Electrical Power and Energy Systems 101 (2018) 311–322

Contents lists available at ScienceDirect

Electrical Power and Energy Systems

journal homepage: www.elsevier.com/locate/ijepes

Grid-connected converter active and reactive power production T

maximization with respect to current limitations during grid faults
⁎
Ivan Todorović , Stevan Grabić, Zoran Ivanović
Department of Power, Electronic and Telecommunication Engineering, Faculty of Technical Sciences, Trg Dositeja Obradovica 6, Novi Sad, Serbia

A R T I C LE I N FO A B S T R A C T

Keywords: During the grid disturbances, particularly voltage sags, if the grid-connected converter’s (GCC’s) power refer-
Grid connected converter ences are kept at the pre-fault level, the excessive currents would flow. This could force the inverter dis-
Current limitation connection, which is in conflict with the grid codes and the tendency to keep the GCC system connected to the
Current references grid as long as possible. Keeping the system connected for a prolonged periods of time offers a possibility of a
Power maximization
more effective grid support, better exploitation of available energy resources and generally more reliable power
supply. The aim of this paper is to address the utilization of active and reactive power production capacities
during unbalanced voltage sags with respect to the current limits. First proposed algorithm gives the grid op-
erator the opportunity to choose whether active or reactive power production is prioritized during voltage
disturbance, with the curtailment done only to the extent that the current limits are not surpassed. Second
approach allows for the power factor to stay the same before and during the sag. Selected hardware-in-the-loop
experiments are presented to validate the developed theoretical background and implemented algorithm.

1. Introduction
Grid-connected converters are affected by grid disturbances, parti-
cularly when connected to the distribution power grids and microgrids,
as they tend to be less stiff and prone to irregularities. With further
proliferation of distributed energy sources based on GCCs, the asso-
ciated grid interconnection problems will be amplified if not addressed
properly [1]. On the other hand, the grid codes demand ever increasing
grid support from these units, even in cases of pronounced grid faults
[2].
Voltage sags are particularly important disturbances to be ad-
dressed, as they are the most frequent and can have the most adverse
effects on the GCCs’ operation. The main reasons for the unbalanced
voltages occurrences are short-circuit faults, connection of the big loads
and unbalanced loads [3]. The first two often result in a transient,
short-lasting, still more serious dips. The load unbalances, intrinsic
characteristic of the distribution networks, usually result in less pro-
nounced, but significantly longer-lasting dips. In any case, the longer
the converter can safely stay connected to the grid the better. If the
voltage sag is serious, the grid support realized by the production of
reactive power could have strong palliative influence during and post-
fault. On the other hand, if the voltage dips last longer, inadequate
control strategies could lead to poor utilization of available energy
supplied by the primary energy sources, to load shedding etc.
Furthermore, keeping the same power references as before the fault will
cause excessive currents that will lead to triggering of the unit’s pro-
tection and disconnection from the grid - trait especially pronounced in
cases of asymmetrical voltage dips. Therefore, flexible GCC control al-
gorithms that secure safe and maximized power production are of great
importance for future power systems.
A number of control strategies with flexible current and power
production have been developed. Still, safe production, with respect to
the current limitations, imposed by the converter switching elements’
characteristics, has seldom been discussed.
Until recently, the grid codes did not require photovoltaic based
distribution generation units to offer grid support and, consequently,
generate reactive power during voltage sags, but were expected to
produce active power. As a consequence, safe power production stra-
tegies were proposed, but only with respect to the active power [4–7].
Some grid codes now demand a grid support feature from PV applica-
tions and thus the previous references become obsolete [8]. Unlike in
PV applications, in wind farm and static compensator applications, only
reactive power is of interest during deep voltage sags [9–11]. The ap-
proach proposed in [12,13] addresses not only current amplitudes, but
also comprehensively examines safe power production possibility con-
sidering the AC and DC side voltage limits. Furthermore, load balancing
feature is introduced. Still, this algorithm also takes into account only
the reactive power. Again, dedicated solutions could become outdated

⁎
Corresponding author.
E-mail address: ivan.todorovic@uns.ac.rs (I. Todorović).

https://doi.org/10.1016/j.ijepes.2018.03.037
Received 17 September 2017; Received in revised form 29 January 2018; Accepted 26 March 2018
0142-0615/ © 2018 Elsevier Ltd. All rights reserved.
I. Todorović et al.
and would not suffice if the grid codes are changed [14]. Having the
possibility of fluid change in power production at disposal would be in
the accordance to the tendency to make the grid more flexible and
responsive to different fault scenarios. Also, the fact that one-sided
approach was assumed limits the number of applications in which the
proposed solutions can be applied.
The paper [15] does offer the simultaneous active and reactive
power production, but it does not account for the negative sequence
current components. Thus, this oversimplified solution results in non-
sinusoidal currents and uncontrolled oscillations in active and reactive
power. The solution proposed in [14] gives general approach to the
problem of the power references calculation in the stationary reference
frame. This should be the base for the flexible peak current limitation
solution for different power curtailment scenarios. Still, final algorithm
output manages only average components of active and reactive
powers. Oscillating components were not taken into account. Also, in
order to function properly, the proposed algorithm necessitates usage of
the sag detector. The sag detector’s dynamics certainly has a degrading
effect on transient processes. Furthermore, for the reactive power re-
ference calculation, active power reference, that is the output of the
higher-layer control loop, is used (and vice versa). It could happen that
this active power reference alone would result in currents that are
above the limits. This situation was not addressed and thus the ex-
pressions for the references would become unsolvable. A similar remark
can be made for the approaches given in [16,17]. The appropriate
curtailment of the power reference that comes from the higher-layer
level loop was not considered. The results supplied in [18] suggest that
the current capacities are not fully used as currents’ amplitudes are well
under the limit. Also, the currents have highly non-sinusoidal wave-
forms.
This paper proposes contemporary calculation frame that yields
maximized converter utilization both in regular and irregular grid
states, taking into account different grid support strategies, currently
valid and beyond standards. The algorithm shown here offers two
paradigms with total of six possible scenarios for safe simultaneous
control of active and reactive power flow with certain distinctive fea-
tures. The first paradigm corresponds to the scheme of output power
maximization while giving production priority to the active power or
reactive power. The second paradigm corresponds to the scheme of
output power factor control, again with maximized total power output.
In both paradigms, active or reactive power oscillating components at
double the grid frequency are optionally minimized. Minimized active
power oscillations would lead to minimized oscillations in the DC link
voltage, while minimized reactive power oscillation could be of primary
interest because of AC side voltage stability. Next, the algorithm does
not require a sag detector. Thus, during algorithm execution, there is no
need for the code structure change. The transition from nominal con-
dition to the fault-ride-through regime progresses seamlessly, implying
the best possible transient behavior. Furthermore, the produced grid
currents have sinusoidal shape and are limited at exactly the set max-
imal value. This guarantees, on the one side, maximal power produc-
tion, and, on the other side, safe operation of the converter and pre-
vention of overcurrent tripping. The above stated features apply all
down to the complete voltage collapse. Finally, the algorithm was set
up to suit practical implementation on a dedicated microcontroller re-
garding structural and computational burden.
The paper documents all theoretical and implementation aspects of
the algorithm structure and is organized as follows. Section 2 describes
the underlying current control algorithm.
In Section 3, the current references calculation scheme that fulfills
respective control and functional objectives is introduced. Section 4
presents selected HIL experimental results obtained on the hardware-in-
the-loop experimental laboratory setup and provides discussion on
feasibility and effectiveness of the proposed control strategies. Finally,
Section 5 gives concluding remarks and proposes possibilities for future
research.
312
Electrical Power and Energy Systems 101 (2018) 311–322
2. Grid-connected converter control algorithm
The connection of renewable energy sources to the existing power
grid has become an interesting and challenging task for researchers in
recent years, especially when the disturbances are taken into account.
As a result, a number of inverter control strategies that ensure safe and
reliable power production even with the perturbation in the grid have
been developed. Most of the strategies are based on symmetrical com-
ponents extraction [19–21]. Those based on direct power control
[22,23] and sliding mode control [24] have also been examined and
successfully validated. A thorough overview of these control strategies
and other less common approaches can be found in [25–27].
Algorithms based on symmetrical sequences extraction employ ei-
ther the proportional-integral (PI) or the proportional-resonant (PR)
controllers. PI current regulators face the existence of the ripple at
double the grid frequency caused by inversely rotating current and
voltage components [28]. This poses fundamental constraint upon the
attainable control bandwidth. On the other hand, the ability of a PR
controller to control oscillating variables in their original form obviates
the need for filtering and simplifies the design of the loop dynamic
characteristics [29]. Still, it is less intuitive and harder to analyze the
alternating variables in stationary reference frame and the process of
references calculation is a more straightforward task in synchronous
reference frame.
As both aspects - control dynamics and handling the references - are
crucial for attaining control goal, the approach taken here was to
combine the control methods - PR regulators were used for the current
control, with the references calculated in the synchronous reference
frame. The outlook of the plant and control algorithm is illustrated in
Fig. 1. Measured three-phase grid currents, iabc, and grid voltages, uabc,
are first transformed to the two-phase stationary αβ system by using
Clarke transformation. The transformed voltages, uαβ, are fed to the-
sequence extractor and phase-locked-loop (PLL) unit that calculates
synchronous frame grid voltage components, udp, uqp, udn and uqn, es-
timated grid angle, θ, and frequency, ω. This stage is based on the dual
second order generalized integrator [30]. Active power reference, Pr, is
usually produced by the DC-link voltage controller. According to the
[31,32], the reactive power reference is generated by either grid vol-
tage control loop, can be specifically defined by the grid operator in
VAr or can be calculated using defined cosφ and available active power.
Power references, grid voltage sequences and angle are used in the
process of current references calculation. The transformation of positive
(p) and negative (n), direct (d) and quadrature (q) current reference
components, idpr, iqpr, idnr and iqnr, into the stationary reference frame is
done according to inverse Park transformation (1) [33].
r r r
⎡iα ⎤ = ⎡ cosθ − sinθ ⎤ ⎡idp ⎤ + ⎡ cos(−θ) − sin(−θ) ⎤ ⎡idn ⎤
⎢ iβ ⎥ ⎣ sinθ cosθ ⎦ ⎢iqp ⎥ ⎢ sin(−θ) cos(−θ) ⎥ ⎢iqn
r r r ⎥
⎣ ⎦ ⎣ ⎦ ⎣ ⎦⎣ ⎦ (1)
After the expression (1) is reorganized, the expressions for stationary
reference frame components, iαr and iβr, are obtained.
iαr = (idp
r r
+ idn r
)·cosθ−(iqp r
−iqn )·sinθ (2)
iβr = (idp
r r
−idn r
)·sinθ + (iqp r
+ iqn )·cosθ (3)
Finally, according to the current references, PR regulators control the
actual currents by providing corresponding inputs to the space-vector
modulator.
3. Current references calculation scheme
The grid current references can be calculated using the system of
equations that stem from the expression for the apparent power ex-
pressed in complex notation:
∗
S ̂ = P + iQ = (eiωt udq
p n p n ∗
+ e−iωt udq )·(e−iωt idq + eiωt Udq ) (4)
I. Todorović et al. Electrical Power and Energy Systems 101 (2018) 311–322

Fig. 1. Control algorithm outline.

with the superscript ‘∗’ noting complex conjugate. Eq. (4) can be de- shown in Table 1. Expressions (t1) are used when the oscillation in
composed in six components forming the system of Eqs. (5)–(10): active power is to be minimized and (t2) when oscillation in the re-
r r r r active power should be minimized.
P r = idp ·udp + iqp ·uqp + idn ·udn + iqn ·uqn (5) The depicted equations provide references that are not limited by
r r r r any means. Grid voltage sag combined with unaltered, pre-fault active
Qr = idp ·uqp−iqp ·udp + idn ·uqn + iqn ·udn (6)
and reactive power set values would lead to excessive current refer-
r
ΔP1 = iqn r
·uqn−iqp r
·udn−idn r
·uqp + iqn ·udp ences. The proposed implementation of the “Reference calculation and
(7)
limitation” block is based on the approach in which power components’
r
ΔP2 = idp r
·udn + iqp r
·uqn + idn r
·udp + iqn ·uqp (8) reference values are modified, at the same time attaining following
objectives:
r r r r
ΔQ1 = idp ·uqn−iqp ·udn + idn ·uqp−iqn ·udp (9)
- Phase currents are bound to maximum permissible values, Imax,
r r r r
ΔQ2 = −idp ·udn−iqp ·uqn + idn ·udp + iqn ·uqp (10) thereby maximizing grid support effect of the plant,
- The frame of the reference components’ calculation is kept the same
with Pr and Qr being average and ΔP1, ΔP2 and ΔQ1, ΔQ2 being oscil- as in regular mode, making practical implementation rather simple,
lating components of active and reactive power, respectively. Super- - Selection of the preferred grid support methodology is para-
script ‘r’ indicates reference value. meterized.
As there are four controlled variables, i.e. grid current components,
idpr, iqpr, idnr and iqnr, only four of these equations can be used as con- The prerequisite for both control paradigms to function properly is
stituents of a solvable system. Two most commonly used systems con- that the highest phase current that would appear is identified. This can
sist of Eqs. (5)–(8) and (5), (6), (9) and (10)}, with (7)–(10) being be done using simple transformation of the current references from
equated with zero. The first system yields active power with minimized synchronous reference frame to the natural abc frame, that is, obtaining
oscillations and the second gives response with reactive power with maximum of the phase currents by executing inverse Park and inverse,
minimized oscillations. Once the active and reactive power references power invariant, Clarke transformations on the calculated current re-
and grid voltage sequences are fed to the references calculation block ferences values.
and the system is solved, four grid current references are obtained, as

Table 1
Current references expressions for the two strategies.
r r r r
idp iqp idn iqn

ΔP = 0 Pr ·udp Qr ·uqp Pr ·uqp Qr ·udp Pr ·udn Qr ·uqn Pr ·uqn Qr ·udn (t1)

+ − − + − −
D1 D2 D1 D2 D1 D2 D1 D2

ΔQ = 0 Pr ·udp Qr ·uqp Pr ·uqp Qr ·udp Pr ·udn Qr ·uqn Pr ·uqn Qr ·udn (t2)

+ − − +
D2 D1 D2 D1 D2 D1 D2 D1

2 2 2 2
D1 = udp + uqp−udn −uqn .
2 2 2 2
D2 = udp + uqp + udn + uqn .

313
I. Todorović et al. Electrical Power and Energy Systems 101 (2018) 311–322

Table 2 Once φa = 0 is used in expressions (11)–(14), expressions (16) are

New references for the powers. obtained. Expression (16a) should be used as the supplementary
iamax ⩾ Imax ibmax ⩾ Imax i cmax ⩾ Imax equation, if it is identified that the phase a current would be the
highest, that is if iamax ≥ Imax. Similarly, (16b) should be used if
ΔP = 0 r
Pnew K·
D1
K·
D1
K·
D1 (t3) ibmax ≥ Imax and so forth.
udp − udn 2·udp + udn + 3 ·uqn 2·udp + udn + 3 ·uqn

r D2 D2 D2 (t4)
2
Qnew K· K· K· ·(idp + idn ) = Imax
udp − udn 2·udp + udn + 3 ·uqn 2·udp + udn + 3 ·uqn 3 (16a)
ΔQ = 0 r D2 D2 D2 (t5)
Pnew K· K· K·
udp + udn 2·udp − udn − 3 ·uqn 2·udp − udn + 3 ·uqn 2
·0.5·(2·idp−idn− 3 ·iqn ) = Imax
r
3 (16b)
Qnew K·
D1
K·
D1
K·
D1 (t6)
udp + udn 2·udp − udn − 3 ·uqn 2·udp − udn + 3 ·uqn
2
·0.5·(2·idp−idn + 3 ·iqn ) = Imax
D1 = udp2
+ uqp 2 2
−udn 2
−uqn . 3 (16c)
2 2 2 2
D2 = udp + uqp + udn + uqn .
Eq. (17) are obtained when φa = pi/2 is used. One of these should
3
K= ·Imax . be used when the reactive power maximization is desired.
2

2
·(−iqp + iqn ) = Imax
The following equations are used: 3 (17a)
r r r r
iαx = idp ·cos(φx )−iqp ·sin(φx ) + idn ·cos(−φx )−iqn ·sin(−φx ) (11) 2
·0.5·(−2·iqp + 3 ·idn−iqn ) = Imax
r r r r 3 (17b)
iβx = idp ·sin(φx ) + iqp ·cos(φx ) + idn ·sin(−φx ) + iqn ·cos(−φx ) (12)
2
with x = a, b, c and φx being ·0.5·(−2·iqp− 3 ·idn−iqn ) = Imax
3 (17c)
φa = atan(Qr / P r )
The new system of equations has five equations. For example, if
φb = φa−2·pi/3 active power maximization is necessary, with minimized oscillations in
φc = φa + 2·pi/3 (13) active power and if it is concluded that iamax ≥ Imax, the new system of
equations would consist of Eqs. (5)–(8) and (16a). Eqs. (6)–(8) should
Now the maximum of phase currents can be calculated using inverse
be equal to zero in this example so that all of the current capacities are
Clarke transformation,
used for active power production and to minimize the oscillations in
active power.
⎡1 0 ⎤
⎡iamax ⎤ 2 ⎢ 0.5 3 ⎥ i αx Accordingly, there are five unknowns, namely four reference cur-
⎢ ibmax ⎥ = 2 ⎥·
⎡ ⎤,
⎢ icmax ⎥ 3⎢ ⎢ iβx ⎥ rents (idprnew, iqprnew, idnrnew and iqnrnew) and a new, redefined power re-
⎢ 0.5 − 3 ⎥ ⎣ ⎦
⎣ ⎦
⎣ 2 ⎦ (14) ference (Prnew or Qrnew). In order words, it is taken that the power re-
ferences are also unknown and are to be calculated, not passed from the
that is, by using the expressions (15a)–(15c): outer loops (DC-link voltage and grid voltage control loops or operator).
2 Once the system is solved offline, new power and current references
iamax = ·iαa expressions are obtained and can be readily used in the specific sce-
3 (15a)
narios. The corresponding expressions are shown in Table 2. It should
2 ⎛ 3 be noted that it is computationally less demanding to calculate only the
ibmax = ·⎜0.5·i αb + ·iβb⎞⎟ new power reference and use it in the original equations shown in
3 ⎝ 2 ⎠ (15b)
Table 1 than to calculate new power and use new expressions for the
2 ⎛ 3 current references. Thus, only new power references are shown in
icmax = · 0.5·i αc−
⎜ ·iβc ⎞ ⎟
Table 2. For the observed example, the first term in expression (t3) in
3 ⎝ 2 ⎠ (15c)
Table 2 should be used to calculate new active power reference.
There is an additional aspect that is crucial for practical im-
A. Active and reactive power production maximization with respect to plementation, but hasn’t been addressed before in the literature. It can
prioritization scheme and power oscillations happen that the selected grid supporting routine has issued active
power production (Prnew) which is larger than the available (or required)
The algorithm developed in this chapter has a goal of maximizing power capacity in the GCC’s DC bus, Pr. The reason is that computed
active and reactive power production when the references are gener- Prnew depends solely on the value of the remaining grid voltage (see
ated by DC-link voltage control (Pr) and grid side voltage control or Table 2), but the primary cause of the current limitation could be re-
defined directly by the operator (Qr). latively high reactive power reference issued prior to the fault, Qr, with
The objective of power maximization during voltage dips is a small Pr (e.g. GCC works in grid voltage regulation mode and the
equivalent to bounding the highest phase currents to the maximum primary energy source is low). If the remaining grid voltage provide-
permissible values. This can be accomplished by appending an addi- shigher margin for the active power production than Pr, it follows that
tional equation to four equations taken form (5)-(10), from which the Prnew > Pr. In this case, in order to insure integrity of the DC bus voltage
current references were originally calculated. This additional equation regulation, the algorithm needs to respond by canceling the new power
ensures that the amplitudes of the phase currents do not surpass and reference and retain original, Pr. At the same time reactive power re-
stay exactly at the defined limit Imax. ference would be unnecessarily brought to zero, as Qr was set to zero
The additional equation can be found by again using expressions previously. In order to utilize full current capacity of the plant and
(11)–(14), but in this case with different φx. If maximization of active enhance its effectiveness in the grid supporting mode of operation,
power production is desired, φa should be set to zero, as currents that reactive power carrying current component ixq, of the phase current
are not phase shifted to the voltages produce only active power. Simi- bounded to Imax (x = a, b or c), has to be recalculated according to the
larly, if reactive power production should be maximized, the φa should (18), where ixp is maximal value of the current that corresponds to Pr;
be set to pi/2. since Pr < Prnew, follows that ixp < Imax and this margin is used to inject

314
I. Todorović et al. Electrical Power and Energy Systems 101 (2018) 311–322

Fig. 3. Current limitation algorithm (constant cosφ).

Fig. 2. Current limitation algorithm (active power production prioritized).

of Eqs. (20) and (21). The ixp and ixq expressions are generated by the
ixq. appropriate expressions from Table 2. Power factor is defined by grid
2 2 operator. It should be noted that Eq. (21) has to be solved only once,
i xq = Imax −i xp . (18) when there is a change in cosφ demanded by the grid operator.
r
Value of ixp can be obtained according to Table 2 where P is used in- Otherwise, it is not recalculated.
stead of Prnew and ixp instead of Imax. For the case in which phase-a
2 2
current has the highest value and ΔP = 0, first form of expression (t3) Imax = i xp + i xq (20)
yields expression (19).
−1 r
2 2 2 2 Qnew
⎛ 3 1 udp + uqp−udn−uqn ⎞ = tan(acos(cos(φ)))
iap = ⎜ · r ⎟
r
Pnew (21)
⎝ 2 P udp−udn ⎠ (19)

The ixq is now used in the first expression (t4) in Table 2, instead of the For example, if the phase-a current would be the highest one and the
Imax so that the new reactive power reference can be calculated, and minimized active power oscillations are desired, the expression (20)
also used in equations in Table 1. The process would be analogous to becomes:
the one shown above, if the reactive power maximization power is
2 2
desired. In this case new reactive power reference would be calculated ⎛ 2 udp−udn r ⎞ ⎛ 2 udp−udn r ⎞
Imax = ⎜ 3 · u 2 + u 2 −u 2 −u 2 ·Pnew ⎟ + ⎜ 3 · u 2 + u 2 + u 2 + u 2 ·Qnew ⎟
and compared to the old one, and then new active power would be ⎝ dp qp dn qn ⎠ ⎝ dp qp dn qn ⎠
calculated, if there are some current capacities left for it. Other com- (22)
binations are possible and can be developed accordingly.Fig. 2 offers
the depiction of the algorithm and implies its simplicity. The case when Once the system of Eqs. (21) and (22) is solved for the observed ex-
the active power production maximization is required is shown, but the ample, new active and reactive power references are calculated (Eq.
algorithm when the reactive power production is prioritized, would be (23)). These power references are used to calculate new current refer-
completely analogous. ences, using Table 1, that will result in safe maximization active and
reactive power production, with power factor being at the defined level,
B. Active and reactive power maximization with respect to constant cosφ both before and during the fault. Similarly, if phase-b current would be
and power oscillations the highest one and if ΔQ = 0 is desired, Eq. (24) should be used. The
rest of the expressions are developed accordingly. The signal flow of
In order to attain control over the system dynamics during the this part of the algorithm is depicted in Fig. 3.
disturbance period, such as adjusting its relative damping coefficient or
stability margin, the algorithm is provided with the functionality to r Imax
Pnew =
2 2
output power references according to the set value of power factor, cosφ ⎛ udp − udn ⎞ udp − udn
⎜
2
· 2 2 2 ⎟ + ⎜⎛ 2
· 2 2 2 ·tan(acos(cos(φ))) ⎟⎞
[34]. Again, in order to maximize the unit’s impact on the system, the ⎝
2
3 u + uqp − u − uqn
dp dn ⎠ ⎝ dp
2
3 u + uqp + u + uqn
dn ⎠
algorithm will seek for the solution with maximized Prnew and Qrnew va-
(23)
lues. Alongside, it can be selected to minimize active power oscillations
or reactive power oscillations. This can be achieved by solving system

315
I. Todorović et al. Electrical Power and Energy Systems 101 (2018) 311–322

Table 4
Power stage data.
Parameter Symbol Value

Filter and equivalent grid resistance Ra, Rb, Rc 0.2 Ω

Filter and equivalent grid inductance La, Lb, Lc 0.01H
Nominal line voltage Ua , U b , U c 230 V
Nominal current Ia, Ib, Ic 6.7 A
DC-link voltage Udc 360 V
Grid frequency F 50 Hz
Braking and precharge resistance Rbr, Rpa, Rpb, Rpc 100 Ω
Breaking chopper turn on voltage 650 V
Breaking chopper turn off voltage 600 V

Fig. 4. HIL experimental setup. operational states and its structural and computational burden. For the
controller, Texas Instruments TMS320F28335 was used. The control
code was developed using Matlab/Simulink code generation toolbox,
Table 3 while the debugging was done using Code Composer Studio. Perun
Controller relevant data. Power Desk was used for scoping functionalities [35]. The micro-
controller implements the complete reference and calculation algo-
Parameter Symbol Value
rithm, covering all grid support strategies, at the PWM frequency
Switching and current loop frequency Tpwm and Ti 10 kHz (Tpwm) and only the DC-link voltage control is executed at the slower
DC-link voltage control frequency Tdc 200 Hz rate (Tdc). Table 3 gives important control code parameters.
Base current Ib ( 3 In) 11.6 A Power stage was emulated on the Typhoon HIL 600 unit. Emulation
Base voltage Ub (Unline) 230 V time step was one microsecond, which is sufficiently low for the PWM
Current controller proportional gain Kpi 7.9
frequency of 10 kHz. Fig. 5 shows the outlook of the power stage used
Current controller resonant gain Kri 3900
Voltage controller proportional gain Kpdc 81.2 for the HIL experimentation, which was developed using Typhoon HIL
Voltage controller integral gain Kidc 14 schematic editor.
The inverter depicted in Fig. 5 is a three phase, two level voltage
source converter. Its model is a hybrid of continuous time dynamics
Imax
r
Pnew = model (continuous-time state-space) and discrete, finite automaton
2 2
⎛ 2 2·udp − udn − 3 ·uqn ⎞
· 2 + ⎜⎛ 2 2·udp − udn − 3 ·uqn
· 2 ·tan(acos(cos(φ))) ⎟⎞
based, model. This modeling framework is based on generalized hybrid
⎜ 2 2 2 ⎟ 2 2 2
3 u + uqp − u − uqn 3 u + uqp + u + uqn
⎝ dp dn ⎠ ⎝ dp dn ⎠ automaton (GHA) with piecewise linear continuous dynamics. The
(24) emulator used here was validated as an experimenting tool [36] and as
such was used for the control algorithm testing both by the academia
and the industry [5,35]. Also, it allows for a variety of testing scenarios
4. Results and discussion to be carried out easily at a reduced cost. Further details about the
mentioned power stage modeling and emulation can also be found in
This section is dedicated to the feasibility and effectiveness ex- [5,36].
amination of the proposed algorithm. It will be shown through selected Table 4 gives the important circuit parameters.
tests that defined power production strategies are realized while phase The breaking chopper (switch S2 and Rbr in Fig. 5), was used as a
currents are not surpassing the limit during the faults. means for managing the DC-link energy surplus during the active power
All the tests were executed in the hardware-in-the-loop (HIL) en- curtailment. The blade pitch angle control mechanism in wind power
vironment shown in Fig. 4. The setup consists of an interface board with plant, or analogous mechanism in other applications, could help with
the controller (1), emulator unit (2) and the PC (3). the excessive energy accumulated during the fault, depending on its
The HIL experimental setup employs real controller which, in con- dynamics, but this mechanism has not been considered. The controlled
junction with the emulator, runs the control algorithm in real time. This current source is used as a primary source of energy. Constant DC-link
enables thorough testing of the algorithm effectiveness in all

Fig. 5. Power stage used for the HIL experiments.

316
I. Todorović et al.
Fig. 6. Converter response during fault without limitation.
current is set so that constant active power production is realized. Si-
milarly, for the sake of clarity, the reactive power production is set to be
constant. In power prioritization scenario, the prefault and postfault
values of both powers are 1500 W and 1500 V-ampere reactive, re-
spectively. Following the remarks in [33] on which types of sags are the
most frequent, the inverter was exposed to four types of faults.
The algorithm was tested against symmetrical voltage drop on the
primary side of the transformer. This is manifested as also symmetrical
voltage sag at the secondary of the transformer (type A), after the delta/
wye transformer. Next, the test with two-phase-to-ground on the pri-
mary side of the transformer was executed. This fault manifests as (type
F) on the secondary side. The third test was single-phase-to-ground on
the primary side of the transformer. This fault manifests as (type C) on
the secondary side. The fourth fault was the complete collapse of one
phase voltage on the secondary side (type B). The closing of the con-
tactor S1, shown in Fig. 4, results in the direct connection of one sec-
ondary side phase to the ground. This would certainly trigger the pro-
tection devices in normal conditions, but the inverter behaves properly,
even during this serious fault.
The behavior during other types of faults is not shown in the paper
as they are similar to the observed ones.
Six strategies are tested:
- The first produces minimized oscillations in active power, with
prioritization of reactive power production.
- The second produces minimized oscillations in reactive power and
in prioritization of reactive power production.
- The third again produces minimized oscillations in active power, but
this time with prioritization of active power production.
- The fourth strategy results in minimized oscillations in reactive
power and active power production being maximized.
- The fifth operating condition ensures minimized oscillations in ac-
tive power with constant cosφ (in the averaged sense - powers
averaged over 0.01 s were used for depicting cosφ during the fault).
- The sixth operating condition results in minimized oscillations in
reactive power with constant cosφ.
The current limit Imax is set to be 11.6 A. This implies that the
current limit is 3 times higher than nominal value, as nominal value is
6.7 A.
317
Electrical Power and Energy Systems 101 (2018) 311–322
Fig. 6 depicts the results that are obtained during the phase-to-
ground fault on the primary side of the transformer (type C fault on the
secondary side) without current references limitation and is included
for the sake of comparison with the algorithm.with the limited currents.
The currents reach 17.1 A during the fault which is almost double the
nominal maximum value. Considering that the overcurrent limits are
usually set to 50% higher values than the nominal value, or less, this
fault would result in the GCC’s disconnection from the grid.
Fig. 7 shows the results obtained using the first strategy during the
abrupt appearance of one phase voltage collapse on the secondary side
of the transformer (type B sag). It can be seen that the active power
production was lowered to 750 W, while the reactive power is left un-
changed, which is in accordance to the power production prioritization
scheme. Furthermore, it is apparent that the oscillations in active power
are negligible, while oscillations in reactive power are significant. This
also suggests that the converter behaves expectedly. After the fault, the
active power has increased above the pre-fault value so that the excess
of energy accumulated during the fault in the DC-link capacitor can be
released and so that the voltage can be returned to the reference value.
Fig. 8 depicts the results obtained using the second strategy during
continuously changed voltage conditions. Firstly, one phase voltage on
the primary of the transformer is changed from nominal value to zero.
Afterwards, the second phase voltage is lowered from nominal value to
close to zero value. It can be seen that active power production is de-
creased first, which is in accordance to the second strategy. Once the
active power production has ceased, the curtailment of reactive power
occurred.
Fig. 9 illustrates the use of the third strategy, as oscillations in active
power are minimized. The sag (type C) occurred at the secondary side
of the transformer.
It can be seen that the oscillations in active power are significantly
smaller than those of the reactive power. Also, as defined, reactive
power curtailment happens as the sag occurs, with the active power
being at the pre-fault level, that is, Prnew = Pr. The currents are at the
defined maximum level, which suggests that all of the producing cap-
abilities are exploited.
Fig. 10 presents the response when the fourth strategy was used.
During the testing, the two phase-to-ground fault was executed, (type F)
fault at the secondary side, and as a result both active and reactive
power were reduced. The figure suggests that the reactive power
I. Todorović et al. Electrical Power and Energy Systems 101 (2018) 311–322

Fig. 7. Oscillating reactive power, active power curtailed first (abrupt change).

production completely ceased and active power diminished only par-
tially, in accordance with the power production requirements. Currents
are again limited at one relative unit, that is, at 11.6 A. As the active
power production was lowered only to 1400 W, the DC-link voltage has
not reached the 650 V limit during the fault.
In Fig. 11 one can see the behavior of the converter when the
complete grid collapse occurs, that is, when all of the phases sag to zero
voltage (type A fault). The first strategy was used, which suggests that
reactive currents are produced. Oscillations in reactive power do not
exist, as the dip is symmetrical. According to the [31], converter must
stay connected to the grid for 150 ms when the voltage sags to the 45%
of the nominal value in all of three phases. Also, this grid code demands
that the converter stays connected to the grid even when complete
collapse of the grid voltages occurs, although in this case short dis-
connection for two seconds is allowed if instabilities occur. Thus, as the
currents are at the limit and no noticeable instabilities occur, this test
shows that the algorithm meets the requirements of even the most
stringent grid codes [31].
Resilience to complete loss of grid voltages was not reported in any
of the cited references.
In Fig. 12, response to the transformer secondary side (type C) fault
can be seen. The fifth operating condition was executed on the DSP. Low
value cosφ was defined (cosφ = 0.1). The active power production was
set to be such that the active power production with reactive power pro-
duction (calculated using cosφ) would demand maximum currents before
fault. It can be seen from the figure that both powers were simultaneously
reduced during the fault so that the current limits are not surpassed. The
power factor was calculated using averaged active power and its value
remains the same to the defined value, i.e. to the pre-fault value (only a
small drop to 0.089 can be seen). This proves that the algorithm allows

Fig. 8. Oscillating active power, active power curtailed first (cont. change).

318
I. Todorović et al. Electrical Power and Energy Systems 101 (2018) 311–322

Fig. 9. Oscillating reactive power, reactive power reduced first.

different reactive power reference generation strategies to be im-
plemented, that are defined by the grid codes [31,37].
Similarly to the previous test, in Fig. 13 can be seen the variables’
responses when the algorithm that can ensure constant power factor
when oscillations in reactive power are allowed (sixth strategy) is used.
This time, high cosφ was set (cosφ = 0.9). Again, it can be seen that the
power factor remains roughly the same before the fault and during the
fault.
Through the shown tests, proper functioning of the proposed algo-
rithm becomes evident, even during the faults that would certainly
trigger the protection and would disconnect the converter from the
grid. The sags are set to last for 500 ms. Thus, the algorithm allows the
longer connection of the GCC, even during the permanent unbalances
and serious faults than those defined by the existing grid codes.
It should be emphasized that the currents do not surpass the limits -
not even transiently. The absence of the sag detector in the algorithm
largely contributes to great transient responses. Also, the currents are
always exactly at the limit throughout the fault conditions. This sug-
gests maximum possible and safe production, unlike in [18] where the
currents are under the limits and production capabilities are not met.
A great feature of the proposed algorithm is the possibility of the
simultaneous production of active and reactive power even during the
fault - feature that many of the proposed algorithms lack. This implies
that the proposed algorithm can be used in conjunction with any dis-
tributed generation sources that uses the three phase inverter. Even if
the production of certain power is not demanded by the grid codes,
having the possibility of doing so gives this algorithm the advantage. It
is plausible that the future grid codes will allow more fluid power flow

Fig. 10. Oscillating active power, reactive power decreased first.

319
I. Todorović et al. Electrical Power and Energy Systems 101 (2018) 311–322

Fig. 11. Oscillating reactive power, active power reduced first.

rules, influenced by local grid characteristics.
Another useful feature of the proposed algorithm is that it considers
the oscillations in powers and allows for the oscillations to be mini-
mized, which is the main shortcoming of the approaches considered in
[14,11].
Different approach that considers positive and negative, active and
reactive power components [14] offers interesting grid support cap-
abilities that could be utilized in future grid codes’ regulations. It is to
be examined if the corresponding functionalities could be devised using
the approach shown in this paper. Arguably, as most approaches shown
in the literature are based on controlling four grid current components,
it is reasonable to expect that these approaches could be adapted to the
calculation frame used in this paper utilizing five equations system. If
so, every control algorithm that is based on symmetric components
could simply be upgraded with the functionalities shown in this paper.
Finally, it should be emphasized that, contrary to the [12], the AC
side voltage limits were not considered. It was assumed that the voltage
limits control is a task for the outer control loops, namely, grid voltage
control loop, and therefore that is a separate problem that is out of
scope of the paper. Properly set current loop and power maximization
procedures themselves cannot cause overvoltage, but can mitigate it (in
accordance to the voltage loop commands). The voltage control stra-
tegies shall be addressed in future studies.

Fig. 12. Oscillating active power, constant cosφ.

320
I. Todorović et al.
Fig. 13. Oscillating reactive power, constant cosφ.
5. Conclusions
Grid connected inverter faces power flow with the increased cur-
rents during voltage unbalances. This increase could compromise
maximized and safe power production with respect to the primary
source available energy, grid support and the current limits. Although
this problem has previously been observed in the literature, the pro-
posed algorithm addresses the shortcomings of the previously devel-
oped algorithms. It offers a new approach to the simultaneous pro-
duction of the active and reactive power, comprehensively addressing
different power production strategies during the grid faults, as de-
manded by the grid operator and the grid characteristics. As such, it can
be used in conjunction with any distributed energy source. The algo-
rithm derives active and reactive power references based on a set of
objectives to maximize power production, limit the output currents and
enable different grid support strategies. The proposed control frame-
work suits practical implementation regarding structural and compu-
tational burden which was extensively tested on the HIL experimental
setup. It is left for the future considerations to adapt other, previously
developed power production control approaches, i.e. those based on the
symmetrical components, to the proposed methodology and to address
grid side voltages control.
Acknowledgments
This work was supported by the Ministry of Education, Science and
Technological Development of the Republic of Serbia under project
III42004.
References
[1] Khani D, Sadeghi Yazdankhah A, Madadi Kojabadi H. Impacts of distributed gen-
erations on power system transient and voltage stability. Int J Electr Power Energy
Syst 2012;43(1):488–500.
[2] Etxegarai A, Eguia P, Torres E, Iturregi A, Valverde V. Review of grid connection
requirements for generation assets in weak power grids. Renew Sustain Energy Rev
Jan. 2015;41:1501–14.
[3] Chen T-H, Yang C-H, Yang N-C. Examination of the definitions of voltage unbalance.
Int J Electr Power Energy Syst Jul. 2013;49:380–5.
[4] Miret J, Castilla M, Camacho A, Vicuña LGd, Matas J. Control scheme for photo-
voltaic three-phase inverters to minimize peak currents during unbalanced grid-
voltage sags. IEEE Trans Power Electron Oct. 2012;27(10):4262–71.
[5] Ivanović ZR, Adžić EM, Vekić MS, Grabić SU, Čelanović NL, Katić VA. HIL
321
Electrical Power and Energy Systems 101 (2018) 311–322
evaluation of power flow control strategies for energy storage connected to smart
grid under unbalanced conditions. IEEE Trans Power Electron Nov.
2012;27(11):4699–710.
[6] Sosa JL, Castilla M, Miret J, Matas J, Al-Turki YA. Control strategy to maximize the
power capability of PV three-phase inverters during voltage sags. IEEE Trans Power
Electron Apr. 2016;31(4):3314–23.
[7] Jiang W, Wang Y, Wang J, Wang L, Huang H. Maximizing instantaneous active
power capability for PWM rectifier under unbalanced grid voltage dips considering
the limitation of phase current. IEEE Trans Ind Electron Oct.
2016;63(10):5998–6009.
[8] Yongheng Y, Blaabjerg F, Low-voltage ride-through capability of a single-stage
single-phase photovoltaic system connected to the low-voltage grid. Int J
Photoenergy, vol. 2013, p. 9.
[9] Rodriguez P, Medeiros G, Luna A, Cavalcanti MC, Teodorescu R. Safe current in-
jection strategies for a STATCOM under asymmetrical grid faults. IEEE Energy
Convers Congr Exposition 2010;2010:3929–35.
[10] Rodriguez P, Luna A, Medeiros G, Tedorescu R, Blaabjerg F. Control of STATCOM in
wind power plants based on induction generators during asymmetrical grid faults.
The International Power Electronics Conference - ECCE ASIA 2010;2010:2066–73.
[11] Camacho A, Castilla M, Miret J, Guzman R, Borrell A. Reactive power control for
distributed generation power plants to comply with voltage limits during grid
faults. IEEE Trans Power Electron Nov. 2014;29(11):6224–34.
[12] Milicua A, Abad G, Vidal MÁR. Online reference limitation method of shunt-con-
nected converters to the grid to avoid exceeding voltage and current limits under
unbalanced operation #x2014; Part I: Theory. IEEE Trans Energy Convers Sep.
2015;30(3):852–63.
[13] Milicua A, Abad G, Vidal MÁR. Online reference limitation method of shunt-con-
nected converters to the grid to avoid exceeding voltage and current limits under
unbalanced operation #x2014;Part II: Validation. IEEE Trans Energy Convers Sep.
2015;30(3):864–73.
[14] Camacho A, Castilla M, Miret J, Borrell A, de Vicuña LG. Active and reactive power
strategies with peak current limitation for distributed generation inverters during
unbalanced grid faults. IEEE Trans Ind Electron Mar. 2015;62(3):1515–25.
[15] Ebrahimi E, Sanjari MJ, Gharehpetian GB. Control of three-phase inverter-based DG
system during fault condition without changing protection coordination. Int J Electr
Power Energy Syst Dec. 2014;63:814–23.
[16] Remus Teodorescu, Marco Liserre, Pedro Rodriguez. Grid converters for photo-
voltaic and wind power systems. Wiley-IEEE Press, 2011.
[17] Shabestary MM, Mohamed YARI. An analytical method to obtain maximum al-
lowable grid support by using grid-connected converters. IEEE Trans Sustain Energy
Oct. 2016;7(4):1558–71.
[18] Zheng T, Chen L, Guo Y, Mei S. Flexible unbalanced control with peak current
limitation for virtual synchronous generator under voltage sags. J Mod Power Syst
Clean Energy May 2017:1–12.
[19] Rodriguez P, Timbus AV, Teodorescu R, Liserre M, Blaabjerg F. Flexible active
power control of distributed power generation systems during grid faults. IEEE
Trans Ind Electron Oct. 2007;54(5):2583–92.
[20] Kabiri R, Holmes DG, McGrath BP. Control of active and reactive power ripple to
mitigate unbalanced grid voltages. IEEE Trans Ind Appl Mar. 2016;52(2):1660–8.
[21] Nejabatkhah F, Li YW, Wu B. Control strategies of three-phase distributed genera-
tion inverters for grid unbalanced voltage compensation. IEEE Trans Power
Electron Jul. 2016;31(7):5228–41.
I. Todorović et al.
[22] Shehata EG, Salama GM. Direct power control of DFIGs based wind energy gen-
eration systems under distorted grid voltage conditions. Int J Electr Power Energy
Syst 2013;53:956–66.
[23] Zhang Y, Qu C. Direct power control of a pulse width modulation rectifier using
space vector modulation under unbalanced grid voltages. IEEE Trans Power
Electron Oct. 2015;30(10):5892–901.
[24] Chen Z, Luo A, Wang H, Chen Y, Li M, Huang Y. Adaptive sliding-mode voltage
control for inverter operating in islanded mode in microgrid. Int J Electr Power
Energy Syst 2015;66:133–43.
[25] Blaabjerg F, Teodorescu R, Liserre M, Timbus AV. Overview of control and grid
synchronization for distributed power generation systems. IEEE Trans Ind Electron
Oct. 2006;53(5):1398–409.
[26] Jaalam N, Rahim NA, Bakar AHA, Tan C, Haidar AMA. A comprehensive review of
synchronization methods for grid-connected converters of renewable energy source.
Renew Sustain Energy Rev Jun. 2016;59:1471–81.
[27] Peña R, Cárdenas R, Asher G. Overview of control systems for the operation of
DFIGs in wind energy applications. In: IECON 2013–39th Annual Conference of the
IEEE Industrial Electronics Society, 2013, pp. 88–95.
[28] Reyes M, Rodriguez P, Vazquez S, Luna A, Teodorescu R, Carrasco JM. Enhanced
decoupled double synchronous reference frame current controller for unbalanced
grid-voltage conditions. IEEE Trans Power Electron Sep. 2012;27(9):3934–43.
[29] Vidal A, et al. Assessment and optimization of the transient response of propor-
tional-resonant current controllers for distributed power generation systems. IEEE
322
Electrical Power and Energy Systems 101 (2018) 311–322
Trans Ind Electron Apr. 2013;60(4):1367–83.
[30] Rodriguez FJ, Bueno E, Aredes M, Rolim LGB, Neves FAS, Cavalcanti MC. Discrete-
time implementation of second order generalized integrators for grid converters. In:
34th Annual Conference of IEEE Industrial Electronics, 2008. IECON
2008;2008:176–81.
[31] Tennet TSO GmbH, Grid Code - High and extra high voltage -. Nov-2015.
[32] Altin M, Göksu Ö, Teodorescu R, Rodriguez P, Jensen BB, Helle L. Overview of
recent grid codes for wind power integration. In: 2010 12th International
Conference on Optimization of Electrical and Electronic Equipment (OPTIM), 2010,
pp. 1152–1160.
[33] Roiu D, Bojoi RI, Limongi LR, Tenconi A. New stationary frame control scheme for
three-phase PWM rectifiers under unbalanced voltage dips conditions. IEEE Trans
Ind Appl Jan. 2010;46(1):268–77.
[34] Fairley P. How rooftop solar can stabilize the grid. IEEE Spectr Feb.
2015;52(2):11–3.
[35] Porobić Vlado B, Adžić Evgenije M, Rapaić Milan R. HIL evaluation of control unit
in grid-tied coverters. Therm Sci 2016;20(2):393–406.
[36] Majstorovic D, Celanovic I, Teslic ND, Celanovic N, Katic VA. Ultralow-latency
hardware-in-the-loop platform for rapid validation of power electronics designs.
IEEE Trans Ind Electron Oct. 2011;58(10):4708–16.
[37] Energinet, “Technical regulation 3.2.5 for wind power plants with a power output
greater than 11 kW”. Sep-2010.