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

net/publication/281304420

Nonlinear Adaptive Control of Grid-Connected Three-Phase Inverters for

Renewable Energy Applications

Article  in  International Journal of Control · August 2015

DOI: 10.1080/00207179.2015.1086026

CITATIONS READS

5 1,493

5 authors, including:

Nima Mahdian Dehkordi Houshang Karimi

Sharif University of Technology Polytechnique Montréal
9 PUBLICATIONS   39 CITATIONS    49 PUBLICATIONS   1,359 CITATIONS   

SEE PROFILE SEE PROFILE

Prasanna Piya
Mississippi State University
6 PUBLICATIONS   22 CITATIONS   

SEE PROFILE

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

Irregularity Detection in Output Power of Distributed Energy Resources View project

Design and Performance Evaluation of Data-Centric Protection Systems for Smart Grids View project

All content following this page was uploaded by Nima Mahdian Dehkordi on 24 September 2016.

The user has requested enhancement of the downloaded file.

 International Journal of Control

ISSN: 0020-7179 (Print) 1366-5820 (Online) Journal homepage: http://www.tandfonline.com/loi/tcon20

Nonlinear adaptive control of grid-connected

three-phase inverters for renewable energy
applications

N. Mahdian-Dehkordi, M. Namvar, H. Karimi, P. Piya & M. Karimi-Ghartemani

To cite this article: N. Mahdian-Dehkordi, M. Namvar, H. Karimi, P. Piya & M. Karimi-

Ghartemani (2015): Nonlinear adaptive control of grid-connected three-phase
inverters for renewable energy applications, International Journal of Control, DOI:
10.1080/00207179.2015.1086026

To link to this article: http://dx.doi.org/10.1080/00207179.2015.1086026

Accepted author version posted online: 25

Aug 2015.
Published online: 04 Oct 2015.

Submit your article to this journal

Article views: 156

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=tcon20

Download by: [nima mahdian dehkordi] Date: 24 September 2016, At: 05:45
INTERNATIONAL JOURNAL OF CONTROL, 
http://dx.doi.org/./..

Nonlinear adaptive control of grid-connected three-phase inverters for

renewable energy applications
N. Mahdian-Dehkordia , M. Namvara , H. Karimi b
, P. Piyac and M. Karimi-Ghartemanic
a
Department of Electrical Engineering, Sharif University of Technology, Tehran, Iran; b Department of Electrical Engineering, Polytechnique
Montreal, Montreal, QC, Canada; c Department of Electrical and Computer Engineering, Mississippi State University, Starkville, MS, U.S.A.

ABSTRACT ARTICLE HISTORY

Distributed generation (DG) units are often interfaced to the main grid using power electronic con- Received  November 
verters including voltage-source converters (VSCs). A VSC offers dc/ac power conversion, high con- Accepted  August 
trollability, and fast dynamic response. Because of nonlinearities, uncertainties, and system parame-
KEYWORDS
ters’ changes involved in the nature of a grid-connected renewable DG system, conventional linear Distributed generation;
control methods cannot completely and efficiently address all control objectives. In this paper, a non- renewable energy;
linear adaptive control scheme based on adaptive backstepping strategy is presented to control the backstepping; adaptive
operation of a grid-connected renewable DG unit. As compared to the popular vector control tech- control
nique, the proposed controller offers smoother transient responses, and lower level of current distor-
tions. The Lyapunov approach is used to establish global asymptotic stability of the proposed control
system. Linearisation technique is employed to develop guidelines for parameters tuning of the con-
troller. Extensive time-domain digital simulations are performed and presented to verify the perfor-
mance of the proposed controller when employed in a VSC to control the operation of a two-stage
DG unit and also that of a single-stage solar photovoltaic system. Desirable and superior performance
of the proposed controller is observed.

1. Introduction 2011; Dasgupta, Sahoo, Panda, & Amaratunga, 2011;

Prodanović, De Brabandere, Keybus, Green, & Driesen,
In the recent years, the use of renewable energy sources
2007).
(RESs) has grown considerably in order to meet the
Three-phase current-source converters (CSCs) and
increasing demand of energy while reducing the air pol-
voltage-source converters (VSCs) are employed by the
lution associated with the fossil fuel-based power gener-
DGs. The VSCs are more popular thanks to their desir-
ation, and lowering the system expansion costs. Domi-
able controllability and dynamic responses. The block
nant types of RESs are the photovoltaic (PV) array and
diagram of a three-phase grid-connected RES interfac-
wind power systems. There are two main modes of oper-
ing to the grid via a VSC is illustrated in Figure 1. In a
ation for a distributed generation (DG) system: the grid-
two-stage structure, the first stage (a dc/dc in PV appli-
connected mode and the islanded mode of operation
cations or an ac/dc in wind applications) is responsi-
(Abeyasekera, Johnson, Atkinson, & Armstrong, 2005;
ble for controlling the amount of power extracted from
Dai, Marwali, Jung, & Keyhani, 2008; Karimi, Davison,
the primary source. Normally, a maximum power point
& Iravani, 2010). In the grid-connected mode of opera-
tracking (MPPT) strategy is adopted to ensure that max-
tion, a DG unit should interact with the grid in a fully
imum power is extracted. In the two-stage structure, the
controlled manner without violating the grid codes and
dc-bus voltage is technically independent from the input
standards. In this mode, the DG unit should maintain the
power. In the one-stage conversion system, the primary
synchronisation with the grid, follow the set-points of the
power is directly converted to ac. In this structure, the
real and reactive powers, and abide by the limits for cur-
dc-bus voltage determines the amount of power extracted
rent distortion (Blaabjerg, Teodorescu, Liserre, & Tim-
from the primary source. The VSI is responsible for dc/ac
bus, 2006; Dasgupta, Sahoo, & Panda, 2011). Moreover,
conversion, power flow control, and power quality. The
the DG may perform ancillary services, such as reactive
popular output filter structures are L-type and LCL-type
power compensation and harmonic compensation of the
filters (Blaabjerg et al., 2006). The use of LCL-type fil-
local nonlinear loads (Bojoi, Limongi, Roiu, & Tenconi,
ter offers better attenuation of high-frequency harmonics
2011; Bracale, Angelino, Carpinelli, Mangoni, & Proto,
and noise (Park, Chen, Lai, & Moon, 2008).
CONTACT H. Karimi [email protected]
©  Taylor & Francis
2 N. MAHDIAN-DEHKORDI ET AL.

Figure . Block diagram of a three-phase renewable energy system.

In a grid-connected VSC, active and reactive power control the inverter output current for the case of vary-
components are regulated through vector control of the ing grid impedance within a specific range. The objective
VSC currents. The vector control (dq-current control) was to minimise the effect of uncertain grid impedance
is usually converted to two decoupled control loops on current tracking. Feedback linearisation technique
where two identical proportional-integrating (PI) con- is used in Khajehoddin, Karimi-Ghartemani, Jain, and
trollers (Blaabjerg et al., 2006) are individually designed. Bakhshai (2011), Zue and Chandra (2009), Bao, Zhuo,
In renewable applications, the dc-bus voltage is a medium Tian, and Tan (2013) within the context of vector con-
through which the d-component of the current is con- trol method. A robust partial feedback linearising con-
trolled. Other type of controllers, such as proportional- troller is used in Mahmud, Hossain, Pota, and Roy (2014),
resonant may be used when the control is done in station- Mahmud, Pota, Hossain, and Roy (2014) by considering
ary reference frame (Blaabjerg et al., 2006; El-Arroudi, uncertainties within the PV system model. Although Lya-
Joos, Kamwa, & McGillis, 2007; Guo & Wu, 2010; Lis- punov approach in Mahmud, Pota, and Hossain (2014),
erre, Teodorescu, & Blaabjerg, 2006; Marei, El-Saadany, Yahya, El Fadil, Guerrero, Giri, and Erguig (2014) based
& Salama, 2003; Prodanovic & Green, 2003; Zmood on the nonlinear model of the system has been presented,
& Holmes, 2003). These controllers have some advan- the uncertainty in system is not considered. A Lyapunov-
tages, such as faster dynamics and effective elimination based controller in Dasgupta, Mohan, Sahoo, and Panda
of harmonics at the expense of implementation com- (2013) is designed directly in the abc frame consider-
plexity (Francis & Wonham, 1976; Freijedo et al., 2011; ing unbalance conditions. In Li, Ruan, Pan, and Wang
Rodriguez, Bueno, Aredes, Rolim, Neves, & Cavalcanti, (2013) feedforward control scheme has been proposed in
2008; Yepes, Freijedo, López, & Doval-Gandoy, 2011; stationary and synchronous reference frames to reduce
Yepes et al., 2010; Yuan, Merk, Stemmler, & Allmeling, the injected grid current harmonics and unbalance con-
2002). The challenges for the control problem of a grid- ditions caused by grid voltages. Other approaches, such
connected VSC arise from the fact that the system has as μ-synthesis-based control and linear matrix inequal-
structural nonlinearities (such as the dc-bus dynamics), ity (LMI) control have been well-developed and widely
components nonlinearities (inductors, transformers, sen- accepted (Kahrobaeian & Mohamed, 2014; Sampaio, de
sors, signal conditioning circuits, dead-band of switch- Brito, de Azevedoe Melo, & Canesin, 2013). In Jain, Behal,
ing signals), and grid distortions and uncertainties (such Zhang, Dawson, and Mohan (2004), an adaptive nonlin-
as grid harmonics, unbalance, impedance, and faults). ear approach was used for fast voltage regulation in the
The existing linear control methods cannot address all presence of load uncertainty in case of static compen-
the control objectives (arising in a practical situation) sators. The controller was shown to possess satisfactory
without having an extremely complicated structure which transient response for step changes in load conductance
is undesirable due to practical limitations. In Yang, Lei, and voltage reference. A composite nonlinear feedback is
Peng, and Qian (2011), linear H theory was used to used in Eren, Pahlevaninezhad, Bakhshai, and Jain (2013)
 INTERNATIONAL JOURNAL OF CONTROL 3

to improve the stability margins of the VSC control loop Pinheiro, 2009). The synchronous reference (dq) domain
when an LCL output filter is used. (Malinowski & Bernet, 2008; Wu & Lehn, 2006) signals
In this paper, a nonlinear control scheme based on the are defined as
adaptive backstepping method is proposed for the VSC
of a grid-connected DG unit. The proposed control sys- vdq = vd + jvq = e− jθ vαβ . (3)
tem is able to regulate the power exchange between the
VSC and the grid, and provides fast and smooth tran- In this paper, we perform the control in the dq domain.
sient response with acceptable level of current distor- Hence, (1) is represented in the dq frame by (Jain et al.,
tions. The proposed controller is further improved to 2004)
adaptively accommodate the system parameters uncer-
tainties, e.g., filter inductance, dc-bus capacitance. To the Li̇dq = vdq − vg,dq − jLωidq , (4)
best of our knowledge, the use of adaptive backstepping
technique for the control of grid connected DG systems where ω = θ˙ is the grid angular frequency. By using the
based on the exact nonlinear model has not been pro- grid voltage phase-angle as the reference phase-angle for
posed/investigated before. To verify the performance of the dq transformation, the q component of the grid volt-
the proposed controller, digital time-domain simulations age, i.e., v gq , becomes zero. Hence,
are carried out for both single-stage and two-stage struc-
tures of the DG conversion system. The simulation results 1 1
of the proposed controller are compared with those of the i̇d = ωiq + vd − vgd (5)
L L
conventional vector control method which confirm the 1
superiority of nonlinear controller. i̇q = −ωid + vq , (6)
L

where v gd represents the magnitude of the grid voltage.

2. System description and control objectives
The power balance equation is given by
Single-line block diagram of a three-phase grid-
connected VSC is shown in Figure 2. The dc-bus Cvc v̇c = pin − pout , (7)
capacitor C provides a constant dc voltage which is
converted to ac by the VSC interfaced with the grid where pin denotes the output power from the primary
through an L-filter. The circuit equation on the ac side of source (Khajehoddin et al., 2011) and pout is the instan-
the system is given by taneous power to the inverter. Neglecting the internal
inverter losses, we obtain
diabc
L = vabc − vg,abc , (1) 3 3 3
dt pout = vd id + vq iq = vgd id . (8)
2 2 2
where v abc and v g, abc denote the output voltage of the con-
verter and the grid voltage, respectively. In a converter with pulse width modulation (PWM) tech-
By applying the abc/αβ transformation, i.e., nique, v d = v c ud and v q = v c uq (Khajehoddin et al.,
2011), and the system equations may be represented by
2 
vαβ = vα + jvβ = va + vbe j2π/3 + vc e− j2π/3 , (2)
3 1 1
i̇d = ωiq + vc ud − vgd (9)
the two-phase representation of the three-phase voltage L L
in the stationary domain is obtained (Gabe, Montagner, & 1
i̇q = −ωid + vc uq (10)
L
 
1 3
p in
ig v̇c = pin − vgd id . (11)
Cvc 2
Inverter
+ + L The following control objectives are desired.
C vg
vc v (1) Regulate the dc-bus voltage v c at the desired value
vc∗ .
- - (2) Regulate the output reactive power Q = − 32 vgd iq
at the desired value Q∗ .
Notice that regulating the dc-bus voltage implies
Figure . Block diagram of the grid-connected VSC. regulation of the real power.
4 N. MAHDIAN-DEHKORDI ET AL.

Further secondary objectives of the controller are that it follows the first-order exponentially stable dynam-
(1) Maintain the quality of current waveform despite ics given by ṽ˙ c = −kṽc , where k  R is a constant positive
the grid voltage distortion, grid voltage unbalance, gain. Assuming v̇c∗ = 0 yields
and other system nonlinearity effects including
the dead band in the switching patterns, nonlin- 2
iideal
d =α= [pin + Ckvc ṽc )]. (13)
earity of the inductor. 3vgd
(2) Maintain robust performance in the presence of
system uncertainties. These include first the value The error between the ideal and actual current is defined
of the inductance, grid impedance, and the grid by z  id − iideal
d . Hence, dynamic of z is obtained by
frequency.
To proceed with developing our method, the reactive 1 1
ż = ωiq + vc ud − vgd − α̇, (14)
current and dc voltage error signals are defined as L L

where α̇ is expressed as a function of available signals by

ĩq = iq − i∗q , ṽc = vc − vc∗ , (12)
2kC   2 ṗ
v˙c ṽc + ṽ˙ c vc +
in
α̇ = .
∗ 3vgd 3vgd
where i∗q = − 3v
2Q
gd
is the reference value for reactive
current.
It should be noted that v gd is assumed constant, i.e., grid
voltage magnitude remains unchanged within the con-
troller response time, which is a practical assumption.
3. Non-adaptive controller Substituting for v˙c and ṽ˙ c from (11) yields
3.1 Derivation of error model   
2k ṽc 3 2 ṗin
α̇ = +1 pin − id vgd + .
Equations (9)–(11) can be regarded as cascade of two sub- 3vgd vc 2 3vgd
systems, as shown in Figure 3. The first subsystem is spec-
ified by (9) and (10) with ud , uq as inputs and id as the Substituting for ṽc from (12) results in
output. The second subsystem is described by (11) where  
id acts as a virtual input and v c is the output. 2k  ∗
 3 2 ṗin
α̇ = 2vc − vc pin − id vgd + . (15)
The cascade consideration of system dynamics per- 3vc vgd 2 3vgd
mits to use nonlinear backstepping approach for control
of system states and eventually take into account param- Equation (10) can be written as
eter uncertainties by an adaptive method (Khalil, 2002).
ĩ˙q = −ωid + vc uq ,
Consider (11) and regard id as its input. Define iideal as 1
d (16)
the ideal value of id that can control v c appropriately, such L

Figure . Plant block diagram.

 INTERNATIONAL JOURNAL OF CONTROL 5

and Equation (11) can be expressed as Proof: The control law of (21) results in the following
   closed-loop equations:
1 3 2 
ṽ˙ c = pin − vgd z + pin + Ckvc ṽc
ĩ˙q = −k2 ĩq
Cvc 2 3vgd
3 vgd z 3 vgd z
= −kṽc − . (17) ṽ˙ c = −kṽc −
2 Cvc 2 Cvc (23)
3 vgd ṽc
Equations (14)–(17) provide a description of the system ż = −k1 z + 2 .
Cvc
that are re-ordered and written below,
Consider the Lyapunov function candidate V as
ĩ˙q = −ωid + vc uq
1
(18)
1 2 
L
3 vgd z V = ṽc + z2 + ĩ2q . (24)
ṽ˙ c = −kṽc − (19) 2
2 Cvc
1 1 Taking the time-derivative of the Lyapunov function and
ż = ωiq + vc ud − vgd − α̇. (20)
L L substituting from (18)–(20) yield

3.2 Control law  

3 vgd z
V̇ = + ṽc −kṽc −
−k2 ĩ2q
Inspired by structure of the open-loop error system (18)– 2 Cvc
   
(20), we propose the control input ud , uq by 3 vgd ṽc
+ z −k1 z + = − k2 ĩ2q + k1 z2 + kṽc2 < 0,
2 Cvc
L  
uq = ωid − k2 ĩq (25)
vc
 
L 1 3 vgd ṽc
ud = −ωiq + vgd + α̇ − k1 z + , (21) which by virtue of (24) implies (22). It should be noted
vc L 2 Cvc that the convergence of error variables to zero is in global
sense. 
where k1 and k2 are arbitrary positive constants.
Assumption 1: It is assumed that v c , vc∗ , id , iq , i∗q , z, L, v gd ,
C, and ω are known to the controller. 3.3 Design guidelines
Note that for implementation of the control law, Equation set (23) indicates that the gain k2 determines the
derivative of pin is needed. However, in most cases, varia- speed of reactive power regulation. The linearisation of
tion of pin is negligible. This is particularly justifiable in a the second and the third equations of (23) around the zero
two-stage conversion system where a first-stage converter equilibrium point results in the following linear model:
is used to perform MPPT. Therefore, the primary power
is controlled by the first stage and is independent from the

−k −
dc-bus voltage. Thus, the two stages are decoupled. More- Ẏ = Y, (26)
 −k1
over, the MPPT (that tracks the input power variations)
is normally slower than the dc-bus controller, which gives
3 vgd
another justification to neglect this term. The dc-bus volt- where  = 2 Cvc∗
. The characteristic equation of (26) is
age is, then, controlled at a given value which should be then
sufficiently higher than the grid voltage and low enough
to avoid efficiency degradation (Eren, Pahlevaninezhad, λ2 + (k + k1 )λ + kk1 + 2 = 0. (27)
Bakhshai, & Jain, 2015). However, in a single-stage con-
verter, the power is directly dependent on the dc-bus volt-
Here, the local stability conditions are k + k1 > 0 and kk1
age and thus ṗin may or may not be negligible depending
+ 2 > 0 which allows the negative values for k or k1 .
on the bandwidth of the dc-bus voltage changes. If ṗin is
Notice that the conditions k > 0 and k1 > 0 in Theorem 1
not negligible, it may be computed by numerical differen-
are sufficient, not necessary. Specifically, the root-locus of
tiation.
this equation is shown in Figure 4 when one of the gains
Theorem 1: The control law of (21) guarantees that
is selected at zero. The roots will√have a damping factor of
0.707 when the gain is equal to 2 (a suggested setting)
lim {z(t ), ĩq (t ), ṽc (t )} = 0. (22)
t→∞ and will have a damping of unity when the gain is 2.
6 N. MAHDIAN-DEHKORDI ET AL.

the dc-bus voltage. This is a model for a two-stage renew-

able energy system shown in Figure 1(a). The following
values are used for the system and control parameters:

C = 470μF, L = 10mH, vgd = 110V, ω = 120πrad/s

3 vgd
vc∗ = 500V,  = = 700,
2 Cvc∗
k = 2, k1 = 0, k2 = 5.

With the above setting, the three poles of the linearised

system are at −, −  (for the real power and dc-bus
dynamics), and −5 (for the reactive power dynamics).
The switching frequency is set at 10 kHz. The control
Figure . Root-locus of the real power (dc-bus voltage) dynamics algorithm is implemented using digital control blocks at a
(). sampling frequency of 20 kHz. Care has been taken when
selecting the parameters to ensure that the loop band-
3.4 Simulation results
width is sufficiently smaller than the switching frequency.
.. Study system I Figure 5 shows the results of a simulation study, where
A controllable power source (generating pin ) is simulated the initial value of the capacitor voltage is set at 480 V,
where power commands can be applied independent of the initial reactive power reference is set to zero, and pin

Figure . Performance of the proposed controller in tracking input power jump, reactive power command, dc-bus voltage command
(Study System I).
 INTERNATIONAL JOURNAL OF CONTROL 7

Figure . Performance of the optimally adjusted conventional vector control method under the same simulation conditions of Figure .

is also initially assumed zero. At t = 0.05 s, pin experi-

ences a jump to 1500 W, and is followed by a jump of
1000 Var in Q∗ at t = 0.1 s and a jump of vc∗ to 520 V
at t = 0.15 s. The real and reactive powers supplied by
the inverter, the capacitor voltage, and the grid currents
are shown in Figure 5. The starting stage of the oper-
ation shows a very smooth response with limited cur-
rents. A negligible steady state error in the amount of
reactive power is observed. Figure 6 shows responses of
the popular vector control method whose parameters are Figure . MPPT algorithm used in Study System II.
optimally set according to the method of Khajehoddin
et al. (2011), such that it can offer a transient response
controller is simulated. This is a single-stage conversion
almost as fast as the proposed method. The starting stage
system as shown in Figure 1(b). The values for the power
as well as the transient responses offered by this method
circuit and the control parameters are the same as those
are not as good. Overall, the desirable and superior tran-
of Study System I. The reference for the capacitor voltage
sient response of the proposed method is confirmed.
(or the PV voltage) is provided by the MPPT algorithm.
The MPPT algorithm used in this simulation is shown in
.. Study system II Figure 7. Variations of power with respect to variations
A 3.5 kW PV generation system directly connected to an of voltage are digitally calculated by taking the samples of
inverter equipped with the digital version of the proposed these two measurements at the right sample times. The
8 N. MAHDIAN-DEHKORDI ET AL.

Figure . Performance of the proposed controller in tracking input power jump, and reactive power command (Study System II).

quantity dP/dV is passed through a PI controller to yield II are also superior to those of the conventional method
the dc-bus voltage reference. The gains of this controller (not shown).
are set at 10 and 1000, respectively. The PSIM software
is used for simulations which facilitates the possibility of
applying changes to the light intensity of the PV system
3.5 Inverter current THD
during the simulation. The light intensity directly changes
the power generated by the solar panels. The grid voltage is contaminated with the following
Figure 8 shows the results of a simulation study with distortions: 2% of the 5th harmonic at phase-angle
the following specifications: the solar intensity is set to 40 degrees, 2% of the 7th harmonic at phase-angle
500 W/m2 initially, the capacitor voltage is 480 V, and 10 degrees, 2% of the 11th harmonic at phase-angle of
the reactive power command is zero. At t = 0.15 s, the 90 degrees, and 2% of grid imbalance at phase-angle
light intensity jumps to 750 W/m2 and subsequently, a 0 degree. The inverter current total harmonic distortion
reactive power command of 2000 Var is applied at t = (THD) is shown in Figure 9 for different power levels in
0.25 s. The real and reactive powers generated by the a 3.5 kW inverter. The THD is below 2% at normal oper-
inverter, and the maximum PV power are shown on the ation (operation above 1 kW). As the power is increased
upper subplot of Figure 8. The MPPT is working well (above 2 kW), the proposed method generates currents
and the reactive power command is followed quickly. with lower THD compared to the conventional vector
The middle subplot shows the PV voltage across the control method (when optimally adjusted to have sim-
capacitor and the lower subplot shows the inverter out- ilar transients). Therefore, the overall distortion power
put currents fed to the grid. Similar to the results of injected to the grid by the proposed VSC control is much
Study System I, the transient responses of Study System lower than the conventional method.
 INTERNATIONAL JOURNAL OF CONTROL 9

this study, the controller is designed based on the nomi-

nal value of 10 mH. As shown in Figure 10, the controller
shows a steady state error of about 5% in the amount of
generated reactive power. Although this amount of error
may be considered within the acceptable limits, we will
present an adaptive controller in Section 4 to overcome
this problem.

4. Adaptive controller

Figure . Output current THD of the proposed controller at differ- 4.1 Derivation of error model
ent powers.
For convenience of notations, we define θ = L1 and β =
1
3.6 Uncertainty in L-filter C
, where L and C are assumed as unknown system param-
eters. Hence, (9)–(11) are transformed into
As mentioned in Study System I, the proposed method
assumes a negligible steady state error in the reactive
power. This error grows with the amount of uncertainty i̇d = ωiq + θvc ud − θvgd
(change) in the value of the series filter L. In Figure 10,
i̇q = −ωid + θvc uq
we simulate a case study, where the filter L experiences a  
change of −50 per cent from its nominal value, i.e., from β 3
v̇c = pin − vgd id . (28)
10 to 5 mH at t = 0.1 s and back to 10 mH at t = 0.15 s. In vc 2

Figure . Performance of proposed non-adaptive controller when the inductance L experiences step changes.
10 N. MAHDIAN-DEHKORDI ET AL.

The error parameters β̃ and θ˜ are defined as Hence, the state error equations become

β̃ = β̂ − β, θ˜ = θˆ − θ, (29) ĩ˙q = −ωid + θvc uq

3β̂vgd z
where β̂ and θˆ are the estimates of β and θ, respectively. ṽ˙ c = −kṽc − − β̃Y
2vc
Substituting β̃ and θ˜ in (28) yields 
2k  1
ż = ωiq + θud vc − θvgd − 2vc − vc∗
    3vgd β̂
β̂ 3 1 3
v̇c = pin − vgd id − β̃ pin − vgd id .
   ˙ 
vc 2 vc 2 β̂ 3 β̂ 2 ṗin
pin − vgd id − β̃Y − ṽc vc − .
(30) vc 2 β̂ 2 3v gd
(37)
Defining
 
1 3 4.2 Control law
Y = pin − vgd id , (31)
vc 2 The control law and the adaptive parameters update laws
are proposed as follows:
and substituting from (31) in (30) yield

  1
β̂ 3 uq = (−k1 ĩq + ωid ),
v̇c = pin − vgd id − β̃Y. (32) ˆ
θ vc
vc 2

1 ˆ 3β̂vgd ṽc 2k
ud = θ vgd − k2 z + − ωiq +
θˆvc 2vc 3vgd
Consider (32), ignore the second term β̃Y for a moment,

  ˙  
and regard id as its input. Define iideal as the ideal value of 1 ∗ 3 β̂ 2 ṗin
d
× (2vc − vc ) pin − vgd id − ṽc vc + ,
id that controls v c according to ṽ˙ c = −kṽc . Then, vc 2 β̂ 2 3vgd
˙ 1
 θˆ = (z(ud vc − vgd ) + vc uq ĩq ),
2 vc γ1
iideal
d =α= pin + (kṽc ) . (33)  
3vgd β̂ ˙ 1 2k (2vc − vc∗ )
β̂ = − Y z + Y ṽc , (38)
γ2 3vgd β̂
By defining z  id − iideal
d , dynamics of z is obtained
from where k, k1 , γ 1 , γ 2 and k2 are positive gains.

ż = ωiq + θud vc − θvgd Assumption 2: It is assumed that v c , vc∗ , id , iq , i∗q , z,


 v gd , and ω are known to the controller, and the unknown
2k  
∗ 1 β̂ 3
− 2vc − vc (pin − vgd id ) − β̃Y parameters L = θ1 and C = β1 are estimated via adaptation
3vgd β̂ vc 2
law (38).
˙ 
β̂ 2 ṗin
− ṽc vc − , Theorem 2: The control and adaptive laws of (38) guaran-
β̂ 2 3v gd
tee that
(34)

where id can be expressed as lim {z(t ), ĩq (t ), ṽc (t )} = 0, and (39)

t→∞

2 vc lim θˆ (t ) = θ. (40)
id = z + pin + (kṽc ) . (35) t→∞
3vgd β̂
Proof: Consider the following radially unbounded Lya-
Substituting for id from (35) in (32) results in punov function:

3β̂vgd z 1 2 
ṽ˙ c = −kṽc − − β̃Y. (36) V = ṽc + z2 + ĩ2q + γ1 θ˜2 + γ2 β̃ 2 . (41)
2vc 2
 INTERNATIONAL JOURNAL OF CONTROL 11

Taking the time-derivative of Lyapunov function and sub- Finally, substituting ud and uq from (38) into (45) yields
stituting from (37) yield
 
 
V̇ = − k1 ĩ2q + k2 z2 + kṽc2 ≤ 0, (46)

3β̂vgd z
V̇ = ṽc −kṽc − − β̃Y + z ωiq + θ ud vc − θ vgd
2vc

 ˙  
which implies that ĩq , z, ṽc ∈ L2 ∩ L∞ and β̃, θ˜ ∈ L∞ ,
2k 1 β̂ 3 β̂ 2 ṗin where L∞ and L2 denote the spaces of bounded and
− (2vc − vc∗ ) (pin − vgd id )−β̃Y − ṽc vc −
3vgd β̂ vc 2 β̂ 2 3vgd
square-integrable signals, respectively (Khalil, 2002). Due
˙ ˙
+ ĩq (−ωid + θ vc uq ) + γ2 β̃ β̃ + γ1 θ˜θ˜. (42) to the smoothness and continuity of the error signals and
by invoking the Barballat’s lemma, (39) is inferred. To
By separating the terms containing θ and β̃, (42) is prove (40), we use LaSalle’s invariance principle presented
in Khalil (2002). Determine the set S = {x ∈ R5 | V̇ = 0}
expressed as  T
where x = z ṽc ĩq . Hence, S = {id = α, vc = vc∗ , iq =
  
i∗q } on which the motion of the system (28) and (38) is
3β̂vgd z 2k   described by
V̇ = ṽc −kṽc − + z ωiq − 2vc − vc∗
2vc 3vgd    
  ˙   θ θ ˙ ˙
1 3 β̂ 2 ṗin i̇d = ωi∗q −1 , i̇q = −ωi∗d −1 , v̇c = θ˜ = β̃ = 0.
pin − vgd id − ṽc vc − ˆ
θ ˆ
θ
vc 2 β̂ 2 3vgd (47)
 
˙ ˙
− ĩq ωid +γ2 β̃ β̃+γ1 θ˜θ+θ
˜ z(ud vc − vgd ) + vc uq ĩq

 Now, the largest invariant set MS is determined when
2k (2vc − vc∗ ) θ
+ β̃ Y z − Y ṽc . i̇d = i̇q = 0 in (47), i.e., when − 1 = 0. Therefore, by
3vgd β̂ θˆ
(43) the LaSalle’s theorem, every solution of the closed-loop
system starting in any compact set
(0 
R5 ) asymp-
totically approaches the invariant set M = {id = α, vc =
Furthermore, using θ = θˆ − θ˜ and factorising the z terms vc∗ , iq = i∗q , θˆ = θ} as t → . In particular, (40) holds,
in (43) yield i.e.,

lim θˆ (t ) = θ.
3β̂vgd ṽc 2k
V̇ = −kṽc2 + z − + ωiq − (2vc − vc∗ ) t→∞
2vc 3vgd
  ˙  
It should be noted that although the boundedness of β̃ is
1 3 β̂ 2 ṗin ˙
pin − vgd id − ṽc vc − −ĩq (ωid )+θ˜ γ1 θ˜ guaranteed by the LaSalle’s theorem, β̂ does not necessar-
vc 2 β̂ 2 3vgd ily track β. 

ˆ
− z(ud vc −vgd )−vc uq ĩq +θ[z(u d vc −vgd ) + vc uq ĩq ] The implementation block diagram of the nominal

 and adaptive nonlinear control scheme is depicted in
˙ 2k (2vc − vc∗ )
+ β̃ γ2 β̃ + Y z − Y ṽc . Figure 11. From Figure 11, it is also seen that the output
3vgd β̂ currents of the three-phase inverter and grid voltages are
(44) transformed into direct and quadrature-axis components
using abc − dq transformation. Then, the control law is
Substituting the parameter update laws from (38), we obtained in the dq frame and, eventually, reverse trans-
obtain formation (i.e., dq − abc) is performed to implement the
controller through the PWM.

3β̂vgd ṽc
V̇ = −kṽc2 +z − + ωiq
2vc 4.3 Design guidelines

  ˙ 
2k ∗ 1 3 β̂ A linear analysis is performed for the proposed adaptive
− (2vc − vc ) pin − vgd id − ṽc vc controller in order to find guidelines as how to adjust
3vgd vc 2 β̂ 2
   the controller gains. The linear analysis confirms that the
2 ṗin
− − ĩq (ωid ) + θˆ z(ud vc − vgd ) + vc uq ĩq . dynamics of z and ṽc variables are coupled to each other,
3vgd and are decoupled from rest of the system. Moreover, they
(45) follow the same dynamics as the non-adaptive method.
12 N. MAHDIAN-DEHKORDI ET AL.

Grid
PV Converter DC Link Inverter L Filter
Supply

ia ib ic vga vgb vgc

PWM abc-dq abc-dq

ua u b u c vgd vgq=0
pin vc dq-abc
id iq
ud uq

Controller
MPPT (21)

PLL Adaptaion Law

Adapve Controller (38)

Figure . Implementation block diagram in adaptive and nominal case.

Figure . Performance of the proposed adaptive controller in tracking input power jump, reactive power command, dc-bus voltage
command, and inductance change (Study System I).
 INTERNATIONAL JOURNAL OF CONTROL 13

Therefore, the same design guideline presented in Sec- Therefore, there will be two poles located at −kg1 for
tion 3.3 is valid for designing k and k1 . k1 = 2kg1 .
The linearisation of the equations related to ĩq and θ˜
around the zero equilibrium point results in the following
linear system: 4.4 Simulation results

 .. Study system I
−k2 −
Ẋ = X, (48) The proposed adaptive controller is used by Study System
/γ1 0
I as defined and simulated in Section 3.4.1. The values of
2ωLpin power circuit and control parameters are the same with
where  = . The eigenvalues of the linear model the additional value of kg1 = 200.
3vgd
(48) are obtained from the following characteristic equa- Figure 12 shows the results of a simulation where the
tion: initial value of the capacitor voltage is set at 480 V, the
initial reactive power reference is set to zero, and pin is
2 initially 500 W. At t = 0.05 s, pin experiences a jump to
λ2 + k2 λ + = 0. (49)
γ1 1500 W, and is followed by a jump of 1000 Var in Q∗ at t
= 0.1 s and a jump of vc∗ to 520 V at t = 0.15 s. Finally, the
2 inductance L steps down to 5 mH from 10 mH, at t = 0.2 s.
The suggested value for γ 1 is γ1 = , where kg1 is a
k2g1 The real and reactive powers supplied by the inverter,
constant number. The root-locus of (49) will be similar the capacitor voltage, the grid currents, the estimated
to Figure 4 with the difference that  is replaced by kg1 . inductance, and the estimated capacitance are shown in

Figure . Performance of the proposed adaptive controller in tracking input power jump, reactive power command, and inductance
change (Study System II).
14 N. MAHDIAN-DEHKORDI ET AL.

Figure 12. The transient and steady state responses of the Bao, X., Zhuo, F., Tian, Y., & Tan, P. (2013). Simplified feedback
system are desirable. The estimated inductance follows linearization control of three-phase photovoltaic inverter
the actual one with a small error. with an LCL filter. IEEE Transactions on Power Electronics,
28(6), 2739–2752.
Blaabjerg, F., Teodorescu, R., Liserre, M., & Timbus, A.V.
.. Study system II (2006). Overview of control and grid synchronization for
The proposed adaptive controller is employed by Study distributed power generation systems. IEEE Transactions on
System II as defined and simulated in Section 3.4.2. The Industrial Electronics, 53(5), 1398–1409.
values of power circuit and control parameters are the Bojoi, R.I., Limongi, L.R., Roiu, D., & Tenconi, A. (2011).
Enhanced power quality control strategy for single-phase
same with the additional value of kg1 = 200. inverters in distributed generation systems. IEEE Transac-
Figure 13 shows the results of a simulation study with tions on Power Electronics, 26(3), 798–806.
the following specifications: the solar intensity is set to Bracale, A., Angelino, R., Carpinelli, G., Mangoni, M., & Proto,
500 W/m2 initially, the capacitor voltage is 480 V, and the D. (2011). Dispersed generation units providing system
reactive power command is zero. At t = 0.15 s, the light ancillary services in distribution networks by a centralised
control. IET Renewable Power Generation, 5(4), 311–321.
intensity is increased to 750 W/m2 and subsequently, a
Dai, M., Marwali, M., Jung, J.-W., & Keyhani, A. (2008). A three-
reactive power command of 2000 Var is applied at t = phase four-wire inverter control technique for a single dis-
0.25 s, and an inductance change from 10 to 5 mH is tributed generation unit in island mode. IEEE Transactions
applied at t = 0.3 s. The real and reactive powers gen- on Power Electronics, 23(1), 322–331.
erated by the inverter, the maximum solar power, the PV Dasgupta, S., Mohan, S., Sahoo, S., & Panda, S. (2013). Lya-
voltage across the capacitor, the inverter output currents, punov function-based current controller to control active
and reactive power flow from a renewable energy source to
the estimated inductance, and the estimated capacitance a generalized three-phase microgrid system. IEEE Transac-
are shown in Figure 13. The desirable performance of the tions on Industrial Electronics, 60(2), 799–813.
proposed adaptive controller is observed. Dasgupta, S., Sahoo, S., & Panda, S.K. (2011). Single-phase
inverter control techniques for interfacing renewable
energy sources with microgrid–part i: Parallel-connected
5. Conclusion inverter topology with active and reactive power flow con-
trol along with grid current shaping. IEEE Transactions on
This paper presents a nonlinear adaptive controller for Power Electronics, 26(3), 717–731.
the VSC of a grid-connected DG unit. The VSC is used Dasgupta, S., Sahoo, S.K., Panda, S.K., & Amaratunga, G.A.
to interface the RES with the utility grid. The controller (2011). Single-phase inverter-control techniques for inter-
is designed based on adaptive backstepping technique facing renewable energy sources with microgrid–part ii:
Series-connected inverter topology to mitigate voltage-
in which the uncertainties of the system parameters are
related problems along with active power flow control. IEEE
considered. The proposed controller presents a fast, yet Transaction on Power Electronics, 26(3), 732–746.
smooth, transient response, and accurately regulates the El-Arroudi, K., Joos, G., Kamwa, I., & McGillis, D.T. (2007).
power exchange between the VSC and the grid. The adap- Intelligent-based approach to islanding detection in dis-
tive controller also estimates the value of the interfac- tributed generation. IEEE Transactions on Power Delivery,
ing inductance. The transient responses of the proposed 22(2), 828–835.
Eren, S., Pahlevaninezhad, M., Bakhshai, A., & Jain, P.K. (2013).
method are superior to those of the conventional vector Composite nonlinear feedback control and stability analysis
control method. The quality of the VSC current is higher of a grid-connected voltage source inverter with LCL filter.
at the higher powers meaning that the amount of distor- IEEE Transactions on Industrial Electronics, 60(11), 5059–
tion injected to the grid by the VSC is significantly small. 5074.
Eren, S., Pahlevaninezhad, M., Bakhshai, A., & Jain, P.K. (2015).
An adaptive droop dc-bus voltage controller for a grid-
Disclosure statement connected voltage source inverter with LCL filter. IEEE
Transactions on Power Electronics, 30(2), 547–560.
No potential conflict of interest was reported by the authors. Francis, B., & Wonham, W. (1976). The internal model principle
of control theory. Automatica, 12, 457–465.
ORCID Freijedo, F., Yepes, A., Malvar, J., López, O., Fernandez-
Comesana, P., Vidal, A., & Doval-Gandoy, J. (2011). Fre-
H. Karimi http://orcid.org/0000-0003-4956-0925 quency tracking of digital resonant filters for control of
power converters connected to public distribution systems.
IET Power Electronics, 4(4), 454–462.
References Gabe, I.J., Montagner, V.F., & Pinheiro, H. (2009). Design and
implementation of a robust current controller for VSI con-
Abeyasekera, T., Johnson, C.M., Atkinson, D.J., & Armstrong,
nected to the grid through an LCL filter. IEEE Transactions
M. (2005). Suppression of line voltage related distortion in
on Power Electronics, 24(6), 1444–1452.
current controlled grid connected inverters. IEEE Transac-
Guo, X.Q., & Wu, W.Y. (2010). Improved current regulation
tions on Power Electronics, 20(6), 1393–1401.
of three-phase grid-connected voltage-source inverters for
 INTERNATIONAL JOURNAL OF CONTROL 15

distributed generation systems. IET Renewable Power Gen- LCL fuel cell inverter. IEEE Transactions on Power Electron-
eration, 4(2), 101–115. ics, 23(4), 1716–1723.
Jain, A.K., Behal, A., Zhang, X.T., Dawson, D.M., & Mohan, Prodanović, M., De Brabandere, K., Keybus, J. Van den, Green,
N. (2004). Nonlinear controllers for fast voltage regula- T., & Driesen, J. (2007). Harmonic and reactive power com-
tion using STATCOMs. EEE Transactions on Power Systems, pensation as ancillary services in inverter-based distributed
12(6), 827–842. generation. IET Generation, Transmission & Distribution,
Kahrobaeian, A., & Mohamed, Y.-R. (2014). Robust single-loop 1(3), 432–438.
direct current control of LCL-filtered converter-based dg Prodanovic, M., & Green, T.C. (2003). Control and filter design
units in grid-connected and autonomous microgrid modes. of three-phase inverters for high power quality grid connec-
IEEE Transactions on Power Electronics, 29(10), 5605– tion. IEEE Transactions on Power Electronics, 18(1), 373–
5619. 380.
Karimi, H., Davison, E., & Iravani, R. (2010). Multivariable Rodriguez, F.J., Bueno, E., Aredes, M., Rolim, L.G.B., Neves,
servomechanism controller for autonomous operation of a F.A.S., & Cavalcanti, M.C. (2008). Discrete-time implemen-
distributed generation unit: Design and performance eval- tation of second order generalized integrators for grid con-
uation. IEEE Transactions On Power Systems, 25(2), 853– verters. In 34th Annual Conference of IEEE Industrial Elec-
865. tronics (pp. 176–181). Orlando, FL: IEEE.
Khajehoddin, S.A., Karimi-Ghartemani, M., Jain, P.K., & Sampaio, L.P., de Brito, M.A.G., de Azevedoe Melo, G., &
Bakhshai, A. (2011). A control design approach for three- Canesin, C.A. (2013). Power flow control in single-phase
phase grid-connected renewable energy resources. IEEE and three-phase grid-connected inverters using LMI, state-
Transactions on Sustainable Energy, 2(4), 423–432. feedback linearization and D-stability. In 2013 15th Euro-
Khalil, H.K. (2002). Nonlinear systems (3rd ed.). Upper Saddle pean Conference on Power Electronics and Applications
River, NJ: Prentice-Hall. (pp. 1–10). Lille: IEEE.
Li, W., Ruan, X., Pan, D., & Wang, X. (2013). Full-feedforward Wu, E., & Lehn, P.W. (2006). Digital current control of a
schemes of grid voltages for a three phase LCL type grid voltage source converter with active damping of LCL res-
connected inverter. IEEE Transactions on Industrial Elec- onance. IEEE Transactions on Power Electronics, 21(5),
tronics, 60(6), 2237–2250. 1364–1373.
Liserre, M., Teodorescu, R., & Blaabjerg, F. (2006). Multiple Yahya, A., El Fadil, H., Guerrero, J.M., Giri, F., & Erguig, H.
harmonics control for three-phase grid converter systems (2014). Three-phase grid-connected of photovoltaic gener-
with the use of PI-RES current controller in a rotating ator using nonlinear control. In 2014 IEEE Conference on
frame. IEEE Transactions on Power Electronics, 21(3), 836– Control Applications (pp. 879–884). Juan Les Antibes: IEEE.
841. Yang, S., Lei, Q., Peng, F., & Qian, Z. (2011). A robust control
Mahmud, M., Hossain, M., Pota, H., & Roy, N. (2014). Robust scheme for grid-connected voltage-source inverters. IEEE
nonlinear controller design for three-phase grid-connected Transactions on Industrial Electronics, 58(1), 202–212.
photovoltaic systems under structured uncertainties. IEEE Yepes, A., Freijedo, F., Doval-Gandoy, J., López, O., Malvar, J.,
Transactions on Power Delivery, 29(3), 1221–1230. & Fernandez-Comesana, P. (2010). Effects of discretization
Mahmud, M., Pota, H., & Hossain, M. (2014). Nonlinear cur- methods on the performance of resonant controllers. IEEE
rent control scheme for a single-phase grid-connected pho- Transactions on Power Electronics, 25(7), 1692–1712.
tovoltaic system. IEEE Transactions on Sustainable Energy, Yepes, A., Freijedo, F., López, O., & Doval-Gandoy, J. (2011).
5(1), 218–227. High performance digital resonant controllers imple-
Mahmud, M., Pota, H., Hossain, M., & Roy, N. (2014). Robust mented with two integrators. IEEE Transactions on Power
partial feedback linearizing stabilization scheme for three- Electronics 26(2), 563–576.
phase grid-connected photovoltaic systems. IEEE Journal of Yuan, X., Merk, W., Stemmler, H., & Allmeling, J. (2002).
Photovoltaics, 4(1), 423–431. Stationary-frame generalized integrators for current con-
Malinowski, M., & Bernet, S. (2008). A simple voltage sensor- trol of active power filters with zero steady-state error for
less active damping scheme for three-phase PWM convert- current harmonics of concern under unbalanced and dis-
ers with an LCL filter. IEEE Transactions on Industrial Elec- torted operating conditions. IEEE Transactions on Industry
tronics, 55(4), 1876–1880. Applications, 38(2), 523–532.
Marei, M.I., El-Saadany, E.F., & Salama, M.M.A. (2003). An Zmood, D., & Holmes, D. (2003). Stationary frame current reg-
intelligent control for the DG interface to mitigate voltage ulation of PWM inverters with zero steady-state error. IEEE
flicker. In 18th Annual IEEE Applied Power Electronics Con- Transactions on Power Electronics, 18(3), 814–822.
ference and Exposition (Vol. 1, pp. 179–183). Miami Beach, Zue, A., & Chandra, A. (2009). State feedback linearization con-
FL: IEEE. trol of a grid connected photovoltaic interface with MPPT.
Park, S.-Y., Chen, C.-L., Lai, J.-S., & Moon, S.-R. (2008). Admit- In 2009 IEEE Electrical Power Energy Conference (pp. 1–6).
tance compensation in current loop control for a grid-tie Montreal, QC: IEEE.

View publication stats