Battery state of health estimation based on real-time data
In some examples, a system may determine route information for a route to be traversed by a vehicle from a start location to a destination location, the vehicle including a battery having a state of health. The system receives discharge data corresponding to traversal of the vehicle along the route, the discharge data indicative of a rate of discharge of the battery at a plurality of locations along the route. Based on the received discharge data, the system at least one of trains or updates a first machine learning model configured for predicting the state of health of the battery. The vehicle may determine an estimated battery state of health based at least on the first machine learning model, and may receive control information while traversing the route based on the estimated battery state of health to at least partially minimize battery degradation during traversal of the route.
1 . A system comprising:
one or more processors configured by executable instructions to perform operations comprising:
providing, to a vehicle, a first machine learning model configured for predicting a state of health of a battery of the vehicle;
determining route information for a route to be traversed by the vehicle from a start location to a destination location;
receiving discharge data corresponding to traversal of the vehicle along the route, the discharge data indicative of a rate of discharge of the battery at a plurality of locations along the route;
receiving information related to charging of the battery along the route, including receiving a plurality of second machine learning models trained based on data from a plurality of charging stations, respectively, each respective second machine learning model having been trained separately based on local charging data associated with charging the battery at a corresponding respective charging station;
using each received respective second machine learning model to generate a respective prediction for generating a plurality of respective predictions;
retraining the first machine learning model configured for predicting the state of health of the battery using training data that includes the plurality of respective predictions generated by the plurality of respective second machine learning models; and
providing at least updated parameters of the first machine learning model to the vehicle, wherein the vehicle determines an estimated battery state of health based at least on the first machine learning model, the vehicle receiving control information while traversing at least one of the route or a future route based at least on the estimated battery state of health, wherein the vehicle is controlled during traversal of the at least one of the route or the future route based at least in part on the received control information to at least in part minimize battery degradation during traversal of the at least one of the route or the future route.
2 . The system as recited in claim 1 , further comprising an operation of aggregating an updated first machine learning model with at least one prior first machine learning model to generate an aggregated first machine learning model whose parameters are sent to the vehicle when providing at least the updated parameters of the first machine learning model to the vehicle.
3 . The system as recited in claim 1 , wherein, based on reaching the destination, the vehicle executes the first machine learning model to determine a current estimated battery state of health, and wherein the vehicle sends feedback regarding the first machine learning model to at least one entity that maintains the first machine learning model.
4 . The system as recited in claim 1 , wherein, during traversal of the route, the vehicle provides the estimated battery state of health to a first charging station of the plurality of charging stations, wherein the first charging station controls a charging profile during charging of the battery based at least on the estimated battery state of health.
5 . The system as recited in claim 1 , the operations further comprising:
while the vehicle is traversing the route, receiving at least speed profile data from a plurality of roadside units positioned along the route, the roadside units transmitting the speed profile data to the one or more processors, wherein the speed profile data is indicative of discharge of the vehicle battery in a vicinity of a respective roadside unit; and
converting the received speed profile data to discharge data for the vehicle based at least on battery profile information associated with the battery.
6 . The system as recited in claim 5 , wherein the training data further includes one or more discharge profiles received from one or more roadside units of the plurality of roadside units.
7 . The system as recited in claim 1 , wherein the training data further includes an ideal charging profile for the battery received from a manufacturer of the battery.
8 . A method comprising:
providing, by one or more processors, to a vehicle, a first machine learning model configured for predicting a state of health of a battery of the vehicle;
determining, by the one or more processors, route information for a route to be traversed by the vehicle from a start location to a destination location;
receiving discharge data corresponding to traversal of the vehicle along the route, the discharge data indicative of a rate of discharge of the battery at a plurality of locations along the route;
receiving information related to charging of the battery along the route, including receiving a plurality of second machine learning models trained based on data from a plurality of charging stations, respectively, each respective second machine learning model having been trained separately based on local charging data associated with charging the battery at a corresponding respective charging station;
using each received respective second machine learning model to generate a respective prediction for generating a plurality of respective predictions;
retraining the first machine learning model configured for predicting the state of health of the battery using training data that includes the plurality of respective predictions generated by the plurality of respective second machine learning models; and
providing at least updated parameters of the first machine learning model to the vehicle, wherein the vehicle determines an estimated battery state of health based at least on the first machine learning model, the vehicle receiving control information while traversing at least one of the route or a future route based at least on the estimated battery state of health, wherein the vehicle is controlled during traversal of the at least one of the route or the future route based at least in part on the received control information to at least in part minimize battery degradation during traversal of the at least one of the route or the future route.
9 . The method as recited in claim 8 , further comprising aggregating an updated first machine learning model with at least one prior first machine learning model to generate an aggregated first machine learning model whose parameters are sent to the vehicle when providing at least the updated parameters of the first machine learning model to the vehicle.
10 . The method as recited in claim 8 , wherein, during traversal of the route, the vehicle provides the estimated battery state of health to a first charging station of the plurality of charging stations, wherein the first charging station controls a charging profile during charging of the battery based at least on the estimated battery state of health.
11 . The method as recited in claim 8 , further comprising:
while the vehicle is traversing the route, receiving at least speed profile data from a plurality of roadside units positioned along the route, the roadside units transmitting the speed profile data to the one or more processors, wherein the speed profile data is indicative of discharge of the vehicle battery in a vicinity of a respective roadside unit; and
converting the received speed profile data to discharge data for the vehicle based at least on battery profile information associated with the battery.
12 . The method as recited in claim 11 , wherein the training data further includes one or more discharge profiles received from one or more roadside units of the plurality of roadside units.
13 . The method as recited in claim 8 , wherein the training data further includes an ideal charging profile for the battery received from a manufacturer of the battery.
14 . A non-transitory computer-readable medium storing instructions executable by one or more processors to cause the one or more processors to perform operations comprising:
providing, to a vehicle, a first machine learning model configured for predicting a state of health of a battery of the vehicle;
determining route information for a route to be traversed by the vehicle from a start location to a destination location;
receiving discharge data corresponding to traversal of the vehicle along the route, the discharge data indicative of a rate of discharge of the battery at a plurality of locations along the route;
receiving information related to charging of the battery along the route, including receiving a plurality of second machine learning models trained based on data from a plurality of charging stations, respectively, each respective second machine learning model having been trained separately based on local charging data associated with charging the battery at a corresponding respective charging station;
using each received respective second machine learning model to generate a respective prediction for generating a plurality of respective predictions;
retraining the first machine learning model configured for predicting the state of health of the battery using training data that includes the plurality of respective predictions generated by the plurality of respective second machine learning models;
providing at least updated parameters of the first machine learning model to the vehicle, wherein the vehicle determines an estimated battery state of health based at least on the first machine learning model, the vehicle receiving control information while traversing at least one of the route or a future route based at least on the estimated battery state of health, wherein the vehicle is controlled during traversal of the at least one of the route or the future route based at least in part on the received control information to at least in part minimize battery degradation during traversal of the at least one of the route or the future route.
15 . The non-transitory computer-readable medium as recited in claim 14 , further comprising an operation of aggregating an updated first machine learning model with at least one prior first machine learning model to generate an aggregated first machine learning model whose parameters are sent to the vehicle when providing at least the updated parameters of the first machine learning model to the vehicle.
16 . The non-transitory computer-readable medium as recited in claim 14 , wherein, during traversal of the route, the vehicle provides the estimated battery state of health to a first charging station of the plurality of charging stations, wherein the first charging station controls a charging profile during charging of the battery based at least on the estimated battery state of health.
17 . The non-transitory computer-readable medium as recited in claim 14 , the operations further comprising:
while the vehicle is traversing the route, receiving at least speed profile data from a plurality of roadside units positioned along the route, the roadside units transmitting the speed profile data to the one or more processors, wherein the speed profile data is indicative of discharge of the vehicle battery in a vicinity of a respective roadside unit; and
converting the received speed profile data to discharge data for the vehicle based at least on battery profile information associated with the battery.
18 . The non-transitory computer-readable medium as recited in claim 17 , wherein the training data further includes one or more discharge profiles received from one or more roadside units of the plurality of roadside units.
19 . The non-transitory computer-readable medium as recited in claim 14 , wherein the training data further includes an ideal charging profile for the battery received from a manufacturer of the battery.