See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/224188668

Voltage Regulation of Photovoltaic Arrays: Small-Signal Analysis and Control

Design

Article  in  IET Power Electronics · December 2010

DOI: 10.1049/iet-pel.2008.0344 · Source: IEEE Xplore

CITATIONS READS

131 1,134

3 authors:

Marcelo Gradella Villalva Thais Gama De Siqueira

University of Campinas Universidade Federal de Alfenas
101 PUBLICATIONS   3,501 CITATIONS    26 PUBLICATIONS   250 CITATIONS   

SEE PROFILE SEE PROFILE

Ernesto Ruppert Filho

University of Campinas
188 PUBLICATIONS   3,998 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Switched Reluctance Generator View project

HTS fault current limiter View project

All content following this page was uploaded by Marcelo Gradella Villalva on 21 May 2014.

The user has requested enhancement of the downloaded file.

 www.ietdl.org

Published in IET Power Electronics

Received on 16th December 2008
Revised on 27th January 2010
doi: 10.1049/iet-pel.2008.0344

ISSN 1755-4535

Voltage regulation of photovoltaic arrays:

small-signal analysis and control design
M.G. Villalva1 T.G. de Siqueira2 E. Ruppert1
1
School of Electrical and Computer Engineering, University of Campinas – UNICAMP, Campinas – São Paolo, Brazil
2
Department of Engineering, University of Alfenas at Poços de Caldas – UNIFAL, Poços de Caldas – Minas Gerais, Brazil
E-mail: [email protected]

Abstract: This study deals with the regulation of the output voltage of photovoltaic (PV) arrays. As a case study,
the DC – DC buck converter is used as an interface between the PV array and the load, but other types of
converters can be used for the same purpose. The input voltage of the converter is controlled in order to
regulate the operating point of the array. Besides reducing losses and stress because of the bandwidth-limited
regulation of the converter duty cycle, controlling the converter input voltage reduces the settling time and
avoids oscillation and overshoot, making easier the functioning of maximum power point tracking (MPPT)
methods. The voltage regulation problem is addressed with a detailed analysis that starts with the modelling
of the PV array and the converter. This analysis is followed by study, design, simulation and practical
experiments of three closed-loop control strategies for the buck converter. Control stability and
implementation considerations are presented.

1 Introduction proposed to enhance MPPT methods in order to obtain

performance improvements. However, the achievement of
A photovoltaic (PV) array converts sunlight into electricity. MPPT strongly depends on the performance of the
The voltage and current available at the terminals of the converter control and its ability to regulate the operating
PV array may directly feed small loads such as lighting point of the PV array.
systems and DC motors. More sophisticated applications
require electronic converters to process the electricity from The PV array operating point can be adjusted by regulating
the array. These converters may be used to regulate the the voltage or current at the terminals of the array. The
voltage and current at the load, to control the power flow voltage control is preferred because the voltage at the MPP
in grid-connected systems and mainly to track the is approximately constant [1]. The PV current, on the
maximum power point (MPP) of the array. other hand, changes greatly when the solar irradiation varies.

Converters with the maximum power point tracking This paper addresses the control problem illustrated in Fig. 1.
(MPPT) feature use an algorithm to continuously detect The PV array feeds the DC–DC buck converter. The output of
the maximum instantaneous power of the PV array. As the the converter is represented by a constant DC voltage source Vo
operating conditions of the array (solar irradiation and that represents a battery or a DC link for another cascaded
temperature) may change randomly during the operation of converter. The buck converter serves as an interface between
the system, an MPPT algorithm is necessary so that the the PV array and the voltage Vo . The MPPT block (not
maximum instantaneous power can be extracted and studied in this paper) provides a voltage reference and the
delivered to the load. voltage controller regulates the PV array voltage.

Many MPPT methods have been proposed in the Some works in the literature present PV systems with the
literature. Many theoretical improvements and even direct duty cycle control of the DC – DC converter [2 – 4],
advanced techniques of artificial intelligence have been where the MPPT block directly actuates on the converter

IET Power Electron., 2010, Vol. 3, Iss. 6, pp. 869– 880 869
doi: 10.1049/iet-pel.2008.0344 & The Institution of Engineering and Technology 2010
www.ietdl.org

of the array, Vt = Ns kT /q is the thermal voltage with Ns cells

connected in series (k is the Boltzmann constant, q is the
electron charge and T is the temperature of the PV cells),
Rs is the equivalent series resistance, Rp is the equivalent
shunt resistance and a is the ideality constant of the diode.
The parameters of (1) may be obtained from practical
measured characteristics of the PV array. An explanation on
the PV array modelling and the determination of the
parameters of the i × v equation may be found in [8, 9].

A linear PV array model is necessary in the analyses that

Figure 1 Input-regulated converter interfaces the PV array will be done in the next sections. The nominal i × v curve
with the output Vo , which may be a battery or a is linearised at the MPP as shown in Fig. 2.
cascaded converter
The derivative of the non-linear i × v curve at a given
duty cycle without employing a voltage or current controller. In point (V, I ) is
this kind of system, no appropriate voltage or current regulation  
is achieved and the converter is subject to increased switching I0 V + IRs 1
g(V , I ) = − exp − (2)
stress and losses [5]. A feedback controller with a Vt N s a Vt N s Rp
proportional and integral (PI) compensator for regulating the
converter voltage is preferred. Besides reducing losses and The linear model is described by the line tangent to the i × v
stress due to the bandwidth-limited regulation of the duty curve at the linearisation point (V, I )
cycle, the presence of the controller reduces the settling time
of the converter and avoids oscillation and overshoot, making ipv = ( − gV + I ) + gvpv (3)
easier the functioning of MPPT methods [6].
and may be represented by the equivalent circuit of Fig. 3,
2 PV array modelling where Req = −1/g and Veq = V − I /g.
Photovoltaic devices present the non-linear i × v characteristic
The circuit of Fig. 3 is valid at the linearisation point (V, I )
illustrated in Fig. 2. This example shows the curve of the
and is a good approximation of the PV array for a small-
KC200GT [7] solar array.
signal analysis at the vicinities of this point. The dynamic
behaviour of the system composed of the PV array and
The equation of the i × v characteristic is [8, 9]
buck converter depends strongly on the point of operation
    of the array. The system is designed and optimised to
vpv + Rs ipv vpv + Rs ipv
ipv = Ipv − I0 exp −1 − (1) operate at the MPP and later the design is checked for
Vt a Rp

where Ipv and I0 are the photovoltaic and saturation currents

Figure 3 Linear equivalent circuit valid at the linearisation

point

Table 1 Parameters of the KC200GT solar array

I0 9.825 × 1028 A
Ipv 8.214 A
a 1.3
Rp 415.405 V
Rs 0.221 V
Veq 50.9007 V
Figure 2 Non-linear i × v characteristic of the KC200GT
Req 3.2327 V
solar array and equivalent linear model at the MPP

870 IET Power Electron., 2010, Vol. 3, Iss. 6, pp. 869– 880
& The Institution of Engineering and Technology 2010 doi: 10.1049/iet-pel.2008.0344
 www.ietdl.org

other operating points. The operation in the current source Finally, the following equation relates the input voltage of
region of the i × v curve (see Fig. 2) presents the worst the converter and the capacitor voltage
and most critical dynamic behaviour [1, 10]. In a later
section, several closed-loop configurations will be analysed d
and they will be evaluated in all regions of the i × v curve. pv = RC
v  +v
v C (8)
dt C
Table 1 lists the parameters of the KC200GT array
obtained through modelling [8, 9]. 3.2 Small-signal model for
voltage control
3 Converter modelling The objective of modelling the converter for voltage control is
to obtain a small-signal transfer function that relates the
In this section, small-signal models of the buck converter fed ′
small-signal voltage v̂pv and the control variable d̂ = −d̂ .
by the PV array are developed. The method of average The minus signal is necessary because negative variations of
variables [11 – 14] is used to obtain small-signal converter duty cycle cause positive increments in the input voltage.
transfer functions. The first and second models describe the
behaviour of vpv and iL with respect to the duty cycle The modelling process is essentially composed of three
of the converter. The third model describes the behaviour steps: inserting the small-signal variables in the state
of vpv with respect to iL . equations, applying the Laplace transformation and
manipulating the equations in order to find the desired
3.1 Average state equations transfer function.
Fig. 4 shows the PV-buck system with average voltages and The small-signal variables are introduced with the following
, i) means the
currents. The bar over a variable name (e.g. v definitions
discrete-time average value of the variable within one
switching period of the converter. By writing the circuit
equations with average variables, the high-frequency C = VC + v̂C
v
components are eliminated and only the natural system pv = Vpv + v̂pv
v
behaviour remains. (9)
i L = IL + î L
The average capacitor state equation is d = D − d̂
Veq − v
pv d
−C  − i 1 = 0
v (4) where DC steady state values are capitalised and small signals
Req dt C
are marked with a hat.
The average inductor state equation is By substituting (7) and (8) in (4) and then replacing the
average variables by the definitions of (9) one finds
d
34 − RLi L − Vo − L
v i =0 (5)
dt L
IL d̂ + d̂ î L + VC /Req + Veq /Req − v̂C /Req − IL D −
The circuit constituting the transistor and diode may be d d
replaced by the average equivalent quadripole with − î L D − C v̂ − RC C/Req v̂C = 0 (10)
dt C dt
terminals 1 –2 – 3 – 4, which is described by the following
equations, where d is the duty cycle of the transistor From (10), by neglecting the non-linear product d̂ î L and
applying the Laplace transformation, the small-signal
34 = v
v 12 d = v
pv d (6) equation is obtained
i 1 = i 3 d = i L d (7)
IL d̂ (s) − v̂C (s)/Req −

− î L (s)D − sC v̂C (s) − sRC C/Req v̂C (s) = 0 (11)

Similarly, from (5), (6), (8) and (9), one gets

−VC d̂ (s) − RL î L (s) + v̂C (s)D − RC CVC sd̂(s) +

+ RC CDsv̂C (s) − sLî L (s) = 0 (12)

Figure 4 Buck converter with PV array linear model From (11), (12) and (8) one gets the voltage to duty cycle

IET Power Electron., 2010, Vol. 3, Iss. 6, pp. 869– 880 871
doi: 10.1049/iet-pel.2008.0344 & The Institution of Engineering and Technology 2010
www.ietdl.org

transfer function
′
Gvd (s) = v̂pv /d̂ (s)
(1 + sCRC )(IL + (VD(1 − sCRC ))/(RL + sL))
=
sC + (1 + sCRC )((1/Req ) + (D2 /(RL + sL)))
(13)

3.3 Small-signal model for current control Figure 5 Equivalent average circuit used to obtain Gvi(s)
The objective of modelling the converter for current control is to
obtain a small-signal transfer function that relates the buck obtain a very useful transfer function for v̂pv . The influence
converter inductor current iL and the converter duty cycle. of d̂ on the system is already embedded in the î L (s)/d̂ (s)
One must notice that in this case, unlike the voltage transfer transfer function and the system is analysed near the
function previously obtained, the control variable is d̂ because operating point corresponding to the steady-state duty cycle
positive duty cycle increments cause positive current increments. D, so one can make i 1 = i L D as shown in the circuit of Fig. 5.

The modelling process has the same steps used in the Veq − v
pv d
−C  − i L D = 0
v (16)
previous section, so the development of the inductor Req dt C
current transfer function will not be presented in detail. For
the current control, the following definitions are used By using (8) and (16) with the following small-signal
definitions
 C = VC + _
v vC
pv = Vpv + v̂pv
v  C = VC + _
v vC
i L = IL + î L (14) pv = Vpv + v̂pv
v
i L = IL − î L (17)
_
d =D+d
_
d =D+d
From the average state equations (4) and (5), with (8) and
(14), the following current to duty cycle transfer function is the desired transfer function is found
obtained
v̂pv (s) D(1 + sCRC )Req
Gid (s) Gvi (s) = ′ = (18)
î L (s) 1 + sCRC + sCRC Req
= î L (s)/d̂ (s)
 
IL D
= − V (1 + sCRC )/ 4 Open-loop analysis
sC(1 + (RC /Req )) + (1/Req )
  With the converter transfer functions Gvd (s), Gid (s) and
D2 (1 + sCRC ) Gvi (s), one can make some analyses of the PV-buck system
−RL − sL − (15)
sC(1 + (RC /Req )) + (1/Req ) that help understand the effects of RL and RC in the
system. Before these analyses are done, the transfer
3.4 Small-signal model for voltage and functions are validated by comparison with the responses of
current control a simulated switching converter built with the parameters
of Tables 1 and 2.
The objective of modelling the converter for voltage and
current control is to obtain a transfer function that relates
the input voltage v̂pv to the inductor current î L .
4.1 Validation of transfer functions
This transfer function will be used in the double-loop Figs. 6 – 8 show simulation results of the buck converter and
control scheme presented later, where both the current and the transfer functions. The converter switching frequency is
the voltage of the converter are controlled. In this scheme, 20 kHz and the circuit was evaluated with the PSIM
the control variable of î L is the duty cycle d̂ and the control simulator. The responses of the switching converter are
′
variable of v̂pv is î L = −î L . superimposed with the transfer function responses so that
one can verify the validity of the small-signal modelling in
′
The v̂pv (s)/î L (s) transfer function is obtained by the time domain. The responses match almost perfectly
assuming that small duty cycle perturbations directly affect and the results are reasonably good considering the
the current î L and the dynamics of v̂C is dictated by î L . In simplifications and assumptions made in the mathematical
this way, v̂C is decoupled from d̂ and this permits to easily modelling.

872 IET Power Electron., 2010, Vol. 3, Iss. 6, pp. 869– 880
& The Institution of Engineering and Technology 2010 doi: 10.1049/iet-pel.2008.0344
 www.ietdl.org

Table 2 Characteristics of the buck converter

L 2 mH
RL 0.1 V
C 3000 mF
RC 0.05 V
Vo 13.15 V
D 0.5

Figure 9 Bode plots of the simulated converter frequency

response and the transfer function Gvd (s) (dashed line)
superimposed

Figure 6 Simulated converter and Gvd(s) transfer function

responses to small-signal steps around the operating
point D ¼ 0.5

Figure 7 Simulated converter and Gid(s) transfer function

responses to small-signal steps around the operating
point D ¼ 0.5

Figure 10 Bode plots of the simulated converter frequency

response and the transfer function G id (s) (dashed line)
superimposed

The comparison of the time- and frequency-domain

responses of the converter and the transfer functions
confirms that the small-signal modelling procedure is
correct and the model transfer functions can be used to
Figure 8 Simulated converter and Gvi(s) transfer function analyse the converter and design closed-loop systems.
responses to small-signal steps around the operating
point D ¼ 0.5
4.2 Influence of RC and RL
Figs. 9 –11 show frequency responses of the switching In [1, 15] the problem of PV voltage regulation was
converter obtained with the ACSWEEP analysis from addressed with a boost converter using simplified converter
10 Hz to 10 kHz. The responses are superimposed on the models, where only RL is considered and RC is neglected.
same axes with the responses of the transfer functions However, both RC and RL are important parameters that
Gvd (s), Gid (s) and Gvi (s). must be considered in the model if one wishes for a

IET Power Electron., 2010, Vol. 3, Iss. 6, pp. 869– 880 873
doi: 10.1049/iet-pel.2008.0344 & The Institution of Engineering and Technology 2010
www.ietdl.org

Figure 13 Step responses of Gvd(s) for different RC and RL

this kind of system. Neglecting RC produces a wrong

converter model that leads to wrong conclusions and wrong
compensator design. Generally, because of the abrupt and
Figure 11 Bode plots of the simulated converter frequency large phase shift observed in the Bode plots of the input-
response and the transfer function Gvi (s) (dashed line) regulated converter, the compensator design is considered
superimposed difficult and a second-order proportional integral derivative
(PID) compensator is suggested in some works [1, 15].
realistic and useful analysis of the input-regulated DC – DC However, an ordinary PI compensator must be enough to
converter fed by a PV array. regulate the PV array voltage if RC is considered in the model.

Figs. 12 and 13 illustrate the effects of RC and RL on the 5 Closed-loop design

system responses. Fig. 13 shows that RL dumps the system
and makes it more stable. No noticeable effect of RC is The design of closed-loop controllers for the DC – DC
present in the responses of Fig. 13, but in Fig. 12 one converter fed by a PV array is based on the transfer
notices that RC introduces a high-frequency zero in the functions developed in previous sections. In the following
system and reduces the system phase lag. subsections, three control strategies for the buck converter
will be studied theoretically and experimentally. The two
Although RC is generally undesired in converter designs, last subsections make brief analyses about the practical
in the input-regulated converter it helps to stabilise the implementation of the compensators and about the stability
system by increasing the phase margin and significantly of the closed-loop systems when the operating point of the
reducing the phase lag at high frequencies. This explains PV array changes. An experimental converter with the
why even a poorly designed compensator may work with parameters presented in Table 2 was built and evaluated.

5.1 Single feedback loop

voltage controller
In the single-loop scheme of Fig. 14, the voltage controller
actuates on the converter duty cycle and directly regulates
the input voltage vpv . Fig. 15 shows how the controller is
constituted.

Fig. 16 shows the Bode plots of Gvd (s) (open-loop system)

and of the loop transfer function Cvd (s)Hv Gvd (s)
(compensated system). The system is compensated with
Cvd (s) = (30s + 750)/s and the feedback gain is
Hv = 1/30. The compensator places the crossover
frequency at v ¼ 6.6 × 103 rad/s. As a rule of thumb, the
crossover frequency can be chosen to be equal or greater
than the break frequency of the zero (1 + sCRC ),
Figure 12 Bode responses of Gvd(s) for different RC and RL v = 1/(CRC ), thus avoiding the critical phase-shift zone

874 IET Power Electron., 2010, Vol. 3, Iss. 6, pp. 869– 880
& The Institution of Engineering and Technology 2010 doi: 10.1049/iet-pel.2008.0344
 www.ietdl.org

Figure 14 Voltage-controlled converter

Figure 17 Experimental result with the single-loop voltage

controller: regulated input voltage and inductor current
∗
(vpv ¼ 20 and 25 V)

Figure 15 Voltage controller with single feedback loop

Figure 18 Voltage- and current-controlled converter

5.2.1 Current control loop: Fig. 19 shows how the

current controller is constituted. Fig. 20 shows the loop
Bode plots of the uncompensated and compensated
systems. In this example, the system is compensated with
Cid (s) = (9.5s + 950)/s. The crossover frequency is placed
at 1 kHz and the phase margin is 908. Fig. 21 shows an
experimental result of the current controller.

Figure 16 Bode plots of the open-loop system Gvd(s) and 5.2.2 Voltage control loop: Fig. 22 shows the control
the compensated system Cvd(s)HvGvd(s) scheme employing an inner current control loop and an
external voltage loop. The closed-loop current controller
transfer function is
highlighted in Fig. 16. This warranties good phase margin
even when the PV array operates at points other than the
point at which the i × v curve was linearised. Fig. 17 Cid (s)Gid (s)
Gicl (s) = (19)
shows an experimental result of the converter with the 1 + Cid (s)Gid (s)Hi
single-loop voltage controller.
With the transfer functions Gicl (s) and Gvi (s) from (19)
and (15), the compensator Cvi (s) may be designed in order
5.2 Double feedback loop to control the input voltage vpv .
voltage controller
In the double-loop scheme of Fig. 18, the converter inductor
current is directly controlled and the input voltage is
indirectly regulated. The voltage controller generates a
current reference for the current controller and the current
controller actuates on the converter duty cycle and regulates
the inductor current. Figure 19 Current controller

IET Power Electron., 2010, Vol. 3, Iss. 6, pp. 869– 880 875
doi: 10.1049/iet-pel.2008.0344 & The Institution of Engineering and Technology 2010
www.ietdl.org

Figure 20 Bode plots of the open-loop transfer function Figure 23 Voltage control with inner current loop: Bode
Gid(s) and the compensated system Cid(s)HiGid(s) plots of the uncompensated and compensated loop
transfer functions

Fig. 24 shows an experimental result with the double-loop

control scheme. Both current and voltage are controlled. The
major advantage of this control scheme is that the inductor
current may be controlled (and limited) in order to avoid
current overshoots that may be potentially dangerous when
large voltage steps occur.

5.3 Digital compensators

The continuous-time compensators designed in previous
sections can be realised as discrete-time digital
Figure 21 Experimental result with current controller: input compensators. There are several ways of discretising
voltage and controlled inductor current (iL∗ ¼ 5 and 12 A) analogue compensators. All of them present similar results
In this example only the inductor current is controlled when the compensator bandwidth is not greater than 10%
of the sampling frequency [16]. The Tustin or bilinear
transformation is used in this work. The compensators and
controllers were implemented with the Texas floating point
digital signal controller TMS320F28335 with sampling
frequency fs = 1/Ts = 10 kHz.

Figure 22 Voltage control with internal current control loop

Fig. 23 shows the Bode plots of the uncompensated loop

transfer function Gicl (s)Gvi (s)Hv and of the system
compensated with Cvi (s) = (32s + 3200)/s. Owing to the
1808 phase shift some care must be taken when placing the
crossover frequency. In this example, the compensated
system bandwidth was chosen to be 500 Hz and the phase
margin is 63.88. If a greater bandwidth is desired, a Figure 24 Experimental result with double-loop controller:
∗
derivative compensator may be necessary to boost the phase controlled input voltage (v pv ¼ 24 and 26 V) and
margin at the crossover frequency. controlled inductor current

876 IET Power Electron., 2010, Vol. 3, Iss. 6, pp. 869– 880
& The Institution of Engineering and Technology 2010 doi: 10.1049/iet-pel.2008.0344
 www.ietdl.org

A discrete-time compensator is a digital filter. The filter

coefficients may be obtained from the s-domain transfer
function with the desired transformation method. In
MATLAB the following command may be used to find
the filter coefficients: [numz, denz] = bilinear
(nums, dens, fs), where numz = [b0 b1] and
denz = [a0 a1] are the numerator and denominator
vectors with the coefficients of the Tustin (or bilinear)
approximation.

In the infinite impulse response (IIR) direct transposed

form, the filter difference equation is [17]

yk = b0 ek + b1 ek−1 − a1 yk−1 (20)

where y is the compensator output and e is the control error.

The difference equation (20) is the simplest way to

implement the PI compensator. The proportional and
integral components are embedded in the equation. In
order to implement the simple anti-windup strategy
proposed in this paper, one wishes to develop another Figure 25 PI compensator with integrator anti-windup
equation. action

The following equation describes a PI compensator

the first-order PI compensator. However, in the form of
yk = Kp ek + Ki ik (21) (21), the simple anti-windup algorithm of Fig. 25 is
possible. This algorithm simply stops the integrator
when the control output exceeds the limits. In practical
with the trapezoidal integrator ik
systems, the compensator output may sometimes exceed
Ts the maximum effective control effort. This causes
ik = [e + ek−1 ] + ik−1 (22) integrator saturation and deteriorates the control
2 k
performance.
The compensator of (21) and (22) is equivalent to
the compensator described by (20), with the difference Fig. 26 shows a simulation result employing a voltage
that in (21) the proportional and integral parts are compensator with the proposed anti-windup scheme. The
separated. simulated system is the double-loop voltage control of
Fig. 22. In contrast, Fig. 27 shows the same system
One can write the past integrator value ik−1 as controlled with a conventional compensator without
integrator anti-windup action.
1 Kp
ik−1 = y − e (23)
Ki k−1 Ki k−1

so ik can be rewritten as

Ts T 1 Kp
ik = ek + s ek−1 + yk−1 − ek−1 (24)
2 2 Ki Ki

Replacing (24) in (21), the difference equation (20)

becomes
   
ki Ts Ki Ts
yk = Kp + ek + − Kp ek−1 + yk−1 (25)
2 2

  
 
b0 b1

Equation (25) shows the correspondence between the

coefficients of (21) and (20), with a1 = −1. Equations Figure 26 Controlled voltage and compensator output with
(20) and (21) may be used indistinctly to implement integrator anti-windup action

IET Power Electron., 2010, Vol. 3, Iss. 6, pp. 869– 880 877
doi: 10.1049/iet-pel.2008.0344 & The Institution of Engineering and Technology 2010
www.ietdl.org

Figure 27 Controlled voltage and compensator output

without integrator anti-windup action Figure 28 Single-loop voltage control: Bode plots of the
closed-loop system compensated with Cvd (s) in three
different operating points of the PV array i × v curve
5.4 Stability analysis
approximately the same. The only exception is for the
At the beginning of the analysis, the PV array was linearised
operation in the voltage source region, which makes the
at the nominal MPP and the closed-loop system was
system more stable due to the increased phase margin in
designed to operate at the vicinities of this point. Although
Figs. 28 and 30.
the system is ideally optimised to operate at the nominal
MPP, the closed-loop system can work satisfactorily at
other operating points also. This can be verified through The conclusion is that placing the closed-loop crossover
simulation and experimentally. From the theoretical point frequency sufficiently far from the critical phase-shift zone
of view, the most important concern is the stability analysis. is enough to warranty the system operation at all points of
Although the closed-loop system is not expected to present the i × v curve. Regarding the operation of the PV array
the same performance when the operating point moves with conditions different of the nominal ones, it may be
along the i × v curve (and when the i × v curve itself shown (this is the subject of a future paper) that changing
changes), the system must be stable independently of the the temperature and solar irradiation produces the same
conditions. effect, from the point of view of the converter dynamic
analysis, as moving the operating point along the nominal
i × v curve. In fact, what matters is the punctual
It was shown in [10] that when the operating point of the
conductance of the i × v curve at the point of operation.
PV array moves towards the voltage source region of the i × v
Changing the i × v curve or moving the operating point
curve, the system becomes more dumped and stable. On the
other hand, when the operating point moves leftward and
enters the current source region of the i × v curve, the
dynamic behaviour of the system is more critical. Even
considering that the system operates at the MPP, the
system dynamics becomes less or more critical when the
temperature or the solar irradiation changes. An in-depth
analysis of the dynamic behaviour of the PV-buck system
in different points of the i × v curve is beyond the
objectives of this paper. However, in Figs. 28– 30 one can
observe the Bode responses of the closed-loop system
when operating in the extreme edges of the current-source
and voltage-source regions of the i × v curve. The
operation in the current-source region demonstrated in
these figures correspond to the worst-case operation of the
PV-buck system. The comparisons provided by these
figures show that if the closed-loop crossover frequency is
sufficiently high, the displacement of the PV array
operating point has no impact on the system stability. As Figure 29 Single-loop current control: Bode plots of the
seen in Figs. 28– 30, above the crossover frequencies the closed-loop system compensated with C id (s) in three
gain and phase characteristics of the systems are different operating points of the PV array i × v curve

878 IET Power Electron., 2010, Vol. 3, Iss. 6, pp. 869– 880
& The Institution of Engineering and Technology 2010 doi: 10.1049/iet-pel.2008.0344
 www.ietdl.org

8 References
[1] XIAO W., DUNFORD W.G., PALMER P.R., CAPEL A.: ‘Regulation of
photovoltaic voltage’, IEEE Trans. Ind. Electron., 2007, 54,
(3), pp. 1365 – 1374

[2] FEMIA N., PETRONE G., SPAGNUOLO G., VITELLI M.: ‘Optimizing
duty-cycle perturbation of P&O MPPT technique’. Power
Electronics Specialists Conf., 2004, PESC 04, 2004 IEEE
35th Annual, 2004, vol. 3, pp. 1939 – 1944

[3] FEMIA N., PETRONE G., SPAGNUOLO G., VITELLI M.: ‘Optimizing
sampling rate of P&O MPPT technique’. Power Electronics
Specialists Conf., 2004, PESC 04, 2004 IEEE 35th Annual,
2004, vol. 3, pp. 1945 – 1949
Figure 30 Double-loop voltage control: Bode plots
[4] KOUTROULIS E., KALAITZAKIS K., VOULGARIS N.C.: ‘Development
of the closed-loop system compensated with Cvi (s)
of a microcontroller-based, photovoltaic maximum power
in three different operating points of the PV array i × v
point tracking control system’, IEEE Trans. Power
curve
Electron., 2001, 16, (1), pp. 46 – 54

[5] KISLOVSKI A.S.: ‘Dynamic behavior of a constant-frequency

to the worst place (i.e. to the extremity of the current buck converter power cell in a photovoltaic battery charger
source region) corresponds to the situations analysed in with a maximum power tracker’. Fifth Applied Power
Figs. 28 –30. Electronics Conf. and Exposition, APEC, 1990

[6] VILLALVA M.G., GAZOLI J.R., RUPPERT E.F. : ‘Analysis and

6 Conclusions simulation of the P&O MPPT algorithm using a linearized
photovoltaic array model’. Tenth Brazilian Power
Regulation of photovoltaic voltage can be achieved with Electronics Conf., COBEP, 2009
ordinary PI compensators. Although the literature shows
many examples of uncompensated converters, controlling [7] Kyocera: ‘KC200GT high efficiency multicrystal
the output voltage of a PV device brings the advantage of photovoltaic module datasheet’, available at: http://www.
significantly increasing the performance of MPPT kyocerasdar.com/pdf/specsheets/KC200GT.pdf
algorithms [6]. This paper has presented a detailed analysis
of the PV voltage regulation problem using a buck [8] VILLALVA M.G., GAZOLI J.R., RUPPERT E.F.: ‘Comprehensive
converter as a PV array interface. The analysis permits the approach to modeling and simulation of photovoltaic arrays’,
solid design (with compensator design criteria) of the IEEE Trans. Power Electron., 2009, 25, (5), pp. 1198–1208
control systems. Three control strategies were proposed and
investigated. The controllers were optimised to operate at [9] VILLALVA M.G. , GAZOLI J.R. , RUPPERT E.F. : ‘Modeling and
the nominal MPP of the PV device. The reasons for this circuit-based simulation of photovoltaic arrays’, Revista
choice were explained in Section 2, where the modelling of Eletrônica de Potência (SOBRAEP), Brazilian J. Power
the PV device was presented. In Section 3 the converter Electron., 2009, 14, pp. 35 – 45
and the PV array were modelled and system transfer
functions were developed. In Section 4, the closed-loop [10] VILLALVA M.G., RUPPERT E.F.: ‘Dynamic analysis of the
systems were analysed theoretically and experimentally input-controlled buck converter fed by a photovoltaic array’,
and a brief consideration about the discrete-time Revista Controle & Automação – Sociedade Brasileira de
implementation of PI compensators was presented. Finally, Automática, Brazilian J. Control Autom., 2008, 19, (4),
Section 4 presented some considerations regarding the pp. 463–474
closed-loop system stability.
[11] CUK S., MIDDLEBROOK R.D.: ‘Modeling, analysis and design
of switching converters’. Technical Report, California
Institute of Technology, 1977
7 Acknowledgments
This work was financially supported by FAPESP and CNPq. [12] MIDDLEBROOK R.D.: ‘Small-signal modeling of pulse-width
Evaluation boards kindly donated by Texas Instruments were modulated switched-mode power converters’, Proc. IEEE,
used in the research. 1988, 76, (4), pp. 343– 354

IET Power Electron., 2010, Vol. 3, Iss. 6, pp. 869– 880 879
doi: 10.1049/iet-pel.2008.0344 & The Institution of Engineering and Technology 2010
www.ietdl.org

[13] MIDDLEBROOK R.D., CUK S.: ‘A general unified approach to voltage MPPT’. Power and Energy Engineering Conf.,
modelling switching converter power stage’. IEEE Power 2009, APPEEC 2009, Asia-Pacific, 27 – 31 March 2009,
Electronics Specialists Conf., 1976, pp. 18– 34 pp. 1 – 4

[14] ERICKSON R.W., MAKSIMOVIC D.: ‘Fundamentals of power [16] FRANKLIN G.F. , POWELL J.D. , EMAMI-NAEINI A. : ‘Feedback
electronics’ (Kluwer Academic Publishers, 2001) control of dynamic systems’ (Prentice Hall, 2002, 4th edn.)

[15] ZHIHAO Y. , XIAOBO W.: ‘Compensation loop design [17] OPPENHEIM A.V., SCHAFER R.W.: ‘Discrete-time signal
of a photovoltaic system based on constant processing’ (Prentice Hall, 1999, 2nd edn.)

880 IET Power Electron., 2010, Vol. 3, Iss. 6, pp. 869– 880
& The Institution of Engineering and Technology 2010 doi: 10.1049/iet-pel.2008.0344

View publication stats