www.ietdl.

org
Published in IET Renewable Power Generation
Received on 5th March 2013
Revised on 16th August 2013
Accepted on 4th November 2013
doi: 10.1049/iet-rpg.2013.0067

ISSN 1752-1416

Power control strategy for photovoltaic system

based on the Newton quadratic interpolation
Yun Liu1, Huanhai Xin1, Zhen Wang1, Taicheng Yang2
1
Department of Electrical Engineering, Zhejiang University, Hangzhou 310027, People’s Republic of China
2
Department of Engineering and Design, University of Sussex, Brighton, UK
E-mail: xinhh@zju.edu.cn

Abstract: To maximise the economic benefit, photovoltaic (PV) systems in general operate in the so-called maximum power
point tracking (MPPT) mode. However, in certain occasions (e.g. in a microgrid or in a weak system), it is beneficial for a
PV system not to always operate in the MPPT mode, but occasionally in the power dispatch mode, because of the top priority
of maintaining system stability. To this end, a Newton quadratic interpolation-based power control strategy for PV system is
proposed to iteratively obtain the required terminal voltage of PV system by approximating the power–voltage characteristic
curve with a quadratic curve. With this control strategy, PV systems can operate in the power dispatch mode to flexibly adjust
the active power output in a wide range, or adaptively switch to the MPPT mode if necessary. Details on the convergence
rate and the way to achieve the fault ride-through capability are also discussed. Simulation is performed based on a detailed
PV dynamical model, illustrating that the proposed method has fast convergence rate and robust performance compared with
a revised perturb and observe method which can attain the same function.

1 Introduction participate in frequency regulation without the need of ESS

With increasing environmental concern, uprising cost for
fossil fuels, together with the desire to increase the diversity
in energy supply, electricity industry focuses more on the
development of renewable energy sources (RES). Among
all forms of distributed generation, photovoltaic (PV)
generation is developing rapidly worldwide because of its
environment-friendly characteristics and economic
incentives from governments. Continued rapid development
can be foreseen in the near future [1, 2].
Since the initial installation costs of PV systems are still
relatively high, PV systems nowadays mostly operate in the
maximum power point tracking (MPPT) mode in order to
maximise the revenue [3]. However, if the penetration of
PV systems is high (e.g. in a microgrid), the MPPT mode
could severely weaken the power system stability because
system inertia and frequency regulation capability are
decreased by replacing synchronous generators [4].
Therefore it can be predicted that PV systems in the future
should be required to provide some necessary ancillary
services, such as frequency regulation. This task can be
attained by corporation of energy storage systems (ESSs),
but it is costly [5]. Another possible solution is to design a
new control strategy which enables PV systems to adjust
the power output temporarily and flexibly as needed, even
sacrificing economically. Some relevant work has been
done recently, for example, a distributed control strategy is
presented to have all PV systems in a system operate in
certain ratio of maximum available power [6]. A simple
fuzzy-based control method is proposed for PV systems to
[7]. However, these studies do not provide details on power
control methods of PV system.
In the past decades, many researchers have conducted
studies on MPPT algorithms (e.g. [8–11] and the references
therein). Esram and Chapman [8] surveyd many different
MPPT algorithms, such as the perturb and observe (P&O)
[11], incremental conductance (IncCond) [9], fractional
short-circuit current [10] etc. As surveyed in [8], these
algorithms have certain limitations, such as slow
convergence, poor stability and low precision. Some revised
methods are provided to eliminate certain limitations and
improve the performance [12–15]. In those algorithms, Pai
and Chao [15] introduced an effective MPPT algorithm
based on a Lagrange quadratic interpolation. This method is
reported to have faster convergence rate and high precision
in various operational scenarios. However, when the three
sampling points are monotonically increasing/decreasing,
the shifting procedure might be time consuming sometimes.
Despite a large number of papers studying MPPT methods,
few have discussed PV systems operating in the power
dispatch mode. Li et al. [16] separately builds the MPPT
mode and power dispatch mode, with the MPPT mode
realised by an existing method, for example, the P&O, and
the power dispatch mode realised by an integral link. Since
the operating point converges at the uphill section of the
power–voltage (P−V ) characteristic curve, it is unstable [3].
Delfino et al. [17] implemented the power dispatch mode
by directly controlling the active power output of the
inverter: the imbalance between the PV array and inverter
power outputs charges/discharges the capacitor parallel to

IET Renew. Power Gener., 2014, Vol. 8, Iss. 6, pp. 611–620 611
doi: 10.1049/iet-rpg.2013.0067 & The Institution of Engineering and Technology 2014
www.ietdl.org
the PV array until the PV terminal voltage reaches a steady
state. This control strategy drives the operating point to
converge at the downhill section of the P−V characteristic
curve, but it malfunctions if the dispatched power command
is more than the maximum PV power output. Besides, both
the control strategies in [16, 17] need a logic signal to
switch between the two modes, which are not discussed in
the papers.
Motivated by the idea of [15–17], a novel power control
strategy based on the Newton quadratic interpolation (NQI)
is presented. With this strategy, the required reference
voltage of the PV array can be iteratively calculated by
approximating the P−V characteristic curve with a quadratic
curve. The proposed strategy enables the PV system to
adaptively switch between the MPPT mode and the power
dispatch mode according to the relationship between the
dispatch command (e.g. the reference power) and maximum
power output. To be specific, when the reference power is
more than the maximum power output, the PV system
operates in the MPPT mode; otherwise, the PV system
operates in the power dispatch mode and tracks the
reference power. Details on measures to guarantee
convergence and the fault ride-through (FRT) capability of
the PV system are also discussed. It is demonstrated that
the proposed strategy can work well in various scenarios.
The paper is organised as follows. Section 2 introduces the
main problem and the layout of the grid-connected PV
system. The power control strategy based on the NQI is
introduced in Section 3, and some discussions are provided
in Section 4. Section 5 validates the proposed strategy
under several scenarios by simulation. Conclusions are
drawn in Section 6.
2 Problem formulation
For a PV array, the current IPV and terminal voltage UPV are
related by [18, 19]
0
IPV = IPV − Is0 [e(UPV +Rs IPV )/(Vt a) − 1] − (UPV + Rs IPV )/Rp
0 by regulating UPV: the outer loop forces UPV to track UPVref
where IPV is the light induced current depending on the
irradiance level G and array temperature T; Is0 is the diode
saturation current depending on T only; Rp is the equivalent
parallel resistance of PV array; Rs is the equivalent series
resistance of PV array; a is the diode ideality constant and
Vt is the thermal voltage depending on T only.
It follows from (1) that the power generating from the PV
array can be expressed as
PPV = UPV · IPV = PV(UPV , a)
Fig. 1 P−V characteristic curve
increased. This trend stops when PPV reaches the maximum
value; beyond this voltage, PPV decreases as UPV is
increased. This behaviour implies that PPV can be
controlled by adjusting UPV [20]. Therefore the problem to
be solved can be described as follows.
Design a power control strategy for the PV system such that
the active power injecting to the utility grid Pgrid can converge
ref
to a given reference Pgrid when the power output of the PV
MPP
array satisfies PPV , PPV (corresponding to UPV1 and
UPV2 in Fig. 1), or converge to the maximum power that
max
can be injected to the utility grid Pgrid such that
MPP MPP
PPV = PPV (corresponding to UPV in Fig. 1), where PPV MPP
represents the maximum available power of the PV array.
Thus, it follows from (2) that the main task is to design an
updating rule of UPVref, such that it converges to UPVref ∗
,
which satisfies
max ref ∗
min (Pgrid , Pgrid ) = PV (UPVref , a) − Ploss (3)
where Ploss denotes the loss of the inverter which connects the
PV array to the grid.
A single-stage inverter topology, which has the merits of
saving components and reducing cost, is adopted in this
study to illustrate the control design, as shown in Fig. 2.
The classical double-loop scheme is applied to the PWM
(1)
control, as shown in Fig. 3. The upper channel controls PPV
and generates the current command Idref , which is delivered
to a dq-frame current-control scheme (inner loop), forcing
Id to track Idref . The lower channel adopts the same
philosophy to directly control reactive power Q, by which
some ancillary functions can be attained, for example,
voltage regulation and oscillation damping (which is
beyond the scope of the paper and thus is skipped).
The feedforward of Id and Iq by a gain ωL is added to the
(2)

where α denotes a vector of parameters such as the

temperature and solar irradiance. Fig. 1 plots a classical
P−V characteristic curve.
As pointed out in the introduction, when the penetration of
PV systems is high, it is necessary and meaningful to flexibly
adjust the PV power output, for example, a PV station must be
dispatched by a distribution system operator (DSO) or provide
frequency regulation service to the utility grid. Therefore a
PV system should be able to operate either in the MPPT
mode or the power dispatch mode to provide necessary
ancillary services.
As shown in Fig. 1, PPV changes with the change of UPV. Fig. 2 Equivalent block diagram of a single-stage PV energy
PPV is zero when UPV is zero and increases as UPV is conversion system

612 IET Renew. Power Gener., 2014, Vol. 8, Iss. 6, pp. 611–620
& The Institution of Engineering and Technology 2014 doi: 10.1049/iet-rpg.2013.0067
 www.ietdl.org
polynomial can be written as [24]

Pout (U ) = C0 + C1 (U − U1 ) + C2 (U − U1 )(U − U2 ) (6)

where the coefficient C0, C1 and C2 are the notations for

divided difference and are defined recursively by

P2 − P1
C0 = [P1 ] = P1 , C1 = [P1 , P2 ] = ,
U2 − U1
(7)
[P2 , P3 ] − [P1 , P2 ]
C2 = [P1 , P2 , P3 ] =
U3 − U1

Fig. 3 Control of a grid-connected inverter Rearrange (6) in descending powers

Pout (U ) = aU 2 + bU + c (8)
PWM control outputs to decouple the dynamics. The
current-control strategy also enhances protection of the
inverter against overload and external faults by limiting where the coefficients are a = C2, b = C1−C2U1−C2U2 and
Idref and Iqref , details of which is to be illustrated in Section c = C0−C1U1 + C2U1U2.
Equation (8) can be used to estimate the PV reference
4 [20, 21]. It is worth mentioning that the control strategy voltage when the reference power Pgrid ref
is given. The
to be developed can be used for various types of PV following quadratic equation can be acquired by equating
systems with different inverter topologies (such as the ref
Pout (U ) with Pgrid
dual-stage, multiple-stage topologies etc. [22]) and PWM
control strategies (such as the feedback linearisation method
in [23]). aU 2 + bU + (c − Pgrid
ref
)=0 (9)
The grid-injected power Pgrid is used to update the reference
voltage of the PV array. Assume that the dynamics of the Define
inverter are very fast compared to the change of reference
voltage, the dynamics of the inverter can be approximately
described as a first-order delay link by ignoring the small D = b2 − 4a(c − Pgrid
ref
) (10)
power loss [16]
MPP ref
Δ < 0 indicates that PPV is less than Pgrid and the PV system
Pgrid = PPV /(1 + Td s) (4) operates in the MPPT mode, as shown in Fig. 4a. Select the
extreme point of the predicted P−V characteristic curve as
As the time constant Td is rather small, it is reasonable to the reference point in the next iteration step, the PV voltage
assume that Pgrid is approximately equal to PPV. Therefore at the maximum power point (MPP) can be approximated as
∗
the required reference voltage of the PV array UPVref
satisfying (4) can be simplified as satisfying b
k
UPVref =− (11)
MPP ref ∗ 2a
min (PPV , Pgrid ) = PV (UPVref , a) (5)
MPP ref
Δ ≥ 0 indicates that PPV is more than Pgrid . It should be noted
Since α is time-varying and there are some unknown that there are two solutions for the dispatch purpose. Figs. 4b
parameters of the PV array, it is difficult to obtain the exact and c show that the PV system operates on the uphill and
MPP
expression of PPV = PV(UPV, α) and the value of PPV . A downhill section of the P−V characteristic curve,
feasible solution is to use an iterative method by respectively. Clearly, both of them can be easily obtained
considering the convex property of the P−V characteristic by solving (9). The left-hand side solution (denoted by
curve, for example, the P&O method for tracking the MPP. NQIL) can be written as
Quadratic interpolation is another good option to improve
the results. This idea will be used to solve the above √


problem and the details will be discussed next. k −b + D
UPVref = (12)
2a
3 Control strategy based on the NQI
whereas the right-hand side solution (denoted by NQIR) can
3.1 Basic idea of the NQI-based power control be written as
strategy for PV system
√


Consider the kth step in the iteration and assume that there are k −b − D
UPVref = (13)
three sampling points on the P−V characteristic curve. Their 2a
voltage and power values are noted as (U1k , P1k ), (U2k , P2k ) and
(U3k , P3k ), satisfying U1k , U2k , U3k . The basic idea of the
proposed strategy is to use the quadratic curve passing the The characteristics of the two solutions will be compared in
three points to approximate the P−V characteristic curve. the next section and in the simulations, which show that the
Once the points are chosen, the Newton quadratic NQIR is generally better than the NQIL.

IET Renew. Power Gener., 2014, Vol. 8, Iss. 6, pp. 611–620 613
doi: 10.1049/iet-rpg.2013.0067 & The Institution of Engineering and Technology 2014
www.ietdl.org

Fig. 4 Successive iteration steps of a PV system operating in

a MPPT mode
b Power dispatch mode (NQIL)
c Power dispatch mode (NQIR)

3.2 Iteration process (14) implies that the three sampling points are so close to
k
each other that the calculation of (7) might result in
By the solutions of (11) and (12) or (13), a new UPVref can be numerical error. A feasible solution to prevent this is: after
k k k
obtained and the corresponding power output PPVref can be (14) is satisfied, point (UPVref , PPVref ) is eliminated in every
k
measured when the inverter makes UPV track UPVref . iteration step and the rest points keep unchanged.
By far, the information of four points, namely (U1k , P1k ), The P−V characteristic curve varies as to the irradiance and
(U2k , P2k ), (U3k , P3k ) and (UPVref
k k
, PPVref ) have been obtained. temperature. The above rule of iteration guarantees the new
To launch the next quadratic interpolation, one point steady state to be reached in a short time as the operating
should be removed from the four. The rule of iteration is to conditions change. The theoretical proof that the NQI
remove (U1k , P1k ) and k k
 (U3 , P3 ) in turn, that is, to remove
converges to the desired operating point can be observed in
literatures: when operating in the MPPT mode, the

k k
U(−1) k +2 , P(−1)k +2 . Then, rearrange the rest of the three
algorithm can be viewed as a maximisation problem, which
is proved to have convergence rate with order 1.32 [25];
points as (U1k+1 , P1k+1 ), (U2k+1 , P2k+1 ) and (U3k+1 , P3k+1 ), and when operating in the power dispatch mode, the algorithm
k+1 k+1
(UPVref , PPVref ) can be obtained following the same can be viewed as a root finding problem, which is proved
procedure. to have convergence rate with order 1.84 [26].
k k
As k increases, the reference point (UPVref , PPVref ) obtained Fig. 5 is the flowchart of the NQI-based power control
in every iteration step gradually approaches the new steady strategy for PV system. Some measures to avoid
state. The iteration are considered converged if the miscalculation if the Newton quadratic polynomial fails to
following condition is satisfied estimate the real P−V characteristic curve and the way to
achieve the FRT capability are also included, which will be
discussed in the next section. The cycle of Fig. 5 is executed
max |Pik − PPVref
k
| , 1 (i = 1, 2, 3) (14)
every other sampling interval. The value of sampling interval
Ts needs to be carefully tuned in coordination with the
where ε is a predefined constant. In the simulation, ε = 2 × characteristics of the PV array and the PI parameters of the
10−4 pu (50 W in the base power of 0.25 MW). Equation inverter control. Specifically, if Ts is too small, the controller

614 IET Renew. Power Gener., 2014, Vol. 8, Iss. 6, pp. 611–620
& The Institution of Engineering and Technology 2014 doi: 10.1049/iet-rpg.2013.0067
 www.ietdl.org
right solution randomly when the PV system operates in the
power dispatch mode. The left-hand side solution converges
slowly whereas the right-hand side solution leads to severe
power oscillation around the steady state, both of which are
not satisfactory. This method is implemented for simulation
in comparison with the proposed strategy.

4 Discussion on the convergence properties

4.1 Comparison between the NQIL and NQIR in the
power dispatch mode

As discussed above, there are two optional solutions when the

PV system operates in the power dispatch mode (as shown in
block 1 in Fig. 5). The main features of the two solutions are
listed as follows.
When the reference power output is very low, the
corresponding PV array voltage of the NQIL may be lower
than the peak voltage of the AC side of the inverter and
Fig. 5 Flowchart of the NQI-based power control strategy for PV thus lead to inverter failure [27]. Although this problem can
system be alleviated by increasing the number of PV panels in
series to raise the DC side voltage, or introducing the
dual-stage inverter topology, it is uneconomic or sometimes
impractical. However, for the NQIR with PV array voltage
may subject to possible mistakes caused by transient behaviour ranging between UPV MPP
and the open-circuit voltage UOC,
of the whole system. On the other hand, a too large Ts may such a problem will never occur.
lower the convergence rate. A proper Ts is approximately When adopting the dual-stage inverter topology, the
long enough for UPV to converge to UPVref. This conclusion traditional DC/DC converters have a negative impedance
will also be verified by simulation. characteristic inherently, due to the fact that their current
It is worth mentioning that the power dispatch function can increases when voltage decrease in order to keep the power
also be realised with a method similar to the P&O MPPT constant. Consequently, if the PV system operates on the
k
method. UPVref is updated every other sampling interval NQIL, which corresponds to the low-impedance (namely,
with the following updating rule low-voltage) section of the P−V characteristic curve, the
solar array terminal voltage may collapse [3].
k+1
UPVref k
= UPVref k+1
+ step × sign(Pgrid k
− Pgrid ) When the reference power changes abruptly, the NQIR
may have faster convergence rate than the NQIL. This can
k+1 k ref k
× sign(UPV − UPV ) × sign(Pgrid − Pgrid ) (15) be explained intuitively that downhill section of the P−V
characteristic curve is steeper than the uphill section.
where step is a predefined perturbation step size and sign is a Therefore, from the above comparison, the NQIR is a better
sign function. choice compared with the NQIL. Those merits will be further
ref
If Pgrid k
.Pgrid , the PV system should operate in the MPPT proved by simulation.
ref k
mode. sign Pgrid − Pgrid ; 1 and (15) degrades to the
4.2 Boundary on voltage of the PV array
conventional P&O MPPT algorithm.
ref k
If Pgrid , Pgrid , the PV system should operate in the power The quadratic curve can approximate the P−V characteristic
dispatch mode. The convergence can be illustrated using curve if P2k is the largest power of the three sampling point,
Table 1 and Fig. 6. that is, (U1k , P1k ), (U2k , P2k ) and (U3k , P3k ) satisfy
However, this method has some obvious drawbacks: (i) it
has to make a compromise between convergence rate and
system stability in choosing the value of ‘step’ [8]; and (ii) P2k − P1k P3k − P2k
· ,0 (16)
disturbances may lead the PV terminal voltage to the left or U2k − U1k U3k − U2k

Otherwise, the approximation may be unacceptable. For

Table 1 Illustration of the perturbation direction of P&O example, when the three sampling points are as shown in
algorithm in power dispatch mode
Fig. 7, the quadratic curve cannot fit the downhill section of
k+1 k ref k
Case sign(Pgrid − Pgrid )× sign(Pgrid − Pgrid ) Direction of the P−V characteristic curve. If the NQIR is selected, the
k+1 k perturbation calculated UPVref may lead UPV to a value much higher
sign(UPV − UPV )
than UOC, which may even damage the devices.
(1) +1 +1 right (towards To prevent this, the following saturation function can
UPV1) be used to keep UPVref within an accepted range
(2) +1 −1 left (towards min
[UPVref max
, UPVref ] (as shown in block 2 in Fig. 5)
UPV1)
(3) −1 −1 right (towards ⎧ min
UPV2) min
⎨ UPVref UPVref , UPVref
(4) −1 +1 left (towards UPVref = UPVref min max
(17)
UPV2)
UPVref≤ UPVref ≤ UPVref
⎩ max max
UPVref UPVref . UPVref

IET Renew. Power Gener., 2014, Vol. 8, Iss. 6, pp. 611–620 615
doi: 10.1049/iet-rpg.2013.0067 & The Institution of Engineering and Technology 2014
www.ietdl.org

Fig. 6 Diagram showing the convergence of the P&O algorithm in

power dispatch mode

Fig. 7 Solution when the estimated voltage is out of limit

min max
where UPVref
and UPVref are the predefined constants. The
min
lower boundary UPVref can be set as the peak voltage of 4.3 FRT capability
the AC side of the inverter plus some margin, whereas the
max
upper boundary UPVref can be set as the maximum When the penetration of PV systems is high, it is necessary
open-circuit PV array voltage in all possible operational for the PV systems not to disconnect from the grid when
min max
conditions. In the simulation, UPV = 0.4 kV and UPV = the terminal voltage of the point of common coupling
1.1 kV are selected. (PCC) is low or high. That is to say, the PV system must

Fig. 8 Responses of the PV system to a sudden change of operating conditions

a PV power output
b PV array voltage

616 IET Renew. Power Gener., 2014, Vol. 8, Iss. 6, pp. 611–620
& The Institution of Engineering and Technology 2014 doi: 10.1049/iet-rpg.2013.0067
 www.ietdl.org
Table 2 Main features of the schemes power output of the PV array cannot be fully delivered to
the utility grid, the extra power will charge the capacitor
Schemes Convergence rate Overshoot Steady error
parallel to the PV array, thus the voltage of the capacitor
P&O slow small large
increases until PPV equals the power limit of the inverter.
NQIL medium medium small As a result, UPV fails to track UPVref and all the sampling
NQIR fast large small points will converge to the same point and the interpolation
will fail when calculating (7).
To prevent the numerical error, an additional block (as
have the fault ride-through capability, which is similar to the shown in block 3 in Fig. 5) is designed, whose function is:
definition of the FRT for wind-turbines [28]. As the proposed the iteration is stopped when the low voltage is detected,
control strategy functions properly in case of overvoltage, and is reinitialised when the system recovers. In this way,
low-voltage ride-through capability is specifically concerned the three sampling points are separated and the system
in this paper. Low voltage limits the grid-injected power could return to normal state. Simulation on the FRT
because of the current protection of the inverter. When the characteristic of the PV system will be shown in the next
section.

5 Simulation results
In this section, a model of PV system is used to validate the
proposed NQI-based power control strategy. The PV system
capacity is 0.3 MW, connecting to a 10 kV external grid

Fig. 10 Responses of the PV system under real operational

Fig. 9 Responses of the PV system to reference power change with scenario
different sampling intervals a Variation of solar irradiance
a PV power output b Grid-injected power
b–d PV array voltage with respect to different sampling intervals c PV array voltage

IET Renew. Power Gener., 2014, Vol. 8, Iss. 6, pp. 611–620 617
doi: 10.1049/iet-rpg.2013.0067 & The Institution of Engineering and Technology 2014
www.ietdl.org
through a transformer. The parameters of the DC/AC inverter,
the PV array, and the PWM controller are listed in the
Appendix. The value of the sampling interval Ts is 0.01 s in
all cases unless otherwise stated. The initial operating
ref
condition is Pgrid = 0.25 MW, irradiance G = 1000 W/m2
MPP
(PPV = 0.24 MW) and temperature T = 25°C, with the PV
system operating in the MPPT mode. As the influence of
temperature on the P–V characteristic curve is trivial, it is
kept constant and only the influence of solar irradiance is
studied.
To analyse the characteristics of the proposed strategy,
three schemes are compared: (i) the NQIR, (ii) the NQIL
and (iii) the P&O. The following scenario is considered: at
ref
0.5 s, Pgrid changes to 0.175 MW, the PV system switches
from the MPPT mode (S1) to power dispatch mode (S2)
ref MPP
(Pgrid , PPV = 0.24 MW). At 1.5 s, the irradiance G
changes to 600 W/m2, the PV system switches back to the
ref MPP
MPPT mode (S3) (Pgrid , PPV = 0.144 MW). Fig. 7 plots
the simulation results, where (a) and (b) show the power
output and terminal voltage of the PV array, respectively.
From Fig. 8a, it can be observed that when the PV system
switches between the MPPT and power dispatch mode, the
convergence rate of the NQIR, the NQIL, and the P&O
schemes are much different: from S1 to S2, it, respectively,
takes the three schemes about 0.08, 0.38 and 0.48 s to reach
the new steady state; from S2 to S3, it, respectively, takes
the three schemes about 0.09, 0.27 and 0.40 s to reach the
new steady state. The enlarged views of Fig. 8 illustrate that
for the P&O scheme, the operating point oscillates around
the steady state, slightly lowering the overall efficiency of
the MPPT mode. However, this phenomenon does not exist
in the NQI schemes. Notice that for the P&O, the step size
of perturbation and the sampling time have been tuned to
obtain a good performance. A larger step size might
accelerate the convergence, but the oscillation around the
steady state will be more obvious. Moreover, the operating
Fig. 11 Responses of the PV system to three-phase short-circuit in
a MPPT mode
b Power dispatch mode
618
& The Institution of Engineering and Technology 2014
point randomly converges to the right-hand side solution in
the power dispatch mode.
It also follows from Fig. 8 that the NQI-based schemes has
larger overshoot than the P&O scheme in the transient from
S1 to S2. This is because the variation of UPVref between
two samples is small and constant for the P&O scheme,
whereas it is unpredictable and sometimes large for the
NQI-based schemes. Specifically, the overshoot for the
NQIR is more obvious than the NQIL, which coincides
with the discussion in Section 4 that the downhill section
of the P–V characteristic curve is steeper than the uphill
section. The features of the three cases are summarised
in Table 2. Of the three features, the convergence rate
is of more importance, for example, when there is
over-frequency in a microgrid, a shorter response time may
easily help to keep the system stable. From this aspect, the
NQIR has the best performance. This case proves that the
proposed strategy with NQIR can operate well in response
ref
to irradiance or Pgrid change and is more robust and
efficient than the P&O scheme.
As discussed in Section 4, the choice of the sampling
interval Ts has profound influence on the transient stability
and convergence rate. Here, the transient responses of PV
system under different sampling intervals are studied and
compared. The PV system switches from the MPPT mode
MPP ref
to the power dispatch mode (PPV . Pgrid = 0.1 MW) at
1 s. Fig. 9 shows the results of the PV system under
the NQIR scheme with Ts = 0.0005, 0.01 and 0.5 s,
respectively. It can be observed that the operating points
converge in about 0.05, 0.1 and 1.0 s for the different
sampling intervals. However, Fig. 9b shows that the
algorithm becomes unstable when Ts = 0.0005 s. The reason
can be explained as follows: in the control design, the
dynamics of the inverter is assumed to be much faster than
that of the proposed PV power control strategy and can be
ignored. So the control design can be considered as an
IET Renew. Power Gener., 2014, Vol. 8, Iss. 6, pp. 611–620
doi: 10.1049/iet-rpg.2013.0067

open-loop problem. However, when Ts is very small, the
above assumption will not hold and the dynamics of the
inverter and PV power control may undergo certain
interactions. By trial and error, it is found that Ts = 0.01 s is
a proper value to achieve the best effect in this study. This
conclusion is also proved by the enlarged view of Fig. 9c
that UPV can almost track UPVref in each sampling interval.
Next, the performance of the proposed control strategy is
studied under real operational scenario. The PV system is
simulated over a 60 s period of a typical cloudy day (real
irradiance data from [4] is used, as shown in Fig. 10a). The
sharp variation of the solar irradiance may be because of a
passing cloud, or the shade of a building. Figs. 10b and c
show the time responses of Pgrid, and UPV when the PV
system operates in the MPPT and power dispatch (Pref grid =
0.13 MW) modes, respectively. It is proved that the PV power
output can follow the change of irradiance immediately,
although some voltage and power spikes can be observed
when the irradiance changes too sharply. When the PV
system operates in the power dispatch mode, if the irradiance
is abundant, the PV power output is exactly 0.13 MW, with
the excess energy discarded by raising UPV, otherwise, the PV
system adaptively switches to the MPPT mode.
Finally, the FRT capability of the proposed strategy is
tested and analysed. Suppose that a three-phase short-circuit
fault (the short-circuit resistance is 0.003 Ω) occurs at the
PCC at 0.5 s and the fault is cleared at 1.5 s. Figs. 11a and
b, respectively, show the current in dq-frame, voltage of the
DC and AC sides of the inverter, and the grid-injected
power of the PV system when the PV system operates in
the MPPT mode and the power dispatch mode
ref
(Pgrid = 0.2 MW, G = 1000 W/m2 ). When the fault
occurs, the AC side voltage of the inverter suddenly
decrease to 0.07 kV. Id must increase to sustain the PV
power output transmitted to the grid. Therefore Idref

















increases until I ref = (Idref )2 + (Iqref )2 reaches the current
limitation IMax = 0.9 kA. Pgrid is limited to about 0.1 MW.
The surplus PV power output charges the capacitor until
UPV reaches 0.95 kV. When the short-circuit fault is
cleared, the PV system returns to normal state in less than
0.5 s. Thus, the simulation results indicate that the proposed
algorithm has the FRT capability in both the MPPT mode
and the power dispatch mode.
6 Conclusions
In this paper, a new strategy to control the power output of
PV system is presented. The strategy is based on the NQI
algorithm and it enables the PV system to adaptively
switch between the MPPT and the power dispatch mode
according to the operating condition. Namely, when the
given reference power is more than the maximum power
output of the PV system, the power output is the
maximum power output; otherwise the PV system follows
the reference power. With this control strategy, various
functions can be achieved, such as the frequency
regulation in an emergency condition, which are important
for a weak grid or a microgrid. Compared with the
conventional P&O scheme, the propose control strategy
has better performance in the respective of convergence
rate and steady-state error. The simulation results validate
the effectiveness of the proposed algorithm, even under
some severe conditions.
IET Renew. Power Gener., 2014, Vol. 8, Iss. 6, pp. 611–620
doi: 10.1049/iet-rpg.2013.0067
www.ietdl.org
7 Acknowledgements
This work is jointly supported by the National High
Technology Research and Development Program of China
(2011AA050204), the National Natural Science Foundation
of China (No. 51177146) and the Development Foundation
of the State Grid (field demonstration of the Luxi renewable
power grid).
8 References
1 Ahmed, S., Mohsin, M.: ‘Analytical determination of the control
parameters for a large photovoltaic generator embedded in a grid
system’, IEEE Trans. Sustain. Energy, 2011, 2, (2), pp. 122–130
2 Sheikh, N., Kocaoglu, D.: ‘A comprehensive assessment of solar
photovoltaic technologies: literature review’. Proc. PICMET ’11:
Technol. Manage. Energy Smart World, August 2011, pp. 1–11
3 Hua, C., Lin, J., Shen, C.: ‘Implementation of a Dsp-controlled
photovoltaic system with peak power tracking’, IEEE Trans. Power
Electron., 1998, 45, (1), pp. 99–107
4 Tan, Y., Kirschen, D.: ‘Impact on the power system of a large
penetration of photovoltaic generation’. IEEE Power Eng. Soc. Gen.
Meeting, June 2007, pp. 1–8
5 Omran, W., Kazerani, M., Salama, M.: ‘Investigation of methods for
reduction of power fluctuations generated from large grid-connected
photovoltaic systems’, IEEE Trans. Energy Convers., 2011, 26, (1),
pp. 318–327
6 Xin, H., Qu, Z., Seuss, J., Maknouninejad, A.: ‘A self-organizing
strategy for power flow control of photovoltaic generators in a
distribution network’, IEEE Trans. Power Syst., 2011, 26, (3),
pp. 1462–1473
7 Datta, M., Senjyu, T., Yona, A., Funabashi, T., Kim, C.: ‘A frequency-
control approach by photovoltaic generator in a PV–diesel hybrid power
system’, IEEE Trans. Energy Convers., 2011, 26, (2), pp. 559–571
8 Esram, T., Chapman, P.: ‘Comparison of photovoltaic array maximum
power point tracking techniques’, IEEE Trans. Energy Convers.,
2007, 22, (2), pp. 439–449
9 Kish, G., Lee, J., Lehn, P.: ‘Modelling and control of photovoltaic panels
utilising the incremental conductance method for maximum power point
tracking’, IET Renew. Power Gener., 2012, 6, (4), pp. 259–266
10 Noguchi, T., Togashi, S., Nakamoto, R.: ‘Short-current pulse-based
maximum-power-point tracking method for multiple photovoltaic-and-
converter module system’, IEEE Trans. Ind. Electron., 2002, 49, (1),
pp. 217–223
11 Femia, N., Petrone, G., Spagnuolo, G., Vitelli, M.: ‘Optimization of
perturb and observe maximum power point tracking method’, IEEE
Trans. Power Electron., 2005, 20, (4), pp. 963–973
12 Piegari, L., Rizzo, R.: ‘Adaptive perturb and observe algorithm for
photovoltaic maximum power point tracking’, IET Renew. Power
Gener., 2010, 4, (4), pp. 317–328
13 Menniti, D., Burgio, A., Sorrentino, N., Pinnarelli, A., Brusco, G.: ‘An
incremental conductance method with variable step size for Mppt:
design and implementation’. Proc. Tenth Int. Conf. Electr. Power
Qual. Utilis., September 2009, pp. 1–5
14 Ahmed, N., Miyatake, M.: ‘A novel maximum power point tracking for
photovoltaic applications under partially shaded insolation conditions’,
Electr. Power Syst. Res., 2008, 78, (5), pp. 777–784
15 Pai, F., Chao, R.: ‘A new algorithm to photovoltaic power point tracking
problems with quadratic maximization’, IEEE Trans. Energy Convers.,
2010, 25, (1), pp. 262–264
16 Li, P., François, B., Degobert, P., Robyns, B.: ‘Power control strategy of
a photovoltaic power plant for microgrid applications’. Proc. ISES
World Congr., April 2009, pp. 1611–1616
17 Delfino, F., Procopio, R., Rossi, M., Ronda, G.: ‘Integration of large-size
photovoltaic systems into the distribution grids: a P–Q chart approach to
assess reactive support capability’, IET Renew. Power Gener., 2010, 4,
(4), pp. 329–340
18 Chatterjee, A., Keyhani, A., Kapoor, D.: ‘Identification of photovoltaic
source models’, IEEE Trans. Energy Convers., 2011, 26, (3), pp. 883–889
19 Keyhani, A.: ‘Design of smart power grid renewable energy systems’
(John Wiley & Sons, 2011)
20 Yazdani, A., Dash, P.P.: ‘A control methodology and characterization of
dynamics for a photovoltaic (Pv) system interfaced with a distribution
network’, IEEE Trans. Power Deliv., 2009, 24, (3), pp. 1538–1551
21 Keyhani, A., Marwali, M., Dai, M.: ‘Integration of green and renewable
energy in electric power systems’ (John Wiley & Sons, 2009)
619
& The Institution of Engineering and Technology 2014
www.ietdl.org
22 Xue, Y., Chang, L., Kjaer, S., Bordonau, J., Shimizu, T.: ‘Topologies of
single-phase inverters for small distributed power generators:
an overview’, IEEE Trans. Power Electron., 2004, 19, (5),
pp. 1305–1314
23 Delfino, F., Denegri, G., Invernizzi, M., Procopio, R.: ‘Feedback
linearisation oriented approach to Q–V control of grid connected
photovoltaic units’, IET Renew. Power Gener., 2010, 6, (5), pp. 324–339
24 Hildebrand, F.: ‘Introduction to numerical analysis’ (Dover Pubns, 1987,
2nd edn.)
25 Sun, W., Yuan, Y.: ‘Optimization theory and methods: nonlinear
programming’ (Springer, 2006)
26 Sharma, J.: ‘A family of methods for solving nonlinearequations using
quadratic interpolation’, Comput. Math. Appl., 2004, 48, (5–6),
pp. 709–714
27 Chen, Y., Smedley, K.: ‘A cost-effective single-stage inverter with
maximum power point tracking’, IEEE Trans. Power Electron., 2004,
19, (5), pp. 1289–1294
28 Mullane, A., Lightbody, G., Yacamini, R.: ‘Wind-turbine fault
ride-through enhancement’, IEEE Trans. Power Syst., 2005, 20, (4),
pp. 1929–1937
620
& The Institution of Engineering and Technology 2014
9 Appendix
The parameters of the DC/AC inverter are: the rated capacity
0.3 MW, the rated voltages of the DC side and the AC side
0.75 and 0.25 kV, respectively.
The PV array adopts the model in [18], of which the
parameters are: short-circuit current Isc = 8.3125 A,
open-circuit voltage UOC = 32.9 V, voltage and current at
MPP MPP
MPP UPV = 26.3 V and IPV = 7.61 A, series resistance
Rs = 0.2 Ω, parallel resistance Rp = 540 Ω, diode quality
factor a = 1.3 pu, number of cells in series Ns = 30 and
number of cells in parallel Np = 40.
The parameters of the double-loop-based PWM
controller are determined by trial and error: in the
outer-loop, the d- and q-axis controls are 10 + 10/s and 1 +
10/s; in the inner-loop, the d- and q-axis controls are both
0.1 + 10/s.
IET Renew. Power Gener., 2014, Vol. 8, Iss. 6, pp. 611–620
doi: 10.1049/iet-rpg.2013.0067