Apparatus and method for estimating battery state of charge

1. A method for training a support vector machine to determine a present state of charge of an electrochemical cell system comprising:

choosing a training data;

preprocessing the training data;

finding an optimal parameter of the support vector machine; and

determining support vectors.

2. The method of claim 1 , wherein the choosing a training data comprises:

obtaining the training data which is different from a test data;

obtaining the training data which represents a continuous flow of a measured battery data;

obtaining the training data covering an expected range of a state of charge of the electrochemical cell;

obtaining the training data covering an expected range of a current of the electrochemical cell; and

obtaining the training data covering a range of a voltage value of the electrochemical cell corresponding to the expected range of current of the electrochemical cell and the expected range of state of charge of the electrochemical cell.

3. The method of claim 1 , wherein the preprocessing the training data comprises:

adding an element to the training data; and

scaling the training data.

4. The method of claim 3 , wherein the element is a change in a voltage of the electrochemical cell during a last one second of a measuring period.

5. The method of claim 3 , wherein the scaling the training data comprises converting each training datum of the training data to a scaled value between 0.0 and 1.0, each the training datum includes a support vector, the support vector includes at least one of:

an instantaneous current;

an instantaneous voltage;

a calculated state of charge from the last one second of the measuring period; and

the change in a voltage of the electrochemical cell during the last one second of the measuring period.

6. The method of claim 1 , wherein the finding the optimal parameter of the support vector machine comprises:

optimizing a variable of a training function.

7. The method of claim 6 , wherein the optimizing the variable further comprises minimizing a loss function under a constraint.

8. The method of claim 7 , wherein the variable includes at least one of a first variable α i and a second variable α i *, the training function is

f

⁡

(

x

)

=

∑

i

=

1

N

⁢

(

α

i

-

α

i

*

)

⁢

⁢

K

⁡

(

x

i

,

x

)

,

K

is a Kernel function K(a,b)=s×(a·b) 2 +r, a and b are vectors, and x is a support vector.

9. The method of claim 7 , wherein the loss function is

L

ε

⁡

(

y

,

f

⁡

(

x

)

)

=

{



y

-

f

⁡

(

x

)



-

ɛ

for

⁢

⁢



y

-

f

⁡

(

x

)



≥

ɛ

0

otherwise

.

10. The method of claim 9 , wherein the constraint is at least one of

∑

i

=

1

N

⁢

(

α

i

-

α

i

*

)

=

0

,

0

≤

α

i

≤

C

,

and

⁢

⁢

0

≤

α

i

*

≤

C

.

11. A method for estimating a present state of charge of an electrochemical cell using a trained support vector machine, the method comprising:

preprocessing test data; and

testing an optimized support vector machine using the test data.

12. The method of claim 11 , wherein the preprocessing test data comprises

adding an element to the test data; and

scaling the test data, the scaling the test data producing a scaled test data.

13. The method of claim 12 , wherein the element is a change in a voltage of the electrochemical cell during a last one second of a measuring period.

14. The method of claim 12 , wherein the testing the test data using the trained support vector machine comprises applying the scaled test data to a test function, the test function providing an estimate of the present state of charge of the electrochemical cell.

15. The method of claim 11 , wherein the scaling the test data comprises converting a test datum of the test data to a scaled value between 0.0 and 1.0, each the test datum includes a support vector, the support vector includes:

an instantaneous current;

an instantaneous voltage; and

the change in a voltage of the electrochemical cell during the last one second of the measuring period.

16. A method for estimating a present state of charge of an electrochemical cell system comprising:

training a support vector machine for a state of charge estimation;

testing the support vector machine; and

calculating an estimate of the present state of charge of the electrochemical cell using the support vector machine.

17. The method of claim 16 , wherein the training the support vector machine comprises:

choosing a training data;

preprocessing the training data; and

finding an optimal parameter of the support vector machine.

18. The method of claim 17 , wherein the choosing a training data comprises:

obtaining the training data which is different from a test data;

obtaining the training data which represents a continuous flow of a measured battery data;

obtaining the training data covering an expected range of a state of charge of the electrochemical cell;

obtaining the training data covering an expected range of a current of the electrochemical cell; and

obtaining the training data covering a range of a voltage value of the electrochemical cell corresponding to the expected range of current and the expected range of state of charge of the electrochemical cell.

19. The method of claim 17 , wherein the preprocessing the training data comprises:

adding an element to the training data, the element being a change in a voltage of the electrochemical cell during a last one second of a measuring period; and

scaling the training data.

20. The method of claim 19 , wherein the scaling the training data comprises converting a training datum of the training data to a scaled value between 0.0 and 1.0, each of the training datum includes a support vector, the support vector includes:

an instantaneous current;

an instantaneous voltage;

a calculated state of charge from the last one second of the measuring period; and

the change in the voltage of the electrochemical cell during the last one second of the measuring period.

21. The method of claim 17 , wherein the finding the optimal parameter of the support vector machine comprises:

optimizing a variable of a training function.

22. The method of claim 21 , wherein the optimizing the variable comprises minimizing a loss function under a constraint.

23. The method of claim 16 , wherein the testing the support vector machine comprises:

preprocessing test data; and

testing the support vector machine using the test data.

24. The method of claim 23 , wherein the preprocessing test data comprises:

adding an element to the test data, the element being a change in a voltage of the electrochemical cell during a last one second of a measuring period; and

scaling the test data, the scaling the test data producing a scaled test data.

25. The method of claim 24 , wherein the scaling the test data comprises converting a test datum of the test data to a scaled value between 0.0 and 1.0, each the test datum includes a support vector corresponding to a specific time, the support vector includes at least one of:

an instantaneous current;

an instantaneous voltage; and

the change in the voltage of the electrochemical cell during the last one second of the measuring period.

26. The method of claim 24 , wherein the testing the test data comprises applying the scaled test data to a test function, the test function providing the estimate of the present state of charge of the electrochemical cell.

27. An apparatus for estimating the state of charge of an electrochemical cell, the apparatus comprising:

a sensing component configured to measure a parameter of the electrochemical cell; and

a trained component configured to estimate the state of charge of the electrochemical cell, wherein the trained component is in electrical communication with the sensing component, and the trained component is a support vector machine.

28. The apparatus of claim 27 , wherein the sensing component configured to measure the parameter of the electrochemical cell comprises at least one of:

a voltage sensor to measure a voltage of the electrochemical cell; and

a current sensor to measure a current of the electrochemical cell.

29. The apparatus of claim 28 , wherein the trained component comprises an arithmetic circuit configured to receive an instruction, the instruction enabling the trained component to find an optimal variable of a first function when a second function is minimized under a constraint.

30. The apparatus of claim 29 , wherein the trained component further comprises the arithmetic circuit configured to estimate the state of charge of the electrochemical cell via a third function.

31. The apparatus of claim 29 , wherein the first function is

f

⁡

(

x

)

=

∑

i

=

1

N

⁢

(

α

i

-

α

i

*

)

⁢

⁢

K

⁡

(

x

i

,

x

)

,

wherein K is a Kernel function K(a,b)=s×(a·b) 2 +r, a and bare vectors, x is a support vector, and the optimal variable is at least one of α i and α i *.

32. The apparatus of claim 29 , wherein the second function is

L

ε

⁡

(

y

,

f

⁡

(

x

)

)

=

{



y

-

f

⁡

(

x

)



-

ɛ

for

⁢

⁢



y

-

f

⁡

(

x

)



≥

ɛ

0

otherwise

and the constraint is at least one of

∑

i

=

1

N

⁢

(

α

i

-

α

i

*

)

=

0

,

0

≤

α

i

≤

C

,

and

⁢

⁢

0

≤

α

i

*

≤

C

.