Claims:We claim:
1. A device (102) to estimate quality parameters of a battery pack (104), said
battery pack (104) used in an Electric Vehicle (EV) (114) in a swappable
5 manner, said device (102) configured to:
a. measure battery parameters during a drive cycle of said EV (114)
to derive a drive profile/usage pattern of said battery pack (104),
said battery parameters comprises at least one of current, voltage,
and temperature, said battery parameters are measured for every
10 drive cycle until said battery pack (104) is swapped;
b. determine a stress factor on said battery pack (104) based on said
measured battery parameters in reference to a map/table prestored
in a memory element (106) of said device (102),
characterized in that,
15 c. estimate, through an estimator module (110), at least one quality
parameter selected from a real time resistance growth and
capacity fade, said estimator module (110) applies a stochastic
particle filtering to said measured battery parameters to account
for irregularities due to a non-linear characteristics of said
20 battery pack (104).
2. The device (102) as claimed in claim 1, said estimator module (110)
configured to,
process said measured battery parameters through a predetermined
25 stochastic model (112) of said battery pack (104) and calculate a gaussian
noise, said predetermined stochastic model (112) is a non-linear second
order equivalent circuit model;
process said gaussian noise using said stochastic particle filtering to
estimate a resistance growth value, said estimated resistance growth value
30 is calculated and stored for every drive cycle;
14
fit a curve of said estimated resistance growth value from said
stochastic model (112) with empirically derived resistance growth curve to
calculate minimum Root Mean Square Error (RMSE), and
add said RMSE to said empirically derived resistance value to
determine re 35 alistic resistance growth value.
3. The device (102) as claimed in claim 2, wherein a capacity loss model
configured to,
receive said output from said estimator module (110) as an input and
40 predict a capacity loss,
calculate a difference of said predicted capacity loss and
predetermined empirical capacity loss based on said predetermined
stochastic model (112), wherein said difference is used to finetune
said predicted capacity loss, and
45 determine a realistic capacity loss by addition of said difference to
said predetermined empirical capacity loss.
4. The device (102) as claimed in claim 1, wherein a remaining life of said
battery pack (104) is estimated based on subtraction of capacity fade during
50 said previous drive cycle from an expected capacity of said battery pack
(104), wherein said expected capacity of said battery pack (104) is energy
throughput of said battery pack (104) over lifetime.
5. The device (102) as claimed in claim 1, wherein a value of said battery pack
55 (104) is determined based on synthetic Depth of Discharge (SDOD) of said
battery pack (104), wherein said SDOD is inversely related to said real-time
capacity fade.
6. A method for estimating quality parameters of a battery pack (104), said
60 battery pack (104) used in an Electric Vehicle (EV) (114) in a swappable
manner, said method comprising the steps of:
15
a. measuring battery parameters during a drive event of said EV
(114) for deriving a drive profile/usage pattern of said battery
pack (104), said battery parameters comprises at least one of
current, voltage, and tem 65 perature and said battery parameters are
measured for every drive cycle until said battery pack (104) is
swapped;
b. determining a stress factor on said battery pack (104) based on
said measured battery parameters in reference to a map/table pre70
stored in a memory element (106) of a device (102),
characterized by,
c. estimating, through an estimator module (110), at least one
quality parameter selected from a real time resistance growth and
capacity fade, said estimator module (110) applies a stochastic
75 particle filtering to said measured battery parameters to account
for irregularities due to a non-linear characteristics of said
battery pack (104).
7. The method as claimed in claim 1, wherein said estimator module (110)
80 comprises the steps of,
processing said measured battery parameters through a
predetermined stochastic model (112) of said battery pack (104) and
calculating a gaussian noise, said predetermined stochastic model (112) is a
non-linear second order equivalent circuit model stored in said memory
85 element (106);
processing said gaussian noise using said stochastic particle filtering
to estimate a resistance growth value, said estimated resistance growth value
is calculated and stored for every drive cycle;
fitting, a curve of said estimated resistance growth value from said
90 stochastic model (112) with empirically derived resistance growth curve,
and calculating a minimum Root Mean Square Error (RMSE), and
16
adding said RMSE to said estimated resistance growth value for
determining realistic resistance growth value.
8. The method as 95 claimed in claim 7, wherein said a capacity loss module
comprises the steps of,
receiving said realistic resistance growth value and estimating a
capacity loss,
calculating a difference of said estimated capacity loss and
100 predetermined empirical capacity loss based on said predetermined
stochastic model (112), and
determining a realistic capacity loss by adding said difference to said
predetermined empirical capacity loss.
105 9. The method as claimed in claim 1, comprises determining a remaining life
of said battery pack (104) by subtracting said capacity loss during said
previous drive cycle from expected capacity of said battery pack (104),
wherein said expected capacity of said battery pack (104) is
predetermined/rated energy throughput of said battery pack (104) over
110 lifetime.
10. The method as claimed in claim 1, comprises determining a value of said
battery pack (104) based on a synthetic Depth of Discharge (SDOD) of said
battery pack (104), wherein said SDOD is inversely related to said real-time
115 capacity fade (energy throughput after said drive cycle). , Description:Complete Specification:
The following specification describes and ascertains the nature of this invention
and the manner in which it is to be performed:
2
Field of the invention:
[0001] The present invention relates to a device and method to estimate quality
parameters of a battery pack.
5 Background of the invention:
[0002] In a battery swapping ecosystem, the battery pack (an asset) is borrowed
from a swapping station on a per trip basis (for ‘x’ no of kms). A driver swaps the
discharged battery pack for a charged one when needed. Each of the swappable
battery over its lifetime resides with a different driver after every instance of swap.
10 Hence, the driver, as an active participant in the swapping ecosystem, significantly
dictates the remaining life of the swappable battery pack. The driver inflicts stress
on the swappable battery basing the utilization pattern of the battery pack during
the trip. Hence it is important to classify driver behavior based on battery stress
factors (BSF) which remarkably changes after every event of swap. Considering
15 this non-linear factor, it becomes increasingly difficult to realistically compute
accurate resistance growth and capacity fade in a given lithium ion swappable
battery pack which determines the remaining usable life. There are various
deterministic methods (empirical/semi-empirical) means to measure the capacity
loss which happens due to increase in internal resistance over lifetime. Most of the
20 deterministic models are equations derived from experimental measurements in
laboratory conditions. These methods generally possess approximation errors as it
is performed in a static set-up, which leads to inaccuracy or failure in accurately
determining the factors influencing the capacity loss, as the real driving conditions
are extremely stochastic(random) in nature.
25
[0003] A patent literature DE102019219436 discloses a method for detecting a
stress-related condition of at least one component, in particular an energy store, in
a motor vehicle. The invention relates to a method for detecting a stress-related
condition of at least one component, in particular an energy store, in a motor vehicle,
30 wherein at least one sensor at least one metric (Ubatt, Ibatt) of the component,
wherein, for the recognition of the state of the component comprises at least two
3
different diagnostics may be used, wherein in one diagnostics and a stress rating
model is used in the further diagnosis a statement for a performance of component
is performed.
5 Brief description of the accompanying drawings:
[0004] An embodiment of the disclosure is described with reference to the
following accompanying drawing,
[0005] Fig. 1 illustrates a system containing a device to estimate quality parameters
of a battery pack, according to an embodiment of the present invention;
10 [0006] Fig. 2 illustrates different plots to estimate quality of the battery pack,
according to an embodiment of the present invention, and
[0007] Fig. 3 illustrates a method for determining quality parameters of the battery
pack, according to the present invention.
15 Detailed description of the embodiments:
[0008] Fig. 1 illustrates a system containing a device to estimate quality parameters
of a battery pack, according to an embodiment of the present invention. The system
100 comprises a cloud 116 and an Electric Vehicle (EV) 114. The battery pack 104
is used in the EV 114 in a swappable manner. The device 102 configured to measure
20 battery parameters during a drive cycle of the EV 114 to derive a drive profile/usage
pattern of the battery pack 104. The battery parameters comprise at least one of
current, voltage, temperature of the battery pack 104 during the drive cycle. Other
parameters are usable which indicates deterioration of the battery pack 104. The
battery parameters are measured for every drive cycle until the battery pack 104 is
25 swapped. The device 102 further determines a stress factor on the battery pack 104
based on the measured battery parameters in reference to a map/table pre-stored in
a memory element 106 of the device 102, characterized in that, the device 102
further configured to estimate, through an estimator module 110, at least one quality
parameter selected from a real time resistance growth and capacity fade (or capacity
30 loss). The estimator module 110 applies a stochastic particle filtering to the
measured battery parameters to account for irregularities caused by factors pertinent
4
to a non-linear characteristics of the battery pack 104. The factors comprise nonlinear
usage pattern of different users/drivers during the course of lifetime of the
battery pack 104, along with different pattern of charging at different charging
stations. In other words, the factors contain usage or discharge of the battery pack
104 and charging of the battery pack 104 during 5 its lifetime is not same and always
varies.
[0009] In accordance to an embodiment of the present invention, the device 102 is
at least one of an Electronic Control Unit (ECU) or controller of the EV 114, a cloud
10 based apparatus 116, a Battery Control Unit (BCU), a communication unit and a
combination thereof. The device 102 refers to computing devices/units comprising
components such as memory element 106 such as Random Access Memory (RAM)
and/or Read Only Memory (ROM), Analog-to-Digital Converter (ADC), Digitalto-
Analog Convertor (DAC), clocks, timers and a processor (such as Central
15 Processing Unit (CPU)) (capable of implementing machine learning) connected
with the each other and to other components through communication bus channels.
The components mentioned are just for understanding and may have more or less
components as per requirement. The memory element 106 of the device 102 is prestored
with logics or instructions or programs or applications or thresholds or values
20 which is accessed by the processor as per the defined routines. The internal
components of the device 102 are not explained for being state of the art, and the
same must not be understood in a limiting manner. The device 102 is capable to
communicate through wired and wireless means such as but not limited to Global
System for Mobile Communications (GSM), 3G, 4G, 5G, Wi-Fi, Bluetooth,
25 Ethernet, serial networks, Universal Serial Bus (USB) cable, micro-USB, and the
like.
[0010] In accordance to the present invention, the device 102 dynamically accounts
for the non-linear behavior change of the swappable battery pack 104 after every
30 instance of swap from swapping station over its lifetime. The system 100 comprises
the swappable battery pack 104 (connected/non-connected), swapping stations, the
5
driver, and the telematics unit of the EV 114. Further, the drive cycle is considered
to completed after any one of every swap or drive session/event or a summation of
many drive cycles.
[0011] In accordance to an embodiment 5 of the present invention, the device 102 is
the BCU and is integrated with the battery pack 104. The BCU independently
monitors the driver profile, estimates the stress factor, the resistance growth, and
the capacity loss. Alternatively, the device 102 is the cloud based apparatus 116
which is connected to the battery pack 104 either directly through a Telematics
10 Control Unit (TCU) of the battery pack 104, or through a TCU of the EV 114. The
cloud based apparatus 116 receives all the data from the battery pack 104 and
estimates the stress factor, the resistance growth, and the capacity loss. The cloud
based apparatus 116 also determines a remaining life or value of the battery pack
104. In yet another alternative, the device 102 is communication unit of the user
15 which is connected to the EV 114 and works similar to the cloud based apparatus.
The examples of communication unit are smartphone, tablet, smartwatch, wearable
apparatuses, and the like. In yet another alterative, the device 102 is an external
dongle which is interfaced with the EV 114 through On-Board Diagnostic (OBD)
port or other wired or wireless connectivity means. In yet another alternative, at
20 least two of the BCU, cloud based apparatus 116, communication unit, external
dongle is used to determine the quality parameters of the battery pack 104 by
sharing the computational resources.
[0012] In accordance to an embodiment of the present invention, a
25 determination/quantification of the stress factor on the battery pack 104 is disclosed.
The quantification of the stress factor for the swappable battery pack 104 is done
by assessing drive profiles of different drivers. The stress factor is computed as a
function of root mean square (RMS) value of current (defined as Irms) based on drive
profile/ usage pattern of the driver. The rate at which state of charge (SOC) drops
30 is directly proportional to severity of stress inflicted on the battery pack 104. Also,
comparing the severity of stress against temperature rise creates a reference to
6
curate a proof of stress on the battery pack 104. The RMS value of current (Irms)
and voltage of the battery (Vbatt) are taken as inputs to quantify a relationship
between stress factor over time. A base plot 118 is provided between State of
Charge (SOC) and time represented by Y axis and X axis respectively. A curve t1,
t2, t3 indicates the time 5 taken by the battery pack 104 to depreciate its energy over
time. 118, t1 indicates a low severity in stress factor when compared to t2 and t3
which signify a steeper drop. The discharge profile (current) at t1 is less aggressive
when compared to t3. A tabular matrix data 108 of stress factor across bands of
current profile is stored based on the discharge profile as depicted in the base plot
10 118, in the memory element 106, where the max severity score is 1.
[0013] In accordance to an embodiment of the present invention, a construction of
a Design of experiment (DOE) to measure resistance growth and capacity loss over
lifetime is briefly explained. The DOE is usually a part of the validation procedure
15 pertaining to the battery pack 104. A few known factors that influence the resistance
increase and capacity fade in real time are discharge current (Irms), temperature rise
(Trise) during operation, ambient temperature (Troom), Depth of discharge (DOD)
and C-rate (rate at which current discharge happens). Based on the test and
experimental analysis, an empirical relationship is established between capacity
20 fade and all the factors, and the data generated out of the empirical equation is used
to train a stochastic model 112.
[0014] According to an embodiment of the present invention, the device 102 is
configured to use the estimator module 110 to, process the measured battery
25 parameters or the stress factor through a predetermined stochastic model 112 of the
battery pack 104 and calculate a gaussian noise. The stochastic model 112 is an
empirically derived non-linear second order stochastic circuit model. The estimator
module 110 processes the gaussian noise using the stochastic particle filtering to
estimate the resistance growth value. The estimated resistance growth value is
30 calculated and stored for every drive cycle. The device 102 fits a curve of the
estimated resistance growth value from the stochastic model 112 with empirically
7
(experimentally) derived resistance growth curve to calculate minimum Root Mean
Square Error (RMSE). The device 102 adds the RMSE (or difference) to the
estimated resistance growth value to determine realistic resistance growth value. In
a very first instance, the estimated resistance growth value is just one value, and the
same is used to fit with the empirically 5 derived resistance growth curve. However,
with regular use of the battery pack 104 by the same and/or different drivers, a
multiple of the estimated resistance growth values are stored and generates a curve.
This generated curve is then fit with the empirically derived resistance growth curve
for determining the error and further processing.
10
[0015] The estimator module 110 with appropriate signal filtering is used to
calculate the realistic resistance increase per drive cycle. The real time estimation
of remaining life of swappable battery pack 104 requires introduction of
randomness (impact of driver behavior on capacity fade and resistance growth)
15 through stochastic processes. The battery parameters such as like voltage, current,
temperature, etc. are measured in real time from the battery pack 104 (if it’s a nonconnected
battery pack 104, a TCU measures the aforementioned battery
parameters). The stochastic model 112 is provided which mimics the diffusion
dynamics like in electrochemical model and represents the battery pack 104 as a
20 non-linear / discrete state space representation. The device 102 estimates the model
parameters (SOC, current) by fitting the model output to experimental data
(power/current profiles) with curve fitting to have minimum root mean square error
(RMSE). The state variables of non-linear and non-gaussian systems is chosen to
preserve non-gaussian noise. The particle filtering is based on sequential Monte-
25 Carlo using recursive Bayesian filtering. The output of particle filtering, i.e. the
estimator module 110, is effective realistic resistance increase for the event and
SOC.
[0016] According to the present invention, the device 102 is configured to
30 quantify/determine the capacity loss during real time conditions. The device 102
uses a stochastic method for the same. The capacity loss varies based on usage per
8
event and is directly proportional to resistance growth, and empirical calculations
establishes a square root relation between the capacity loss and the resistance
growth. The output of the particle filtering, i.e. the effective resistance growth and
SOC, is fed into the capacity loss model (Qloss). In other words, the device 102
receives the 5 output from estimator module 110 as an input and predicts the capacity
loss using the capacity loss model. The equation derived from the relation is
predicted capacity loss (Qloss(pred)) = A +vR.B,
where A and B are approximation constants.
10 [0017] An error between the predicted capacity loss and the empirical capacity loss
is then calculated. A normal distribution for this error is the “stochastic term” ?
which is then used to finetune the predicted capacity loss.
Estimated capacity loss: Q(loss(pred)) = Q((loss(emp)) + ?
where ?= error
15 The device 102 then determines a realistic capacity loss by adding the differences
or error to the predetermined empirical capacity loss.
[0018] Fig. 2 illustrates different plots to estimate quality of the battery pack,
according to an embodiment of the present invention. The quality of the battery
20 pack 104 is determined in terms of two indicators, comprising remaining life of the
battery pack 104 and a value of the battery pack 104. According to an embodiment
of the present invention, the remaining life of battery pack 104 is estimated by the
device 102 based on the quantified/determined realistic capacity loss. The
remaining life of the battery pack 104 is estimated based on subtraction of capacity
25 loss during the previous drive cycle from the expected capacity of the battery pack
104. The expected capacity of the battery pack 104 is the rated energy throughput
of the battery pack 104 over lifetime. The remaining life of the battery pack 104 is
estimated in terms of unit of Kwh. Below is a generic equation for better
understanding.
30 Energy throughput over lifetime(kWh) – Energy throughput after the event has
occurred(kWh)
9
[0019] According to an embodiment of the present invention, the device 102
determines/estimates the value of the battery pack 104 based on synthetic Depth of
Discharge (SDOD) of the battery pack 104. A first plot 210 represents relation
5 between energy throughput of the battery pack 104 over lifetime in X-axis 202 and
capacity loss in Y axis 204, in respective suitable units. A second plot 220
represents the same as the first plot 210 but for a single point. A third plot 230
represents a curve between SDOD and a value of the battery pack 104 in X-axis
206 and Y-axis 208 respectively in suitable units. A Depth of Discharge (DOD) of
10 the battery pack 104 is inversely related to the cycle life of the battery pack 104.
The DOD dictates the amount of energy throughput and is fixed in order to have
homogenous performance across lifetime of the battery pack 104. Considering the
first plot 210, a first curve 212 represents theoretical/ideal capacity fade to energy
throughput as a baseline. The first curve 212, when co-related to a second curve
15 214, which denotes a real-time capacity loss, the energy throughput remains the
same, but the rate at which each of the battery pack 104 degrades is non-linear. A
difference of real-time capacity fade to ideal capacity fade is done based on SDOD
calculation as depicted in second plot 220. The SDOD is introduced in the algorithm
calculation to estimate the theoretical drop in DOD from the reference value, i.e.
20 what is the expected/touted performance of the battery pack 104 in ideal conditions
and what has it dropped to over usage as a function of the driver behavior. The
SDOD is inversely related to the real-time capacity loss/ (energy throughput after
the drive cycle). Larger the SDOD value, lesser is the value/quality of the battery
pack 104. The co-relation of SDOD vs life of battery is established and shown in
25 the third curve 216 of the third plot 230. The life of battery pack 104 is expressed
in “number of kms”, “useful life of battery in years” etc.
[0020] A working of the device 102 is explained. Consider the device 102 is the
cloud based apparatus 116. The cloud based apparatus 116 denotes the use of cloud
30 computing architecture, as known in the art, to determine the quality parameters of
the swappable battery pack 104. A driver drives the EV 114 aggressively which
10
causes high stress factor on the battery pack 104. The battery parameters are
measured and then transmitted to the cloud based apparatus 116. The stress factor
is accumulated after every drive cycle by the cloud based apparatus 116. Once the
driver swaps the battery pack 104 at the swapping station, the quality parameters of
the battery pack 5 104 are calculated and stored in the cloud based apparatus 116. The
quality parameters are then processed to derive quality indicator of the battery pack
104, i.e. remaining life or value to decide on further activities with the battery pack
104 such as but not limited to changing the fee structure for using the battery pack
104, discarding the battery pack 104, etc.
10
[0021] Fig. 3 illustrates a method for determining quality parameters of the battery
pack, according to the present invention. The battery pack 104 used in the Electric
Vehicle (EV) 114 in the swappable manner. The method comprises plurality of
steps, of which a first step 302 comprises measuring battery parameters during a
15 drive event of the EV 114 for deriving the drive profile/usage pattern of the battery
pack 104. The battery parameter comprises at least one of current, voltage, and
temperature. The battery parameters are measured for every drive cycle until the
battery pack 104 is swapped. A step 304 comprises determining the stress factor on
the battery pack 104 based on the measured battery parameters in reference to the
20 map/table pre-stored in the memory element 106 of the device 102. The method is
characterized by, a step 306 which comprises estimating, through the estimator
module 110, at least one quality parameter selected from a real time resistance
growth and capacity fade. The estimator module 110 applies the stochastic particle
filtering to the measured battery parameters to account for irregularities due to the
25 non-linear characteristics of the battery pack 104. The method is executed by the
device 102 as described in Fig. 1 and Fig. 2.
[0022] According to the present invention, the estimator module 110 comprises
steps to determine realistic resistance growth value. A step 308 comprises
30 processing the measured battery parameters (or stress factor) through the
predetermined stochastic model 112 of the battery pack 104 and calculating the
11
gaussian noise. The predetermined stochastic model 112 is the non-linear second
order equivalent circuit model stored in the memory element 106. A step 310
comprises processing the gaussian noise using the stochastic particle filtering to
estimate the resistance growth value. The estimated resistance growth value is
calculated and stored for every 5 drive cycle. A step 312 comprises fitting, the curve
of the estimated resistance growth value from the stochastic model 112 with
empirically derived resistance growth curve and calculating the minimum Root
Mean Square Error (RMSE). A step 314 comprises adding the RMSE to the
estimated resistance growth value for determining realistic resistance growth value.
10
[0023] According to the present invention, method steps followed by the capacity
loss model is described. A step 316 comprises receiving the realistic resistance
growth value and estimating the capacity loss. A step 318 comprises calculating the
difference of the estimated capacity loss and predetermined empirical capacity loss
15 based on the predetermined stochastic model 112. A step 320 comprises
determining the realistic capacity loss by adding the difference to the predetermined
empirical capacity loss.
[0024] The method further comprises determining quality through quality
20 indicators comprising the remaining life and value of the battery pack 104. The
method further comprises determining the remaining life of the battery pack 104 by
subtracting the capacity loss during the previous drive cycle from expected capacity
of the battery pack 104. The expected capacity of the battery pack 104 is
predetermined/rated energy throughput of the battery pack 104 over lifetime. The
25 method for determining the value of the battery pack 104 is based on the synthetic
Depth of Discharge (SDOD) of the battery pack 104. The SDOD is inversely related
to the real-time capacity fade (energy throughput after the drive cycle).
[0025] According to an embodiment of the present invention, the system 100, the
30 device 102 and method to ascertain the value of a swappable battery pack 104 based
on the rate of battery energy deterioration in real driving conditions is disclosed.
12
Thereby, a solution is proposed to dynamically estimate the realistic remaining life
of a swappable lithium ion battery pack 104 by accounting the irregularities caused
by various actors (mainly the driver) pertinent to a non-linear ecosystem. Based on
the dynamic estimation of remaining life of battery pack 104, the method uses an
algorithm to ascertain v 5 alue to the asset (battery pack 104) over its lifetime. The
device 102 and method is aimed to introduce the stochastic nature of real driving
conditions into a parametrized non-linear system and to quantify the impact of real
driving conditions on a swappable battery pack 104 in terms of capacity loss.
Further, the estimation of capacity loss is used to ascertain the remaining life of the
10 battery left at any given point of time and thus a method to ascertain the value of
the swappable battery pack 104.
[0026] It should be understood that embodiments explained in the description
above are only illustrative and do not limit the scope of this invention. Many such
15 embodiments and other modifications and changes in the embodiment explained in
the description are envisaged. The scope of the invention is only limited by the
scope of the claims.