prognostics for microgrid components - nasa · pdf fileprognostics for microgrid components...
TRANSCRIPT
Ames Research Center
Prognostics for Microgrid Components Batteries, Capacitors, and Power Semiconductor Devices
Abhinav Saxena Stinger Ghaffarian Technologies, Inc.
NASAAmes Research Center, Moffett Field CA 94035
Presented at
Advanced Microgrid Concepts and Technologies Workshop Beltsville, MD
June 7 - 8,2012
prOgnOSlicsJD!7 ~
PROGNOSTICS CENTER OF EXCELLENCE
STINGER
GHAFFARIAN
, ___ , TECHNOLOGIES
https://ntrs.nasa.gov/search.jsp?R=20120015367 2018-04-21T02:02:13+00:00Z
Ames Research Center
Prognostics and Health Management
In Perspective
PROGNOSTICS CENTER OF EXCELLENCE
Goals for Prognostics What does prognostics aim to achieve?
Improved mission planning
Ability to reassess mission feasibility
Avoid cascading effects onto healthy
subsystems
Maintain consumer confidence,
product reputation
More efficient maintenance
planning
Reduced spares
Ames Research Center
Service only specific aircraft
which need servicing
Service on Iy when it is needed
Contingency Management View Maintenance Management View
• Prognostics goals should be defined from users' perspectives
• Different solutions and approaches apply for different users
PROGNOSTICS CENTER OF EXCELLENCE 3
Integrated Logistics information
Health Management Ames Research Center
Contingency Management View
Condition Based Mission Planning
~m 1 Reconfigurntion
•
Control Reconfiguration
l
Condition Monitoring Safety and Risk
Analyses
Prognostic Control
Data Comm - Sensors
- Reporting - Scheduled Inspections
Troubleshooting and Repair
Maintenance Management View
Tech Support
T03;1;09 / ~_ • V // '. ~ r:O~l ~!-- ;;;;:------a::::;;;::;::~ ~ ..... \ ~ondition Based
Knowledg~~~se \ Maintenance
e·g·T\ ,. · Portable
Maintenance Aids
'~-
Feedbackto Production
Control
Predictive Mairitenance
~.~;.,.~ ......
Condition Monitoring Reliability Analysis
Schematic adapted from: A. Saxena, Knowledge-Based Architecture for Integrated Condition Based Maintenance of Engineering Systems, PhD Thesis, Electrical and Computer Engineering, Georgia Institute of Technology, Atlanta May 2007. Liang Tang, Gregory J. Kacprzynski, Kai Goebel, Johan Reimann, Marcos E. Orchard, Abhinav Saxena, and Bhaskar Saha, Prognostics in the Control Loop, Proceedings of the 2007 AAAI Fall Symposium on Artificial Intelligence for Prognostics, November 9-11,2007, Arlington, VA.
PROGNOSTICS CENTER OF EXCELLENCE 4
Prognostics for Microgrids
• Key components Power storage
• Batteries • Capacitors and SuperCapacitors
Power components and devices
}!fIIlf41 Wm', liKrtJrilld
• Power switches (semiconductor switches and packagl" ""-
• Passive components (inductors, capacitors, high frequency transformers)
• Controllers and Gate drivers
• Microgrids PHM - Potential benefits* Advanced inverter controls for microgrids Robust operation during fault conditions More informed decision support
* Key R&D Areas for micro grid reliability as identified in 2011 DoE Microgrid Workshop, San Diego CA
PROGNOSTICS CENTER OF EXCELLENCE
Ames Research Center
5
Grid PHM Framework
•
. 11 '\..
Classification Algorithms
• • Take Corrective Action
• Prognostic Framework
• u --~ J
• Grid Map
High load (peak hours)
PROGNOSTICS CENTER OF EXCELLENCE
Ames Research Center
•
1i!"-
6
Ames Research Center
Fundamentals of Predicting Remaining Useful Life
Understanding the Prognostic Process
PROGNOSTICS CENTER OF EXCELLENCE
Prognostics Categories Ames Research Center
• Type I: Reliability Data-based - Use population based statistical model - These methods consider historical time to failure data which are used to model
the failure distribution. They estimate the life of a typical component under nominal usage conditions
- Example: Weibull Analysis
• Type II: Stress-based - Use population based fault growth model - learnt from accumulated knowledge - These methods also consider the environmental stresses (temperature, load,
vibration, etc.) on the component. They estimate the life of an average component under specific usage conditions
- Example: Proportional Hazards Model
• Type III: Condition-based - Individual component based data-driven model - These methods also consider the measured or inferred component degradation.
They estimate the life of a specific component under specific usage and degradation conditions
- Example: Cumulative Damage Model, Filtering and State Estimation • For more details please refer to last year's PHM09 tutorial on Prognostics by Dr. J. W. Hines: [http://WININ.phmsociety.org/events/conference/phm/09/tutorialsj
PROGNOSTICS CENTER OF EXCELLENCE 8
Faul
t Dim
ensi
on (a
)Prognostics Framework
~------------------------------------------------------------------------------------.:~ ! / ...
/ ..
••
// ....
••
// .....
•• / _ ..... .
--- .---~~ -----------------~...- •
/
/ ____ r
~
Ames Research Center
I'Decision
1iQ,f'T = PH(a = aFT)' f PEoLCtla = aFT)dt o
)
n is adjusted with
probability of failure at the given damage size
Decision Point
../-,.....-./"
~ ~ Time (t) to
Risk is now a compound function of chosen failure threshold and the decision point
PROGNOSTICS CENTER OF EXCELLENCE 9
Faul
t Dim
ensi
on (a
)Prognostics Framework
Failure Threshold (aFT)
Probability of damage size being greater than
the critical value at time t3
Ames Research Center
00
7[ts = J Pa (ait = t3)da OFT
OFT
= 1 - J Pa(ait = t3)da o
= 1 - Pa (ap1' it = t3 )
Damage size pdf at a given time can be useful when planning a mission (usage) profile Answers how risky it is to go on a mission of known duration?
,--------------------- ----~
to t1 t2 13 Eol • Ti me (t)
We can figure out if the system would withstand by the time mission is completed
PROGNOSTICS CENTER OF EXCELLENCE 10
Particle Filters • Allows model adaptation • State estimation, tracking and prediction • Nice tradeoff between Me and KF • Useful in both diagnostics and prognostics • Represents uncertainty • Manages uncertainty
• actual state value
x measured state value
• stateparticlevalue
• statepdf(belief)
P(x)
x
actual state trajectory
estimated state trajectory
particle propagation
particle weight
/_._._.-.-._.-/
tk tk+l
Measurement
.......
t
PROGNOSTICS CENTER OF EXCELLENCE
Ames Research Center
Initialize PF Parameters
Propose Initial Population, <xo,wo>
Propagate Particles using State Model, Xk_l -7Xk
Update Weights, W k-l -7Wk
No
11
Particle Filter-Based Prognostics
System Sensors
1 Raw Data
Feature Extraction/ Direct Measurement
Model Parameters are part of the State Vector X == rxk,Okl
•••••••••••••••••••• ~
• ••••••••••••••••••• ~
Failure Threshold
• Parameter Identification •
Measurement Zk
System Model
Prior p(xklxk_1)
Particle Filter
Posterior p(xklxk_lZk)
.··········1·········. • '" +. • • Tuned
System Model
• •• +. •• Id t"f" d M d I- • ••••••••••••••••••••••••••••••••••••••••• - en I Ie 0 e· ••..................• -
State Tracking Loop Prediction Loop
PROGNOSTICS CENTER OF EXCELLENCE
Ames Research Center
Prediction
12
.-...... ns u .-c. C
. <C tn U
.-...... tn o
en c:
()) C
) E
0 en
a... Lii
D..
w
U
Z
w
w
U
x w
~
o '" w fZ
w
U
"' U
f"' o z OJ
o '" "-
Volta
ge (V
)
volta
ge
Volts
Actu
al E
OD
Star
t Pre
dict
ion
Pred
icte
d EO
D
min
s
Model Validation in Realistic Environments
• Battery discharge algorithm was used in e-UAV BHM • More than 3 dozen successful flights
Prediction update rates at 1 Hz Limited onboard computational power
23 L I I ,~
22
21
20
19
18
17
16
150
EOD Threshold
5 10 15 mins
20
r==:J Scaled EOD pdf -~Fwd Motor Battery 1 -~Fwd Motor Battery 2 -~Aft Motor Battery 1 -~Aft Motor Battery 2
600FE------C-------r----
25 900 1000 1100 1200 1300 1400
~ffi
PROGNOSTICS CENTER OF EXCELLENCE
Ames Research Center
15
Cap
acity
(Ahr
)
volta
ge
z o iii'
~
Power Electronics Failure - MOSFETs Ames Research Center
• Objective: predict abnormal functioning of power electronics devices Prediction approach validated in data from 1 OOV power MOSFETs
The failure mechanism for the stress conditions is determined to be die-attachment degradation
Change in ON-state resistance is used as a precursor of failure due to its dependence on junction tem perature
~.
A il/ S ~~~
.1., .I,I < . ...;;~~ .' - . . ~.- ~ ~., .... ~~~.( . ~ ~ . ~ ,~ ,,- ,::::::-
II{ 7 ,,, """ . . .::::;:-I /.~"' o " ' ~ '~."\{' ~}: ... : /%
". ,'·, . .Ii
Increase in RD5(ON) due 10 [)e,,.;ce Degradation
-- Rre{a-J) De_ceB
- Rre{a-J) De_ceg
- Rre{a-J) De·.ce1 1 ' - - - - - - - - - -
- Rre{a-J) C".ce12 ' - - - - -
- Rre{a-J) [".ce14 ,
obe:~~-----------100 200 300 400 500
aging time (m mites I 600 700 800
RoS(on) Prediction
0.1 f-- --i
0.08 f-
I O.oSf-. .. a: <l
0.04 f-
0.02f-.-- -_.
o - ; : I I
50 100 nl { u } n ! J lEOL~ J _______ " ~J,ft - ~_ -1--
Aging Time (minutes)
Celaya, J., Saxena, A., Wysocki, P., Saha, S., Goebel, K., "Towards Prognostics of Power MOSFETs: Accelerated Aging and Precursors of Failure" Annual conference of the PHM Society, Portland OR, October 2010.
PROGNOSTICS CENTER OF EXCELLENCE 17
Text
Volu
me
(mm
3 )
RU
L
RU
L(i)
RU
L(i)
Met
ric M
(i)
RU
L(i)
RU
L
Prognostic Performance Metrics Ames Research Center
• Metrics Hierarchy
PROGNOSTICS CENTER OF EXCELLENCE 20
21
Challenges in Prognostics Ames Research Center
-• Requirements Specification
- How can a requirement be framed for prognostics considering uncertainty?
- How to define and achieve desired prognostics fidelity
• Uncertainty in prognostics - Quantification, representation, propagation and management
- To what extent the probability distribution of a prediction represent reality
• Validation and Verification - How can a system be tested to determine if it satisfies specified
requirements?
- If a prediction is acted upon and an operational component is removed from service, how can its failure prediction be validated since the failure didn't happen?
- Prognostics performance evaluation - offline and online?
- Verifiability of prognostics algorithms
PROGNOSTICS CENTER OF EXCELLENCE 21
~ c en
..c r-
('tn c: o .-...... tn Q
) :::l a
u z w
w
u x w
o fZ
w
"' u f"' o z OJ
o '"