10 1109@tia 2018 2863652

This article has been accepted for publication in a future issue of this journal, but has not been
fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2018.2863652, IEEE
Transactions on Industry Applications
Performance Improvement of Grid Integrated Solar

PV System using DNLMS Control Algorithm
Subarni Pradhan, Student Member, IEEE, Ikhlaq Hussain, Member, IEEE, Bhim Singh, Fellow, IEEE
and Bijaya Ketan Panigrahi, Senior Member, IEEE
Abstract—An integration of renewable sources based DBWF Decomposition based wavelet filters
distributed generating systems, encounters various power quality FC Power factor correction
(PQ) issues due to uncertain loads at the distribution end. These ZVR Zero voltage regulation
uncertainties arise due to nonlinearity, disturbances or DO Decorrelation operator
unbalanced loads. A 3-phase grid integrated solar photovoltaic
(PV) system incorporating a control technique based on modified
DPO Decorrelated posteriori error
decorrelation normalized least mean square (DNLMS) algorithm, PLL Phase locked loop
aiming to enhance its overall performance under adverse THD Total harmonic distortion
conditions, is presented in this work. The 3-phase, grid tied, MSD mean square deviation
single stage solar PV system comprises of solar PV array with a
suitable maximum power point tracking (MPPT) method, filters, LIST OF SYMBOLS
loads, and a capacitor fed voltage source converter (VSC). The Lfi, Cpdc interfacing inductors, DC link capacitor
key objectives of the solar PV integrated structure with adaptive
Vpdcrf, Vpdca Reference DC link voltage, Sensed DC
law based control algorithm is to attain unity power factor (UPF)
at the grid end ensuring harmonics mitigation from the grid
link voltage
currents. Moreover, this structure effectively transfers active Γpv, Γiv Proportional and integral gains of PI
power from the PV array to the local loads and the grid. These controller controlling the DC link voltage
aforesaid objectives are achieved through providing controlled φpa, φpb, φpc Estimation of active weight components of
switching pulses to the insulated gate bipolar transistor (IGBT) load currents
based VSC using a modified DNLMS control algorithm with fast φqa, φqb, φqc Estimation of reactive weight components
convergence rate. Harmonics free, sinusoidal reference grid of load currents
currents, are obtained by using a modified DNLMS algorithm. A φcp, φcq, φpv DC link loss component, AC loss
simulation model developed in MATLAB/ Simulink, is used for
component, PV feed-forward component
validation of a modified DNLMS based control approach. In the
laboratory, an experimental prototype is developed and proposed
φavp, φavq Average active and reactive weight
algorithm is implemented to verify its performance. components of load
φap, φrq Resultant active and reactive weight
Keywords— Photovoltaic (PV) Array, MPPT Technique, Total components
Harmonics Distortion, DNLMS Algorithm and Power Quality. vpa, vpb, vpc Phase voltages of POI
LIST OF ACRONYMS vsab, vsbc Line voltages of POI
û spa , û spb , û spc Decorrelated in-phase unit template
PQ Power quality
û sqa , û sqb , û sqc Decorrelated quadrature unit template
POI point of interconnection
PV Photovoltaic iLa, iLb, iLc Three phase load currents
DNLMS Decorrelation normalized least mean square Vtprf, Vtpa Reference terminal voltage, terminal
MPPT Maximum power point tracking voltage
VSC voltage source converter Γpc, Γic Proportional and integral gains of PI
UPF unity power factor controller controlling the POI voltage
IGBT insulated gate bipolar transistor Vpva, Ipva, Ppva, PV voltage, current, power
UNFCCC United Nations framework convention on µ Variable step size
climate change ȇspa, ȇspb, ȇspc, Prediction errors
INC Incremental conductance ȇsqa, ȇsqb, ȇsqc
P&O Perturb and observe α, b Decorrelation operator, Bias factor
PSO Particle swarm optimization i*spa , i*spb , i*spc Active part of reference grid currents
VSSIR Variable step size incremental resistance i*sqa , i*sqb , i*sqc Reactive part of reference grid currents
LIA Lock in amplifier Reference grid currents
i*sa , i*sb , i*sc
GWO Grey wolf optimization
isa , isb , isc Sensed grid currents
SMC Sliding mode controller
RSEGS Renewable sources based energy generating x(r), x̂(r) , η(r) input signal, decorrelated input signal,
systems noise
LMF Least mean fourth φ(r) Weight parameter associated with the
LMS Least mean square adaptive law
VSS-LMS Variable step size least mean square ê(r) Decorrelated output error
NLMS Normalized least mean square ϑ, θ Deviation of weight, forgetting factor
1
0093-9994 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2018.2863652, IEEE
I. INTRODUCTION amplitude of signals ensuring harmonics elimination and

Global warming as a result of climate change, has been improved performance of grid integrated distribution system.
established beyond doubt. In fact, the fossil fuel based energy Moreover, various filters and observer structures have been
sources emitting greenhouse gases, are the lead cause of the proposed for harmonics and fundamental extraction [18-25].
effect. Coordinated global effort under the aegis of United Some of them are adaptive vectorial filter (AVF) [18], Kaman
Nations framework convention on climate change (UNFCCC), filter, decomposition based wavelet filters (DBWF), concise
is being attempted to contain the rise of global atmospheric discrete adaptive filter (CDAF) [19], wiener filter [20],
temperature below 20 C of pre-industrial level, which are to be frequency adaptive observer [21-22] etc. Estimation of weight
reflected in the intended nationally determined contributions parameters in many fields like signal processing [23-26] and
(INDCs) of each country under UNFCCC framework. This power system applying adaptive controller, has become a key
underscores the centrality of renewable energy sources as area of research owing to the merits of the method. Simplicity
dominant alternative to the fossil fuel based sources of of computation, has encouraged the use of simple LMS by
electrical power generation especially when electrical energy is researchers [23]. In LMS, the squared error is taken as the cost
being envisaged to replace petroleum as energy sources for function. Further enhancement has been made in the LMS
vehicles, which are among the major sources of environmental technique such as normalized LMS (NLMS) [24], LMF,
pollutants. While renewable energy includes many existing and adaptive LMS (ALMS), Filter-X LMS (FXLMS) [25],
evolving resources in its fold, solar power leads the pack quantized kernel least mean square (QKLMS). Some
currently. The extensive research and development work have publications have proposed decorrelation normalized LMS
been done on the technology making it commercially more (DNLMS) [26] where the error is calculated by considering a
competitive. Power engineers exhibiting interest toward decorrelation operator. Minimizing l2 norm of the decorrelated
development of advanced information technologies in this field posteriori error (DPE) with decorrelation operator (DO),
by modeling, designing and implementing novel techniques. contributes equations of the DNLMS based control algorithm,
An extensive literature work has been reported to model solar thus resulting in fast convergence rate and good performance.
photovoltaic (PV) array [1] and for harnessing its peak power This paper deals with a modified DNLMS based control
[2-3]. Among the peak power extraction techniques, perturb algorithm for a single stage solar PV system integrated to the
and observe (P&O), incremental conductance (INC), variable national grid. The modified DNLMS algorithm is used to
step size incremental resistance (VSSIR), digital lock in estimate the active weight component of the load current,
amplifier (LIA), fuzzy logic method, Grey wolf optimization which is used to generate the reference grid currents. A
(GWO), particle swarm optimization (PSO), sliding mode decorrelation operator is taken for improvement of the filter
controller (SMC) etc. are some adopted conventional methods. performance. This filter works through the estimation of
Integration of renewable sources based energy generating weight component of the input signal. This filter depends on
systems (RSEGS) to the national grid at the point of the weight updating law, which depends on a step size. A
interconnection (POI), is performed through solid state devices proper step size has to be chosen for fast convergence and
based power converters like voltage source converters (VSCs), accurate results. Here, the step size has the advanced
buck and boost converters. Double stage and single stage proficiency to regulate itself to fulfill the need, thereby proving
systems are the most common arrangements adopted to its behavior under steady as well as dynamic conditions.
integrate the PV array with the national grid. A boost converter Hence, the estimated weight has the capability to tune itself
responsible for peak solar power tracking, is incorporated according to the requirement. Stability of the algorithm is
between the PV array and VSC, executing power conversion in analyzed through the convergence of weight deviation vector in
two stages. In the contrary, the boost converter is absent in the the sense of mean square error. Moreover, dynamic behavior of
later one [4] performing single stage of power conversion. By the system is studied under sudden change in the unknown
mounting the solar PV array at the distribution end, provides system. Main objectives of the proposed system are as follows.
alleviated burden on the conventional grid at the POI by  Estimation of the active and reactive fundamental weights
meeting energy demand of the customer at higher level of of the load current, is done by using modified DNLMS
satisfaction. However, an integration of RSEGS encounters algorithm. Necessity of the PLL is avoided by unit template
challenging issues due to uncertainties in the distribution estimation from the grid voltage.
system [5]. These uncertainties arise due to nonlinearity,  Power factor correction (PFC), grid currents balancing and
disturbances or unbalanced loads. Moreover, grid tied RSEGS reduced total harmonic distortion (THD) at grid side
come across challenges in maintaining reliability, stability, parameters at nonlinear load under steady state and
synchronization and performance of the overall system. Hence, dynamic conditions.
the grid tied RSEGS have to satisfy the restrictions imposed by  Capability of meeting peak solar power at different solar
the grid codes [6-7]. Grid tied distributed generators suffer irradiation.
from various power quality (PQ) issues at the load as well as  This structure effectively transfers active power from the
grid side [8]. Hence, it is mandatory to maintain the reliability, solar PV array with P&O based MPPT technique, to the
stability and quality of energy supplied to the utility. The local loads as well as the grid.
aforementioned issues are addressed in the literature and
The grid tied solar PV system, including the proposed
further improvement in these techniques have been going on
modified DNLMS based control is simulated at nonlinear load
[9-23] to enhance the system dynamic performance.
under various dynamic conditions describing its working and
Many adaptive techniques such as as least mean fourth system performance. A comparative study on the performance
(LMF) [9], direct power control [10], Discrete Fourier of modified DNLMS algorithm with conventional algorithms,
transform (DFT), droop control [11], DNLMS [12], modified is presented to study the superiority of the given logic. A
variable step size least mean square (VSS-LMS) [13], dual-tree laboratory prototype of the 3-phase grid integrated PV system
complex wavelet transform (DT-CWT) [14], phase locked loop incorporating P&O based MPPT technique and a modified
(PLL) technique [15-17], etc. have been reported in the DNLMS algorithm, is developed. Further, its behavior under
literature that can be used to control, compute phase and dynamic conditions is demonstrated through test results.
2
II. MODIFIED DNLMS ALGORITHM α
The modified DNLMS algorithm is an advanced version of x(r)=i L (r) ˆ = î L (r)  u(r)×
e(r) ˆ  (r)  (r)
îL (r) = i L (r)  α i L (r  1)
the basic DNLMS algorithm with improved performance, u(r) ˆ
u(r)= u(r)  α u(r  1)
stability and robustness in its behavior. In the modified
ˆ
u(r) eˆ (r)
DNLMS algorithm, the step size of the weight equation  (r)
η(r) x(r) e(r) Z 1
adjusts itself to have improved performance. Fig. 1 shows the
schematic of adaptive filter based system identification and
weight estimation structure of the modified DNLMS ˆ
û(r)e(r)
 (r+1) =  (r)  μ  (r+1)
algorithm. Here, φopt is the set of unknown weights and φ is  opt (r)
2
û(r)  b
the set of estimated weight. The input signal (x(r)) for rth u(r)
 (r)

sample time is expressed as, b
(a) (b)
x(r) =  T u(r) + η(r) (1) Fig.1 Schematic of adaptive filter based system identification and weight
where, η(r) and φ are noise and weight parameters, estimation Structure.
respectively. The error between the desired and the actual A. Stability Analysis
signals, decorrelated output error and decorrelated posteriori
From (12) and (13), the weight equation is modified as,
error after introducing the DO (α), are estimated as [26],
e(r) = x(r)   T (r)u(r) (2)  (r  1)   (r) u(r)
ˆ sign e(r) ˆ
 (r+1) =  (r)  (14)
e(r) ˆ   T (r)u(r)
ˆ = x(r) ˆ (3)  
1   ˆ
u(r) 2
ˆ   T (r+1)u(r) Let the error between φopt and φ is denoted as ϑ. Hence,

eˆ p (r) = x(r) ˆ (4)
 (r)  opt   (r) (15)
where, û(r) and x̂(r) are the signals after the introduction of
The deviation vector of weight parameter (ϑ (r+1)) is given as,
‘α’ into the formulation and are expressed as [26],
 (r  1)   (r) u(r) ˆ
ˆ sign e(r)
x̂(r) = x(r)  α x(r  1) (5)  (r  1)   (r)  (16)
û(r) = u(r)  α u(r  1) (6)
1    ˆ 2
u(r)
Hence, the error can also be written as, Multiplication of ϑT(r+1) to (16) yields,
ê(r) =  x(r)   T (r)u(r)   α  x(r  1)   T (r) u(r  1)  (7)  (r  1) T (r  1)   (r) T (r) 
 (r  1)   (r)
where, φ(r) is the weight parameter associated with the
1   
adaptive law, derived from the cost function defined as,  u(r)
ˆ  T (r) sign e(r)ˆ ˆ 
 (r)uˆ T (r)sign e(r)
   (17)
min eˆ p (r) =  x(r) ˆ 
2
ˆ   (r+1)u(r)
T
(8) ˆ
u(r) ˆ
u(r)
 (r +1)  2 2 
Such that,  (r+1)   (r) 2  
2
(9)  (r  1)   (r) u(r)u
ˆ ˆ (r)
T

Depending upon its robust behavior, the weight at each sample 1    û(r)
2
2
time is updated using the adaptive law, The expectation of (17) gives,
ˆ
û(r)sign e(r)   (r  1)   (r)
 (r+1) =  (r)  μ' (10) E  (r  1) T (r  1)  E  (r) T (r)  E 
û(r) 2  b  1   
where, µ is the step size responsible for adjusting the weight in  u(r)
ˆ  T (r) sign e(r) ˆ ˆ  
the adaptive law such as,     (18)
 ˆ 2
u(r) ˆ 2
u(r)
  
 ê(r) 
μ' = min   (r ),μ  (11)   (r  1)   (r) u(r)u
ˆ ˆ T (r) 
 û(r) 2 
 E  2 
 1    û(r) 2 
The basic and complete structure of the weight estimation
equation is written as, When the system reaches steady state (i.e. r →  ), one can
ˆ
û(r)e(r) write,
 (r+1) =  (r)  μ 2
or E  (r  1) T (r  1)  E  (r) T (r) and  (r)  E  (r) in (18)
û(r) 2  b
(12) as,
 ê(r)  û(r)sign e(r)
ˆ
 (r+1) =  (r)  min   ,μ   u(r)
ˆ  T (r) sign e(r)
ˆ ˆ 
 ˆ 2  u(r)
u(r) ˆ 2 b E  
 ˆ 2
u(r) ˆ 2
u(r)
 
For stability analysis, a delta sequence is chosen as, (19)
 u(r)u
ˆ ˆ T (r) 
2  E  () E  2 
 (r  1)   (r)  (1   )  (r+1)   (r) 2  û(r) 2 
 ê(r)
2
 (13) Trace of both the sides results in,
=  (r)  (1   )min  μ 2 ,  (r) 
   (r) u(r)sign ˆ 
2
û(r) 2
T
ˆ e(r)
  2E    E  () (20)
 û(r) 2 
where, the forgetting factor (θ) is chosen with the operating
range of 0 < θ < 1. Taking assumption of  T (r) u(r) ˆ for sufficiently long filter,
3
 T (r)u(r) ˆ 
ˆ sign e(r) 2 E  (r  1)   (r) ˆ
E
û(r) 2
 K ( r  1)  K (r )  a
 1   

RK (r )  K (r ) Rˆ 
 
(21) (28)
  
 1  2  1  ˆ 2    p 1  p  E  (r  1)   (r)
 E  2 eâ ,  Eˆ  exp   2     Y
 û(r) 2   2 eâ ,
2  2 eˆ ,
 a

    eâ ,r  (K  1) b
2 2
 eˆ2a ,r   b2  1   
 
The detailed derivation of each equation is given in [27]. To
where,  eˆ2a ,  lim E  T (r)uˆ (r) 
2
(22) deal with the effects of a sudden change in the unknown
r 
system, a parameter reset method is used [26] by following the
Applying (21) and E  ( )  0 in (20) results in, steps given here.
ˆ
 eˆ2 ,  Tr RV
a

  0  (23) eˆ f , r  sign(eˆ(r )) max( eˆ( r )  t , 0)
where, Rˆ  E uˆ (r)uˆ T (r) and V  lim E  (r) T (r) .  2f , r 1   2f , r  (1   )eˆ 2f , r

r 
Hence, the steady state mean square deviation (MSD) if  2f , r 1 ( b2   2f , r 1 )  

expressed as,
r 

lim E opt   (r)
2
2 0 proves the  (r)   (0)

else
convergence in the sense of mean square [26].
B. Dynamic Performance
 (r  1)   (r)  (1   ) min  2 eˆ(r )  2 2
uˆ ( r ) 2 ,  (r ) 
For studying of dynamic behavior of modified DNLMS end
algorithm, (18) is modified by considering where, 0    1 ,  is a threshold value and  2f , r 1 is defined
a  E 1 u(r)
ˆ 2  and Y  E u(r)u
  as,
 .
2
ˆ ˆ T (r) u(r)
ˆ as,  2f , r 1  E  u (r ) T ( r ) 
2
 2
E  (r  1)   E  (r)
E  (r  1) T (r  1)  E  (r) T (r)  a The parameter t is preferred to attain robustness against the
1    impulsive noise as,
 2
   1     Y
E  (r  1)   E  (r) eˆ(r )
 ˆ  (r)uˆ T (r)  u(r)
E sign e(r) ˆ  T (r) 
t   ˆ eˆ,r 1  eˆ(r )
, eˆ(r )  ˆ eˆ,r 1
(29)
(24)  2
  b , eˆ(r )  ˆ eˆ,r 1
By defining K(r  1)  E  (r  1) T (r  1) , (24) is written as,
and
K(r  1)  K(r)  a
E  (r  1)   E  (r)
1   
 E sign e(r)
ˆ 
ˆ e,ˆ2 r 1  ˆ e,ˆ2 r  (1   )median eˆr2 , eˆr21 ,..., eˆr2 N 1 
T
 (30)
(25) where, 0    1 , N ,   0 and 0    1 .
   1     Y
E  (r  1)   E  (r)
 (r)uˆ T
ˆ  (r)
(r)  u(r) T
The system at steady state, possesses  2f , r 1   b2 . Under the
circumstance of sudden disturbance in the system,  2f , r 1 tends
By considering the noise sequence (η(r)) is formed by
combining the background noise (b(r)) and an impulsive part to increase and reach  2f , r 1   b2 . In this case, the parameter
(i(r)), presented as,  (r )  b(r )  i (r ) (26) reset method [26], is used.
An assumption is made that b(r) is Gaussian with zero mean
III. SOLAR PV SYSTEM WITH P&O MPPT TECHNIQUE AND A
and E b(r ) 2    b2 ) and an impulsive part is described as, MODIFIED DNLMS BASED CONTROL ALGORITHM
i (r )   (r ) N(r ) , where ω(r) is the Bernouli with P(  1)  p Fig. 2 depicts the schematic of solar PV system consisting
and N(r) is Gaussian with E  N (r ) 2
  K 2
, K  1 and zero of a solar PV array, integrated to the grid at POI through a VSC
b
and interfacing inductors (Lfi) feeding power to a load of
mean. Including the effect of noise to the analysis, an nonlinear nature. Lfi gives advantage of grid current ripples
ˆ  (r)uˆ T (r)  u(r)
expression for E sign e(r) ˆ  T (r)  is derived  mitigation. Moreover, mitigation of switching ripples at the
POI is accomplished using RC filter. Single stage power
as [27], conversion is achieved through the DC link capacitor (Cpdc)
 associated with the VSC. An uncontrolled diode bridge
 1 p
ˆ  (r)uˆ T (r)  u(r)
E sign e(r) ˆ  T (r)   2 rectifier together with RL load, is taken to realize a nonlinear
  eâ ,r   b load. An extensive simulation work is carried out using
2 2
(27) MATLAB Simulink model under possible dynamic conditions

 considering the load as nonlinear. Verification of the control

p 
 eâ ,r  (K  1) b 
2 2

  K ( r ) Rˆ  RK
ˆ (r )  scheme is done by assembling a laboratory prototype. Fig. 3
illustrates the control logics used for the system. Two control
methodologies are involved in the PV system operating at grid
 
ˆ (r ) and p are the probability of impulse
where,  eˆ2a ,r  Tr RV integrated mode. P&O MPPT scheme helps in achieving the
noise. The final model describing the transient behavior of the goal of maximum PV power extraction through estimating the
reference voltage required at DC link (Vpdcrf), which is to be
modified DNLMS control logic is written as,
maintained at desired value.
4
Fig.2 Schematic diagram of 3- phase grid integrated PV system

The power loss in the DC link is also included in the response, can be attained. Moreover, the phase crossover
control scheme with the modified DNLMS algorithm that frequency (ωpm) is greater than the gain crossover frequency
estimates the load active and reactive components. A feed- (ωgm). Therefore, the behaviour of the system is justified under
forward term is included in the control scheme to improve the various operating scenario as all the parameters are satisfying
system performance under dynamic conditions. The reference the stability limits.
current of the grid, depends upon the resultant weight Similarly, in order to achieve the performance under zero
component. It is made sure to get a harmonics free reference voltage regulation (ZVR) mode, the POI voltage can be
grid currents synchronized with the grid voltages without the maintained through AC loss component (φcq). The output of a
help of any PLL. The modified DNLMS based control logic is PI controller that takes the difference between the required
used to estimate the load active and reactive weights. The terminal voltage (Vtprf) and the actual terminal voltage (Vtpa),
system is operated in PFC mode. Hence, the estimated active such as,
weight components of load currents, are taken into account.   
The amplitude of current flow of the system is controlled cq (r)    pv  iv   Vtprf (r)  Vtpa (r)  (34)
through these estimated weights that results in control over the  s 
active power flow from the PV array to the grid, after meeting where, Γpv and Γiv are the PI controller gains.
the local load power demand. Moreover, the PV system with Vtpa denotes the terminal voltage amplitude calculated as [8],
the proposed modified DNLMS based control is capable of 2
addressing unbalanced load, grid currents harmonics and power Vtpa =  v 2pa  v 2pb  v 2pc  (35)
3
factor. Further, the stability analysis of the system is done by
where, the phase voltages (vpa, vpb and vpc) are calculated from
considering the stability of DC link voltage. For achieving
the sensed line voltages as [28],
satisfactory injection of active power to the grid, it is required
to accomplish regulated DC link voltage. This objective is  v pa   1 1 2
  2 v 
achieved by using a PI controller as shown in Fig. 4(a), which v =
 pb  3   1 2 1 2   sab  (36)
v
takes the difference between Vpdcrf and voltage sensed at Cpdc  v pc 
    1 2 1   sbc 
(Vpdca). The PI controller output denotes the power loss at DC
link (φcp) calculated as, The unit-templates (in-phase and quadrature) of the phase
voltage, are determined from following equations,
  
cp (r)    pc  ic   Vpdcrf (r)  Vpdca (r)  (31)  u spa   v pa 0 0  1 Vtpa 
 s       (37)
where, the PI controller gains are denoted as Γpc and Γic. The  u spb  =  0 v pb 0  1 Vtpa 
 u spc   0 0 v pc  1 Vtpa 
plant gain ( G p ( s) ) including the PI controller gain ( G c ( s ) )   
serves the forward loop gain is presented as,  u sqa   0 2 2   u spa 
1      1    (38)
G f ( s )  G p ( s )  Gc ( s )    pc  ic  (32) u
 sqb  =  3 1 1  u spb 
sC pdc  s   u sqc 
 
2 3
 3 1 1  u spc 
A low pass filter (LPF) of cut-off frequency of 10 Hz, is used
in the feedback path of the DC voltage controller. The A feed-forward term (φpv) is added to include the effect of PV
feedback gain is given as, system, which is calculated as,
c 2P
H(s)  (33) pv (k)= pva (39)
s  c 3Vtpa
By considering values given in Appendix, a Bode diagram of This PV feed-forward term is responsible for reducing the
the open loop transfer function  G f (s) H(s)  , is plotted. By burden on the PI controller. The weight estimation at each
instant r, is performed through a modified DNLMS concept for
analyzing the Bode plot, shown in Fig. 4(b), one can comment
nonlinear load current of phase ‘a’, which is calculated as [26],
on the stability of the system. The gain margin (GM) and phase
margin (PM) are observed as 76.4dB and 52.720, respectively. û (r)eˆ spa (r)
pa (r+1) = pa (r)  μ' spa (40)
When the PM is in the range of 300 to 600, a good trade-off û spa (r)  b
2
between the stability of the system and the speed of the
5
Where, φpa (r) is the phase ‘a’ estimation of active weight of at of 0.01. This error follows a certain form within the hysteresis
rth instant. A bias factor (b) having a very small value, is band that results in tracking of actual current according to the
considered to avoid its denominator to go to zero. μ' is the step- behavior of the reference one through a controlled switching
size, the prediction error (ȇspa(r)) with decorrelatin operator is pattern of the VSC. As, i*sa , i*sb , i*sc are harmonic free sinusoidal
calculated as, eˆ spa (r) = î La (r)  uˆ spa (r)×pa (r) (41) signals synchronized with the grid voltages, the grid currents
follow the pattern, thereby achieving a current control process
îLa (r) = i La (r)  α i La (r  1) (42) of VSC.
û spa (r) = u spa (r)  α u spa (r  1) (43) The Bode plot of the control system is drawn by considering
where, iLa(r) is the phase ‘a’ load current and uspa(r) is the in the weight adaptation law (40) for some sampling instants. The
phase unit template. The step size of the weight estimation law weight (φpa(r)) depends on the step size ( μ' ) (40), the step size
( μ' ) is described as, depends on the squared error between consecutive weights,
 delta sequence (δ(r)) and the forgetting factor (θ). Hence, step
eˆspa 
μ'  min   (r ),μ  (44) size is updated after each sampling instant. This value adjusts
 uˆspa  itself according to the knowledge of present and previous
 2 
weights. For analysis point of view, the unit templates (u(r))
Similarly, estimation of active weights for other phases at rth
instant, are determined by following equations given as, are considered as 1. For the Bode plot, (40) to (43) are solved
to get the transfer function as,
û (r)eˆ (r)
pb (r+1) = pb (r)  μ' spb spb (45) îLa (r)  i La (r) (1  αz 1 ) (63)
û spb (r)  b
2
û spa (r) = u spa (r)(1  α z 1 ) (64)
eˆ spb (r) = î Lb (r)  uˆ spb (r)× pb (r) (46)
êspa (r)  (1   z )(i La (r)  u spa (r)× pa (r))
1
(65)
îLb (r) = i Lb (r)  α i Lb (r  1) (47)
z pa (r) =  pa (r)  μ'eˆ spa (r)
û spb (r) = u spb (r)  α u spb (r  1) (48)
z pa (r) = pa (r)  μ'(1  αz 1 )(i La (r)  u spa (r)× pa (r)) (66)
û spc (r)eˆ spc (r)
pc (r+1) =  pc (r)  μ' (49) pa (r) zμ'  μ'
û spc (r)  b 
2
i La (r) z 2  z (μ'  1)  μ'
eˆ spc (r) = î Lc (r)  uˆ spc (r)×pc (r) (50) Fig. 5 shows a Bode plot of DNLMS based control algorithm
îLc (r) = i Lc (r)  α i Lc (r  1) (51) with different step size ( μ' = 0.5, 0.35, 0.33, 0.3, 0.2, 0.1, 0.05).
û spc (r) = u spc (r)  α u spc (r  1) (52) It is observed that, for step size in the range (0, 1), the system is
At each rth instant, the phase ‘a’ reactive weight component stable. Fig. 6 shows the pole/zero plot of the transfer function
(φqa (r)) for load current is estimated using a modified DNLMS (66) for different step size ( μ' = 0.5,0.3,0.2,0.05 ). It is
algorithm as, observed that poles and zeros are inside the unit circle, hence,
û (r)eˆ sqa (r) the system is stable. For large step size ( μ' = 0.5), the right-side
qa (r+1) = qa (r)  μ' sqa (53)
pole is with smaller magnitude and for small step size ( μ' =
û sqa (r)  b
2
0.05), the pole is with higher magnitude (nearer to the unit
eˆ sqa (r) = î La (r)  uˆ sqa (r)×qa (r) (54) circle). Thus, for a large step size, the dynamic response is
û sqa (r) = u sqa (r)  α u sqa (r  1) (55) faster and for a small step size, the steady state response is
Similarly, reactive weight components of other two phases (φqb better. In this algorithm, the step size is initially (when error is
(r), φqc (r)) are calculated. Deducting φpv from the addition of high) a large value for fast dynamics and then reduces to a
φavp and φcp, results in resultant active weight component (φap) small value for accurate weight estimation. Here, μ' = 0.5 is
as, ap = avp  cp   pv (56) chosen to show its response with a large value and μ' = 0.05 is
The equivalent active weight of the load (φavp) is obtained as; chosen to show the response with a small value.
1
avp =  pa  pb  pc  (57) IV. VALIDATION THROUGH CONTROL IMPLEMENTATION
3
The load reactive weight (φrq) is calculated as; The MATLAB/ Simulink toolbox is used for simulating the
rq = cq  avq (58) grid interfaced 50kW PV system with a control logic based on
modified DNLMS algorithm. System behaviour is tested in the
1
avq =
3
qa  qb  qc  (59) laboratory after developing a prototype. Both steady state and
dynamic performances at nonlinear load in PFC mode, have
The reference grid currents ( i*sa , i*sb , i*sc ) are resulted from the been tested. The PV array is realized with the help of the PV
two components (active and reactive). φap and φrq are simulator (AMETEK ETS600x17DPV). Hall-Effect voltage
responsible for determining active (i*spa, i*spb, i*spc) and reactive and current sensors (LV-25P and LA-55P) are taken for
reference grid currents (i*sqa, i*sqb, i*sqc) calculated as, sensing purpose. A DSP (dSPACE-1202 Micro Lab Box, real
i*spa = ap × u spa ;i*spb = ap × u spb ;i*spc = ap × u spc (60) time controller), is used to implement the DNLMS logic based
current control technique. For capturing the behaviour of the
i*sqa =  rq × u sqa ;i*sqb =  rq × u sqb ;i*sqc =  rq × u sqc (61) grid integrated system at steady state and dynamic conditions,
*
i =i
sa
*
spa  i ;i = i
*
sqa
*
sb
*
spb  i ;i = i
*
sqb
*
sc
*
spc i *
sqc (62) power analyser (Fluke 43B) and four channel digital storage
* * * oscilloscope (DSO), are used. Detailed list of system
The error between i , i , i and sensed grid currents
sa sb sc parameters considered for simulation and the prototype, are
( isa , isb , isc ) is given to hysteresis controller operating at width depicted in Appendix.
6
û spa (r)eˆ spa (r)

φ pa (r+1) = φ pa (r)  μ'
û spa (r)  b
2
eˆ spa (r) = î La (r)  uˆ spa (r)×φ pa (r)
î La (r) = i La (r)  α i La (r  1)
û spa (r) = u spa (r)  α u spa (r  1)
i *spa = φ ap × u spa i *sa
i *sb
û sqa (r)eˆ sqa (r) i *sc

φqa (r+1) = φ qa (r)  μ'
û sqa (r)  b
2
eˆ sqa (r) = î La (r)  uˆ sqa (r)×φ qa (r)
û sqa (r) = u sqa (r)  α u sqa (r  1)
î La (r) = i La (r)  α i La (r  1)
i *sqa = φ aq × u sqa
Fig. 3. Modified DNLMS based control structure
Magnitude (dB)
G c ( s)
Γ pc
1
CpdcS
1
Phase (deg)
Γ pc G p ( s)
S
ωc
S + ωc
(a) (b)
Fig. 4 DC link voltage Regulation (a) block diagram (b) Bode plot for stability analysis
on performance under steady state and dynamic conditions
between these algorithms, are summarized in Table I. LMF
algorithm exhibits fast tracking as compared to LMS and
NLMS algorithms but it lacks in achieving accuracy in
estimated weight component. NLMS algorithm provides
accurate results under both the conditions but it fails to give
fast tracking. Modified DNLMS exhibits accurate estimation
with fast convergence and almost zero oscillation.
Fig.5 Bode plot of modified DNLMS based control
A. Comparative Analysis between Modified DNLMS with

Conventional Algorithms
A comparative study of performance of modified DNLMS
algorithm with least mean square (LMS), least mean fourth (a) μ' = 0.5 (b) μ' = 0.3
(LMF) and normalized LMS (NLMS) algorithms, is presented
to prove the superiority of given control logic. Fig. 7 shows the
simulated results describing the performance through the
estimated average active weight component (φavp) at steady
state and unbalanced load. Phase ‘a’ load is disconnected at
0.5s and again reconnected at 0.7s. Less oscillation in the
estimated φavp is observed when the system is operated with
modified DNLMS algorithm as compared to the LMS, LMF
and NLMS algorithms under steady state and unbalanced load.
The LMS, LMF and NLMS algorithms use a fixed step size,
whereas the step size of modified DNLMS algorithm adjusts (c) μ' = 0.2 (d) μ' = 0.05
itself for achieving fast convergence. A comparative analysis Fig.6 Pole/zero plot of modified DNLMS based control with different step size.
7
TABLE I The performance of the control scheme is corroborated through

COMPARATIVE STUDY WITH CONVENTIONAL ALGORITHMS the estimated weights (φcp, φpv, φavp, φap) as illustrated in Fig.
Control Steady State Dynamic condition 8(c). No variation is observed in φpv as it is supplying a
Technique Accuracy Oscillation Tracking Oscillation Deviation constant power with perceptible variation in φavp and φap.
LMS Poor High Moderate More High
Performance under transition in solar irradiation from
LMF Poor High Fast More High
1000W/m2 to 500W/m2 is observed (Figs. 9 (a-c)) as, no
NLMS good Less Slow Less Less
Modified Better Almost zero Fast Almost zero Less
variation in vabcs and load current (iLa), dip in PV current (Ipva)
DNLMS and power (Ppva) reflects in VSC currents (ipvs), grid currents
(iabcs) and power (Ps) with constant PL. Moreover, the
performance of the control scheme is validated through φcp,
φpv, φavp, φap as denoted in Fig. 9(c). Moreover, the transition in
solar irradiation, is proved through φpv. Under both dynamic
conditions, the PV system supplies active power to the grid
after meeting the demand by local loads, grid side power factor
is maintained at near to unity, Vpdca is maintained at desired
value. The estimated Vpva becomes the reference DC link
voltage, thereby satisfying the single stage operation of the grid
integrated PV system.
C. Steady State Performance through Test Results
A photograph of the system prototype developed in the
laboratory, is depicted in Fig. 10. Different components of the
Fig. 7 Comparison of estimated φavp using modified DNLMS algorithm with
conventional LMS, LMF and NLMS algorithms system, are highlighted. The steady state behaviour of the
prototype is presented in Figs. 11(a-i), recorded by using the
B. System Investigation through Simulation Results power analyser presenting, vsab (POI voltage) with (a) grid
The grid tied 50kW PV system is simulated using current (isc), (b) load current (iLc), (c) VSC current (ic_pvs). Fig.
MATLAB Simpower systems blocks at nonlinear load under 11(d) demonstrates the harmonic spectrum of grid voltage
dynamic conditions. Other system parameters are given in (THD of 1.8%). The harmonic spectrum of grid current is
Appendix. Unbalanced loading and variation in solar depicted in Fig. 11(e) showing THD of 3.4 %, which satisfies
irradiation (G) cases, are studied for performance verification the IEEE-519 standard [29] whereas the THD of load current is
as shown in Fig. 8 and Fig. 9. Phase ‘a’ load is disconnected at 27.4 % (Fig. 11(f)), thereby proving its working under steady
0.74 s. Under load unbalance, performance of the system is state. The grid power (Ps), load power (PL) and PV-VSC power
observed (Figs. 8(a-b)) as, phase ‘a’ current (iLa) becomes zero (Ppvs) are demonstrated in Figs. 11(g-i), justifying the active
with current flowing in other two phases, no variation is power flow to the grid and the load from the PV array. From
observed in vabcs, sinusoidal and balanced grid currents (iabcs) these results, it is clear that, out of 5.3kW power from the solar
are obtained by tracking the sinusoidal reference currents PV array, 5.04kW is available at the AC side of the 3-phase
generated from the control law, currents from the PV system converter, 1.95kW is fed to the nonlinear load and 3.04 kW is
flows to the grid and the load, hence feeding power to the load fed to the grid. The total power loss of 310W is observed in
(PL) as well as the grid (Ps) maintaining Vpdca at desired value. different parts of the system.
(a) (b) (c)

Fig. 8 Simulation results under unbalanced load
G(W/m2)
φ
φ avp
cp
pv
φ
ap
φ
(a) (b) (c)
Fig. 9 Simulation results under variable solar insolation
of AC side of VSC (ipvs), isa, control signals the weights (ȋLa,

ȗspa, ȇspa, φpa, φcp, φpv, φavp, φap). These results demonstrate that
the system behaves well under both the dynamic conditions by
achieving the steady state at fast convergence rate, thereby
improving the overall performance of the grid integrated PV
system. Moreover, 99.75% MPPT is achieved with the P&O
technique as presented in Fig. 14.
(a) (b) (c)

Fig. 10 Photograph of system prototype developed in the laboratory
D. Dynamic Performance Analysis through Test Results

A prototype is tested under two dynamic conditions
(unbalanced load and transition in solar irradiation (G)). Phase
‘a’ load is disconnected to examine the unbalanced load.
Behaviour of the system under this condition, is recorded using
DSO, demonstrated in Figs. 12(a-f). The terminal voltage (Vtpa)
and unit template (uspa and uspb) from the grid line voltages (vsab (d) (e) (f)
and vsbc), are verified from Figs. 12(a-b). The behaviour of
voltage at DC link (Vpdca), phase ‘a’ currents in VSC (ia_pvs),
load (iLa) and grid (isa) under unbalanced load and reconnection
of balanced load, is depicted in Fig. 12(c). Vpdca, phase ‘b’
currents in VSC (ib_pvs), load (iLb) and grid (isb) under
unbalanced load and reconnection of balanced load, are
presented in Fig. 12(d). All the intermediate signals ȋLa, ȗspa,
ȇspa, φpa, φcp, φpv, φavp, φap are demonstrated in Figs. 12(e-f).
Balanced increased grid currents are achieved during
disconnection of phase ‘a’ load. Fig. 13 describes the (g) (h) (i)
performance of the system prototype under the transition in Fig. 11 Performance validation through steady state analysis describing (a-c)
solar irradiation from 700W/m2 to 1000 W/m2 and again from grid voltage (vsab) with (a) grid current (isc) (b) Load current (iLc) (c) PV current
(ic_pvs) (d) harmonic spectrum of grid voltage (1.8%) (e) harmonic spectrum of
1000W/m2 to 700W/m2. Effect of this transition is grid current (3.4%) (f) harmonic spectrum of load current (27.4%) (g) Grid
demonstrated through PV voltage (Vpva), current (ipva), current power (Ps) (h) Load power (PL) (i) VSC power.
(a) (b) (c)
(d) (e) (f)

Fig. 12 Dynamic performance under unbalanced load (a) Line voltages (vsab, vsbc) and phase voltages (vsa, vsb) (b) Phase voltage (vsc), terminal voltage (Vtpa) and unit
templates (uspa, uspb) (c) DC link voltage (vpdca), ia_pvs, iLa, isa (d) vpdca, ib_pvs, iLb, isb (e) (îLa, ûspa, êspa, φpa) (f) (φcp, φpv, φavp, φap).
(a) (b)
(c) (d)
Fig. 13 Dynamic performance under transition in G describing (a) PV voltage (Vpva), current (ipva), ipvs, isa for variation from 700W/m2 to 1000 W/m2 (b) Vpva, ipva, ipvs,
isa for variation from 1000W/m2 to 700 W/m2 (c) Intermediate signals (φcp, φpv, φavp, φap) for variation from 700W/m2 to 1000 W/m2 (d) (φcp, φpv, φavp, φap) for
variation from 1000W/m2 to 700 W/m2 .
V. CONCLUSION
V-I Curve The proposed modified DNLMS algorithm at nonlinear
Operating load under various dynamic conditions has been demonstrated
Point though the simulation results. The validation of proposed
modified DNLMS based control structure has been performed
on a developed prototype. The prototype has been tested under
rve
possible dynamic conditions to prove its efficiency. A modified

Cu
DNLMS based control has been adopted for active load weight
P-V
estimation. P&O MPPT algorithm has been used for maximum

MPPT 99.75% solar power extraction through estimation of reference DC link
voltage that has to be tracked by the actual value. The need of
PLL has been eliminated through the weight estimation with a
guarantee of synchronization of the grid currents with the grid
Fig. 14 MPPT tracking of solar PV system at 1000 W/m2.
10
voltages. The grid current THD has been observed as 3.4% Control Algorithm,” 7th India Inter. Conf. on Power Electronics
fulfilling the requirement according to the IEEE-519 standard (IICPE), Nov. 2016.
[29]. Active power flow has been observed from the test [13] S. Pradhan, I. Hussain, B. Singh, B. K. Panigrahi. "Modified VSS-LMS
based adaptive control for improving the performance of a single-stage
results. A satisfying performance has been recorded under PV-integrated grid system", IET Science, Measurement & Technology,
dynamic conditions by achieving the steady state condition at vol. 11, no.4, pp. 388-399, Jan. 2017
fast convergence rate with the proposed modified DNLMS [14] R. Garg, B. Singh, D. T. Sahani and C. Jain, “Dual-Tree Complex
control algorithm. Unity power factor has been achieved at the Wavelet Transform Based Control Algorithm for Power Quality
grid side even when the load is nonlinear in nature. Improvement in a Distribution System,” IEEE Trans. on Ind. Electr.,
vol. 64, no.1, pp. 764-772, Jan 2017.
[15] F. D. Freijedo, J. Doval-Gandoy, O. López, and E. Acha, “Tuning of
APPENDIX phase-locked loops for power converters under distorted utility
System Parameters Considered for Simulation: PV array conditions,” IEEE Trans. Ind. Appl., vol. 45, no. 6, pp. 2039–2047,
Nov./Dec. 2009.
voltage (Vpva) = 700 V; array power (Ppva) = 50 kW; ripple
[16] A. Luna, J. Rocabert, J. I. Candela, J. R. Hermoso, R. Teodorescu, F.
filter (Rfr = 5 Ω and Cfr = 10 μF); DC link Capacitance Blaabjerg and P. Rodríguez,“Grid Voltage Synchronization for
(Cpdc)=5000μF; DC link Voltage (Vpdca) = 700 V; interfacing Distributed Generation Systems Under Grid Fault Conditions,” IEEE
inductor (Lfi) = 3.8 mH; grid line voltage (Vsab) = 415 V (rms); Trans. on Ind. Appl., vol. 51, no. 4, pp. 3414-3425, July 2015.
DC PI controller gain (proportional Γpc = 0.3 and integral Γic = [17] P. R. Sanchez, X. T. Garc´ıa, A. P. Torres, and V. Feliu, “Fundamental
0.01); sampling time (Ts) = 10 μs. Positive and Negative-Sequence Estimator for Grid Synchronization
Under Highly Disturbed Operating Conditions,” IEEE Trans. on Power
System Parameters Considered for Laboratory Prototype: Electr., vol. 28, no. 8, pp. 3733-3746, August 2013.
PV simulator array of voltage (Vpva) = 340 V; current (Ipva) = [18] S. Vazquez, J. A. Sanchez, M. R. Reyes, J. I. Leon, and J. M. Carrasco,
15.755 A and power (Ppva) = 5.36 kW; interfacing inductor “Adaptive Vectorial Filter for Grid Synchronization of Power
(Lfi) = 3.4 mH; DC link Voltage (Vpdca) = 340 V; DC link Converters Under Unbalanced and/or Distorted Grid Conditions,” IEEE
Trans. on Ind. Electr., vol. 61, no. 3, pp. 1355-1367, March 2014.
Capacitance (Cpdc)=2200μF; ripple filter (Rfr = 5 Ω, Cfr = 10
[19] X. Quan, X. Dou, Z. Wu, M. Hu, and F. Chen, “A Concise Discrete
μF); grid voltage (Vsab) = 227 V (rms); DC PI controller gains Adaptive Filter for Frequency Estimation Under Distorted Three-Phase
(proportional Γpc = 0.3 and integral Γic = 0.005); sampling time Voltage,” IEEE Trans. on Power Electr., vol. 32, no. 12, pp. 9400-9412,
(Ts) = 30 μs. December 2017.
[20] M. Badoni, A. Singh, and B. Singh, “Comparative Performance of
Wiener Filter and Adaptive Least Mean Square-Based Control for Power
ACKNOWLEDGMENT Quality Improvement,” IEEE Trans. on Ind. Electr., vol. 63, no. 5, pp.
The authors are extremely grateful to Department of 3028-3037, May 2016.
Science and Technology (DST), Govt. of India (RP02979) and [21] J. Kukkola and M. Hinkkanen, “State Observer for Grid-Voltage
Sensorless Control of a Converter Equipped With an LCL Filter: Direct
J.C. Bose fellowship (RP03128) for their financial support. Discrete-Time Design” IEEE Trans. on Ind. Appl., vol. 52, no. 4, pp.
3133-3145, July 2016.
REFERENCES [22] Y. Park, H. S. Kim, and S. K. Sul, “Frequency-Adaptive Observer to
Extract AC-Coupled Signals for Grid Synchronization,” IEEE Trans. on
[1] M. G. Villalva, J. R. Gazoli and E. R. Filho, “Modeling and circuit-
Ind. Appl., vol. 53, no. 1, pp. 273-282, Jan. 2017.
based simulation of photovoltaic arrays,” in Proc. Brazilian Power
Electronics Conference, 2009, pp. 1244-1254. [23] S. M. Kuo and D. R. Morgan, Active Noise Control Systems, Algorithms
and DSP Implementations. New York: Wiley, 1996.
[2] T. Esram and P. L. Chapman, “Comparison of Photovoltaic Array
Maximum Power Point Tracking Techniques,” IEEE Trans. on Energy [24] J. Zhang, H.M. Tai, “Adaptive noise cancellation algorithm for speech
Convers., vol. 22, no.2, pp. 439-449, June 2007. processing,” IECON 2007 - 33rd Annu. Conf. IEEE Ind. Electron. Soc.
pp. 2489-2492, Nov. 2007.
[3] W. Libo, Z. Zhengming and L. Jianzheng, “A single-stage three-phase
grid-connected photovoltaic system with modified MPPT method and [25] S. Haykin, Adaptive Filter Theory. 4th ed. Upper Saddle River, NJ, USA:
reactive power compensation,” IEEE Trans. on Energy Convers., vol. Prentice-Hall, 2002.
22,no. 4, pp. 881–886, Dec. 2007. [26] S. Zhang, J. Zhang, and H. Han, “Robust Variable Step-Size
[4] C. Jain, and B. Singh, “Single-phase Single-stage Multifunctional Grid Decorrelation Normalized Least-Mean-Square Algorithm and its
Interfaced Solar Photo-Voltaic System under Abnormal Grid Application to Acoustic Echo Cancellation,” IEEE/ACM Trans. on
Conditions,” IET Generation, Transmission & Distribution, vol. 9, no. Audio, Speech, and Language Processing, 2016.
10, pp. 886-894, 2015. [27] L. Rey Vega, H. Rey, J. Benesty and S. Tressens, “A New Robust
[5] Y.Yang, F. Blaabjerg and Z. Zou, “Benchmarking of Grid Fault Modes Variable Step-Size NLMS Algorithm,” IEEE Trans. on Signal
in Single-Phase Grid-Connected Photovoltaic Systems,” IEEE Trans. on Processing, vol. 56, no. 5, pp. 1878-1893, May 2008.
Ind. Appl., vol. 49, no. 5, pp. 2167-2176, Sept. 2013. [28] H. Akagi, E. H. Watanabe, and M. Aredes, Instantaneous Power Theory
[6] IEEE Standard for Interconnecting Distributed Resources With Electric and Applications to Power Conditioning. New York, NY, USA: Wiley
Power Systems, IEEE Std. 1547-2003, 2003. IEEE Press, 2007.
[7] The Grid Code: Revision 31, National Grid Electricity [29] IEEE recommended practice and requirements for harmonic control in
Transmission,Warwick, U.K., Oct. 2008, no. 3. electrical power system, 519-1992.
[8] B. Singh, A. Chandra, and K. Al-Haddad, Power Quality: Problems and
Mitigation Techniques. John Wiley & Sons, 2015. Subarni Pradhan (S’15) was born in Odisha, India, in
[9] R. Agarwal, I. Hussain, and B. Singh, “LMF Based Control Algorithm 1989. She received B. Tech. degree in electrical
for Single Stage Three-Phase Grid Integrated Solar PV System,” IEEE engineering from Siksha ‘O’ Anushandhan University,
Trans. on Sust. Energy, vol. 7, no. 4, pp. 1379-1387, 2016. Bhubaneswar, India, in 2011 and the M. Tech. degree
in power control and drives from National Institute of
[10] L. A. Serpa, S. Ponnaluri, P. M. Barbosa, and J. W. Kolar, “A modified Technology, Rourkela, India, in 2013. She is currently
direct power control strategy allowing the connection of three-phase
working toward the Ph.D. degree in Department of
inverters to the grid through LCL filters,” IEEE Trans. on Ind. Appl.,
Electrical Engineering from Indian Institute of
vol. 43, no. 5, pp. 1388–1400, Sep./Oct. 2007.
Technology, Delhi. Her areas of research interests
[11] C. T. Lee, R. P. Jiang, and P. T. Cheng,“A Grid Synchronization Method include power electronics, renewable energy, micro-grid, power quality,
for Droop-Controlled Distributed Energy Resource Converters,” IEEE electrical machines and drives, and application of adaptive and robust control
Trans. on Ind. Appl., vol. 49, no. 2, pp. 954-962, March 2013. techniques in power system.
[12] S. Pradhan, I. Hussain, B. Singh, and B. K. Panigrahi, “Performance
Improvement of Grid Integrated Solar PV System using DNLMS
11
Ikhlaq Hussain (M’14) was born in Doda, Jammu and Techniques published by John Wiley & Sons Ltd. 2015. His areas of interest
Kashmir, India, in 1986. He received his B. E. include solar PV grid interface systems, microgrids, power quality monitoring
(Electrical) from University of Jammu, Jammu, India, in and mitigation, solar PV water pumping systems, improved power quality
2009, M. Tech. (Gold Medalist) in Electrical Power AC-DC converters, power electronics, electrical machines, drives, flexible
System Management from the Jamia Millia Islamia (A alternating transmission systems, and high voltage direct current systems.
Central University), New Delhi, India, in 2012 and Prof. Singh is a Fellow of the Indian National Academy of Engineering
Ph.D. in the Department of Electrical Engineering, (FNAE), The Indian National Science Academy (FNA), The National
Indian Institute of Technology Delhi, New Delhi, India, Academy of Science, India (FNASc), The Indian Academy of Sciences, India
(FASc), The World Academy of Sciences (FTWAS), Institute of Electrical
in 2018. From September 2012 to December 2012, he
and Electronics Engineers (FIEEE), the Institute of Engineering and
was a Lecturer with the Department of Electrical
Technology (FIET), Institution of Engineers (India) (FIE), and Institution of
Engineering, National Institute of Technology Srinagar, India. On 19th April Electronics and Telecommunication Engineers (FIETE) and a Life Member of
2017, he joined as Assistant Professor in the Department of Electrical the Indian Society for Technical Education (ISTE), System Society of India
Engineering, Institute of Technology, University of Kashmir, Srinagar, India. (SSI), and National Institution of Quality and Reliability (NIQR).
His areas of research interests include power electronics, power quality, He has received Khosla Research Prize of University of Roorkee in the
custom power devices, renewable energy systems, power system management year 1991. He is recipient of JC Bose and Bimal K Bose awards of The
and microgrid. Institution of Electronics and Telecommunication Engineers (IETE) for his
Mr. Hussain was a recipient of the POSOCO power system award (PPSA) contribution in the field of Power Electronics. He is also a recipient of
from Power System Operation Corporation (POSOCO) Limited, India and Maharashtra State National Award of Indian Society for Technical Education
Foundation for Innovation and Technology Transfer (FITT) at Indian Institute (ISTE) in recognition of his outstanding research work in the area of Power
of Technology Delhi in 2017, Gandhian Young Innovations Award in 2018, Quality. He has received PES Delhi Chapter Outstanding Engineer Award for
Outstanding Faculty researcher in the field of energy for 2017-2018. He has the year 2006. Professor Singh has received Khosla National Research Award
participated as young scientist in 3rd BRICS young scientist conclave from of IIT Roorkee in the year 2013. He is a recipient of Shri Om Prakash Bhasin
25th – 29th June 2018 at Durban, South Africa by DST, Govt. of India. He was Award-2014 in the field of Engineering including Energy & Aerospace.
recipient of IEEE INDICON Best Paper Award in 2015 and the IEEE UPCON Professor Singh has received IEEE PES Nari Hingorani Custom Power
Best Paper Award in 2016. He has published more than 100 papers in reputed Award-2017. He is also a recipient of “Faculty Research Award as a Most
journals and conferences. He is a contributing reviewer of a number of Outstanding Researcher” in the field of Engineering-2018 of Careers-360,
IEEE/IET/Elsevier journals and conferences India.
He has been the General Chair of the 2006 IEEE International Conference
Bhim Singh (SM’99, F’10) was born in Rahamapur, on Power Electronics, Drives and Energy Systems (PEDES’2006), General
Bijnor (UP), India, in 1956. He has received his B.E. Co-Chair of the 2010 IEEE International Conference on Power Electronics,
(Electrical) from the University of Roorkee, India, in Drives and Energy Systems (PEDES’2010), General Co-Chair of the 2015
1977 and his M.Tech. (Power Apparatus & Systems) IEEE International Conference (INDICON’2015), General Co-Chair of 2016
and Ph.D. from the Indian Institute of Technology IEEE International Conference (ICPS’2016) held in New Delhi, General Co-
Delhi, India, in 1979 and 1983, respectively. In 1983, Chair of 2017 National Power Electronics Conference (NPEC) held in Pune.
he joined the Department of Electrical Engineering,
University of Roorkee (Now IIT Roorkee), as a Bijaya Ketan Panigrahi (SM’06) received the Ph.D.
Lecturer. He became a Reader there in 1988. In degree in power systems from Sambalpur University,
December 1990, he joined the Department of Sambalpur, India, in 2004. Since 2005, he has been an
Electrical Engineering, IIT Delhi, India, as an Associate Professor with the Department of Electrical
Assistant Professor, where he has become an Associate Professor in 1994 and Engineering, Indian Institute of Technology (IIT)
a Professor in 1997. He has been ABB Chair Professor from September 2007 Delhi, New Delhi, India, where he has become a
to September 2012. He has also been CEA Chair Professor from October 2012 Professor in 2017. Prior to joining IIT, he was a
to September 2017. He has been Head of the Department of Electrical Lecturer at the University College of Engineering,
Engineering at IIT Delhi from July 2014 to August 2016. Since, August 2016, Sambalpur, for 13 years. His research interests include
he is the Dean, Academics at IIT Delhi. He is JC Bose Fellow of DST, intelligent control of flexible ac transmission system
Government of India since December 2015. devices, digital signal processing, power quality assessment, and application
Prof. Singh has guided 69 Ph.D. dissertations, and 167 of soft computing techniques to power system planning, operation, and
M.E./M.Tech./M.S.(R) theses. He has been filed 33 patents. He has executed control.
more than eighty sponsored and consultancy projects. He has co-authored a
text book on power quality: Power Quality Problems and Mitigation
12

10 1109@tia 2018 2863652

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

10 1109@tia 2018 2863652

Hochgeladen von

Copyright:

Verfügbare Formate

This article has been accepted for publication in a future issue of this journal, but has not been

Performance Improvement of Grid Integrated Solar

I. INTRODUCTION amplitude of signals ensuring harmonics elimination and

II. MODIFIED DNLMS ALGORITHM α

ˆ   T (r+1)u(r) Let the error between φopt and φ is denoted as ϑ. Hence,

where, Rˆ  E uˆ (r)uˆ T (r) and V  lim E  (r) T (r) .  2f , r 1   2f , r  (1   )eˆ 2f , r

Hence, the steady state mean square deviation (MSD) if  2f , r 1 ( b2   2f , r 1 )  

2 0 proves the  (r)   (0)

(27) MATLAB Simulink model under possible dynamic conditions

Fig.2 Schematic diagram of 3- phase grid integrated PV system

û spa (r)eˆ spa (r)

û sqa (r)eˆ sqa (r) i *sc

Fig. 3. Modified DNLMS based control structure

Fig.5 Bode plot of modified DNLMS based control

A. Comparative Analysis between Modified DNLMS with

TABLE I The performance of the control scheme is corroborated through

(a) (b) (c)

of AC side of VSC (ipvs), isa, control signals the weights (ȋLa,

(a) (b) (c)

D. Dynamic Performance Analysis through Test Results

(a) (b) (c)

(d) (e) (f)

possible dynamic conditions to prove its efficiency. A modified

estimation. P&O MPPT algorithm has been used for maximum

Das könnte Ihnen auch gefallen