US 2026/0016539 A1Published application

System and method for efficient battery capacity estimation

AssigneeGM Global Technology Operations LLC

InventorsInsu Chang, Shobhit Gupta, Su-Yang Shieh, Wesley Gerard Zanardelli

Life of this patent

Estimating the term…

Loading patent…

Abstract

A system for efficient battery capacity estimation includes vehicle with a battery, sensors, and a human-machine-interface (HMI). The vehicle has a controller that executes a battery capacity estimation (BCE) application. The BCE application performs local data collection from the sensors, transmits the data collected by the sensors to the remote servers where data conversion occurs and a data-driven BCE model is trained. Data conversion substantially reduces a size of and removes a time-dependency of the sensor data. The local vehicle portion of the BCE application receives, from the server, the data-driven model and estimates a capacity of a battery of the battery-operated device. Upon determining a battery capacity estimate, the system notifies vehicle users, via the HMI, of the current battery capacity estimate, and shares the current battery capacity estimate with additional vehicle sub-systems.

Claims20

Dependent claims

A system for efficient battery capacity estimation in a battery-operated device, the system comprising:

a battery-operated device having one or more batteries;
one or more sensors disposed on the battery-operated device and collecting real-time information about the battery-operated device;
a human-machine interface (HMI) disposed in the battery-operated device and transmitting information to battery-operated device users;
the battery-operated device has a controller with a processor, a memory, and one or more input/output (I/O) ports, the I/O ports communicating with the one or more sensors, the battery-operated device, and the HMI;
the processor executing programmatic control logic stored in the memory;
the programmatic control logic including a battery capacity estimation (BCE) application, the BCE application comprising:
a first control logic for performing local data collection from the one or more sensors of the battery-operated device;
a second control logic for data conversion of the data collected by the one or more sensors, and training a data-driven BCE model with data converted from the data collected by the one or more sensors, wherein the data conversion reduces a size of the data collected by the one or more sensors over time from a first size to a second size substantially smaller than the first size, wherein the data conversion removes a time-dependency of the data from the one or more sensors;
and
a third control logic for estimating a capacity of a battery of the battery-operated device, wherein upon determining an estimate of the capacity of the battery, the system generates a notification to the battery-operated device users, via the HMI, including the current battery capacity estimate, and shares the current battery capacity estimate with additional battery-operated device sub-systems utilizing accurate battery state-of-charge (SoC) estimations.

7 more elements

A method for efficient battery capacity estimation in a vehicle, the method comprising:

collecting, with one or more sensors disposed on a vehicle, real-time information about the vehicle, including real-time information about one or more batteries equipped to the vehicle;
transmitting, via a human-machine interface (HMI) disposed in the vehicle, information to vehicle occupants and to vehicle subsystems;
utilizing one or more cloud-computing servers;
executing programmatic control logic including a battery capacity estimation (BCE) application stored in memory of a controller of the vehicle and within memory of controllers of the one or more cloud computing servers, the controllers of the vehicle and the one or more cloud computing servers each having a processor, a memory, and one or more input/output (I/O) ports, the I/O ports communicating with the one or more sensors, the one or more remote servers, the vehicle, and the HMI, wherein the BCE application further includes control logic comprising:
performing local data collection from the one or more sensors of the vehicle, including:
determining a charging status of the battery of the vehicle;
upon determining that the battery is currently being charged, continuing to monitor the charging status of the battery;
and
upon determining that the battery is not currently being charged, causing the one or more sensors to begin measuring dynamic battery information during an operation or cycle of the vehicle;
and
at a completion of the operation or cycle of the vehicle, causing the one or more sensors to stop measuring dynamic battery information, and storing an operation or cycle time defining a duration of the operation or cycle, wherein the dynamic battery information further comprises:
a voltage (V), a current (I), and a temperature (T) of the battery;
determining whether a state of charge (SoC) of the battery is zero;
upon determining that the SoC of the battery is greater than zero, continuing to monitor the SoC of the battery until the SoC of the battery is equal to zero;
upon determining that the SoC of the battery is equal to zero, charging the battery and integrating current (I) until a state of charge of the battery is equal to 1;
and
upon determining that the state of charge of the battery is equal to one, generating a full original battery capacity based on integration of the current (I);
transmitting the data collected by the one or more sensors to the one or more cloud-computing servers, and for performing within the one or more cloud-computing servers, data conversion of the data collected by the one or more sensors, and training a data-driven BCE model with data converted from the data collected by the one or more sensors, wherein the data conversion reduces a size of the data collected by the one or more sensors over time from a first size to a second size substantially smaller than the first size, wherein the data conversion removes a time-dependency of the data from the one or more sensors, wherein the data conversion and training further comprise:
determining a quantity of data sets available for training;
defining sizes of partitions for a voltage (V), a current (I), and a temperature (T) of the battery;
and
calculating bins of the voltage (V), current (I), and temperature (T) of the battery, wherein the bins of the voltage (V), current (I), and temperature (T) of the battery are sized based on minimum and maximum values for each variable voltage (V), current (I), and temperature (T) of the battery, including control logic for obtaining voltage minimums and maximums (V_min, V_max) from battery specifications, current minimums and maximums (I_min, I_max) from battery performance data, temperature minimums and maximums (T_min, T_max) from battery operating condition information;
determining an optimal size of the bins by minimizing a performance index according to:
J
=
∑
k
=
1
N
⁢
W
1
(
h
⁡
(
z
k
-
z
k
-
1
)
)
2
+
W
2
⁢
N
;
such
⁢
that
⁢
⁢
z
0
=
V
min
,
z
N
=
V
max
⁢
for
⁢
voltage
⁢
(
V
)
;
that
⁢
z
0
=
I
min
,
z
N
=
I
max
⁢
for
⁢
current
⁢
(
I
)
,
and
⁢
that
⁢
z
0
=
T
min
,
z
N
=
T
max
⁢
for
⁢
temperature
⁢
(
T
)
,
z
k
-
z
k
-
1
>
0
;
N
≤
N
max
∈
ℕ
;
where N is the number of bins, W is a weighting factor;
and
stacking the bins of voltage (V), current (I), and temperature (T) of the battery with the time duration for the operation or cycle, and the full original battery capacity;
augmenting new data to existing data;
determining that data conversion is complete;
and
training a data-driven BCE model by:
defining hidden layers and a quantity of step delays in the data-driven model;
utilizing one or more of a recurrent neural network (RNN), an autoregressive with non-linear input (ARX) model, and a nonlinear autoregressive exogenous (NARX) model;
and
determining whether a battery capacity estimation error is less than a threshold error;
and
wherein upon determining that the battery capacity estimation error is greater than or equal to the threshold, continuing to define hidden layers and quantities of step delays in the data-driven model and continuing to train the data driven model through one or more of RNN, ARX and NARX;
and
wherein upon determining that the battery capacity estimation error is less than the threshold error, determining that the data-driven model design is complete;
and
receiving, from the cloud-computing server, the data-driven model and
beginning a current battery capacity estimation including:
determining a charging status of the battery of the vehicle;
upon determining that the battery is currently being charged continuously monitoring the charging status of the battery;
upon determining that the battery is not currently being charged, causing the one or more sensors to begin measuring dynamic battery information during an operation or cycle of the vehicle;
at a completion of the operation or cycle of the vehicle, causing the one or more sensors to stop measuring dynamic battery information, and store an operation or cycle time defining a duration of the operation or cycle;
calculating bins of voltage (V), current (I), and temperature (T) of the battery according to the data-driven model design;
stacking the bins of voltage (V), current (I), and temperature (T) of the battery with the time duration for the operation or cycle, and the full original battery capacity;
predicting the current battery capacity with the data-driven model;
and
wherein upon predicting the current battery capacity with the data driven model, generating a notification to the vehicle occupants, by the controller via the HMI, including the current battery capacity estimate, and sharing the current battery capacity estimate with the additional vehicle sub-systems utilizing accurate battery SoC estimations.

165 more elements

Dependencies are inferred from the claim text.