Seminar Seminar Sept.
2014
On EMR “Energetic Macroscopic Representation”
Sept. 2014
«Merged Control Loops for the
Stability of Input Filters »
P. BARRADE*, A. BOUSCAYROL**, P. DELARUE**, W. Lhomme**
* LEI, EPFL, Switzerland ** L2EP, Univ. Lille 1, France
« Merged control loops »
- Outline -
2
Seminar on EMR, Sept. 2014
• Introduction
• Input filter oscillations
– Structure and conventional control scheme
– From oscillations rejection to instabilities
• Merged control
– Objectives and tuning parameters
– Merged control loops
– Design of the controllers
• Main results
– Simulation and Experimental results
• Conclusions
Seminar Seminar Sept. 2014
On EMR “Energetic Macroscopic Representation”
Sept. 2014
« Introduction »
« Merged control loops »
- Objectives of the study -
4
Seminar on EMR, Sept. 2014
• General frame
– DC/DC or 3 phases DC/AC converters
– Feeding from a DC source via a 2nd order low pass filter
• Depending on the control of the converter
– Some input filter oscillations can occur
– In extreme cases, instabilities can be observed
• Depending on the operating point
– Reference output power, DC feeding voltage
• Main goal
– Allow the input filter stabilization by the appropriate control of the converter
• Obtained from an Energetic Macroscopic Representation (EMR)
Seminar Seminar Sept. 2014
On EMR “Energetic Macroscopic Representation”
Sept. 2014
« Input filter oscillations»
« Merged control loops »
- Structure and conventional control scheme -
6
Seminar on EMR, Sept. 2014
• Example of a battery charger – Simplified control
• A resistor R is needed to damp the input filter
• The input filter oscillations impact on the output current if not taken into
account by the control of the converter
« Merged control loops »
- From oscillations rejection to instabilities -
7
Seminar on EMR, Sept. 2014
• The control must enable the regulation of the input current, by rejecting the input
filter oscillations
– Use of an Energetic Macroscopic Representation (EMR), leading to an
inversion based control identification
U1 (1) i (2) u (3) u (4) i
L1 C ch L2
S1 S2
iL1 uC ich iL2 U2
D
(6) uch-ref (5) iL2-ref
« Merged control loops »
- From oscillations rejection to instabilities -
8
Seminar on EMR, Sept. 2014
• The input filter oscillations are now rejected
– The input filter becomes unstable
– The input filter oscillations still impact on the output current ripple
By rejecting the input filter oscillations, the
converter absorbs an average constant power:
Po = Uc I e
This makes the converter to behave as a negative
impedance which loads the input filter:
€
δU c Uc Uc 2
0 = I eδU c + U cδI e ⇒ Z d = =− =−
δI e Ie Po
Zd versus R defines the unstability conditions of
the input filter:
€ • function of the operating point Po
Seminar Seminar Sept. 2014
On EMR “Energetic Macroscopic Representation”
Sept. 2014
« Merged Control»
« Merged control loops »
- Objectives and tuning parameters -
10
Seminar on EMR, Sept. 2014
• 2 solutions for the input filter stabilization
– Adapt the value of the damping resistor
– Stabilization by the correct control of the converter
• Control for the input filter stabilization
input filter chopper inductor
(1) (2) (3) (4)
U1 iL1 uC uch iL2
S1 S2
iL1 uC ich iL2 U2
D
uch iL2
uC ich
D D
• Only 1 tuning parameter (D) for two objectives
– Regulation of the input current + Stabilization of the input filter
« Merged control loops »
- Merged Control Loop -
11
Seminar on EMR, Sept. 2014
• One defines 2 independant control loops (one per objective)
U1 (1)i (2) u (3) u (4) i
L1 C ch L2
S1 S2
iL1 uC ich iL2 U2
(19)
D1-ref D2-ref
kw
uC-ref ich-ref uch-ref iL2-ref
(16) (17) (18) (7)
– 2 duty cycles are then defined
• Du for the input filter stabilization loop
• Di for the output current control
– The two duty cycles are merged to obtain the required duty cycle D
– Weightning factor kw
« Merged control loops »
- Modeling and sizing rules -
12
Seminar on EMR, Sept. 2014
• Close-loop average model
" d
$ L1 iL1 = U1 − uC − R1iL1
$ dt
$ d
# C uC = iL1 − kwich−ref − (1− kw ) K1
$ dt
$ d
$ L2 iL 2 = kw K 2 + (1− kw ) uch−ref −U 2
% dt
uch−ref ich−ref
with K1 = iL 2 K2 = uC
uC iL 2
• Linearization along a given operating point
# Ui U i
% K1 ≈ − 2 2L 2o uC + 2 iL 2 + L 2o uch−ref
% uCo uCo uCo
$
% K ≈ iL1o u − U 2 i + uCo i
2 C L2 ch−ref
% i i i
& L 2o L 2o L 2o
« Merged control loops »
- Modeling and sizing rules -
13
Seminar on EMR, Sept. 2014
• Closed-loop description in the Laplace domain
! U $
! I $ # 1
&
# L1 & # U2 &
s.I.X = AL X + BLU X = # Uc & U= # &
## && # I L 2−ref &
" IL2 % # u &
" C−ref %
! $
# R 1 &
- 1 − 0
# L1 L1 &
# &
# 1-k w U 2 I L 20 k p kw (k is + sk ps ) I L 2o − sU 2 &
AL = # (1− kw ) − (1− kw ) 2 &
C CUCo C sCUCo
# &
# kwUco kw kwU 2 ( sk ps + kis ) &
#
# L2 I L 2o L2 I L 2o
( I L1o + k pUCo ) −
L2 I L 2o
− (1− kw )
sL2
&
&
" %
! $
# 1 &
0 0 0
# L1 &
# &
# (1− kw ) I L 20 (1− kw ) (kis + sk ps ) I L 2o k p kw &
BL = # 0 − − &
CUCo sCUCo C
# &
# kw ( sk ps + kis ) k pUco &
#
#
0 − (1− kw ) −kw &
&
" L2 sL2 L2 I L 2o %
« Merged control loops »
- Modeling and sizing rules -
14
Seminar on EMR, Sept. 2014
• Stability analysis
– Considering the transfer function
U c (s)
Fs (s) =
I so (s)
– Focusing on its denominator
D(s) = a 0 s 4 + a1s 3 + a 2 s 2 + a3s + a 4
€
• The coefficients are functions of
– Main
€ parameters (L,C,R, etc…)
– Operating point (Ue, Uso, Is_ref=Iso)
• Dominant poles only are considered
– Equivalent 2nd order system: oscillation frequency (ωn) and damping
(ζ)
« Merged control loops »
- Input filter stability -
15
Seminar on EMR, Sept. 2014
• Defined by the weightning factor kw
• Defined by the input filter stabilization control loop
– Simple proportional controler kp
1
=0.1 35
0.9 =0.4
=0.7
30
0.8
0.7 25
0.6
(A)
20
kw
0.5
ref
iL2
15
0.4
0.3 10
0.2
5
=0.4
0.1 =0.7
0
0 0 10 20 30 40 50 60 70 80 90 100
10
−2 −1
10 10
0
10
1 2
10 U2 (V)
kp
Seminar Seminar Sept. 2014
On EMR “Energetic Macroscopic Representation”
Sept. 2014
«Main results »
« Merged control loops »
- Experimental set-up-
17
Seminar on EMR, Sept. 2014
• From simulation to real-time implementation
L1
iL1 ich
L2
C iL2
U1
uch U2
iL1 uC s s iL2
dSPACE 1005
uC-ref
iL2-ref kw
« Merged control loops »
- Results evaluation -
18
Seminar on EMR, Sept. 2014
• Simulation:
• With/without stabilization loop (kw=0.5, Kp=4)
without with
20 iL2 20 iL2
i (A)
iLi (A)
10
Li
10
0
0 5 10 15 iL1 20 0
0 5 10 15 i 20
Time (s) L1
Time (s)
115
115
uC (V)
110
u (V)
110
105
C
105
100
0 5 10 15 20 100
Time (s) 0 5 10 15 20
4 Time (s)
9 4
iL1 (A)
iL1 (A)
3.5 9
8.5
iL1 (A)
iL1 (A)
3.5
3 8.5
8
3
7.5 7.6 7.7 7.8 7.9 16.5 17 17.5 8
Time (s) Time (s)
7.5 7.6 7.7 7.8 7.9 16.5 17 17.5
Time (s) Time (s)
« Merged control loops »
- Results evaluation -
19
Seminar on EMR, Sept. 2014
• Experimental results:
• With/without stabilization loop (kw=0.5, Kp=4)
without with
!
« Merged control loops »
- Results evaluation -
20
Seminar on EMR, Sept. 2014
• Experimental results for different values of kw:
kw=0 kw=0.3 kw=0.5
Seminar Seminar Sept. 2014
On EMR “Energetic Macroscopic Representation”
Sept. 2014
« Conclusion »
A new and original method for the stabilization of input filter has
been proposed
This method has been identified thanks to an energetic macroscopic
representation
– Allows the identification of efficients control structures
The model of the system and its control has shown:
– The control structure based on an inversion method leads to the effective
input filter stabilization
– One can moreover adjust the dynamic behavior