Regulation and Compensation of Source Harmonics

for the Boost Converter–Based Power Factor

Precompensator
G. Escobar, A.M. Stanković D.J. Perreault
Dept. of Electrical Engineering and Computer
Dept Electrical and Computer Eng
Science
Northeastern University
fgescobar,[email protected] Massachusetts Institute of Technology
[email protected]

Abstract—In this paper we present an adaptive controller for the boost– tor can be made very close to unity as long as the tracked cur-
based power factor precompensator which guarantees fast regulation of the rent signal is a scaled replica of the input voltage. Hence, the
output voltage towards a desired constant value with a power factor close
to unity. This twofold control objective is fulfilled even in the presence of source will see the controlled system as the same equivalent re-
harmonics on the voltage source and uncertainties on the system param- sistor at each harmonic frequency. Our solution considers the
eters and load. The key for the solution to this problem is to express the main parameters of the system (the capacitance and the induc-
model in terms of the input current instead of the inductor current. The
resulting controller is reduced, through transformations, to the cascade in-
tance) and of the applied load as unknowns; we also allow for
terconnection of two controllers, namely the inner and the outer control harmonics in the voltage source. While in the case of known
loop. It is shown that while the latter turns out to be a simple lead–lag plus system parameters the problem can be solved with conventional
integration, the former is composed mainly of second order filters tuned control techniques, the required bandwith of the current control
at the frequencies of the considered harmonics and with transfer functions
that follow a well defined pattern. Simulations are provided to assess the loop can easily become excessive. As an alternative, we devel-
performance of the proposed controller. oped a controller that utilizes the information about the structure
Keywords— AC-DC power conversion, power supplies, reactive power, of the system and disturbances to improve performance, and to
dissipative systems, adaptive control, nonlinear systems. significantly reduce the bandwith of the current loop.
The use of a system model representation in terms of the in-

R
I. I NTRODUCTION put current, instead of the usual inductor current, is instrumental
EGULATION of switched power converters is an active for our developments. This allows us to treat the problem of
area of research, both in the power electronics area [2], harmonic contents in the input voltage in a more natural way.
[3] and in automatic control theory [4], [5]. This is due to the The input voltage can then be expressed in the form of Fourier
fact that power converters are, generally speaking, a ubiquitous series, where the coefficients are unknown constants. The re-
power source whose applicability ranges from electrodomestics sulting controller will have a familiar and simple form which is
and digital computers to industrial electronics and sophisticated suitable for implementation, where the most relevant feature is
communications equipment. From the theoretical viewpoint, the introduction of a bank of second order filters, with resonant
they also constitute an interesting class of discontinuous non- frequencies corresponding to the harmonic under consideration.
linear systems regulated by means of a commanded switch po-
sition function. These features make switched power converters
attractive for both theoretical and practically–oriented studies. II. S WITCH - REGULATED BOOST CONVERTER AS A PFP
In this paper we explore the performance enhancement of
In this section we formulate the control problem of the PFP
Power Factor Precompensators (PFP) via adaptive nonlinear
whose circuit is shown in Fig. 1.
control techniques. The topology of the PFP circuit studied in
the present work consists of a diode bridge and associated boost
converter. This circuit is the most widely employed of the fam- L
ily of PFP’s even though it exhibits certain drawbacks such as
the slight deformation in the signal around the zero crossing.
i
We propose an nonlinear feedback controller designed fol- + L

lowing a dissipativity approach to which adaptation has been C

vs (t)
added to cope with parametric uncertainties. The closed loop ii vi (t) +
vC P0
performance accomplish the twofold control objective: first, to -
achieve a nearly unit power factor at the input of the converter,
and second, to achieve efficient load voltage regulation to a de- -

sired constant level. By defining a reference signal tracking

problem on the input current of the converter, the power fac-
 Corresponding author. Fig. 1. Switch–regulated PFP Boost circuit
 vC2
The differential equations describing the circuit dynamics are z2 = ; e = uvC
d
2
L iL = uvC + vi

dt  Thus, in open sets excluding the zero crossing points1 , i.e., 8t
d vC2 such that ii (t) 6= 0, the model can be rewritten as
C
dt 2
= uvC iL P0 (1)
d
where iL and vC are the inductor current and the capacitor volt- L ii
dt
= sign(ii )e + vS (4)
age variables, respectively; notice that iL = jii j with ii the input d
current (the current on the ac side); vi (t) = sign(ii )vS is the C z = ejii j P0 (5)
dt
voltage measured at the diode bridge output; P0 represents the
output power load, it may be a simple constant power source, where e, the voltage across the transistor, represents the actual
it may also include the effect of a load resistance or simply a control input. Moreover, to accomplish the control objective, z
constant current source; C and L are the capacitance and in- should be driven towards Vd2 =2. We will assume that the dy-
ductance of the circuit, respectively; u, which takes values in namics of ii is much faster than the dynamics of z , and we will
the discrete set f0; 1g, denotes the switch position function and thus treat both dynamics separately for control design purposes.
acts as the control input. For the controller design purposes we
will consider the averaged model, i.e., the signal u, originally of A. Inner control loop
discrete nature, will be considered as a continuous signal repre- In this subsection we design a controller which guarantees
senting the duty ratio of a PWM switching sequence generated tracking of ii towards its desired reference ii computed as in
at a relatively high frequency. (2). It is straightforward to show that the following controller
The control objective is twofold. First, in order to guarantee stabilizes subsystem (4), and guarantees that ii tracks its desired
a power factor near unity, the input current ii should follow a reference ii
signal proportional (same shape and phase) to vS , i.e.,
 
ii ! ii = gvS (2) e = sign(ii )
d
L ii + vS + K1~ii (6)
dt
where g is a gain yet to be defined. This gain represents the
conductance of the equivalent resistor seen by the voltage source where ~ii = ii iid and K1 > 0 is a design parameter.
for a given load P0 under a unitary power factor functioning. Notice that, both the time derivative of ii and the parame-
Second, the dc component of output vC should be driven to ter L are required in order to implement the controller above.
some constant desired value Vd > V . Here and in what fol- In what follows we will show how this term can be estimated
lows we consider the dc component as the average of a sig- by means of adaptation using the description of source voltage
nal taken over a period of the fundamental, that is, hx(t)i0 =
4 vS in Fourier series (its harmonic components) to simplify this
Rt
T
1
(t T )
x( )d . computation.
We will assume that the system parameters L, C and the load Using (2) and (3) we can develop the term containing the time
power P0 are unknown quantities that may vary slowly or in derivative as follows
steps due to changes in the system. Moreover, we will assume d X
that the source voltage can be described with Fourier series L ii = L (g v_ S + gv
_ S) = >
k L (_
g kwg J ) VS;k (7)
dt k2H
X
vS = >
k
VS;k (3)
k 2H where we have used the fact that
X h i
where     v_ S = kw> J VS;k ; J = J> = 0 1
= cos (kwt) Vr k 1 0
k
sin(kwt) ; VS;k = V
S;k
i k 2H
S;k

numbers VS;kr
, VS;k
i
2 IR are the kth harmonic coefficients Now, we define the vector
of the Fourier series description of the source voltage. They
are also assumed unknown constants (or slowly varying) and  = L (_g
k kwg J ) VS;k ; k 2 H (8)
H = f1; 2; 3; :::g is the set of indices of the harmonic compo- which for each k 2 H practically converges towards a constant2 .
nents considered. Superscripts (
)r and (
)i are used to distin-
guish the coefficients associated with cos (kwt) and sin(kwt), Eq. (7) can be further reduced to
respectively. d X
L ii = > k
III. C ONTROLLER DESIGN dt k2H
k

To design a controller that considers a vS with harmonic con- 1 We point out that not much attention is given at the zero crossing points ( =
i i
tents we find it more convenient to rewrite the model above us- 0) in the control design since, as will be shown later, the controller has no other
ing the following coordinate transformations option than taking a zero value at all these points due to physical limitations.
2 Ideally and _ should vary slowly and take constant values in the steady
ii = sign(ii )iL vi = sign(ii )vS
g g
; state.
where vector k is unknown. Thus, we propose to use an es- B. Outer control loop
^ k in the control expression (6) above, this yields the
timate 
Direct substitution of controller (9) and (11) in the second
controller
! subsystem (5) yields the following system (in terms of the in-
X crements of z )
e = sign(ii ) ^ + v + K1~i
> 
k k S i (9)
2H X
k
C z~_ = ii >
k
^ k + jii jvi + ii K1~ii P0 (14)
Subsystem (4) in closed loop with controller (9) yields the k 2H
following error dynamics
2
d X where z~ = z
V

L ~ii = > ~ K1~i

d

k k
2
.
(10)
dt k2H
i
As pointed out before, we consider that the dynamics of the
subsystem (10) are much faster than the dynamics of subsystem
where 
4
~ k = ^ k k . (14), and moreover, that the controller e is bounded, which is
~ k we
To deal with the terms associated with the error signals  ^ k (8k 2 H) are bounded. Thus, in a relatively
true if all terms 
propose the following energy storage function ^ = , and the model reduces
short time, practically ~ii = 0 and 
 2 
= L2 ~i2 +
X 1 ~
2
+ ~
 to X
W r i
C z~_ = gvS > k + gvS2 P0
i

k 2H 2 k
k k

k 2H
k

whose time derivative along the trajectories of (10) is given by where we have used the fact sign(ii ) = sign(vS ).
X X _>
~ ~ Moreover, since we are mainly interested in the behavior of
W_ = K1~i2i + ~ii > ~
k k +
k k
the dynamics of the dc component of z~, we should neglect the
k 2H k 2H k higher order harmonics at the right hand side of the equation
which is forced to be negative semidefinite if the error on the above, this yields
X
C z~_ = hgv > k i0 + ghvS2 i0
estimates is constructed according to the following adaptive law
P0
^_ ~i cos (kwt) ; k 2 H
r S k

k = k i k 2H
^_ ~i sin(kwt) ; k 2 H Notice that hvS2 i0 is nothing else than the square of the RMS
i
= k i

= hvS2 i0 .
k

or in a more compact form value of vS , i.e., vS;RM

2
S
The first term on the right hand side (RHS) can be rewritten,
^_ =
k k i
~i  k ; k2H (11) using (3) and (8), as
where k > 0, k 2 H are design parameters. X X X
It is easy to see that ~ii ! 0 as t ! 1 as long as e is well
hgv S >
k
k i0 = h >
k
gVS;k
>
k
Lg_ VS;k i0
defined for all t. Moreover,  ~ ! 0 as t ! 1 as long as ~iL ! k 2H k 2H k 2H
0. Unfortunately, as will become clear later, this can only be X X
guaranteed in open sets of time, since e, which is restricted to h >
k
gVS;k
>
k
LgkwJ VS;k i0 (15)
take values only in a positive interval, attempts to take negative k 2H k 2H
values at the beginning of every half cycle. This problem can be We observe that the second term at the right hand side of (15)
alleviated if, for instance, L is chosen very small. will contain the products of orthogonal rotating vectors at the
The controller (9) with adaptive laws (11) can be further sim- same angular speed, plus harmonics components of higher or-
plified using the following transformations der, thus its dc component will be zero. On the other hand, the
r
k
= >
k
^ k ; k 2 H first term on the RHS contains harmonic components of higher
i
k = >
k J  ^k ; k 2 H order plus products of colinear rotating vectors which will pro-
duce squares of sinusoidal functions (and thus a dc component)
The controller (9) is reduced to plus higher order harmonics. Thus (15) is reduced to
!
X X
2
> >
X
e = sign(ii ) r
+ v + K1~i
S i (12) hgv S k
k i0 = Lgg_ h k
VS;k i
0
2H 2H
k
k 2H k k

2
and the adaptive laws can be rewritten as
_ ~i
but notice that h > VS;k i =j Vk j2 and thus
kw ; k 2 H
0
r
k = k i
i
k
k

X
2

_ i
k
= kw k ; k 2 H
r
h > V
k S;k

2
i =v
0
2
S;RM S

Expressing the dynamic extension (the adaptations) in the k 2H

form of a transfer function ~ii 7! rk , since in the controller It is common practice in applications to obtain ii as follows
above only the terms rk (k 2 H) appear, this yields
r
= s2 + ks2 w2 ~i
k
; k2H (13) ii =
GvS
(16)
vS;RM
k i 2
S
this is equivalent to make the following transformation in our e +
e
developments vC
G = gvS;RM
2
S
u ii
sgn(.)
+ vS
ii* +
This simple useful transformation keeps the values of most vari-
ables on the same order of magnitude, and thus reduces the error _
+
in numerical computations. Notice that the value of g is usually vC ~ K1
very small. ii +
Finally, the error model can be written as (.) 2
s γ1 +

LGG_
_ s2+w 2
C z~_ = +G G
P0 (17) K is+K i b-Kpb 1st Harmonic
+
vS;RM
2
S
2s(s+b)
+
We propose to compute G as (.) 2 vS vS
s γ2 +

G_ Ki0 z~ Kp0 
= vS2,RMS s2+4w 2
(18) +
2nd harmonic
sa
 =
s+b
z~ (19) Vd vS2,RMS

where s is the complex variable and Kp0 , Ki0 , a and b are positive
Outer control loop Inner control loop

design parameters.
The form of this controller is motivated from the form of a
simple PI, where  represents the signal z~_ filtered by means of
Fig. 2. Block diagram of the proposed controller composed by the inner and
outer control loops
a first order filter which is denoted in control literature as the
“dirty derivative”. We observed that direct use of z~ in the com-
putation of g (using a normal PI) causes the introduction of more IV. E FFECTS OF SWITCHING ON THE POWER FACTOR
harmonics which will in principle deform the shape of reference In this section we study system limitations due to constraints
ii causing the degradation of the power factor. The controller on the control input u 2 [0; 1]. For simplicity we consider
(18)–(19) can be rewritten in a more convenient form as follows a sinusoidal voltage source, i.e., vS = V sin(wt) (vS;RM
2
S
=
G_= K z~ + K  (20) V =2) and we assume that in the steady state
2

 vC = Vd ,
i p

_
= b(~z  ) (21)
 ii = ii (at least at the end of every half cycle)
where Ki = K 0 + K 0 a and K = K 0 a.
i p p p  g = 2 20 P

 ^ 1 = [2wLP0 =V; 0]>

V
It can also be written in the form of a transfer function having
as input z~ and output G The equivalent controller, i.e., the controller that keeps zero
G Ki s + (Ki Kp )b tracking error is given by
z~
= s(s + b)
which turns out to be a simple lead–lag type controller plus an ueq = VV j sin(wt)j 2wLP
VV
0
sign(sin(wt)) cos(wt)] (24)
d d
integrator.
Let us study now the local stability of the closed loop system Let us focus only on the first half cycle, i.e., 0  wt   . We
composed by (17), (20) and (21). Linearization of these equa- observe from (24) that ueq takes negative values at the beginning
tions around the equilibrium point [~ z ; G;  ]T = [0; P0 ; 0]T , of each half period. The controller is thus maintained in u = 0
gives and we can solve for iL from (1) considering iL (0) = 0, this
2 3 2 32
z~_
3 yields
z~
LP0 Kp
LP0 Ki 1
V
6 ~ _ 75 = 64 Cv 2 C C v2
74 ~5 iL (t) =
cos (wt)) (1
0
S;RM S S;RM S
4 G Ki Kp 5 G (22) wL
_ b 0 b  The controller is maintained at u = 0 until the trajectory of
4 iL reaches the tracking reference signal that occurs at wt =
where G~ = G G.  
System (22) is stable provided the following conditions on the
design parameters are fulfilled
= 2 arctan 2wLP0 (25)
V2
CbvS;RM
2
Ki > Kp ; Ki < S The trajectory of the inductor current, in steady state, is ap-
LP0 proximately
CvS;RM
2 
S
Kp > Ki (C + Ki ) 
= Ki2 (23)
iL (t) =
V
(1 cos (wt)); 0  wt 
LP0 b wL
2P0
sin(wt); < wt  
(26)
V
where the last condition is a little conservative.
Fig. 2 presents a block diagram of the inner control loop in- The ac–line current ii (t) given by
tegrated by (12), the bank of second order filters (13) and the
outer control loop composed by a lead–lag plus integrator filter. ii (t) = iL (t)sign(sin(wt)) (27)
 xi To test the robustness of the proposed controller against dis-
turbances in the load we have applied a step change in both,
the current load and the load resistance. The system starts with
π+β 2π R = 2000 and i0 = 0 Amp, then at t = 1:5 we change
R = 2000 to R = 1000 preserving i0 = 0 Amp, finally at
β π
t = 3:5sec we introduce i0 = 0:2 Amp, preserving R = 1000 .
wt

450

Fig. 3. Input current time response.

vC [Volt]
400

has the alternate symmetrical form shown in Fig. 3

Therefore perfect tracking of the current can only be guaran-
teed in open intervals, remaining thus a pulsating alternate signal
on the tracking error ~ii .
350
1.5 2 2.5 3 3.5 4 4.5 5

To avoid possible errors in the estimation of k ; k 2 H due to

0.025
the unavoidable pulsating current tracking error, we propose to
0.02
freeze the adaptation on these intervals, where, as stated before,
u attempts to take negative values. This can be carried out by 0.015

g [Ω−1]
selecting k as follows 0.01

; u > 0; k 2 H
=
0.005
k

0 ; u  0; k 2 H
k (28)
0
1.5 2 2.5 3 3.5 4 4.5 5
Time [s]
whose effect, according to the actual implementation presented
in the point (i) above, consists in disconnecting the second order Fig. 5. Time responses for (Top) voltage vC and (Bottom) g , starting with
filters (13) from the controller of Fig. 2. R = 2000 and i0 = 0, then at t = 1:5sec we change R = 2000 to
R = 1000 preserving i0 = 0 Amp and finally at t = 3:5 we introduce
i0 = 0:2 Amp preserving R = 1000 .
V. S IMULATION RESULTS

Computer simulations were performed to evaluate the pro-

posed feedback controller. We used a resistor and a current
1
source as the output load, as shown in Fig. 4 (controller deriva-
tion is completely analogous to the case of constant power load).
(a)

The system parameters were L = 1mH, C = 450F. The

0

source voltage is composed of the fundamental, 2nd and 3rd −1 0.0075*vS

1.4 1.41 1.42 1.43 1.44 1.45 1.46 1.47 1.48 1.49 1.5
harmonic

vS (t) = 162:6 sin (wt) 15 cos (2wt 0:25) 10 cos (3wt 0:2) 2
(b)

where w = 2  60 rad/s and its RMS value is vS;RM S = 115:7.

0

The desired output voltage is fixed to Vd = 400 Volts with a −2 0.015*v

S

maximum output power P0;max = 250 W. The design parame- 2.9 2.91 2.92 2.93 2.94 2.95 2.96 2.97 2.98 2.99 3

ters were selected as follows: K1 = 15, Kp0 = 2:5 (Kp = 3:75), 4

Ki0 = 0:1 (Ki = 3:85), a = 1:5, b = 450, 1 = 100, 2 = 200, 2

3 = 300. Notice that these parameters largely fulfill conditions

(c)

(23). −2
0.02*vS
−4
4.9 4.91 4.92 4.93 4.94 4.95 4.96 4.97 4.98 4.99 5
L Time [s]

iL Fig. 6. Current ii (t) and scaled vS (t) in dotted line, from top to bottom:
(a) Starting condition R = 2000 and i0 = 0 Amp, (b) Changing load
+

vs (t) C R i resistance R = 1000 with i0 = 0 Amp and (c) Introducing a current

i0 = 0:2 Amp with R = 1000 .
0
ii +
vi (t)
vC
-

-
Fig. 6 shows the time responses of voltage vC and the ampli-
tude g of the reference current under the proposed dissipativity
control for the conditions of the test mentioned above. For im-
plementation we have used G as described before, thus, we have
Fig. 4. Simulated circuit. divided G by vS;RM
2
S
to recover g which appears in Fig. 6. Then
in Fig. 5 we observe the steady state responses of the current troller guarantees fast regulation of the output voltage towards a
iL together with the scaled input voltage vS for the same three desired constant value with a power factor close to unity. These
conditions previously mentioned. In Fig. 7 we exhibit the pro- objectives are fulfilled even in the presence of harmonics on the
portional relationship that is established between ii (t) and vS (t) voltage source and uncertainties on the system parameters and
for the three situations shown in Fig. 5, where the magnitude of load. The overall controller consists of a cascade interconnec-
the slope coincide with the value of g , i.e., (a) g = 0:006 1, tion of two compensators, namely the inner and the outer control
(b) g = 0:012 1 and (c) g = 0:018 1 . In Fig. 8 we com- loop. It is shown that while the latter turns out to be a simple
pare the tracking errors between the proposed controller (top) lead–lag plus integration, the former is composed of second or-
and a controller where the bank of 2nd order filters are substi- der filters tuned at the frequencies of the considered harmonics.
tuted by a conventional PI controller (bottom). We observe that Several simulations are provided to assess the performance of
the proposed controller has significantly smaller error; the er- the proposed controller.
ror observed here is due mainly to the unavoidable distortion
at zero crossings. This distortion increases as the load demand R EFERENCES
increases, as predicted by the analysis. [1] D. Chevreau and C. Marchand. Pollution harmonique du réseau: comparai-
son de deux redresseurs monophasés. In Proc. 9th International Colloquium
on CEM, pages F11–F16, Brest, France, 8-11 June 1998.
4 [2] D. Czarkowski and M. K. Kazimierczuk. Energy-conservation approach to
modeling PWM DC-to-DC converters. IEEE Trans. on Aero. and Elect.
3 Syst., 29:1059–1063, 1993.
[3] M. E. Elbuluk, G. C. Verghese, and D. E. Cameron. Nonlinear control of
2 switching power converters. Control Theory and Advanced Technology,
5(4):601–617, 1989.
ii [Amp]

1
[4] G. Escobar and H. Sira-Ramirez. A passivity based-sliding mode control
0 approach for the regulation of power factor precompensators. In Proc. IFAC
NOLCOS, Enschede, The Netherlands, 1998.
−1 (a) [5] R. Ortega, A. Loria, P. J. Nicklasson, and H. Sira-Ramirez. Passivity–based
control of Euler–Lagrange systems. Springer-Verlag, 1998.
−2 (b)
(c)
−3
−200 −100 0 100 200
vS [Volt]

Fig. 7. Current ii (t) versus vS (t) for the three situations: (a) Starting condition
R = 2000 and i0 = 0 Amp, (b) Changing load resistance R = 1000
with i0 = 0 Amp and (c) Introducing a current i0 = 0:2 Amp with R =
1000 .

0.15

0.1

0.05
(ii−ii ) [Amp]

0
*

−0.05

−0.1

1.5 2 2.5 3 3.5 4 4.5 5

0.15

0.1

0.05
(ii−ii ) [Amp]

0
*

−0.05

−0.1

1.5 2 2.5 3 3.5 4 4.5 5

Time [s]

Fig. 8. Current tracking error ~ii (t) for (Top) proposed controller and (Bottom)
replacing the bank of 2nd order filters by a conventional PI.

VI. C ONCLUSIONS
The paper presents an adaptive controller for the power fac-
tor precompensator based on boost converter topolgy. The con-