Method and device for monitoring an on-board electrical system of a vehicle

1 . A device for monitoring an on-board energy system of a vehicle that comprises different on-board system components and that is operated based on a machine-learning energy management system, the machine-learning energy management system having been taught, for a reference on-board system, via reinforcement learning, the device comprising:

one or more processors configured to:

ascertain, for a state of an electrical energy store of the on-board energy system of the vehicle and for an action effected by the electrical energy store of the energy management system of the vehicle on the basis of the state, a reference reward associated with the electrical energy store that would be obtained during operation of the reference on-board system;

ascertain, for the state and for the action, an actual reward associated with the electrical energy store that is obtained during operation of the on-board energy system of the vehicle;

monitor at least the electrical energy store of the on-board energy system of the vehicle on the basis of the actual reward associated with the electrical energy store and on the basis of the reference reward associated with the electrical energy store, including by a comparison of the actual reward and the reference reward;

determine, based on the actual reward associated with the electrical energy store and the reference reward associated with the electrical energy store, whether the electrical energy store is impaired; and

output advice relating to the electrical energy store when it is determined that the electrical energy store is impaired;

wherein the one or more processors is configured to learn, in a course of a teaching process for the machine-learning energy management system, on the basis of rewards associated with the electrical energy store that have been obtained for different combinations of states and actions with respect to the electrical energy store in the course of the teaching process for the machine-learning energy management system.

2 . The device according to claim 1 , wherein:

the reference reward and/or the actual reward each comprise one or more reward components; and

the one or more reward components comprise:

a reward component relating to a current and/or relating to a voltage within the on-board energy system and/or on an on-board system component;

a reward component relating to a load and/or relating to a loading of an on-board system component; and/or

a reward component relating to a state of charge of the electrical energy store of the on-board energy system.

3 . The device according to claim 1 , wherein the one or more processors is configured to:

ascertain a divergence of a reward component of the actual reward from a corresponding reward component of the reference reward; and

take the divergence as a basis for determining, by way of comparison with a divergence threshold value, whether or not an on-board system component is impaired.

4 . The device according to claim 3 , wherein:

the actual reward and the reference reward each comprise a reward component for a specific on-board system component; and

the one or more processors is configured to take the divergence of the reward components of the actual reward and the reference reward for the specific on-board system component as a basis for determining whether or not the specific on-board system component is impaired.

5 . The device according to claim 3 , wherein

the divergence threshold value has been ascertained by way of simulation and/or by way of tests in advance.

6 . The device according to claim 5 , wherein

the divergence threshold value has been ascertained by way of simulation and/or by way of tests in advance specifically for a plurality of different on-board system components of the on-board energy system and/or for an applicable plurality of different reward components.

7 . The device according to claim 1 , wherein

the actual reward and the reference reward are dependent on one or more measurable variables of the on-board energy system;

the actual reward and the reference reward comprise one or more reward components for the applicable one or more measurable variables of the on-board energy system; and

the one or more processors is configured to:

ascertain measured values for the one or more measurable variables that are obtained as a result of the effected action during operation of the on-board energy system; and

ascertain the actual reward on the basis of the measured values for the one or more measurable variables.

8 . The device according to claim 1 , wherein

the machine-learning energy management system comprises at least one controller designed to regulate a measurable variable of the on-board energy system to a setpoint value; and

the actual reward and the reference reward are dependent on a divergence of an actual value of the measurable variable from the setpoint value during operation of the reference on-board system, or during operation of the on-board energy system.

9 . The device according to claim 1 , wherein the reference unit comprises at least one neural network.

10 . The device according to claim 1 , wherein

the reference on-board system corresponds to the on-board energy system with error-free and/or unimpaired on-board system components.

11 . A computer-implemented method for monitoring an on-board energy system of a vehicle that comprises different on-board system components and that is operated based on a machine-learning energy management system, the machine-learning energy management system having been taught, for a reference on-board system, by way of reinforcement learning; the method comprising:

ascertaining, with one or more controllers of the vehicle, for a state of an electrical energy store of the on-board energy system of the vehicle and for an action effected by the electrical energy store of the energy management system on the basis of the state, a reference reward associated with the electrical energy store that would be obtained during operation of the reference on-board system;

ascertaining, with the one or more controllers of the vehicle, for the state and for the action, an actual reward associated with the electrical energy store that is obtained during operation of the on-board energy system;

monitoring, with the one or more controllers of the vehicle, at least the electrical energy store of the on-board energy system of the vehicle on the basis of the actual reward associated with the electrical store and on the basis of the reference reward associated with the electrical energy store, including by a comparison of the actual reward and the reference reward;

determining, with the one or more controllers of the vehicle, based on the actual reward associated with the electrical energy store and the reference reward associated with the electrical energy store, whether the electrical energy store is impaired; and

outputting, with the one or more controllers, advice relating to the electrical energy store when it is determined that the electrical energy store is impaired

wherein the one or more controllers learn, in a course of a teaching process for the machine-learning energy management system, on the basis of rewards associated with the electrical energy store that have been obtained for different combinations of states and actions with respect to the electrical energy store in the course of the teaching process for the machine-learning energy management system.