Method for estimating a state of charge of batteries
View Patent ↗A method estimates a state-of-charge (SoC) of a battery by constructing a set of two or more battery models. Each battery model is associated with an adaptive SOC estimator. A set of intermediate SOCs is estimated using the models and the associate adaptive SoC estimators. Then, the set of intermediate SoCs are fused to obtain a final SoC of the battery.
1. A method for estimating a state-of-charge (SoC) of a battery, comprising the steps of:
constructing a set of two or more battery models, wherein each battery model is associated with an adaptive SoC estimator;
estimating a set of intermediate SoCs using the models and the associate adaptive SoC estimators; and
fusing the set of intermediate SoCs to obtain a final SoC of the battery, wherein the method is performed in a processor.
2. The method of claim 1 , further comprising:
measuring conditions of the battery; and
determining weights for the conditions; and
fusing the set of intermediate SoCs according to the weights.
3. The method of claim 2 , wherein the conditions include voltages, currents and temperatures of the battery.
4. The method, of claim 1 , wherein the battery, is one of: a lithium (Li) ion, a nickel metal hydride (NiMH), a LiMn 2 O 4 and LiCoO 2 battery.
5. The method of claim 1 , wherein each SoC estimators determines parameters for the associated model.
6. The method of claim 1 , wherein the adaptive SoC estimator is a Kalman filter (KF) for a linear model, or an extended KF for a nonlinear model.
7. The method of claim 2 , wherein M={M 1 , M 2 , . . . , M N } denotes the set of models, an unknown SoC of the battery at time k is x k , measurement of the conditions are z k , and s k denotes a current status, and a probability of the state s k for model M i is denoted as p(s i k ), which indicates an a priori probability that a true model is M i at time k, such that a sum of the probabilities is
∑
i
=
1
N
p
(
s
k
i
)
=
1.
8. The method of claim 7 , wherein x k and z k are continuous random variables and s k is a discrete variable.
9. The method of claim 7 , further comprising:
performing a minimum-mean-square-error (MMSE) estimation or maximum a posteriori (MAP) estimation of x k .
10. The method of claim 7 , wherein μ k i =p(s) k i |Z k and w k i =p(z k |s k i , Z k ), and weights μ k i are
μ
k
i
=
w
k
i
p
(
s
k
i
)
∑
j
=
1
N
w
k
i
p
(
s
k
i
)
.
11. The method of claim 10 , wherein the fusing is according to
{circumflex over (x)} k|k =Σ i=1 N {circumflex over (x)} k|k i μ k i ,
where {circumflex over (x)} (k|k) i is an estimate of x k based on the model M i .
12. The method of claim 1 , wherein the fusing is a linear weighted combination of the intermediate SoCs.
13. The method of claim 1 , wherein the models include a set of submodels that fix parameters of the adaptive SoC estimator.
14. The method of claim 1 , wherein the models include a set of submodels by assuming different sets of values for parameters of the models.
15. The method of claim 1 , wherein the models include an equivalent-circuit model.
16. The method of claim 1 , wherein the models include an electrochemical-principles-based model.
17. The method of claim 1 , wherein the models capture different characteristics of batteries, including charging and discharging processes, cycling and aging effects, or combinations thereof.