2020 7th International Conference on Signal Processing and Integrated Networks (SPIN)

Backstepping Control of Three-Phase Grid-

Connected PV System
N C Sahoo Akshaya K. Pati
Nirbhay Kumar
School of Electrical Sciences School of Electrical Sciences
School of Electrical Sciences
Indian Institute of Technology Indian Institute of Technology
Indian Institute of Technology
Bhubaneswar Bhubaneswar
Bhubaneswar
Bhubaneswar, India Bhubaneswar, India
Bhubaneswar, India
[email protected] [email protected]
[email protected]

Abstract—In this work, a backstepping control approach by using the linear controllers, the desired reference values
is applied on a three-phase grid-integrated PV system. In may not be tracked under a wide range of operating regions.
a conventional two-stage power conversion system, a These controllers are operated in limited range [5]. Thus, a
boost and a DC-AC converters are used which are highly suitable controller should be selected to overcome the
nonlinear systems. In conventional controller, these drawbacks of linear controller. In literature, attempts were
nonlinearities are ignored; they works only in linear made to control the converters using some nonlinear
region and hence operates in limited range. For efficient controllers such as feedback linearization (FBL) and sliding
power transfer, tight regulation of DC-link voltage is mode control (SMC). The FBL method often cancels out
required under generation and demand-side variations; some useful nonlinearities which can enhance the system
so an effective controller is required. Backstepping stability. In the SMC sliding surface is designed to achieve
control is a nonlinear control technique which takes care the control objective. SMC produces discontinuous
of useful nonlinearities in the system, helping to enhance controller, which suffers from chattering in the presence of
system stability. The controller is designed by using unmolded high-frequency dynamics [6]. In a grid-integrated
dynamic equations of the system. A suitable Lyapunov PV system, several control blocks are required to get the
function is chosen to verify the stability of the controller. desired operation. For extraction of maximum power from
The performances of the controller are evaluated under the PV system, a maximum power point tracking (MPPT)
different dynamic conditions such as varying solar algorithm is required. In industry and residential systems,
insolation, changing reactive and active power demand. perturb and observe (P&O), and incremental conductance
The robustness of the presented controller is validated in (INC) methods are widely used MPPT algorithms [7]. The
simulation studies using MATLAB/Simulink. DC-link capacitor voltage regulation is required for power
Keywords—Voltage source inverter, Backstepping controller, transfer from the PV system to the grid. To maintain the
Control Lyapunov function, Nonlinear controller, Photovoltaic power quality, reactive power management is also required.
array. The main purpose of this paper is to design a controller for
the two-stage grid-connected PV system. Here, the
I. INTRODUCTION
backstepping controller is implemented on VSI to control
In recent years, the energy demand is growing day by day. It DC-link capacitor voltage and inject maximum power to the
leads to increase in power generation capacity, which causes grid. The performance of the backstepping controller is
depletion of fossil fuels, emission of greenhouse gases. To evaluated by simulating many realistic scenarios.
minimize these effects, use of renewable energy sources is In this paper, the system modelling is discussed in
increasing. The most widely available renewable energies Section II. The backstepping control of three-phase VSI is
are solar, wind, and fuel cell. Among the available sources, presented in Section III. In Section IV, simulation results are
solar energy is the most prominent one. Annually, earth presented. The conclusion of this work are discussed in last
receives nearly 1, 73,000TW energy from the sun. By Section.
suitable use of this much solar energy, the power demand at
II. SYSTEM MODELING
present and future can be fulfilled [1-2]. To utilize the solar
energy, PV module and power converters are required [3]. A typical two-stage grid-integrated PV system consists of a
The building blocks of the power converter are power DC-DC converter which is used for extracting maximum
diodes and switches which are highly nonlinear in nature. In power from PV system and a three-phase VSI which is used
a grid-integrated PV system, the nonlinearities are also due for DC to AC conversion [8]. The basic blocks of complete
to parametric uncertainties, nonlinear load, sudden system is shown in Fig. 1.
disturbances such as load change or atmospheric conditions, I o I dc Ia
and constrains applied on control structure. Hence the Va R L V ga
complete grid-integrated PV system is a nonlinear system. C Vdc Vb V gb
Thus, designing of a suitable controller for these converters Vc
Ib
became more complex and challenging task [4]. Therefore, Ic Vgc
an appropriate controller should be developed to minimize Cf
the effects of intrinsic nonlinearities with faster response VPV

and also ensures the stability against the power variation I PV

from the generation and load demand. The conventional
linear controllers are widely used due to its simple structure Fig. 1: A typical three-phase grid-integrated PV system
and ability to solve many practical control issues. However,

XXX-X-XXXX-XXXX-X/XX/$XX.00 ©20XX IEEE

978-1-7281-5475-6/20/$31.00 ©2020 IEEE 923

Authorized licensed use limited to: Auckland University of Technology. Downloaded on May 28,2020 at 15:00:01 UTC from IEEE Xplore. Restrictions apply.
 2020 7th International Conference on Signal Processing and Integrated Networks (SPIN)

In this study, the three-phase VSI is controlled by using The power-voltage (P-V) characteristic waveform and its
backstepping control. Here, the backstepping control is used maximum power point (MPP) of a typical PV array is
for below mentioned control objectives: presented in Fig. 2. In industry or residential system, the
(i) Maximum power transfer. P&O MPPT algorithm is widely used. In this MPPT
(ii) DC-link capacitor voltage regulation. algorithm, the PV array voltage is perturbed in order to
(iii) Reactive power control for power quality improvement. reach voltage corresponding to the MPP of the PV array. In
this algorithm, the perturbation of the operating point is
By applying Kirchhoff’s law at the DC-side of the inverter, based on the initially selected parameter. If the MPP is on
dVdc the right side of the initial operating point then perturbation
C = I o − I dc (1)
dt should be positive else perturbation will be negative. This
process will continue till the MPP of the PV system is
In the above expression, C is DC-link capacitor and Vdc is reached. A flowchart of P&O MPPT algorithm is shown in
the voltage across the capacitor. I o is the boost converter Fig. 3.
output current, I dc is the input current to VSI. The DC-link
capacitor voltage regulation is required for smooth power
transfer from PV side to inverter.
The active and reactive powers at the grid side are expressed
as:

3
Pg =
2
(Vgd I d + Vgq I q )
(2)
3
Q g = (V gq I d − V gd I q )
2

In the above expressions, ª¬Vd Vq º¼ and ª¬Vgd Vgq º¼ are the

inverter voltages and grid voltages in the synchronously Fig. 2: Typical P-V characteristic of a PV array

rotating frame, respectively. ª¬ Id I q º¼ are the grid currents.

By considering the q-axis component ( Vgq ) as zero, the
power flow to the grid can be expressed as:
3 3
Qg = − Vgd I q ; Pg = Vgd I d (3)
2 2
ǻP = 0
Here, by assuming a lossless power conversion system, the
power at both AC and DC side should be equal which can
be expressed as:
ǻP > 0
3
Vdc I dc = Vgd I d ; I dc = 3 Vgd I d (4) ǻV > 0
2 2Vdc ǻV > 0

By substituting equation (4) in equation (1)

dVdc I o 3
= − Vgd I d (5)
dt C 2CVdc

Mathematical modeling of three-phase inverter integrated to

grid in synchronous rotating frame can be expressed as [9]: Fig. 3: Flow chart diagram of P&O algorithm
dI d Vd Vgd RI d III. BACKSTEPPING CONTROLLER FOR THREE-PHASE VSI
= + ωIq − − (6)
dt L L L Backstepping controller is designed by using a recursive
process which is a systematic method of nonlinear control
dI q Vq Vgq RI q approach [10]. In this control design, the system is
= − ωId − − (7)
dt L L L decomposed into subsystems and a suitable stabilizing
function (which is a virtual control) is to be selected. The
Where L is filter inductance, R is intrinsic resistance, stability of each subsystem is obtained by using the
and ω = 2π f , where f is the grid frequency. The equation Lyapunov stability approach at different stages of the design
(5),(6) and (7) describes the complete dynamics of the process [11]. At the end, the overall structure of the
system. These equations are used for designing the complete system is formulated by using the control
backstepping controller. Lyapunov function (CLF). Because of the nonlinearities, a
linear controller (such as PI controller) is generally not very

924

Authorized licensed use limited to: Auckland University of Technology. Downloaded on May 28,2020 at 15:00:01 UTC from IEEE Xplore. Restrictions apply.
 2020 7th International Conference on Signal Processing and Integrated Networks (SPIN)

much effective under widely changing operating conditions. The Lyapunov candidate function for the subsystem defined
Here a backstepping controller, which is inherently by Eqn. (5) is defined as:
nonlinear, is designed for this purpose. 1
The main objective of the controller to meet the following V1 = z12 (11)
requirements. 2
• • ªI 3 º
(i) Tight regulation of DC-link capacitor voltage. V1 = z1 z1 = z1 « o − Vgd I d » (12)
(ii) Transfer of maximum power to the grid. ¬ C 2CVdc ¼
•
(iii) Transfer of the desired reactive power to the grid.
For this subsystem to be stable V1 should be negative semi-
(iv) Faster transient response.
definite function. To attain that the state variable I d acts as
This two-stage grid-connected PV system is expressed by a
a virtual control.
set of three differential equations (as described earlier).
These equations can be represented by three subsystems. 2CVdc ª I o º
I d* = « + k1 z1 » (13)
The design process starts from equation (5), where the 3Vgd ¬ C ¼
objective is to track the desired DC-link voltage reference.
The virtual control I d* insures that the equation (12) results
This subsystem will be stabilized completely by using state
variable I d * (virtual control) as a control input, which is in negative semi-definite function as shown in equation
(14).
considered as a reference value for the inner control loop. •
The subsystem corresponding to equation (6) will be V1 = − k1 z12 (14)
stabilized to track I d * by using V d as a control input, which
Step 2: I *
d
is not the actual control input, now the error
is the final control parameter. Similarly, the reactive power
variable z 2 is defined as:
can also be regulated using equation (7) by using V q as a
z 2 = I d − I d* (15)
control input. The control structure for the complete system •
• •
is shown in Fig. 4. The detail blocks of the backstepping z 2 = I d − I d* (16)
controller is shown in Fig. 5, where U is the six switching
pulses to the VSI.
For the subsystem represented by Eqn. (6) the Lyapunov
candidate function is expressed as:
1
C V2 = z22 (17)
2
V is the Lyapunov candidate function defined for system
8

defined by Eqn. (5) and (6) is.

V pv I abc Vabc
Vdc Vgd Vgq V = V1 + V2 (18)
I pv
•
Id Iq
Io For this system to be stable V should be negative semi-
Qref Vdc* definite.
• •
Fig. 4: Complete control structure for three-phase grid- V = −k1 z12 + z2 z2 (19)
integrated PV system.
Vdc
Substituting equation (16) in equation (19)
• • •

Vdc Z1 Id *
Z2 Vd V = −k1 z12 + z2 ( I d − I d* ) (20)
ª I o z2 k1 z1 z 2 2k1 z1 I o 2k z C º 2 2

«− − − − 1 1
Vgd Io Z1 Vgd I o »
« Vdc Vdc 3Vgd 3Vgd »
*
Vdc Id •
*
Iq
Z3 Vq Id = « • » (21)
« 2Vdc I o 2k12 CVdc z1 »
« + 3V − k1 z 2 − 3V »
I qref I q I qref ¬ gd gd ¼
Fig. 5: Controller block diagram for VSI Substituting equation (6) into equation (20)
Step 1: The DC-link capacitor voltage tracking error can be
expressed as: • ªV Vgd RI d •* º
V = −k1 z12 + z2 « d + ω I q − − − Id » (22)
z1 = Vdc − Vdc* (8)
¬L L L ¼
Where Vdc* is the DC-link voltage reference. •
• • • For V to be negative semi-definite final control law Vd is
*
z1 = Vdc − V (9)
dc
ª −R Vgd I o z 2 k1 z1 z2 kzI º
By using equation (5) the dynamics of z1 can be expressed « Vdc + ω I q − + + +2 1 1 o »
« L L V V 3Vgd »
as: dc dc
Vd = − L « • »
• dV I 3 2 2
« 2k1 z1 C 2Vdc I o 2k12 CVdc z1 »
z1 = dc = o − Vgd I d (10) + − + k z + + k z
dt C 2CVdc « 3V 3V
1 2
3 V
2 2»
¬ gd gd gd ¼
Derivative of Vref is 0 as it is considered as a constant value.
(23)

925

Authorized licensed use limited to: Auckland University of Technology. Downloaded on May 28,2020 at 15:00:01 UTC from IEEE Xplore. Restrictions apply.
 2020 7th International Conference on Signal Processing and Integrated Networks (SPIN)

• shown in Fig. 7. In this figure, during the transient

V = −k1 z12 − k2 z22 (24) condition, the inverter output power is negative, it means the
Step 3: The reactive power injected to the grid, can be power is drawn from the grid to maintain the DC-link
expressed as voltage. To maintain the DC-link voltage at reference
3 voltage, the initial stored energy in the capacitor is also
Qref = − Vgd I qref (25) transferred to the grid during transient condition. Hence the
2
2Qref power balance during transient condition is obtained by
I qref = − (26) considering the power generated by the PV system plus
3Vgd extra energy stored in the DC-link capacitor and the power
The reactive power demand should be supplied by the injected to the grid with converter switching losses. The
inverter. Thus, the error variable corresponds to the reactive inverter output current for both the controller are shown in
power can be defined as: Fig. 8(a) and (b). The three-phase current is less distorted in
z 3 = I q − I q ref (27) case of backstepping control where the THD is 5%, where
For the subsystem defined by Eqn. (7), the Lyapunov as in case of PI controller THD is 14% as shown in Fig.
candidate function V 3 can be defined as: 8(b).
1
V3 = z32 (28)
2
•
ª• •
º ªVq Vgq RI q • º
V3 = z3 « I q − I qref »¼ = z3 « − ω I d − − − I qref » (29)
¬ ¬L L L ¼
•
For this subsystem to be stable V3 should be negative semi-
definite function. The final control law Vq can be expressed
as
ª −R •
º
Vq = − L « I q − ω I d − I qref + k3 z3 » (30) Fig. 6: DC-link voltage
¬ L ¼
In these expressions, the tunable parameters
k1 , k 2 and k3 should be positive. After getting V d , V q three
modulating signals can be generated. These three
modulating signals will generate six switching pulses for the
inverter using PWM generator, as shown in Fig. 5. All the
Lyapunov expressions are negative semi definite function
for positive value of gain parameters.
IV. SIMULATION RESULTS
To validate the effectiveness of the backstepping control (a)
structure, the simulation of the complete system with
controller is carried out by using MATLAB/Simulink. The
simulation parameters are presented in Table-1 of the
Appendix.
Case 1: Steady-state condition
In this case, the controller performance is evaluated under
standard test condition (STC), solar insolation of 1000W/m2
and temperature of 25Ԩ. The main objective is to inject
maximum available PV system power to the grid at unity
power factor. It depends on the voltage tracking (b)
performances of the DC-link capacitor. The voltage Fig. 7: Active power injected to the grid using (a)
regulation performance across the DC-link capacitor for Backstepping control; (b) PI controller.
both backstepping and PI controller are shown in Fig. 6. The
backstepping control (settling time 0.05s) has faster
response in comparison to the PI controller (settling time
0.1s), but have higher peak overshoot in the DC-link voltage
as shown in Fig. 6. Which can be sustained by the capacitor.
The simulation of both PI and backstepping controller are
studied under same set of system parameters. The power
injected from the PV system to the grid by using both the
backstepping and PI controller are shown in Fig. 7(a) and
(b), respectively. The transient response of the backstepping (a)
controller is better in comparison to the PI controller as

926

Authorized licensed use limited to: Auckland University of Technology. Downloaded on May 28,2020 at 15:00:01 UTC from IEEE Xplore. Restrictions apply.
 2020 7th International Conference on Signal Processing and Integrated Networks (SPIN)

(b) (a)
Fig. 8: Three-phase current injected to the grid using (a)
Backstepping control; (b) PI controller.

Case 2: Varying solar irradiance

In this case, initial solar irradiance is assumed to be
1000W/m2 and steps down to 800W/m2 at t=0.44s and
maintains this level up to t=1s, then after 1s, the solar
irradiance again steps down to 500W/m2. Controller
performance is evaluated for both backstepping control and
PI controller under this scenario. The distortion in three-
phase current is more in the case of PI controller in (b)
comparison to the backstepping control, as shown in Fig. 9. Fig. 10: Active power injected to the grid using (a)
Active power transferred to grid experiences undershoot at Backstepping control; (b) PI controller.
the point of irradiance change in case of PI controller as
shown in Fig. 10(b), which is not the case with backstepping
control as shown in Fig. 10(a). With the change in reactive
power demand there is not much difference observed in the
performance of the two controller except a small amount of
ripple in case of PI controller as shown in Fig. 11(b). The
DC-link experiences voltage dip with the sudden change in
solar insolation in case of PI controller as shown in Fig.
12(b), but in case of backstepping control these voltage dips
are insignificant as shown in Fig. 12(a).
(a)

(a)

(b)
Fig. 11: Reactive power injected to the grid using (a)
Backstepping control; (b) PI controller.

(b)
Fig. 9: Current injected to the grid under changing insolation
using (a) Backstepping control; (b) PI controller.
(a)

927

Authorized licensed use limited to: Auckland University of Technology. Downloaded on May 28,2020 at 15:00:01 UTC from IEEE Xplore. Restrictions apply.
 2020 7th International Conference on Signal Processing and Integrated Networks (SPIN)

k1 105
k2 105
k3 8.34 *10 7
k pDC 7

kiDC 800
Current loop k p , ki 0.3,20

(b)
Fig. 12: DC-link voltage in case of (a) Backstepping control; V. CONCLUSION
(b) PI controller.
This paper presents a backstepping control approach for a
Case 3: Changing load condition three-phase VSI based two-stage grid-connected PV system.
In this case, the load at PCC (point of common coupling) is The backstepping controller presents a fast and smooth
varied from 0 to 1000W at t= 0.2s. This sudden change in transient response for voltage regulation across the DC-link
load acts as a disturbance to the PV system as shown in Fig. capacitor, as a result the power transfer to the grid side are
13(a). Due to the control action of the backstepping accurately regulated. From the simulation results, the
controller, the DC-link capacitor voltage remains performance of the backstepping control is found to be
unaffected. However, in the case of PI controller, the superior in terms of transient response, tracking
transient are observed during changing load conditions, as performances and disturbance rejection in comparison to the
presented in Fig. 13(a). The active power shared between PI controller.
the grid and PV system is shown in Fig. 13(b).
REFERENCES
[1] J. P. Benner and L. Kazmerski, "Photovoltaics gaining greater
visibility," in IEEE Spectrum, vol. 36, no. 9, pp. 34-42, Sept. 1999..
[2] Yeo, S. M., Cho, H. I., Kim, C. H., Funabashi, T., & Senjyu, T.
(2009). Modeling and analysis of photo-voltaic system using EMTP.
In Transmission and Distribution Conference and Exposition: Asia
and Pacific, T and D Asia 2009.
[3] X. Guo, J. Chen and Q. Liu, "Real-time and grid-connected control of
PV power system," 2011 International Conference on Advanced
Power System Automation and Protection, Beijing, 2011, pp. 923-
928.
(a) [4] T. K. Roy, M. F. Pervej, S. Rahman and F. K. Tumpa, "Nonlinear
adaptive controller design for PWM based DC-DC boost converters
using bankstepping control scheme," 2016 5th International
Conference on Informatics, Electronics and Vision (ICIEV), Dhaka,
2016, pp. 608-613.
[5] Krafes, S., Chalh, Z., & Saka, A. (2016). Review: Linear, nonlinear
and intelligent controllers for the inverted pendulum problem.
In Proceedings of 2016 International Conference on Electrical and
Information Technologies, ICEIT 2016 (pp. 136–141).
[6] H. K. Khalil, Nonlinear Systems, Mcmillan 1992; Prentice Hall 1996
(second edition) and 2002 (third edition). The second edition received
the IFAC Control Engineering Textbook Award in 2002.
[7] A. Kchaou, A. Naamane, Y. Koubaa and N. K. M'Sirdi, "Comparative
(b) study of different MPPT techniques for a stand-alone PV
Fig. 13: (a) DC-link capacitor voltage under load change; system," 2016 17th International Conference on Sciences and
(b) Power shared under load change. Techniques of Automatic Control and Computer Engineering (STA),
Sousse, 2016, pp. 629-634.
[8] N. Altin, S. Ozdemir, H. Komurcugil, I. Sefa and S. Biricik, "Two-
Table I. Simulation Parameters stage grid-connected inverter for PV systems," 2018 IEEE 12th
Parameters Values International Conference on Compatibility, Power Electronics and
PV array Maximum power ( Pm ) 620 W Power Engineering (CPE-POWERENG 2018), Doha, 2018, pp. 1-6.
[9] Rolan, A., Gimenez, P., Yague, S. J., & Bogarra, S. (2018). Modeling
Open circuit voltage (Voc) 98.7 V and Control of a Three-phase Grid-connected Inverter under Fault
Voltage at MP(VMPP) 79 V conditions. In SPEEDAM 2018 - Proceedings: International
Short circuit current(Isc) 8.21 A Symposium on Power Electronics, Electrical Drives, Automation and
Motion (pp. 479–484). Institute of Electrical and Electronics
DC-link voltage reference ( Vref ) 300 V Engineers Inc.
Grid voltage ( V g ) 110 V ine- [10] J. J. Slotine, W. Li, “Applied nonlinear control,” Englewood Cliffs
line (NJ), Prentice-Hall, 1991.
DC link capacitor (C) 220 μ F [11] Salimi, M., Soltani, J., Markadeh, G. A., & Abjadi, N. R. (2013).
Indirect output voltage regulation of DC-DC buck/boost converter
Filter inductance ( L ) 10 mH operating in continuous and discontinuous conduction modes using
10 μ F adaptive backstepping approach. IET Power Electronics, 6(4), 732–74
Filter capacitance ( C f )

928

Authorized licensed use limited to: Auckland University of Technology. Downloaded on May 28,2020 at 15:00:01 UTC from IEEE Xplore. Restrictions apply.