Upload
ngothien
View
225
Download
0
Embed Size (px)
Citation preview
EMR’17
University Lille 1
June 2017
Summer School EMR’17
“Energetic Macroscopic Representation”
« EMR for a Multi-objective Benchmark for
Energy Management of Hybrid Energy
Storage in Electric Vehicles »
Bảo-Huy NGUYỄN 1,2,3, João P. TROVÃO1,
Ronan GERMAN2,3, Alain BOUSCAYROL2,3, 1 e-TESC Lab., Université de Sherbrooke, Canada
2 L2EP, Université Lille 1, France3 MEGEVH, France
EMR’17, University Lille 1, June 20172
« A multi-objective benchmark for EMS of H-ESS for EVs »
- Global context -
Tiny, tiny amount
Badly
misunderstood
our research
Am departing
presidential
councils
Climate change is real.
Even though someone tries to ignore it
Electric and hybrid vehicles are one of the keys to solve the problem
Make our planet
great again
… and EMR ?
EMR’17, University Lille 1, June 20173
« A multi-objective benchmark for EMS of H-ESS for EVs »
- Outline -
1. Context and objective
2. EMR for problem formulation
3. Multi-objective energy management
4. Results and discussion
5. Conclusion and perspective
EMR’17
University Lille 1
June 2017
Summer School EMR’17
“Energetic Macroscopic Representation”
« Context and objective »
EMR’17, University Lille 1, June 20175
« A multi-objective benchmark for EMS of H-ESS for EVs »
- Context -
Studied system: battery/SC H-ESS
tracibati
chi
SCi
Batteries
Supercapacitors
batu
SCu chu
Chopper
Parallel
connection
Inductor
Inverter Electrical
machine
Mechanical
part
Traction subsystem
batu
traci
Traditionally: mono-objective EMS to extend battery life-time
What about a multi-objective approach considering SCs system ?
What kind of optimal benchmark for multi-objective EMS ?
EMR’17, University Lille 1, June 20176
« A multi-objective benchmark for EMS of H-ESS for EVs »
- Objective of the study -
1) Develop a methodology for
multi-objective EMS
2) Generate a Pareto benchmark
Jmain
Jaux.
Pareto
front
Dynamic programming (DP)EMR formalism
+
multi-objective approach
(Pareto optimality)
This is the
optimal
Try to
reach it
How to do?
Which method
at which step?
EMR’17
University Lille 1
June 2017
Summer School EMR’17
“Energetic Macroscopic Representation”
« EMR for problem formulation »
EMR’17, University Lille 1, June 20178
« A multi-objective benchmark for EMS of H-ESS for EVs »
- Modeling and local control -
Bat.
SC
batu
bati
chm
traciTract.
chiSCu
SCi chu
Parallel connection
ChopperInductor
batu
batu
SCi
ch refu
ch refiSC refi
Strategy
bat refi
chm
ch refu
ch refiSC refi
chiSCu
SCi chu batu
SCi
ch refi
chiSCu
SCi batu
Reduction of local control loop
EMR’17, University Lille 1, June 20179
« A multi-objective benchmark for EMS of H-ESS for EVs »
- Reduced model for problem formulation-
Bat. Tract.
Strategy
batu
bati
SCu
SCi
traci
chi
batu
batu
ch refi
bat refi
SC
• Control variable ibat ref
• State variable uSC
• Disturbance ubat & itrac
Variable definition
EMR’17
University Lille 1
June 2017
Summer School EMR’17
“Energetic Macroscopic Representation”
« Multi-objective energy management »
EMR’17, University Lille 1, June 201711
« A multi-objective benchmark for EMS of H-ESS for EVs »
- Benchmark generation-
0
ws bat/SCJ
bat cal SC cal
TJ J J
*
1J
*
nJ
*
iJ
batJ
SCJ
10
1
n
Weighting factor
Pareto front
benchmark
Multi-objective scalarization
DP DP DP
0 1
i
ws bat/SC
i
J 1
ws bat/SCJ
sum bat/SC bat SC1J J J
EMR’17, University Lille 1, June 201712
« A multi-objective benchmark for EMS of H-ESS for EVs »
- Multi-objective scalarization -
bat peakbat rms batbat 0 0 0
bat rms bat peak bat
1
3
IIJ
I I
f
o
2
bat cal bat
t
tJ i dt
SC peakSC lossSC 1 1
SC loss SC peak
1
2
IEJ
E I
f
o
2
SC cal SC
t
tJ i dt
f f
o o
2 2
bat SCsum cal bat/SC 0 1
bat rms SC rms
1t t
t t
i iJ dt dt
I I
Battery stresses cost function SCs efficiency cost function
Battery calculating cost function SCs calculating cost function
Multi-objective calculating cost function
Simplification for calculation:
For performance evaluation:
EMR’17, University Lille 1, June 201713
« A multi-objective benchmark for EMS of H-ESS for EVs »
- Dynamic programming for problem solving -
Bat. Tract.
batu
bati
SCu
SCi
traci
chi
batu
batubat refi
SC
Driving cycle
known in advance
Multi-objective scalarization
Dynamic programming
ws bat/SCJ
Tactical
layer
Strategic
layer
EMR-based
backward model
Strategy
SC measu
Backward representation: [Horrein 2015]
Multi-level hierarchical structure: [Trovão 2013]
EMR’17
University Lille 1
June 2017
Summer School EMR’17
“Energetic Macroscopic Representation”
« Results and discussion »
EMR’17, University Lille 1, June 201715
« A multi-objective benchmark for EMS of H-ESS for EVs »
- Simulation -
Variable Value Step size No. of steps
Time t 1028 [s] 0.1 [s] 10280
State variable uSC 22.5 – 45 [V] 0.25 [V] 90
Control variable ibat ref 0 – 250 [A] 2 [A] 125
ARTEMIS total urban driving cycle
Tazzari Zero
EMR’17, University Lille 1, June 201716
« A multi-objective benchmark for EMS of H-ESS for EVs »
- Results: Pareto front as a benchmark -
EMR’17, University Lille 1, June 201717
« A multi-objective benchmark for EMS of H-ESS for EVs »
- Examples of 2 cases -
= 0.75 (trade-off) = 1 (mono-objective)
EMR’17
University Lille 1
June 2017
Summer School EMR’17
“Energetic Macroscopic Representation”
« Conclusion and perspectives »
EMR’17, University Lille 1, June 201719
« A multi-objective benchmark for EMS of H-ESS for EVs »
- Conclusion and perspectives -
Perspectives
Near future works:
• Real–time strategies development
• HIL experimental validation
1) Develop a methodology for
multi-objective EMS
2) Generate a Pareto benchmark
This work has achieved the objectives:
• Systematically
• Can be extended for the
other system
• Global optimal benchmark
• Can be used to evaluate the
on-going real-time EMS
EMR’17
University Lille 1
June 2017
Summer School EMR’17
“Energetic Macroscopic Representation”
« Biographies and references »
EMR’17, University Lille 1, June 201721
« A multi-objective benchmark for EMS of H-ESS for EVs »
- Authors -
Bảo-Huy NGUYỄN
PhD student since 2015
Université Lille 1, L2EP, MEGEVH, France
Université de Sherbrooke, Sherbrooke, QC, Canada
MSc in Electrical Engineering at Hanoi Univ. Sci. Tech., Vietnam (2015)
Research topics: EVs and HEVs, control in power electronics and electrical drives
Prof. João P. TROVÃO
Université de Sherbrooke, Sherbrooke, QC, Canada
PhD in Electrical Engineering at University of Coimbra, Portugal (2012)
Research topics: EVs, renewable energy, energy management, power quality,
and rotating electrical machines
EMR’17, University Lille 1, June 201722
« A multi-objective benchmark for EMS of H-ESS for EVs »
- Authors -
Prof. Alain BOUSCAYROL
Université Lille 1, L2EP, MEGEVH, France
Coordinator of MEGEVH, French network on HEVs
PhD in Electrical Engineering at University of Toulouse (1995)
Research topics: EMR, HIL simulation, tractions systems, EVs and HEVs
Dr. Ronan GERMAN
Université Lille 1, L2EP, MEGEVH, France
PhD in Electrical Engineering at University of Lyon, France (2013)
Research topics: Energy Storage Systems, EMR, HIL simulation, EVs and HEVs
EMR’17, University Lille 1, June 201723
« A multi-objective benchmark for EMS of H-ESS for EVs »
- References -
L. Horrein, “Gestion d’energie decomposee d’un vehicule hybride integrant les aspects thermiques via larepresentation energetique macroscopique,” PhD thesis, Université Lille 1, 2015.
J. P. Trovão, P. G. Pereirinha, H. M. Jorge, and C. H. Antunes, “A multi-level energy management system formulti-source electric vehicles - An integrated rule-based meta-heuristic approach,” Appl. Energy, vol.105, pp. 304–318, 2013.
and
the missing book on EMR by Prof. Alain Bouscayrol !
EMR’17
University Lille 1
June 2017
Summer School EMR’17
“Energetic Macroscopic Representation”
« Appendix »
EMR’17, University Lille 1, June 201725
« A multi-objective benchmark for EMS of H-ESS for EVs »
- Dynamic programming solving procedure -
Final
state
constraint
k 1k 0 N
t
( )x t
0x
1ix
ix
Mx
minx
maxx
( )ix k ( 1)ix k
1( 1)ix k
1( 1)ix k
2 ( 1)ix k
(0)x ( )x N
Discretization
Qu
an
tiza
tio
n
Limitation (constraint)
Infeasible
Feasible
Backward computation
Bellman equation
* *
, 1,
Cost-to-go from the current stage Optimal cost-to-go from the next stage 1to the next stage 1 to the final stage
min , ,k N D k Nu k
k kk N
J x k g x k u k J f x k u k