples and the plurality of weighting factors.

18. Computer program with a program code for carrying out the method according to claim 17, when the computer program runs on a computer or microcontroller.