View
4
Download
0
Category
Preview:
Citation preview
PWT Embedded EV | January 2018
Renault Nissan Confidential C
Battery Parameter Estimation
Presented by
Shanmugam,Thayalan
Udayakumar, Praveenkumar
Nissan Leaf
2Renault Nissan Confidential C
Established in 2007
More than 6,300 employees
Based in Chennai, India
R-N’s only Alliance R&D center
Competitive alliance center
RNTBCI – Brief Introduction
ENGINEERING• Product Engineering
• Production Engineering
• Research and Advanced
Engineering
IS/IT• Information System/
Information Technology
3Renault Nissan Confidential C
OBJECTIVE
BATTERY TESTS
BATTERY MODEL
BATTERY DYNAMIC CHARACTERISTICS
METHODOLOGY & OPTIMIZATION
DATA REDUCTION
BENEFITS
4Renault Nissan Confidential C
OBJECTIVEFactors that influence the capacity
▪ Rate of charge and discharge
▪ Temperature
▪ Self discharge (internal resistance)
▪ Cycle number
▪ Over charge
Proper battery model is needed
• Robust testing of BMS at HIL.
• Developing Proper controlling algorithm
(ASW).
• Analyzing battery performance w.r.t
different chemistry
• To prevent accelerated aging effect
5Renault Nissan Confidential C
Pulse Current Discharge Test Hybrid Pulse Power Characterization Test
BATTERY TESTING
To check the aging effect, degradation and state of health
Batteries under go some current tests
▪ Pulse Current Discharge
▪ Hybrid Pulse Power Characterization
▪ Static Capacity
▪ Self Discharge
▪ Energy Efficiency
▪ Cold Cranking
6Renault Nissan Confidential C
Batteries
• Non linear Systems
DIFFERENT TYPES OF MODELS
✓ Electrical models
▪Thermal models
▪Electrochemical models
▪Interdisciplinary
models(electro – thermal)
Real-time determination of the parameter is challenging
EQUIVALENT CIRCUIT BASED MODELS
i. Thevenin equivalent model
ii. First, second and third order model
iii. Linear electric model
iv. Non-linear electrical model
v. Impedance based models
BATTERY MODEL
Second order model is used
v
iii
iv
7Renault Nissan Confidential C
What are parameters?
Instantaneous ResponseDue to ohmic effects
Delayed Response
𝑅0=𝑉0
𝐼𝑑𝑖𝑠𝑐ℎ𝑎𝑟𝑔𝑒
𝑅1 =ሻ𝑉2(𝑡3
1 − 𝑒−𝑇𝑐ℎ𝑎𝑟𝑔𝑒
𝜏 ∗ 𝐼𝑐ℎ𝑎𝑟𝑔𝑒
𝜏 =RC
𝐶1 =𝜏
𝑅1
Internal Resistance
Parallel Resistance
Parallel Capacitance
DYNAMIC CHARACTERISTICS OF A BATTERY
How does parameters effect in battery voltage?
Figure: 1
Figure: 2
Due to charge transfer
Due to
diffusion
8Renault Nissan Confidential C
Ucell representation using 2 RC
Only Urc curve fitted using fmincon function from Matlab
fmincon
HPPC data
Open Circuit Voltage
Voltage due to
Internal Resistance
Characteristics
remaining
9Renault Nissan Confidential C
OPTIMIZATIONfmincon(FUN,X,A,B,Aeq,Beq,LB,UB,NONLCON,options,varargin)
Objective func
Constraint func
Fmincon solver
10Renault Nissan Confidential C
METHODOLOGY PARAMETER ESTIMATION PARAMETER UPDATION
Estimator used here is OPTIMIZATION TOOL
Estimation method Fmincon
➢ Parameters (Ro, Rn, Cn) depends on the
Input current, Capacity, SOC and
Temperature
Objective func
Constraint func
Fmincon solver
SOC
CurrentSOH
T1
T2
T3
Ro,Rn,Cn
Assumption
• No Self discharge
• R1,C1 and R2, C2 changes w.r.t SOC in
a pulse ON and assumed throughout
estimation
• 1RC parameters derived from 2 RC
components.
• Over/Under voltage pulses are
eliminated.
Equivalent Circuit Modelling is the most
common approach for battery analysis
Parameters
• Series Resistance (Ri)
• Parallel Resistance (Rn)
• Parallel Capacitance (Cn)
11Renault Nissan Confidential C
WORK FLOW
Parameter Estimation
Ro R1
C1
Drive Cycle
0 2000 4000 6000 8000 10000 12000 140003
3.5
4
4.5
in [S]
in [
V]
Ucell
Actual
Simulated
0 2000 4000 6000 8000 10000 12000 14000-0.1
0
0.1
in [S]
in [
V]
Ucell diff (Actual-Simulation)
Ucell Difference
Lower limit(-50mV)
Upper limit(50mV)
0 2000 4000 6000 8000 10000 12000 14000-200
-100
0
100
in [S]
in [
A]
Ibat
Ibat
0 2000 4000 6000 8000 10000 12000 140000
50
100
in [S]
in [
%]
SOC inst
SOC
0 2000 4000 6000 8000 10000 12000 140000
50
100
in [S]
in [
C]
Temperature
Temperature
Current
HPPC
SOC OCV
Drive CyclesBattery Plant model
SOC
CurrentSOH
T1
T2
T3
Ro,Rn,Cn
12Renault Nissan Confidential C
Data reduction Centered Moving Avg filtering
100 pnt
250 pnt
Remove non boundary
values
Out layer rejection
Raw data
Averaging filter
Out layer rejection
Normal distribution
Remove : > Three standard
distribution (99.73%) 3σ
99.73%
Raw Data vs Reduced Data
Ucell difference
Ucell Difference - Averaged
SOC
CurrentSOH
T1
T2
T3
Raw Data
Ro,Rn,Cn
0
1
13Renault Nissan Confidential C
BENEFITS
Less Complex Model
Faster Parameter Estimation
95% of accuracy in tuning parameters
Effective capturing of nonlinear effects
No manual tuning of parameters required for validation
Battery Plant model
Plant model with the estimated
parameters are used in the HIL test
bench
75%
95%
80%
methods
Co
nve
rge
nc
e,
tim
e, a
cc
ura
cy
System
identification
Lsq curve fit fmincon
1 R
C
2 R
C
1 R
C
14Renault Nissan Confidential C
Battery Pack Generation
MATLAB ASSISTANCE➢ Helped in handling a huge data
(>10GB)
➢ Easy to analyze data
➢ Make robust automation
➢ Estimated results used in HIL/MIL
➢ Reduced data are used ASW
calibration
Automated
Battery Pack Generation (Simulink & m-script)
Battery Test Bench
Excel : Data Collecting
Data Extraction : Pulses & Validation Data (m-script)
Valid Pulse Check (data validity) (m-script)
Estimation of Parameters (Optimization toolbox)
Data Reduction (statistics toolbox )
Validation with Plant Model (Simulink:-Simscape)
Battery Pack Comparison with HV Test Bench (m-script)
15Renault Nissan Confidential C
• “Battery testing procedures” - Maciej Swierczynski
• “How to model a battery, is a source and a resistor enough?” - J.M. Timmermans
• https://www.google.fr/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&cad=rja&uact=8&ved=0ahU
KEwjGpamq2t7YAhUGYlAKHYw8BtEQFggnMAA&url=https%3A%2F%2Fetd.ohiolink.edu%2F!etd
.send_file%3Faccession%3Dosu1306937891%26disposition%3Dinline&usg=AOvVaw3o10OzmcB
Ne7YSk33hHkF_
REFERENCES
16Renault Nissan Confidential C
17Renault Nissan Confidential C
R N
THANK YOU
GO GREEN
Recommended