You are on page 1of 60

# Fuzzy Logic-Based Battery SOC/SOH Estimation Methods

## Pritpal Singh, Ph.D.

Dept. of Electrical and Computer Engineering Villanova University Villanova, PA 19085 singh@ece.vill.edu

## David Reisner, Ph.D.

US Nanocorp, Inc. 74 Batterson Park Rd. Farmington, CT 06032 dreisner@usnanocorp.com
MIT Presentation, December 5, 2002 1

Outline
Definitions of SOC/SOH Overview of SOC/SOH Measurement Techniques Introduction to Fuzzy Logic Examples of Applications of Fuzzy Logic to Battery SOC/SOH Estimation Fuzzy Logic-Based SOC/SOH Measurement Hardware

## MIT Presentation, December 5, 2002

SOC/SOH Definitions

## MIT Presentation, December 5, 2002

State-of-Charge (SOC)
SOC is a percentage of the available capacity. However, available capacity depends on temperature, discharge rate, battery design, etc. By the term available capacity do we mean nominal capacity, theoretical capacity, available capacity considering aging of the battery ? Is the available capacity being considered as the battery is being charged or discharged ?

## MIT Presentation, December 5, 2002

State-of-Health (SOH)
SOH or state-of-condition generally means the ability of a battery to perform the task required for the application. SOH can be defined by the capacity available from the battery (in the UPS/Telecom industry). SOH may be defined by the power delivery capability of the battery (EVs, HEVs, some medical devices). If a battery is not in a good SOH, what is the failing condition? Is this a recoverable failure mechanism?
MIT Presentation, December 5, 2002 5

## SOC/SOH Measurement Methods

Coulomb Counting - widely used in portable
products, e.g. laptop computers (and in electric vehicles). (primarily SOC)

## Transient Voltage Response - commonly used

in testing of UPS and EV batteries (SOC/SOH)

## Impedance Measurements - not widely

employed because of difficulty in analyzing impedance data (used in Midtronics automotive battery testers). (SOC/SOH)

## Fuzzy Logic vs. Aristotelian Logic

In Aristotelian logic, a quantity is either a member of a set or is not a member of a set. The set has sharp (or crisp) boundaries. In fuzzy logic, a quantity may be a member of a set to some degree or not be a member of a set to some degree. The boundaries of the set are fuzzy rather than crisp.
MIT Presentation, December 5, 2002 9

Fuzzy Systems
A fuzzy system is a rule-based mapping of inputs to outputs for a system. It can be theoretically proven that a fuzzy system is a universal approximator*.

## see Fuzzy Engineering by Bart Kosko

MIT Presentation, December 5, 2002 10

11

## Complete Fuzzy System

Knowledge Database

12

## MIT Presentation, December 5, 2002

13

Fuzzy Rules
If temperature is WARM and Discharge Rate is LOW then available capacity is HIGH If temperature is COLD and Discharge Rate is HIGH then available capacity is LOW

14

15

## MIT Presentation, December 5, 2002

16

Mamdani Approach
Most commonly used approach to developing fuzzy logic models for control applications. Uses expert knowledge to generate rule set. Uses membership functions for both input and output variables. Computationally intensive compared to Sugeno approach.
MIT Presentation, December 5, 2002 17

Sugeno Approach
Output membership functions are singletons (zero order) or polynomials (first order). The rule in a first order Sugeno model may be expressed as: if x is A and y is B then z=p*x+q*y+r where p, q, and r are constants. Computationally efficient. Well suited to optimization/adaptation.
MIT Presentation, December 5, 2002 18

## Example: Two input, two rule Fuzzy Model

n1 m1 F1 A1 S1 m2 n2 F2 A2 S2
MIT Presentation, December 5, 2002 19

Rule1

Rule2

## Techniques for Inference

Mamdani Implication: F1 = min (m1, n1) F2 = min (m2, n2) Defuzzification Conversion: Weighted Average
S S iAiFi i y s t e u m =t O= 1n u t p AiFi
i=1

A1, A2: areas of output membership functions S1, S2: singletons of output membership functions.
MIT Presentation, December 5, 2002 20

Supervised Learning
In supervised learning, an initial set of membership functions and rules are generated. The model is then optimized using neural network algorithms (e.g. back propagation) to minimize the error between training data and model-generated data.

## MIT Presentation, December 5, 2002

21

Unsupervised Learning
In unsupervised learning, the initial set of membership functions and rule set are selfgenerated using clustering algorithms. Optimization can then be performed using neural network algorithms.

## MIT Presentation, December 5, 2002

22

Example of using Coulomb Counting and Fuzzy Logic to Estimate Battery SOC

23

## Coulomb Counting Approach for Primary Li/SO2 Cells

D c ag e ie c o Pimr L O c lls is h r e ffic ny f r ay i/S 2 e 1 0 .9 0 .8 D isch arge efficien cy 0 .7 0 .6 0 .5 0 .4 0 .3 0 .2 0 .1 0 0 2C 0 0C - 0C 2 - 0C 3 - 0C 4 1 2 3 4 D c ag r te( A m2 is h r e a m/c ^ ) 5 6

24

## Mamdani Model - Membership Function for Current

L w L wMd m o o _ e iu MMmmH h e iu iu _ ig de d Hh ig V r _ ig ey H h

D egree of m bersh em ip

0 .8

0 .6

0 .4

0 .2

0 0 1 2 3 4 c re t ur n 5 6 7 8

25

## Mamdani Model - Membership Function for Temperature

vr _ o ey c ld c ld o co ol m ild

D egree of m bersh em ip

0 .8

0 .6

0 .4

0 .2

0 -0 4 -0 3 -0 2 -0 1 te p ra r me tue 0 1 0 2 0

26

## Mamdani Model - Membership Function for Discharge Efficiency

Vr - o eyL w lo w L wMdmdd a_o t_ ig H h o _ e iueem lmig h h ig miuiu H h M m s

D egree of m bersh em ip

0 .8

0 .6

0 .4

0 .2

0 0 2 0 4 0 e ie c ffic ny 6 0 8 0 10 0

27

## Model-Predicted Discharge Efficiency vs. Measured Discharge Efficiency

D c ag e ie c o Pimr L O c lls is h r e ffic ny f r ay i/S 2 e 1 0 .9 0 .8 D isch arge efficien cy 0 .7 0 .6 0 .5 0 .4 0 .3 0 .2 0 .1 0 0 2C 0 0C - 0C 2 - 0C 3 - 0C 4 1 2 3 4 D c a er te(m/c ^ ) is h rg a A m2 5 6

28

## Response Surface Associated with Fuzzy Logic Model of Li/SO2 Cells

80 70 60 efficiency 50 40 20 0 -20 temperature -40 0 4 2 current 8 6

29

## Membership Functions for Current and Temperature for Sugeno Model 1

i n 1 m f 1 i n 1 m f 2 i n 1 m f 3 i n 2 m f 1 i n 2 m f 2 i n 2 m f 3 i n 2 m f 4 i n 2 m f 5

1 Do bp e fe s g mh r e m e e r i

1 Do bp e fe s g mh r e m e e r i

0 . 8

0 . 8

0 . 6

0 . 6

0 . 4

0 . 4

0 . 2

0 . 2

0 1 2 3 D ia s c h a r^ g e R t e ( m A / c m 2 ) 4 5

0 4 0 3 0 2 0 1 0 T e m p e r a t u r e 0 1 0 2 0

## MIT Presentation, December 5, 2002

30

Model-Predicted Discharge Efficiency vs. Measured Disharge Efficiency for Sugeno Model 1
1 0 .9 0 .8 D isch arge efficien cy 0 .7 0 .6 0 .5 0 .4 0 .3 0 .2 0 .1 0 0 xc e kd ta hc a otr ind ta a a s .er 7 8 e4 q r.= .0 5 1 2 3 4 D c ag r te(m/c ^ ) is h r e a A m2 5 6

31

## Response Surface for Sugeno Model 1

1.5 1 Efficiency 0.5 20 0 -20 Temperature(C) -40 1 2 3 4 5

Discharge Rate(mAmps/cm^2)

32

## Membership Functions for Optimized Sugeno Model

in m 1 f2 in m in m 1 f 1 f4 1 in m in m 1 f 1 f3 5

inm 2f 3 1

inm 2f 5 2

inm 2f 4

0 .6

D g e o m m e ip e re f e b rsh
1 2 3 D c ag Rte( A m2 is h r e a m /c ^ ) 4 5

D egree of m bersh em ip

0 .8

08 .

06 .

0 .4

04 .

0 .2

02 .

0 -0 4 -0 3 -0 2 -0 1 Tmeaue( ) e pr t r C 0 1 0 2 0

33

## Discharge Efficiency Curves for Optimized Sugeno Model

1 0 .9 0 .8 D isch arge efficien cy 0 .7 0 .6 0 .5 0 .4 0 .3 0 .2 0 .1 0 0 xc e kd ta hc a otr ind ta a a s .er 1 .0 q r.= 5 e4 1 2 3 4 D c ag r te( A m2 is h r e a m/c ^ ) 5 6

34

## Response Surface of Optimized Sugeno Model

0 .8 0 .7 0 .6 0 .5 0 .4 2 0 0 -0 2 T mea r ( ) e p r tue C -0 4 1 2 3 4 5

## D isch arge Efficien cy

D c ag Rte( A m2 is h r e a m/c ^ )

## MIT Presentation, December 5, 2002

35

Examples of Combining Impedance Measurements and Fuzzy Logic to Estimate Battery SOC/SOH

## MIT Presentation, December 5, 2002

36

Impedance Parameters
Applied small signal AC voltage: v(t)= vmaxsin(t) (=2f) Current response of cell: i(t) = imaxsin(t- ) Impedance of cell: Z= v(t) i(t) = Z + jZ = (vmax/imax)ej and tan = Z/Z
MIT Presentation, December 5, 2002 37

## Typical Nyquist Plot (Z vs. Z) for an Electrochemical Cell

Z Diffusion Capacitive behavior

## Anode Cathode Rs 0 1kHz 100Hz inductive tail

MIT Presentation, December 5, 2002

10 mHz

Inductive behavior

Z
38

## Equivalent Circuit for this Cell

Ranode RS Rcathode L

Canode

Ccathode

39

## How can Impedance be used to Determine Battery SOC ?

1. Find a single frequency at which cell impedance varies linearly with battery SOC (e.g. Midtronics tester). 2. a) Measure cell impedance at many frequencies b) Derive cell equivalent circuit parameters c) Find an equivalent circuit parameter that varies linearly with battery SOC. 3. Use intelligent inspection of cell impedance at one or two frequencies to find a measureable parameter that varies monotonically with battery SOC (New fuzzy logic-enhanced method).
MIT Presentation, December 5, 2002 40

## EIS Measurements on Pb Acid Batteries

Spirally-wound, Hawker, 2.5 Ah D-size cells measured EIS measured over a frequency range 1 Hz65kHz, potentiostatically, 10mVamplitude signal Discharged at 1A for 15 minutes; 1 min. rest at open circuit before measurement; continued until 1.69V cutoff voltage reached.
MIT Presentation, December 5, 2002 41

## Bode plot of |Z| vs. frequency

0.07
A 1pred is.z A _015m 1d1 .z A _030m 1d1 .z A _045m 1d1 .z A _060m 1d1 .z A _075m 1d1 .z A _090m 1d1 .z A _105m 1d1 .z A _120m 1d1 .z A _135m 1d1 .z

0.06

0.05

|Z|
0.04 0.03 0.02 10 0

10 1

10 2

1 3 0

10 4

F requency (H z)

42

## Bode plot of Phase Angle vs. frequency

-7 5 -5 0
A p d .z 1 re is Ad _ 1 m 1 1 0 5 .z Ad _ 3 m 1 1 0 0 .z Ad _ 4 m 1 1 0 5 .z Ad _ 6 m 1 1 0 0 .z Ad _ 7 m 1 1 0 5 .z Ad _ 9 m 1 1 0 0 .z Ad _ 0 m 1 1 1 5 .z Ad _ 2 m 1 1 1 0 .z Ad _ 3 m 1 1 1 5 .z

theta

-2 5

2 5

1 0 0

1 1 0

1 2 0

1 3 0

1 4 0

F q e c (H ) re u n y z

43

## Fuzzy Logic Model

3-input, 1 output model Input 1 - |Z| @ 10 Hz; input 2 - |Z| @ 100Hz; input 3 - @ 10 Hz Fuzzy Logic Model has 6 Rules Each of the three inputs has 5 membership functions

44

## Accuracy of Fuzzy Logic Model

2 0

A c t u a l S O C - P re d ic t e d S O C

## % Error between model-predicted and measured battery SOC

1 5

1 0

-5

-1 0

-1 5

-2 0 0 1 0 2 0 3 0 4 0 5 0 6 0 7 0 8 0 9 0 1 0 0

A c t u a l S t a t e - o f- C h a r g e

Battery State-of-Charge
MIT Presentation, December 5, 2002 45

## Defibrillator Battery SOH

Performed (EIS) measurements
- EIS data has been collected on Medtronic 2.5 Ah, 8 V,
Lead Acid batteries at various SOCs throughout the cycle life of the batteries.

## Developed Fuzzy Logic Model

- Based on measured impedance data and cycle number,
a 3-input, 1-output fuzzy logic model has been developed that predicts the number of remaining simulated defibrillator pulses a battery can deliver to a load within +/- 1 pulse.
MIT Presentation, December 5, 2002 46

## Bode plot of |Z| vs. frequency

Medtronic 2.5 Ah, 8 V, Lead Acid, cycle 9
0 2 .1 5
I IT L N IA p lse u -1 p lse u -2 p lse u -3 p lse u -4 p lse u -5 p lse u -6 p lse u -7 p lse u -8 p lse u -9 p lse 0 u -1 p lse 1 u -1 p lse 2 u -1 p lse 3 u -1 p lse 4 u -1 p lse 5 u -1 p lse 6 u -1 p lse 7 u -1 p lse 8 u -1

0 0 .1 0

|Z |

0 7 .0 5

0 5 .0 0

0 2 .0 5

1 0 0

1 1 0

1 2 0

1 3 0

1 4 0

F ue req ncy (H z)

47

## Phase Angle vs. frequency

Medtronic 2.5 Ah, 8 V, Lead Acid, cycle 9
-4 0
IN IA IT L p ls u e1 p ls u e2 p ls u e3 p ls u e4 p ls u e5 p ls u e6 p ls u e7 p ls u e8 p ls u e9 p ls - 0 u e1 p ls - 1 u e1 p ls - 2 u e1 p ls - 3 u e1 p ls - 4 u e1 p ls - 5 u e1 p ls - 6 u e1 p ls - 7 u e1 p ls - 8 u e1

-3 0

theta

-2 0

-1 0

0 1 0 0

1 1 0

1 2 0

1 0

Fe u n y ( z r q e c H)

48

## Bode plot of |Z| vs. frequency

Medtronic 2.5 Ah, 8 V, Lead Acid, cycle 57
01 5 .2

01 0 .0

|Z |

I IT L N IA p ls -1 ue p ls -2 ue p ls -3 ue p ls -4 ue p ls -5 ue p ls -6 ue p ls -7 ue p ls -8 ue p ls -9 ue

00 5 .7

00 0 .5

00 5 .2

1 0 0

1 0

1 2 0

1 3 0

1 0

Fe u n y( z r q e c H)

49

## Bode plot of |Z| vs. frequency

Medtronic 2.5 Ah, 8 V, Lead Acid, cycle 57
-0 4
IN I L I A T p ls u e1 p ls u e2 p ls u e3 p ls u e4 p ls u e5 p ls u e6 p ls u e7 p ls u e8 p ls u e9

-0 3

theta

-0 2

-0 1

0 1 0

1 0

1 2 0

1 0

Fe u n y( z r q e c H)

50

## Fuzzy Logic Model

3-input, 1 output model Input 1 - |Z| @ 160 Hz; input 2 - @ 16Hz; input 3 - cycle number Output - number of simulated defibrillator pulses that may be delivered by the battery Model has 67 rules

51

## Accuracy of Fuzzy Logic Model

5 4 E rro r, M o d e l p re d ic t e d - A c t u a l 3

## Model Error (# of pulses)

-1

-2

-3

-4

RMSE = 0.629

-5 0 2 4 6 8 10 12 14 16 18 20

## Number of Remaining Pulses

MIT Presentation, December 5, 2002 52

53

## SOC Meter Based on Coulomb Counting

Current (analog) Load A/D A/D Temp. Battery (analog) LCD Display
MIT Presentation, December 5, 2002 54

## Software Modules for Coulomb Counting Meter

- acquires the sensed current and temperature values and stores them in memory.

## 2. Fuzzy Logic Module - calculates the discharge efficiency for the

the sensed values of current and temperature.

## 3. Coulomb Counting Module

- accumulates the charge flow

4. Display Driver
- driver software that updates the LCD display with the present value of the cell SOC

5. Runtime Module

55

56

## Advantages of Fuzzy Logicbased Battery Management

Fuzzy Logic model is very code efficient - easily embedded into an 8-bit microcontroller or even on-chip FLASH memory. Robust, accurate approach with nonlinear interpolation capability. Impedance approach may be used - especially useful for SOH of battery. Fast model development time for new battery designs. Can be integrated with Fuzzy Logic-Based charge controller
MIT Presentation, December 5, 2002 57

Bibliography
1. A.J. Salkind, C. Fennie, P. Singh, T. Atwater, and D.E. Reisner, Determination of State-of-Charge and State-of-Health of Batteries by Fuzzy Logic Methodology, J. Power Sources, 80 pp. 293-300 (1999) 2. P. Singh, C. Fennie, Jr., A. J. Salkind, and D. E. Reisner, A Fuzzy Logic Methodology to Determine State-of-Charge (SOC) in Primary and Rechargeable Batteries, Procs. 38th Power Sources Conf., Cherry Hill, NJ, June 8-11, 1998, p. 295 3. P. Singh and C. Fennie, Jr., A Method for Determining Battery State of Charge Using Intelligent Systems, U.S. Patent No. 6,011,379, Jan 4, 2000.

## MIT Presentation, December 5, 2002

58

Bibliography (contd)
4. P. Singh and C. Fennie, Jr., A Method for Determining Battery State of Health Using Intelligent Systems, Patent Pending 5. B. Kosko, Fuzzy Engineering (Prentice-Hall, Upper Saddle River, NJ, 1997). 6. J.-S.R. Jang, C.-T. Sun, and E. Mizutani, Neuro-Fuzzy and Soft Computing, Prentice -Hall, Upper Saddle River, NJ, 1997. 7. The Mathworks, Inc., (Natick, MA) www.mathworks.com (Fuzzy Logic Toolbox for Matlab)

## MIT Presentation, December 5, 2002

59

Bibliography (contd)
8. P.Singh, C. Fennie and D.E.Reisner, Fuzzy Logic Enhanced Electrochemical Impedance Spectroscopy (FLEEIS) to Determine Battery State of Charge, Procs. 15th Ann. Batt. Conf., Long Beach, CA, Jan 11-14, 2000. 9. V.R. Gaddam, X.Q. Wang, S. Arey, Z.J. Yang, P. Singh, C. Fennie, and D.E. Reisner, Fuzzy Logic-Based SOC Hardware for Li/SO2 Cells, Procs. 39th Power Sources Conf., Cherry Hill, NJ, June 12-15, 2000. 10. F. Huet, A Review of Impedance Measurements for Determination of State-of-Charge and State-of-Health of Secondary Batteries, J. Power Sources 70 pp. 59-69 (1998).
MIT Presentation, December 5, 2002 60