DETERMININGTHE STATE OFHEALTH OFA BATTERY
The present invention relates to a method of determining the state of health of a battery, particularly but not exclusively that of a lead-acid battery, and to related apparatus.
Safety critical vehicle systems such as power assisted steering and braking are increasingly being powered electrically. There is a requirement to establish that the condition of the battery allows the operation of these systems with an adequate safety margin. Many methods are known for estimating the performance of an isolated battery but most of these methods are not applicable to a battery in use on a vehicle.
The most common means for providing electrical power on automotive vehicles is the lead-acid battery. Such a battery comprises a number of cells in series, each cell comprising a lead cathode plate and a lead (IV) oxide (PbO2) anode plate, the plates being immersed in a sulphuric acid
(H2SO4) solution. A potential difference is thereby generated at the output terminals of the battery. Discharging of the battery causes insoluble lead (II) sulphate (PbSO4) to be generated on both the plates, whilst charging the battery causes the reverse reaction to occur.
Such a battery generally comprises sufficient cells to generate a nominal 12 volts.
Methods utilising a mathematical model of a battery are known for estimating battery performance while a vehicle is in use, as described in US patent 6 362 598, for example. The models use measurements of battery voltage, current and temperature to calculate changes in the state- of-charge as the battery charges and discharges. Such models can also be
run in isolation to predict the battery voltage, using as input a defined load that corresponds to some safety critical operation.
Mathematical models of batteries require parameters that are specific for each battery. Examples include battery capacity and ohmic resistance. However, the parameters may change as a battery ages. Capacity may reduce if active material becomes electrically isolated (for example, if the lead or lead oxide sheds from the plates) or if lead sulphate forms that cannot be reformed to metallic lead and lead oxide on charging. Similarly, the ohmic resistance may increase as the grills in the positive plates corrode. If mathematical models are to continue to predict the discharge performance accurately, the relevant parameters in the model have to be adapted as a battery deteriorates with age and use.
At the heart of any mathematical model of a battery are algorithms that describe the non-linear way that the voltage drops with discharge current. Four physical mechanisms have been suggested, each one of which makes a contribution to the total voltage drop: Transfer (or Activation) , Concentration, Crystallization and Resistance (or Ohmic). The relative size of the contribution of each of these four parameters to the total voltage drop will depend on the current, the state-of-charge and the battery design.
In the "Crystallization" mechanism the current is proportional to the difference between the actual lead ion concentration in solution and the concentration of a saturated lead sulphate solution whereas the voltage is given by the well-known Nernst equation involving the ratio of the actual and saturated lead ion concentrations. Ohm's law governs the "Resistance" mechanism.
The method described in US Patent number 6 362 598 determines the state-of-charge and the discharge performance of a battery by using a mathematical model of the well-known equivalent circuit type. Measurements are made of the battery voltage, current and temperature as the battery undergoes a series of charges and discharges. The model then simulates the pattern of charges and discharges using current and temperature measurements as inputs and outputting estimates of the battery voltage. The voltage estimates are compared to the voltage measurements and the errors summed over the charge and discharge period. The model parameters are varied, the simulation repeated and a new sum of errors is calculated. Thus the values of model parameters that minimise the sum of errors can be found and the model can adapt to battery ageing and deterioration.
However, the method described has two serious limitations. Firstly, if the vehicle battery has been exchanged for one with different characteristics the model does not adapt the characteristic of the battery model before the vehicle is used. The method described relies on measuring whatever pattern of charging and discharging results from the vehicle's use. Secondly, the equivalent circuit mathematical model described does not involve summing the voltage drops corresponding to different physical processes thus restricting the accuracy of performance predictions even after the adaptation process.
According to a first aspect of the invention, there is provided a method of determining the state of health of a battery, comprising the steps of discharging the battery through a known test load across the battery, measuring the voltage across the battery when the battery is discharging through the test load and from the measured voltage determining at least one parameter relating to the state of health of the battery.
The method provides a convenient way of determining the state of health of a battery. Providing a known test load across the battery and measuring the voltage is an improvement over the prior art method of determining the current passing from the battery through a sense resistor in that the voltage across the battery in use does not vary over so wide a range. In use of a typical nominal 12V vehicle lead-acid cell, the voltage may only vary over a few volts, whereas a current sensor would need to accurately measure from milliamps to possibly several hundred amps.
Herein, the state of health of a battery preferably means the capability of the battery to store charge and to supply current given a certain state of charge.
Preferably, by known we mean that the resistance of the test load is known; in an especially preferred embodiment the resistance is used in the determination of the, or each, parameter. This may be by direct inclusion in a calculation, or by controlling how the test load is connected to the battery.
The method is advantageous as it removes the need for a current sensor that can measure the range of currents expected from a typical vehicle battery. The method may, however, include the step of determining the current when the battery is discharging through the test load, typically by measuring the voltage across a sense resistor; as the discharge is controlled then the current is likely to be with controllable bounds.
In the preferred embodiment, the battery is a lead-acid battery. However, it may be any other suitable battery or cell. Typically, the battery is installed in a vehicle, such as a car, and provides electrical power therefor. Preferably, the method comprises the step of testing the battery whilst it is installed in a vehicle.
The method may be carried out when the ignition of a vehicle powered by the battery is off. It may be carried out only when the ignition is off. This is convenient, as relatively little power is required from the battery when the ignition is off - there is less effect from other circuits. Alternatively, the method may be carried out when the ignition is on; the method may include the step of determining the effect of the superposition of discharging the battery through the test load on the effect of other circuits on the measured circuits.
The step of discharging the battery through the test load may comprise discharging the battery through an electric apparatus installed on the vehicle. The electric apparatus may be any suitable piece of electric equipment, and preferably one that is commonly found on a vehicle, such as the element of a heated windscreen. Advantageously, this uses the pre¬ existing vehicle electrics to test the battery; it is particularly efficient where the method is carried with the ignition off as the user of a vehicle will most likely not be using the vehicle at such a time - they will not notice such equipment as the heated rear windscreen being used, especially if it is only momentarily.
Preferably, the method includes using an engine management computer of a vehicle to determine the, or each, parameter. This is convenient as it allows use of pre-existing resources and allows the method to be carried out automatically by the vehicle.
The voltage across the battery may be measured and sampled periodically during discharge through the test load. The method may include the step of varying the test load whilst the battery is discharging through the test load.
Advantageously, the or each parameter relating to the state of health of the battery is used to model of the state of charge of the battery. From this model, the current-providing capability of the battery may be predicted. In the case where the battery is a vehicle battery, the method may model the state of charge when an ignition of the vehicle is on. It may model the state of charge whilst the ignition is off, typically at a reduced rate.
The parameters may model physical characteristics of the battery, such as, in a lead-acid battery, the concentration of sulphuric acid, the amount of lead or lead oxide shed from the plates, the ohmic resistance of the battery or the level of irreversible crystallisation of the lead sulphate on the battery plates.
The period during which the battery is discharged through the test load may be very small compared with a period when the battery is not being discharged through the test load. The ratio of times may be less, preferably much less, than Is in 10s. The discharge is preferably of the form of a pulse, which may be modulated by a pulse waveform and may be pulse width modulated. The measurement of voltage to use in the determination of the, or each, parameter may also occur when the battery is not being discharged through the test load. The discharging of the battery through the test load may be periodically repeated.
Using a short discharge interval may allow the calculation of relatively complex mathematical functions in the period between discharges. It also reduces the amount of battery charge used in determining the state of health - the discharge of the battery through the test load is very preferably only partial.
The method may iteratively model the, or each, parameter; the method may calculate an estimated voltage given the, or each, calculated parameter and the test load, and from this calculate a revised value of the, or each, parameter. The iterative step may be repeated.
The method may include the step of determining the time constant of the voltage response to a discharge pulse. This may be used in the calculation of the internal resistance of the pulse. Advantageously, this measurement is taken at the beginning of the pulse, such that the internal resistance is measured as that due to ohmic effects, rather than other effects which take some time to develop. In an especially preferred embodiment, the time constant is calculated over a period of the first 10 milliseconds or less of the pulse.
Alternatively, the method may include the step of determining the internal resistance of the battery by determining the current flowing through the battery (typically by measuring the voltage across a sense resistor) and measuring the instantaneous voltage drop as the test load is applied to the battery.
The method may include the step of determining an initial state of charge of the battery. This is advantageous when using a model of the state of charge of the battery that converges over time to the estimate of the state of charge.
According to a second aspect of the invention, there is provided a computer program which, when running on a suitable processor, causes that processor to carry out the method of the first aspect of the invention.
According to a third aspect of the invention, there is provided a data carrier carrying the program of the second aspect of the invention.
According to a fourth aspect of the invention, there is provided an apparatus for monitoring the state of health of a battery, comprising:
a test load, arranged to be connected to the battery and of known resistance;
a voltage sensor, arranged to sense the voltage across the battery;
a switch, arranged to selectively connect the test load to the battery to at least partially discharge the battery through the test load;
a control means arranged to control the switch; and
calculation means, arranged to determine, from the sensed voltages during discharge of the battery through the test load, at least one parameter relating to the state of health of the battery.
Advantageously, the battery is that of an automotive vehicle. Preferably, the apparatus is arranged to be installed within such a vehicle.
The apparatus may be arranged to carry out the method of the first aspect of the invention.
The test load may be a resistor of known value, typically in series with the switch. The test load may comprise an electric apparatus of the vehicle, such as an element of a heated windscreen.
The test load may comprise two parts arranged as a potential divider; a first part of accurately known resistance, across which a voltage measurement is taken, and a second part having a second part resistance,
generally larger than that of the first part and typically known with less accuracy. The first part may be provided with a first part voltage sensor, and the control means may be arranged to determine the resistance of the second part from the ratio of the voltage across the first part to the voltage across the battery. The second part may comprise the electric apparatus of the vehicle.
The apparatus may further comprise a sense resistor and a sense resistor voltage sensor which is arrange to sense the voltage across the sense resistor, in which the calculation means is arranged to calculate the current passing through the sense resistor when the battery is being discharged through the test load. It may be arranged to determine the current only when the battery is being discharged through the test load.
Preferably, the calculation and/or control means comprise an engine management computer of the vehicle.
There now follows, by way of example only, a description of an embodiment of the invention described with reference to the accompanying drawings, in which:
Figure 1 is a block diagram showing the functions carried out in an apparatus according to the present invention;
Figure 2 is a block diagram showing a real time routine of the apparatus of Figure 1;
Figure 3 is a block diagram showing a performance prediction routine of the apparatus of Figure 1, in a current prediction mode;
Figure 4 is a block diagram showing the performance prediction routine in a voltage predication mode;
Figure 5 is a block diagram of a battery characterisation routine of the apparatus of Figure 1;
Figure 6 is a block diagram showing a model adaptation routine of the apparatus of Figure 1 ;
Figure 7 is a flow diagram showing the operation of the battery characterisation of Figure 5;
Figure 8 is a flow diagram showing the recursive operation of the model adaptation routine of Figure 6; and
Figure 9 shows the control loop of the model adaptation routine of Figure 6.
The apparatus 2 shown in the Figures is provided in a vehicle such as a car (not shown) provided with a lead-acid battery 1 so as to provide indications as to the state of health and of charge of battery 1 as described below. Conveniently, it makes use of the vehicle electronics such as an on-board microcomputer but it may form a discrete unit in its own right.
The apparatus comprises a hardware component 3 and a microprocessor 4. Taking the hardware component 3 first, the apparatus is provided with a switch 5 and a test load resistor 6 in series with one another. The switch is controlled by driver means 7, which acts to both control the switch 5 and to determine the current passing through the switch. The apparatus is, in use, connected in series with the battery 1.
The apparatus is also provided with means for determining the temperature 8 of and the voltage 9 across the battery 1.
The measurements are passed to the microprocessor 4, which provides a number of functions which are described below in more detail. A state of charge (SOC) model 10 models the amount of charge stored in the battery. An inverse SOC model 11 models the voltage at the battery terminals given a certain SOC. A battery characterisation routine 12 determines the characteristics of the battery 1. A state-of-health routine 13 determines the health of the battery 1. A model adaptation routine acts to revise the models 11 , 12 based on the determined state-of- health of the battery 1. Storage means 15 is arranged to store a measurement history of the measured currents and voltages. Finally, a performance prediction algorithm 12 predicts how the performance of the battery 1 will change.
The apparatus outputs various signals 17, 18, 19, 20, 21 that may be displayed to a driver or which, more typically, are used by a microprocessor of the vehicle alter the behaviour of the vehicle. For example, if the battery is dangerously low on charge on a vehicle provided only with electrically-actuated brakes, it may refuse to allow the vehicle to move.
In normal use of the vehicle with the ignition on (Figure 2 of the accompanying drawings) , the apparatus uses the temperature 8 and voltage 9 measured to run the SOC model 10. This is any suitable model that estimates the state of charge of the battery 1 from the given inputs.
A typical example would be the model described in European Patent
Publication Number EP 1 373 915. This model estimates the equilibrium open-circuit voltage of the battery and determines the ratio by which the actual voltage of the battery differs from the estimate. This gives a rate
of change of the proportional charge of the battery, which may be used to generate an estimate of the state of charge of the battery by convergence.
This model, as with most models, has a number of parameters that can be changed depending on the characteristics and state of health of the battery, such as the level of shedding of the lead or lead oxide of the plates (so-called "paste shedding") , the concentration of the sulphuric acid within the battery 1, the level of crystallisation of the lead (II) oxide, the internal resistance of the battery 1 and so on.
From the model 10, the apparatus outputs a signal 17 indicative of the state of charge of the battery, and of the "120" value, indicative of the current that the battery 1 could supply for 20 hours before discharging.
The model is repeatedly executed during operation of the vehicle so that an up-to-date indication of the state of charge of the battery 1 is available. It is also executed at a lesser frequency when the vehicle is not in use with the ignition off, so as to monitor the state of charge of the battery so as to ensure that the effect of quiescent drains or self-drainage is measured.
The apparatus also provides a performance prediction algorithm 16, which (as in Figure 3 of the accompanying drawings) uses the SOC model 10 to predict from a given temperature 8a, battery voltage 9a, state of charge 17a and general battery characteristics what current can be drawn from the battery for a given period. This is provided as an output signal 19, which may be expressed as an 120 value. This is a current prediction mode, and can be used to determine the "cold cranking amperage" (CCA) or current available from the battery to start an engine from cold.
The performance prediction algorithm 16 may also have a voltage prediction mode (Figure 4 of the accompanying drawings) in which it uses the inverse 11 of the SOC model 10 to calculate from a given state of charge 17a, temperature 8a and current demand 19a (typically expressed as an 120 value over a given test period) the value of the minimum voltage of the battery. This is expressed as an output signal 18.
Either of the outputs 18, 19 of the performance prediction algorithm 16 can be used by a microprocessor of the vehicle to determine whether certain circuits of the vehicle will function correctly.
The data from which the state of health of the battery 1 is determined is measured using a series of brief discharges of the battery through load resistor 6. The load resistor itself is of known value, and may be part of the apparatus, be external to the apparatus or even be vehicle circuit of known characteristics, such as the heater for a windscreen of the vehicle. The discharges are generated by the switch being controlled by the driver 7 so as to generate a pulse, typically pulse- width modulated, through the load. The pulse is repeatedly generated at a plurality of frequencies, modulations and durations so as to provide a range of data for the state of health routines.
The load resistor may, in an alternative embodiment, comprise two parts arranged as a potential divider; a first part of accurately known resistance, across which a voltage measurement is taken, and a second part of generally larger resistance known with less accuracy. If this is connected across the battery and the battery voltage taken, the resistance of the second part can be determined from the ratio of voltages measured.
The brief discharge data is used by two similar routines. Both routines use the data to model the battery 1 and hence alter how the state of charge
model 10 models the state of charge of the battery 1. The battery characterisation routine 12 is concerned with the present electrical characteristics of the battery 1 - particularly the state of charge and the capacity, whilst the model adaptation routine 14 is concerned with modelling the changes within the battery (such as paste shedding, increased ohmic resistance due to corrosion of the plates and so on) and adapting the model to account for those changes. The two routines are expressed herein as two separate routines, although their function could be provide by a single routine if it is so desired.
Taking the battery characterisation routine 12 first (as demonstrated in Figures 5 and 7 of the accompanying drawings) , the routine takes the battery open circuit voltage 22 (that is, when the load is not connected across the battery and the ignition of the vehicle is off) , battery temperature and the brief discharge voltages and currents. A series of measurements are stored in a memory 15 so as to provide historical data. From the open circuit voltage 22 the state of charge of the battery 1 can be estimated.
Being able to provide an estimate of the state of charge is important when using a state of charge model 10 such as that described above with reference to European Patent Publication Number EP 1 373 915 that uses convergence to estimate the state of charge of the battery - it is desired that the initial state of charge of the battery is known when it is installed in the vehicle or should the apparatus for some reason forget the present state of charge. The routine uses a function 23 of the battery open circuit voltage Vocv and the battery temperature T to do so.
From the brief discharge data 24, comprising data on the battery voltage Vbat, battery temperature T and time t over the brief discharges, the battery characterisation can calculate the capacity of the battery and
internal resistance. The capacity is estimated as an 120 current 25, which is multiplied by 20 at step 26 to determine a capacity in amp hours. The internal resistance is calculated 27 dependent on the time constant of the change in voltage at the beginning of a brief discharge (within the first few milliseconds) before other non-ohmic effects have initiated. Accordingly, sufficiently fast sampling must be used to determine the time constant over this range.
In an alternative embodiment, the battery characterisation routine may determine the internal resistance of the battery by determining the current flowing through the battery (typically by measuring the voltage across a sense resistor) and measuring the instantaneous voltage drop as the test load 6 is applied to the battery 1.
The outputs 28 of the routine are both output as a signal 20 and fed back into the state of charge routine 10 to modify the model so as to provide more accurate results. The values fed back into the model by this routine are especially useful in the situation where a new battery is installed where the characteristics of the battery upon which the model 10 depends are unknown.
The other routine to use the brief discharge data, the model adaptation routine 14 is shown in more detail in Figures 6, 8 and 9 of the accompanying drawings. The routine uses the measurement history stored in memory 15 of current during brief discharges, battery voltage and temperature to determine how the battery is changing over time - the level of paste shedding, crystallisation, increases in ohmic resistance and so on.
These model parameters are calculated using a model 30. Some parameters are found by direct mapping of the brief discharge data and
others by an identification routine such as a regressive optimisation routine based on past data.
The routine is recursive as shown in Figure 8 of the accompanying drawings, wherein the model is simulated using current and temperature as inputs. A calculation of the battery voltage is made using the model parameters and the error between the modelled and actual voltages compared 31. If the error is small, then these parameters can be used, otherwise another iteration is made using the revised model, based on for example non-linear or multi-objective optimisation, data fitting or least- squares techniques. This can be repeated until a desired level of accuracy is reached.
The model adaptation routine may also be run in real time (when the battery is in use with the vehicle ignition on) as shown in Figure 9 of the accompanying drawings. The routing in this case does not use the real time current, as only the brief discharge current is used. The routine may use a slowly adapting parameter observer/estimator loop, where a model 34 of the battery parameters given the battery voltage and temperature is used so as to modify the state of charge model 10, the predicted state of charge and capacity of the state of charge model 10 being used to modify the new model 34.
The parameters output from the model adaptation routine 14 are used to run a state of health routine 13. This calculates from the parameters of the model adaptation routine the changes and type of changes occurring in the battery 1 and to what extent they are detrimental to the operation of the battery 1. The changes can include shedding, crystallisation, increases in ohmic resistance and so on. The state of health routine may also extrapolate from the rate at which these effects are occurring the
expected life of the battery, and may raise a flag if from this the battery 1 is likely to expire within a certain times cale.