Duty cycle independent real time dynamic programming implementation for hybrid systems controls optimization

• US 8,909,507 B2
• Filed: 01/31/2011
• Issued: 12/09/2014
• Est. Priority Date: 11/23/2010
• Status: Active Grant

• Current Assignee: Eaton Intelligent Power Limited (Eaton Corporation PLC.)
• Original Assignee: Eaton Corp. (Wisconsin) (Eaton Corporation PLC.)
• Inventors: Tsourapas, Vasilios, Patil, Chinmaya B.
• Primary Examiner(s): PATEL, SHAMBHAVI K
• Application Number:
  US13/018,088
• Publication Number:
  US 20120130695A1
• Time in Patent Office: 1,408 Days
• Field of Search: None
• US Class Current: 703/8
• CPC Class Codes: G05B 13/048 using a predictor
• CPC Class Codes: G06F 11/261 by simulating additional hardware, e.g. fault simulation
• CPC Class Codes: G06F 13/105 where the programme performs an input/output emulation function
• CPC Class Codes: G06F 30/00 Computer-aided design [CAD]

Abstract

Predicted environment conditions are received for time steps within a prediction horizon N. Information is accessed regarding expected system responses generated by a system model, each expected system response related to a different combination of control parameters and starting environment conditions applied to the model.

An optimal response is designated for each combination of starting environment conditions for each time step from time step N−1 to the current time step based in part on the expected system response and the predicted environment conditions, each optimal response in a time step based at least in part on an optimal response determined for a later time step.

For the current time step, for starting environment conditions substantially similar to the current environment conditions, the combination of control parameters that results in the system response designated as the optimal response is determined from the expected system responses and is applied to the system.

Claims

1. A method, comprising:

receiving from a prediction mechanism predicted environment conditions for a time step N within a prediction horizon;

accessing from a data store information regarding a first set of expected system responses generated by a physical system model provided with a first combination of starting environment conditions, each expected system response generated by the model being a response to applying a different combination of control parameter states to the model;

for time step N−

1, for each expected system response in the first set of expected system responses,

accessing a response cost and a response trajectory from the information regarding the first set of expected system responses;

determining an endpoint for the response trajectory at time step N, the response trajectory starting from time step N−

1;

determining an error cost for the response trajectory related to the difference between the environment conditions at the endpoint of the response trajectory and the predicted environment conditions for time step N; and

calculating a total cost for the expected system response by adding the response cost and the error cost; and

designating the expected system response of the first set of expected system responses with minimum total cost as the first optimal response for time step N−

1, the first optimal response related to the first combination of starting environment conditions and the predicted time step N environment conditions.

2. The method of claim 1, further comprising:

storing in the data store information regarding the first optimal response for time step N−

1, including information regarding the total cost of the first optimal response for time step N−

1, the first combination of starting environment conditions, and the combination of control parameter states that caused the model to generate the system response designated as the first optimal response for time step N−

1.

3. The method of claim 2, further comprising:

accessing from the data store information regarding a second set of expected system responses generated by the system model, the model being provided with a second combination of starting environment conditions, and each expected system response generated by the model being a response to a different combination of control parameter states being applied to the model;

for time step N−

1, for each expected system response in the second set of expected system responses,

accessing a response cost and a response trajectory from the information regarding the first set of expected system responses;

determining an endpoint for the response trajectory at time step N, the response trajectory starting from time step N−

1;

determining an error cost for the response trajectory related to the difference between the environment conditions at the endpoint of the response trajectory and the predicted environment conditions for time step N; and

calculating a total cost for the expected system response by adding the response cost and the error cost;

designating the expected system response of the second set of expected system responses with minimum total cost as the second optimal response for time step N−

1, the second optimal response related to the second combination of starting environment conditions and the predicted time step N environment conditions; and

storing in the data store information regarding the second optimal response for time step N−

1, including information regarding the total cost of the second optimal response for time step N−

1, the second combination of starting environment conditions, and the combination of control parameter states that caused the model to generate the system response designated as the second optimal response for time step N−

1.

4. The method of claim 3, further comprising:

determining a set of M optimal responses for time step N−

1, each optimal response relating to one combination of starting environment conditions, the set of M optimal responses relating to the predicted environment conditions at time step N, the set of M optimal responses including the first optimal response for time step N−

1 and the second optimal response for time step N−

1.

5. The method of claim 4, further comprising:

receiving from a prediction mechanism predicted environment conditions for a time step N−

1 within the prediction horizon;

for each expected system response in the first set of expected system responses,

accessing the response cost and the response trajectory;

determining an endpoint for the response trajectory at time step N−

1, the response trajectory starting from time step N−

2;

determining an error cost for the response trajectory related to the difference between the environment conditions at the endpoint of the response trajectory and the predicted environment conditions for time step N−

1;

determining from the set of M optimal responses for time step N−

1 the combination of starting environment conditions nearest to the response trajectory endpoint environment conditions;

accessing the total cost of the optimal response for time step N−

1 related to the nearest combination of starting environment conditions; and

calculating a total recursive cost for the expected system response by adding the response cost, the error cost, and the total cost of the optimal response for time step N−

1 related to the nearest combination of starting environment conditions;

designating the expected system response of the first set of expected system responses with minimum total recursive cost as the first optimal response for time step N−

2, the first optimal response related to the first combination of starting environment conditions and the predicted time step N environment conditions; and

storing in the data store information regarding the first optimal response for time step N−

2, including information regarding the total recursive cost of the first optimal response for time step N−

2 and the combination of control parameter states and starting environment conditions that caused the model to generate the system response designated as the first optimal response for time step N−

2.

6. The method of claim 5, further comprising:

determining a set of K optimal responses for time step N−

2, each optimal response relating to one combination of starting environment conditions, the set of K optimal responses relating to the predicted environment conditions at time step N, the set of K optimal responses including the first optimal response for time step N−

2.

7. The method of claim 6, further comprising:

determining a set of optimal responses for each time step N−

3 to the current time step relating to predicted environment conditions of time steps to time step N; and

applying to the system the combination of control parameters indicated by the optimal response for the current time step for the current environment conditions.

8. The method of claim 1, the system model being a model of the structure of a vehicle, at least a portion of the response cost being fuel consumed, the environment conditions including speed, and an expected system response trajectory being acceleration.

9. The method of claim 8, the vehicle being a hybrid vehicle with at least two power sources.

10. The method of claim 9, the at least two power sources including at least two from a group of power sources including internal combustion engine, electric motor, hydraulic motor, and fuel cell.

11. The method of claim 9, a control parameter being applied torque.

12. The method of claim 8, an endpoint for the response trajectory at time step N being an expected speed resulting from the acceleration from time step N−

1, and at least a portion of the error cost being a difference between the expected speed resulting from the acceleration from time step N−

1 and the predicted speed for time step N from the prediction mechanism.

13. A system, comprising:

a prediction mechanism predicting environment conditions for time steps up to a time step N within a time horizon;

a data store including information regarding a first set of expected system responses generated by a system model where each expected system response has a response trajectory, the model being provided with a first combination of starting environment conditions, and each expected system response generated by the model being a response to a different combination of control parameter states being applied to the model;

a computing device including at least one processor, the computer device determining the expected system response of the first set of expected system responses with a minimum total cost as the first optimal response for time step N−

1 where the minimum total cost is based at least in part on the expected system response trajectories starting from time step N−

1 and having an endpoint at time step N, the first optimal response related to the first combination of starting environment conditions and predicted environment conditions for time step N.

14. The system of claim 13, further comprising:

the computing device determining a set of M optimal responses for time step N−

1, each optimal response relating to one combination of starting environment conditions, the set of M optimal responses relating to the predicted environment conditions at time step N, the set of M optimal responses including the first optimal response for time step N−

1.

15. The system of claim 14, further comprising:

the computing device determining a set of optimal responses for each time step N−

2 to the current time step relating to the predicted environment conditions to time step N and applying to the system the combination of control parameters indicated by the optimal response for the current time step for the current environment conditions.

16. The system of claim 13, the system model being a model of the structure of a vehicle, at least a portion of the response cost being fuel consumed, the environment conditions including speed, and the expected system response trajectory being acceleration.

17. The system of claim 16, the vehicle being a hybrid vehicle with at least two power sources.

18. The system of claim 17, the at least two power sources including at least two from a group of power sources including internal combustion engine, electric motor, hydraulic motor, and fuel cell.

19. The system of claim 17, a control parameter being applied torque.

20. The system of claim 16, the endpoint for the response trajectory at time step N being an expected speed resulting from the acceleration from time step N−

1, and at least a portion of the error cost being a difference between the expected speed resulting from the acceleration from time step N−

1 and the predicted speed for time step N from the prediction mechanism.

21. A method, comprising:

receiving from a prediction mechanism predicted environment conditions for time steps up to a time step N within a prediction horizon;

accessing from a data store information regarding expected system responses generated by a system model, each expected system response having a response trajectory, and each expected system response being a response to a different combination of control parameters and starting environment conditions being applied to the model;

designating an optimal response for each combination of starting environment conditions for each time step within the prediction horizon from time step N−

1 to the current time step, each optimal response based in part on the expected system response and the predicted environment conditions, each optimal response in a time step based at least in part on an optimal response determined for a later time step and the expected system response trajectories starting from the time step to the later time step;

identifying the combination of control parameters causing the model to generate the system response designated as the optimal response at the current time step, the starting environment conditions of the model similar to current environment conditions; and

applying the identified control parameters to the system.

22. The method of claim 21, the system model being a model of the structure of a vehicle, and an environment condition being speed.

23. The method of claim 22, the vehicle being a hybrid vehicle with an internal combustion engine and an electric motor, and a control parameter being applied torque.

24. The method of claim 22, an environment condition being battery state of charge.