You are on page 1of 6

2019 IFAC Workshop on

2019 IFAC
Control Workshop
of Smart Gridon
and RenewableAvailable
Energy Systems
2019 IFAC Workshop on online at www.sciencedirect.com
Control
2019 of Smart
Jeju, IFAC
Korea, JuneGrid
Workshop and
on
10-12, Renewable Energy Systems
2019
Control
Jeju, of Smart
Korea, June Grid and
10-12, Renewable Energy Systems
2019
Control of Smart Grid and Renewable Energy Systems
Jeju, Korea, June 10-12, 2019
Jeju, Korea, June 10-12, 2019 ScienceDirect
IFAC PapersOnLine 52-4 (2019) 200–205
Robust Control of DC-DC Boost Converter
Robust Control of DC-DC Boost Converter
Robust Control
Robustby of DC-DC
DC-DC Boost
using µ-Synthesis Converter
Approach
byControl of
using µ-Synthesis Boost Converter
Approach
by using
by using µ-Synthesis Approach
B. Alharbi* M. Alhomim*Approach
µ-Synthesis R. McCann*
B. Alharbi* M. Alhomim* R. McCann*
B.
B. Alharbi*
Alharbi* M. M. Alhomim*

Alhomim* R.
R. McCann*
McCann*
* University of Arkansas, Fayetteville, AR 72701 USA (e-mail: ba001@uark.edu)
* University of Arkansas, Fayetteville, AR 72701 USA (e-mail: ba001@uark.edu)
** University
University ofof Arkansas,
Arkansas, Fayetteville,
Fayetteville, AR AR 72701
72701 USAUSA (e-mail:
(e-mail: ba001@uark.edu)
ba001@uark.edu)
Abstract: Boost converters are used in many renewable energy systems in order to interface a lower
Abstract: Boostsuch
voltage source converters
as a solararearray
used orin battery
many renewable
to the dc input energyofsystems in order toinverter.
a grid-connected interface
Thea lower
boost
Abstract:
voltage Boost
Boostsuch
source
Abstract: converters
converters
as a solarare
are used
used or
array in
in many
many
battery renewable
renewable
toforthe energy
energy
dceffects
input ofsystems
systems in
in order
order
a grid-connected to interface
toinverter.
interface
Theaa lower
lower
boost
converter
voltage must
source be controlled
such as a solarin order
array orto correct
battery to the
the dc input of
of variable
a input
grid-connected voltages and
inverter. changing
The boost
voltage
output source
converter must
loads. such
be
This as a solar
controlled in
paper develops array
order ortobattery
correct
an improved tofor the
the
controller dc input
effects of a grid-connected
variable
designofmethod input
using inverter.
voltages
a voltages and
µ-synthesis The boost
changing
approach.
converter
converter
output must
must
loads. be
be
This controlled
controlled
paper in
in
develops order
order
an to
to correct
correct
improved for
for the
the
controller effects
effects
designof variable
variable
method input
input
using a voltages and
and
µ-synthesis changing
changing
approach.
The goal
output is toThis
loads. achieve
paper adevelops
low order an controller
improved that can bedesign
controller implemented
method at a reduced
using cost. This
aa µ-synthesis is an
approach.
output loads.
The goal
improvementis toThis
over paper
achieve
H adevelops
low order
controllers an improved
controller
that achieve controller
that bedesign
can performance
robust method
implemented withusing
at more µ-synthesis
a reduced
complex approach.
cost.realizations.
This is an
The
The goal
goal
improvementis
is to
to achieve
over H
∞ a low order controller that can be implemented at a reduced cost. This is an

achieve a low order
controllers controller
that achieve that can
robust be implemented
performance withat a reduced
more complexcost. This is an
realizations.
Detailed simulation ∞
results for a wide range of parametric changes in the boost circuit environment
improvement
improvement
Detailed over
over
simulation H
H ∞ controllers that achieve robust performance with more complex realizations.
controllers
results for a that
wide achieve
range of robust performance
parametric changes with
in the more
boost complex
circuit realizations.
environment
confirm the benefits of∞ the proposed method. The results are compared with a PI controller
Detailed simulation results for aa wide range of parametric changes in the whereby it is
Detailed simulation
confirmthat
shown the results
benefits of
µ-synthesis for improved
the proposed
provides wide range
method. of results
The
performance parametric changes
are compared
compared in thea boost
with boost circuit
PI controller
to a conventional
environment
circuit whereby
environment
controller. it is
confirm
confirm the benefits
theµ-synthesis
shown that of the proposed
benefits of provides
the proposed method.
method.
improved The results are
The resultscompared
performance compared
are compared with a PI controller
with a PI controller
to a conventional whereby
whereby it
controller. it is
is
© 2019,that
shown
shown that
IFAC
Keywords: µ-synthesis
µ-synthesis
Boost provides
(International
provides
converter, robustimproved
Federation
improved
control, performance
of Automatic
performance
structured compared
Control) Hosting
compared
singular to a
to by
value, conventional controller.
Elsevier Ltd. controller.
aµ-synthesis.
conventional All rights reserved.
Keywords: Boost converter, robust control, structured singular value, µ-synthesis.
Keywords: Boost
Keywords: Boost converter,
converter, robust
robust control,
control, structured 
structured singular
singular value,
value, µ-synthesis.
µ-synthesis.

1. INTRODUCTION  2. ROBUST CONTROLLER DESIGN

1. INTRODUCTION 2. ROBUST CONTROLLER DESIGN
Robustness in control 1.
1. INTRODUCTION
INTRODUCTION
systems has been a focal point for 2.
2. ROBUST
ROBUST CONTROLLER
CONTROLLER DESIGN DESIGN
Robustness
designers in
since control
the 1980s systems
when has been
George a
Zamesfocal point
and Brucefor 2.1 Robust Design Method
Robustness
Robustness
designers since in
in control
control
the 1980s systems
whenhas
systems has been aa focal
been Zames
George focaland point
pointBrucefor
for 2.1 Robust Design Method
Francis
designers first
since introduced
the 1980s thewhen importance
George of the and
Zames topicBruceand 2.1 Robust Design Method
designers
Francis firstsince the 1980sthe
introduced when George Zames
importance of the and topicBruceand State-space Design
2.1 Robust Method models provide advantages
time-domain
extended
Francis it
first quickly
introduced to more
the general
importance problems.
of the Currently,
topic and State-space time-domain models provide advantages
Francis
extendedfirst introduced
it quickly to morethe importance
general problems.of the Currently,
topic and for compared to time-domain
frequency domain descriptions. In particular,
robust
extended control
it design totends
quickly more to general
emphasize H∞∞ techniques
problems. Currently, State-space
State-space
compared to time-domain
frequency domain models
models provide
provide In advantages
descriptions. advantages
particular,
extended
robust
optimization it
control quickly
design
ofdesign
performanceto
tendsmore to general
emphasize
objectives. problems.
The H ∞ Currently,
techniques for analysis
process is based compared and
to design
frequency of
domain multiple-input
descriptions. multiple-output
In particular,
robust
robust control
controlof
optimization design
performance tends to emphasize
tends single-input
to emphasize
objectives. The H ∞ techniques for
techniques
H∞process is(SISO) compared
analysis
for (MIMO)
based to
and frequency
design domain
of descriptions.
multiple-input In particular,
multiple-output
on the theory that a suitable single-output analysis can
and often
design be handled
of in a more compact
multiple-input manner.
multiple-output
optimization
optimization
on the theory of
of
thatperformance
performance
a suitable objectives.
objectives.
single-input The
The process
process
single-output is based
is(SISO)
based analysis
(MIMO) and
can design
often be of
handled multiple-input
in a more multiple-output
compact manner.
controller provides sufficient gain and phase margin. The Designing
(MIMO) a system that has robust performance to structured
on
on the
the theory
theory
controller that
that
provides aa suitable
suitable
sufficient single-input
single-input
gain and single-output
single-output
phase margin. (SISO)
(SISO)
The Designingcan
(MIMO) can often
a system be
be handled
often that handled
has robust in aa more
more compact
in performance compact manner.
manner.
to structured
closed-loop
controller achievessufficient
provides robust stability
gain and byphasesolving a linear
margin. The and unstructured
Designing a systemuncertainty
that has if generally
robust performance very difficult.
to There
structured
controller
closed-loop provides
achieves sufficient
robust gain
stabilityand by phase
solvingmargin.
a The
linear Designing
and
has been a system
unstructured
considerable thatresearch
uncertaintyhas robust on performance
if generally
this topic. to structured
very difficult.
Lin’s There
defined
matrix inequality
closed-loop achievesconstraintrobust with a defined
stability by optimization
solving cost
aa linear and
and
has unstructured
unstructured
been uncertainty
uncertainty
considerable research if generally
if generally
on this were very difficult.
veryimportant
topic. difficult. There
There
Lin’s defined
closed-loop
matrix
function (Doyle,achieves
inequality constraint
1982). robust withstability
a by
defined solving
optimization linear
cost specifications
has been and
considerablemethods research(Lin, 2000)
on this topic. Lin’s in this
defined
matrix
matrix inequality
functioninequality
(Doyle, 1982). constraint with a defined optimization
constraint with a defined optimization cost study: cost has been
specifications considerable
and methods research(Lin, on
2000)this topic.
were Lin’s
important defined
in this
H
specifications ∞
∞ , Loop
and Transfer
methods Recovery
(Lin, 2000) (LTR)
were method,
important in µ-
this
function
function (Doyle,
Robust (Doyle,
control 1982).1982).
has been developed over the previous synthesis, specifications
study: H ∞ , and
Loop methods
Transfer (Lin, 2000)
Recovery were
(LTR) important
method, in this
µ-
study: H and Quantitative Feedback
∞, Loop Transfer Recovery (LTR) method, µ-
Theory (QFT) methods.
Robust
decades control
to guarantee has been developed
stability and high over the
performanceprevious study:
synthesis, H ,
and Loop Transfer
Quantitative Recovery
Feedback
of synthesis, and Quantitative Feedback Theory (QFT) methods.
∞ Theory (LTR) (QFT) method,
methods. µ-
Robust
Robust
decades control
control
to has
has
guarantee been
been developed
developed
stability and highover
over the
the
performanceprevious
previous of synthesis, and Quantitative Feedback Theory (QFT) methods.
electric
decades power
to converters with and respect toperformance
uncertainties,
decades to guarantee
electric power guarantee
converters stability
stability
with and high
highsuch
respect toperformance
uncertainties, of
of 2.2 Perturbation Models
disturbance
electric power effects, and nonlinearities
converters with respect to as magnetic
uncertainties, 2.2 Perturbation Models
electric
disturbance
saturation.power
Using converters
effects,mixed with respect
andsensitivity,
nonlinearitiesH∞∞ loop toshaping,
such uncertainties,
as magnetic
and µ- 2.22.2 Perturbation
Perturbation ModelsModels
disturbance
disturbance
saturation. effects,
effects,
Using mixed and
and nonlinearities
nonlinearities
sensitivity, such
such
loop as magnetic
as magnetic
shaping,
H∞ Postlethwaite, and µ- Although nominal models are generally not accurate for
synthesis approaches (Skogestad and 2005), Although nominal
saturation.
saturation. Using
Using mixed
mixed sensitivity,
sensitivity, H ∞ loop shaping, and µ-
loop shaping,
H∞ Postlethwaite, µ- actual operation, theymodels
can be are usedgenerally
to effectively not accurate
illustrate thefor
synthesis
one
synthesis
approaches
can model uncertainty
approaches
(Skogestad
(Skogestad
and
to find satisfactory
and solutions.and
Postlethwaite,
2005),
2005),
Although
Although
actual nominal
nominal
operation, they models
models
can be are
are
usedgenerally
generally
to effectively not
not accurate
accurate
illustrate for
for
the
synthesis
one can modelapproaches
uncertainty (Skogestad and Postlethwaite,
to find satisfactory solutions.2005), actualmodelsoperation,
of the plant theywherein
can be input
used voltage
to and load
effectively are related
illustrate the
one can model uncertainty to find satisfactory solutions. actual
models
to dc operation,
of the
gain, poles, they
plantand can
wherein
RHP bezero usedofvoltage
input tothe effectively
and load
system. illustrate
At are
this the
related
point,
one can model uncertainty
approach to is find satisfactory solutions.
control and models of the plant wherein input
The µ-synthesis a model for robust
The µ-synthesis approach is a model for robust control and models to ofand
dc gain,
inductor the plantand
poles,
capacitorwherein
RHP
values input
zero ofvoltage
willvoltage
the system.
have
and
and load
little load At are
impact. are related
related
thisSimple
point,
one
The of the successful
µ-synthesis approach techniques
is a modelfor designing
for robust an efficient
control and to
to dc
dc
inductor gain,
gain,
and poles,
poles, and
and
capacitor RHP
RHP
values zero
zero of
of
will the
the
have system.
system.
little At
At
impact. this
this point,
point,
Simple
The
one ofµ-synthesis
the successful approach is a model
techniques for for robust an
designing control and full
efficient inductor blockand perturbation
capacitor valueswill willinclude
have all the
little impact. effects
Simple of
controller
one of thewithin
successfulstructured and
techniques unstructured
for designing uncertainty
an and
efficient inductor
full blockand capacitor
perturbation valueswill will have
include little
all impact.
the Simple
effects of
one of thewithin
controller successfulstructured techniques for designing
and unstructured an efficient
uncertainty and electromagnetic
full block properties
perturbation at
will low
includeand high
all thefrequency.
effects To
of
disturbance
controller conditions.
within structured Thisand method was proposed
unstructured by Doyle
uncertainty and full block
electromagnetic perturbation
properties will
at include
low and all
high the effects
frequency. of
To
controller
disturbancewithin structured
conditions. Thisand unstructured
method was proposed uncertainty
by Doyleand achieve this, an input
electromagnetic multiplicative
properties at low perturbation
and high is appliedTo
frequency. to
in 1982, and
disturbance helps toThis
conditions. assess
methodthe wasimpact
proposedof parametric
by Doyle electromagnetic
achieve this, an properties
input at
multiplicative low and high
perturbation frequency.
is applied To
to
disturbance
in 1982, and
uncertainty. conditions.
In helps
this toThis
research, method
assess the was
µ-synthesis proposed
impact controller by design
of parametricDoyle aachievenominal model:
this, an As
input shown in
multiplicative (1) where ∆(s)
perturbation is aisunity norm
applied to
in
in 1982,
1982,
uncertainty. and
and In helps
helps
this to
to
research,assess
assess the
the impact
impact
µ-synthesis of
of
controller parametric
parametric
design achieve
aperturbation
nominal this, an
model: input
As multiplicative
shown in (1) perturbation
where ∆(s) is ais applied
unity norm to
methods are used as aresearch,
structuredµ-synthesis
singular value approach. aaperturbation
nominal model:and As w(s)
shownis in an(1)uncertainty
where ∆(s) weight
is a unity whose
norm
uncertainty.
uncertainty.
methods are used In this
In this as aresearch,
structuredµ-synthesis
singular value controller
controller
approach. design nominal
design perturbation model:and As shown
w(s) is in
an (1) where
uncertainty ∆(s) is a
weightunity norm
whose
magnitude displays and the isuncertainty bound weight (Baskinwhose and
methods
methods
This paper are used
used as
are develops as aa structured
structured
an improved singular
singular
robust value
value approach.
approach.
controller design perturbation
magnitude
Bulent, 2015), and w(s)
displays w(s) an
an uncertainty
the isuncertaintyuncertainty bound weight (Baskinwhose and
This paper develops an improved robust controller design magnitude magnitude
Bulent, 2015), displays
displays the uncertainty bound (Baskin and
the uncertainty bound (Baskin and
method
This paperusing
develops µ-synthesis
an to achieve optimal voltage
This
method paper develops
using an improved
µ-synthesis improved robust
robust controller
to achieve controller
optimal voltage design
design Bulent,
Bulent, 2015),
2015), G ( s)  G( s)(1  ( s) w( s)) .
regulation
method of a boost converter used in renewable energy (1)
regulationusing
method using
of a boostµ-synthesis
µ-synthesis converter to achieve
to used
achieve optimal
optimal voltage
in renewable voltage
energy G ( s)  G( s)(1  ( s) w( s)) . (1)
applications.
regulation
regulation
applications. of Boost
aa boost
of Boost boost
converters
converter
converterare
converters
are
used
used
particularly
in
in renewable
renewable
particularly
challenging
energy
energy The nominal model
challenging G
G (( ss)) 
 G
G (( ss)(1
)(1 
 
 (( ss)) w
w (( ss))
)) .. (1)
(1)1
due to a tendency toconverters
become unstable at high boost ratios.  and possible plant set is shown in Table
applications.
applications.
due to a tendency Boost
Boosttoconverters
become unstable are particularly
are particularly
at high boost challenging
challenging
ratios. The nominal model and possible plant set is shown in Table
Using
due to an averagingtotechnique,
aa tendency become the evaluation
unstable at high of the method and the nominal setand corresponding these plants is in
shown in1
due
Usingto an
considers tendency
averaging
power totechnique,
conversionbecome and unstable
the
weighting highofboost
at functions
evaluation boost
thethat ratios.
ratios.
method
will
The
The
and nominal
nominal
the nominal
Fig. the1. nominal
model
modelset and
The uncertainty
possible
possible
corresponding plant
plant
boundthese
set
set
these is
is shown
shown
plants
andplants is
possible in Table
Table
shown in1
input 1
Using
Using an
an
considers averaging
averaging
power technique,
technique,
conversion and the
the evaluation
evaluation
weighting of
of
functions the
the method
method
that will and
and
Fig. the1. nominal set corresponding
set corresponding
The uncertainty boundthese andplants is shown
is shown
possible in
in
input
improve
considers robustness,
power stability, and performance.
considers power conversion
improve robustness, stability,and
conversion andandweighting
weighting
performance. functions
functions that that will
will Fig.Fig. 1.1. The The uncertainty
uncertainty bound bound and and possible
possible input input
improve
2405-8963 robustness, stability,
© 2019, IFACstability,
improve robustness, (Internationaland performance.
Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.
and performance.
Copyright
Peer review©under
2019 responsibility
IFAC 223Control.
of International Federation of Automatic
Copyright © 2019 IFAC 223
10.1016/j.ifacol.2019.08.261
Copyright © 2019 IFAC 223
Copyright © 2019 IFAC 223
2019 IFAC CSGRES
Jeju, Korea, June 10-12, 2019 B. Alharbi et al. / IFAC PapersOnLine 52-4 (2019) 200–205 201

multiplicative perturbations is shown in Fig. 2 where the transfer functions; loops can be shaped by weighting
uncertainty bound covers all possible values. functions. Closed-loop systems require stability and
performance and H∞ has the ability to assure that. The H∞
2.3 H-infinity Method method requires that the system transfer function being
evaluated as the norm (Tewari, 2002)
A stable robust system meets specific criteria even in the G 
 sup w{ G( jw) } (2)
existence of defined uncertainties. The H∞ optimization
method, first defined in the 1980s, has proven itself to be This functions to decrease the peak of the bode diagram
exceedingly efficient as a successful robust control design that then raises the margin of robust stability. This study
method in the field of linear time-invariant control systems as uses Matlab software to minimize the norm from the
Ioannou and Sun explained (Ioannou and Jing, 1996). H∞ weighted mixed-sensitivity functions.
methods enable robust stable performance based on norm-
based optimization theory and are intended to perform well 2.4 Mixed Sensitivity Problem
even in worst-case disturbance events.
For robust stability a controller will be defined by a mixed
Table 1. Boost converter parameters sensitivity problem
Parameters Values Units S Wp ( I  GK )1
Input Voltage, Vi 20 ± 20% Voltage, V min  min (3)
K KS 
K KWn ( I  GK )1 
Output Voltage, Vo 45 Voltage, V
Inductor, L 560  10-6 Henry, H which describes the closed-loop system with model
uncertainties and possible perturbations shown in Fig. 3 (Pilat
Capacitor, C 100  10-6 Farad, F
and Wlodarezyk, 2011).
Resistance load, R 10 ± 20% Ohm, Ω
Plant, P

y1
W1

y2
W2

y3
u G W3
u1
-

Fig. 3. Mixed sensitivity control design.


The system G refers to a transfer function of the boost
converter. P is the open-loop transfer function. W1, W2 and
W3 are the performance, control and noise weighting
Fig. 1. Effect of the parameters variation on gain and phase. functions respectively. The input signals are denoted by u, y
is the output vector, and K is the controller that can be found
by using MATLAB commands
[K,CLP,gamma,info]=mixsyn(G,W1,W2,W3)
This penalizes the error signal W1, control signal,W2, and
output signal,W3, so that the closed-loop transfer function
matrix is weighted mixed sensitivity as

W1S 
Ty1u1  W2 R  (4)
W3T 

where the sensitivity, S, is equal to (I+GK)-1, the


Fig. 2. Bound of uncertainty and input multiplicative complementary sensitivity, T, is equal to (GKS), and the
perturbations. control effort, R, is equal to (KS). The three functions of
sensitivity, control effort, and complementary sensitivity will
Design specifications, including tracking and robustness be shaped using mixed sensitivity method.
performances, used constraints on the singular values of loop

224
2019 IFAC CSGRES
202
Jeju, Korea, June 10-12, 2019 B. Alharbi et al. / IFAC PapersOnLine 52-4 (2019) 200–205

The control purpose is to design a stable controller that  ( N (s))  sup  ( N ( jw)), w  R (8)
minimizes the norm of generalized transfer function Tyw such
that 1
where   ( N ) is the lowest singular value of ∆. This
W1S indicates a frequency dependent stability margin.
Tyw  1 or 1 (5)

W2T 

2.5 µ-synthesis

Consider the general plant and controller configuration


u y
presented in Fig. 4. Here P(s) is the generalized plant model
that contains all of known elements encompassing the
nominal plant model and performance and uncertainty
weighting functions. K(s) is a stabilizing controller, and ∆(s)
is the norm-bounded block-diagonal perturbation matrix, w N
represents an external disturbance, v is the measurement
available to the controller, u is the output from the controller,
and ∆y is an error signal. Fig. 5. N∆ structure for robust performance analysis.
Next, to perform robust stability analysis, matrix C(s) is
defined such that
C(s) = N11(s) (transfer function from the output to the input
u y
∆(s)).

u v
u y
K

Fig. 4. General control configuration with uncertainty C

(s)  {diag i ( s) : i ( s)  C, i  1, 2,..n} . (6)


Assume that ∆(s) is stable and Fig. 6. C∆ structure for robust stability analysis
Since C(s) is stable, µ is defined as,
(s) s  jw is normalized such that ( jw) 
 1.
 (C( jw))  [min{ (i ( jw)) : det( I  C( jw)( jw)  0}]1 (9)
The output of the general control configuration are an Here µ (C(jw)) is the structured singular value of
observation output vector v, and a performance scale vector z,
that is desired to be small. The inputs of the configuration are C(s) s  jw , measuring the smallest structured uncertainty
a control input u, and a system disturbance vector w that ∆(s). To maintain robust stability, two conditions should be
includes all inputs external to the system such as scaled met:
reference signals, disturbance, and sensor noise signals. ∆y
and ∆u are the input and output signals of the dynamic N (s) s  jw is stable and (ii) µ(C(jw))<1, w .
uncertainties, respectively.
For robust performance, the conditions are (i) robust stability,
For a given controller K(s), the system is consolidated as
shown in Fig. 5, where and (ii) µ(N(jw))<1, w .

The goal of µ-synthesis design methodology is to find the


N (s)  ( P( s), K (s)) P11 (s)  P12 ( s) K ( s)( I  P22 ( s) K ( s)) 1 P21 ( s) . (7) stabilizing controller K that minimizes the influence of the
external disturbances w, on the performance output z, in the
Definition 1: When N(s) is an interconnected transfer matrix case of a perturbation ∆ and that satisfies
as in Fig. 5, the structured singular value with respect to ∆ is
defined by: Fl [ Fu ( P,), K  1. (10)

225
2019 IFAC CSGRES
Jeju, Korea, June 10-12, 2019 B. Alharbi et al. / IFAC PapersOnLine 52-4 (2019) 200–205 203

3. CONTROLLER DESIGN FOR BOOST CONVERTER

3.1 Controller Concept

The controller K can be designed by solving Riccati


equations (Skogestad et al., 2005), or by a series of LMIs
(Dullerud and Fernardo, 2013), which mimics the standard µ-
synthesis procedures in robust control theory. After the
design process is done entirely, the robust controller is
fulfilled using µ-synthesis method by solving the mixed
sensitivity problem.

3.2 Weighting Functions Selection


Fig. 7. Weighting function performance.
Designing the appropriate weights for the plant in order to According to (5), the significance of the magnitude response
achieve a desired attenuation at certain frequencies is the of S lies beneath the magnitude response of 0.8, and the
challenging task in robust controller design. magnitude response of T should lie less than the response of
0.75, Fig. 8 reveals that these conditions are proved using H∞
If G( jw) K ( jw) is large enough at frequencies of controller. The objective of the weighting functions is to
expected disturbance, the sensitivity of the output of those manifest the importance of fulfilling performance criteria
frequencies will be very small and the disturbance will be within significance frequency range.
rejected. The weight factor is needed to have high enough
gains at disturbance frequencies in order to achieve the 3.3 Robust Analysis
desired loop shape. The general guidelines for weighting
functions choice used in this paper and proposed in The robustness properties can be evaluated by implementing
(Skogestad and Postlethwaite, 2005) and (Gu et al., 2005) the appropriate µ-tests for the uncertain feedback system
s displayed in Fig. 9. The transfer functions Wdel and ∆
 wb
Mp (11) parameterize the multiplicative uncertainty at the converter
W1 
s  wb  e p input. Presume the transfer function Wdel is known, and the
transfer function ∆ is stable.
where W1 is the bound for the sensitivity function and reflects
external disturbance rejection, a steady state error. ep is the
maximum steady-state offset, wb is the demanded bandwidth
and Mp is the peak value of the sensitivity, therefore
0.7 s  6300
W1 
s 1

W2 scales KS that mirror the control input which affects the


converter slightly. To avoid this impulsive input effect on the
converter, W2 is chosen as W2=0.001.
The complementary sensitivity function T can be shaped by
W3 which must be large at high frequencies to limit T for a
stable system,
wb
s
Mp . (12) Fig.8. Mixed sensitivity control design results.
W3 
e p s  wb

To shorten the closed-loop bandwidth, Mp is selected small


Delta*Wdel
and to maintain the stability of the system, the
complementary sensitivity, needs to be low at high frequency
range. For that, ep is selected as 0.1µ and thus r y
+ Controller + System
s  6200 . -
W3 
s  120

Fig. 7 shows the frequency response of weighting functions


W1 and W3. Fig. 9. Closed-loop system with input multiplicative
uncertainty.

226
2019 IFAC CSGRES
204
Jeju, Korea, June 10-12, 2019 B. Alharbi et al. / IFAC PapersOnLine 52-4 (2019) 200–205

The uncertainty weight Wdel is chosen as


s  7 107
Wdel 
s  7 108
3.4 Robust Stability and Performance

Theorem 1: (Robust stability with µ)

Suppose N  Fl ( P, K ) is robustly stable with reference to


 i with i 
 1 , if and only if supw i ( N ( jw))  1 .
Robust stability of the plant was checked based on theorem 1.
The µ-value for robust stability is (0.4005) which means the
allowable structured perturbations with norm less than Fig. 11. Performance analysis with worst case conditions.
(1/0.4005). MATLAB code robstab reports that the closed-
loop system remains stable up to 4 times of the possible
values. 4. MATLAB/SIMULATION RESULTS

Theorem 2: (Robust performance with µ) Applying ‘dksyn’ from MATLAB robust control toolbox and
parameters stated in Table 1, the controller is given by
Suppose N  Fu ( N , i ) with stable lower fractional
A Bk 
transaction (N) that satisfies the performance condition K ( s)   k (13)
Ck Dk 
Fu ( N , i )   1 for all  i with i 
 1, if and only if
supw  ( Fl ( P, K )( jw))  1 . where

The robust performance is achieved based on Theorem 2, if  1 0 0 0 


and only if, the closed-loop frequency response is less than  0 120 64 0 
one for each frequency calculated. The robust performance Ak   
 2.463 109 3.008 106 3.008 107 0.01202
test for the controller is shown in Fig. 10, which shows robust  
performance is achievable with peak µ value (0.4005) which  0 0 4444 1000 
shows that the allowable structured uncertainties with norm
less than 1/0.3002. 0.1508
 0 
These finding correspond with the worst case procedure as Bk   
shown in Fig. 11, which ensures the stability of the plant in  0 
existence of parameter variations.  
 0 

Ck  5.854 108 7.149 105 2.183 106 188.6


Dk  0

Four iterations were performed resulting peak µ-value and


gamma. Table 2 shows the summery of iteration.

Table 2. DK-iteration summery


Iteration # 1 2 3 4
Controller Order 7 51 17 21
Total D-Sclae Order 0 44 10 14
Gamma Achieved 669.6634 24.908 4.200 35.695
Peak mu-Value 0.778 0.654 0.726 11.415

The robust stability and robust performance characteristics of


the closed-loop system can be evaluated. According to the
Fig. 10. Assessment of robust performance.
results, the controller achieves robust performance with µ=
0.4005. The step response of µ-synthesis controller is shown

227
2019 IFAC CSGRES
Jeju, Korea, June 10-12, 2019 B. Alharbi et al. / IFAC PapersOnLine 52-4 (2019) 200–205 205

in Fig. 12. The µ-synthesis controller shows better dynamic


responses over PID controller as shown in Fig. 13.
The model is simulated by using SIMULINK (Fig.14) that
compares both controllers. The output voltages is displayed
in Fig. 15. It is obviously shown that µ-synthesis controller
has less overshoot and less time settling. The load is changed
at 0.04 and 0.08 seconds while the system still stable.

Fig. 15. Simulation result for output voltage.

5. CONCLUSION
A robust controller design method using µ-synthesis for
uncertain models of a boost converter circuit was derived. It
exhibits several improvements compared to traditional design
methods. In particular, a fast step-response to input voltage
disturbances and rapid settling times were achieved. The
effectiveness of the design method were demonstrated
through detailed simulation results using Simulink-SimPower
Systems. It was found that the structured singular value, µ, is
useful for controlling boost converter circuits.
REFERENCES
Fig. 12. Step response of the controller.
Abdlrahem, A., Saraf, P., Hadidi, R., Karimi, A., Sherwali,
H., and Makram, E. (2016). Design of a fixed-order
robust controller using loop shaping method for
damping inter-area oscillations in power systems, 2016
IEEE Power and Energy Conference at Illinois (PECI),
Urbana, IL, pp. 1-6.
Balas, G.J., Doyle, J., Glover, K., Packard, A., and Smith, R.
(2001). µ-Analysis and Synthesis toolbox For Use with
MATLAB. MUSYN Inc, and The MathWorks, Inc.
Baskin, M., and Bulent, C. (2015). μ-Approach based robust
voltage controller design for a boost converter used in
photovoltaic applications. IEEE 41st Annual Conference
IECON 2015.
Doyle, J. (1982). Analysis of feedback systems with
structured uncertainties, IEE Proceedings D-Control
Fig. 13. Comparison of PID and µ-synthesis controllers. Theory and Applications.
Dullerud, G., and Fernando, P. (2013). A course in robust
control theory: a convex approach. Vol. 36. Springer
Science & Business Media.
Gu, D. W., Petkov, P.H., and Konstantinov, M.
(2005). Robust Control Design with MATLAB®.
Springer Science & Business Media.
Ioannou, I., and Jing, S.(1996). Robust adaptive control. Vol.
1. Upper Saddle River, NJ: PTR Prentice-Hall.
Lin, F. (2000). An optimal control approach to robust control
design. International Journal of Control 73.3.
Pilat, A., and Wlodarezyk P. (2011). µ-synthesis and Analysis
of the Robust Controller for the Active Magnetic
Fig. 14. Boost closed loop schematic simulation with Levitation system, Automatica.
Matlab/Simulink. Skogestad, S., and Postlethwaite, I. (2005). Multivariable
Feedback Control: Analysis and Design. New York, NY,
USA: Wiley.
Tewari, A. (2002). Modern control design. NY: John Wiley
& Sons.

228

You might also like