Grid Parameter estimation using Model Predictive

Direct Power Control

Bilal Arif*, Luca Tarisciotti, Pericle Zanchetta, Jon Clare, Marco Degano
Department of Electrical and Electronics Engineering,
The University Of Nottingham, Nottingham, UK
*[email protected]
Abstract – This paper presents a novel Finite Control Set literature for grid connected converters; some solutions
Model Predictive Control (FS-MPC) approach for grid- implement just a current control [12], [13], [14], while others
connected converters. The control performance of such investigate a more complete direct power control [5], [15],
converters may get largely affected by variations in the supply [16]. In [5],[15], the authors propose a Model Predictive
impedance, especially for systems with low Short Circuit Ratio Direct Power Control (MP-DPC) approach; it allows to
(SCR) values. A novel idea for estimating the supply eliminate the external PI based DC-link voltage controller,
impedance variation, and hence the grid voltage, using an hence avoiding windup issues or linearization of the system
algorithm embedded in the MPC is presented in this paper. for proper tuning of the PI controllers. However, as any
The estimation approach is based on the difference in grid
model based control technique, MP-DPC is sensitive to
voltage magnitudes at two consecutive sampling instants,
model parameter variations. In Figure 1, while the converter
calculated on the basis of supply currents and converter
voltages directly within the MPC algorithm, achieving a fast inductance, Lc, and its parasitic resistance, r, are usually
estimation and integration between the controller and the known, the grid parameters (usually inductance dominant,
impedance estimator. The proposed method has been verified, Ls), are unknown and can have highly varying values
using simulation and experiments, on a 3-phase 2-level depending on the grid load conditions. Since MP-DPC is
converter. based on the knowledge of model parameters, any small
variations in these parameters will disturb performance and
I. INTRODUCTION stability of the control system.
The extended use of Renewable and Distributed Variation of transmission line inductance is very
Generation (DG) systems in the past decade is widely common and is due to environmental changes, long distance
contributing to modify the old concept of electrical transmission cables or presence of harmonics generated by
distribution network towards an “active” model where power variable grid loads, such as DG systems. This variation in
can flow in any direction [1]. This has demanded an supply inductance also affects the ripple on the measured
improvement in power electronics systems and control in voltage at the Point of Common Coupling (PCC). A
order to implement systems grid interface, effective and mismatched impedance value alongside a distorted PCC
intelligent power flow control and to avoid grid instability voltage, can affect the predictive control performance. Thus,
[2]. The application of power electronics converters and in it is important that the variation in supply inductance is taken
particular grid connected technology or Active Front End into account in the control implementation. Also it would be
(AFE) has become a key enabling technology in all a great advantage if a good online estimation of the supply
renewable and distributed generation systems like, for inductance, Ls, is provided, using that to update the total AC
example, photovoltaic and wind power generation systems side inductance value in the MP-DPC. Several methods have
[1], [3]. AFE can be also used as active filters where they are been proposed in literature for estimation of the supply
connected in parallel or series to a non-linear load and offer impedance with both offline and online implementations. In
reduction in harmonics production [4], [5]. In these scenarios [17], using a Fast Fourier Transform to calculate different
the converter control is a key element to achieve optimal harmonic impedances, authors used a voltage transient
generation systems grid interface, the required power flow injection at PCC resulting in a transient variation in current
and to avoid grid instability [6]. In contrast with traditional and voltage at the PCC. However, the use of an online FFT
control techniques applied to AFE, such as Voltage Oriented algorithm requires a high computational effort. Other grid
Control (VOC) [7] and Direct Power Control (DPC) [8], impedance estimation methods include variations of active
Finite Control Set Model Predictive Control (MPC) has been and reactive powers [18], or uses a Virtual-Flux based
recently proved to represent a promising solution to control control method to estimate inductance, [19]. An analytic
of power electronics converters. In fact, it provides several approach for estimation of the coupling inductance in direct
advantages including better dynamic response, flexible power control of active rectifiers is instead proposed in [20]
control action and digital implementation [9], [10], [11]. In and [21]. The proposed method works on the principle of
particular, several MPC strategies have been proposed in assuming an equal grid voltage magnitude at two
Manuscript received November 18, 2014; revised March 12, 2015; revised
May 21, 2015, accepted June 25, 2015. Paper 2014-IPCC-0846.R2,
presented at the 2014 IEEE Energy Conversion Congress and Exposition,
Pittsburgh, PA, September 14–18, and approved for publication in the IEEE
TRANSACTIONS ON INDUSTRY APPLICATIONS by the IEEE-IAS
Industrial Power Converter Committee of the IEEE Industry Applications
Society.
consecutive sampling instants, [21], [24], and it is integrated the MPC works in discrete-time, the following equations
within the MPC algorithm. The supply resistance is not represent the discrete-time model of the continuous-time
considered in this work, since its effect can be considered system given in (1) and (2):

' ∙() ∙* 3
negligible with respect to the supply inductance which
heavily affects the magnitude and phase of the grid voltage, $%& $ $ $
thus reducing the performance of the MPC. However, it is
( , * 1 ∙ , 0 1 , 3 45, 67 (4)
always possible to expand the proposed estimation method to ,- ,- ,-
include the supply resistance estimation. . . 1∙2

$
∙ 8 ) ∙0 5
The combination of a finite set MP-DPC and the
$%& $
!
proposed estimation algorithm results in a high-bandwidth
control robust to supply impedance variations without
increasing excessively the complexity of the control system.
The predictions are made for a future sampling instant
II. MODEL PREDICTIVE - DPC APPROACH k+1, using the information provided at the present instant k.
In the practical implementation, due to the calculation time
The equivalent circuit in α β reference frame calculated
required by the digital control system, a delay of one
using [25], of the AC side of the AFE system in Figure 1 is
sampling instant is taken into consideration by making
shown in Figure 2.
predictions at instant k+2. Hence the model becomes:

$%:
' $%& $%&
∙() $%&
∙* 6

$%&
$%:
∙ 8 ) $%&
∙0 7
< =
!

where, vck+1 and idck+1 are calculated using the optimized

switching signals from the predictive controller at the
previous sample interval k. Having the supply current
Figure 1. Grid-connected active front-end rectifier
predictions at k+2, the active and reactive powers,
predictions respectively at this sampling instant are

3
?< $%:
@' $%:
∙ $%:
)' $%:
∙ $%:
C 8
=
2 A A B B

3
E< $%:
' $%:
∙ $%:
' $%:
∙ $%:
9
=
2 B A A B

Equations (7)-(9) are calculated for every possible

converter states; between all of them, the state to apply is
Figure 2. Equivalent AFE model in α and β reference frames selected using a user defined cost function, which represents
DC-Link voltage, Active and Reactive power errors. The
The continuous-time representation for this equivalent cost function and the required references calculation are
circuit is given in (1), while the continuous time model for described in the following subsections.
the system DC side is given in (2).
A. DC-link voltage reference
1
∙ ∙ 1
The method proposed in [5] and [15] was adopted to
obtain suitable references for the active power and the dc-
link voltage. The reference dc-link voltage is calculated

2
using (10) where N* is the reference prediction horizon. A
! "∙!
suitable selection of N* results in a better dynamic response
of the dc-link voltage.

1
In (1)-(2) vgrid, vc and Vdc are, respectively, grid, converter
$%& $
) @ $ $
C 10
=G
H∗ =G
and DC-link voltage; is and idc are supply and converter DC
current, the total inductance L, is the sum of converter filter
inductance LC and supply inductance LS, r is the converter
input filter resistance, R is the load resistance and C is the Since the DC-link voltage reference does not vary
dc-link capacitance. In order to obtain the desired prediction significantly at two consecutive sampling instants, the DC-
of the supply current, the current derivative is approximated link voltage reference at k+2 is assumed approximately
using the Euler forward method as mentioned in [11]. Since equal to the reference at k+1, i.e. Vdc-refk+2 ≈ Vdc-refk+1.
B. Power reference C. Cost function
The reactive power reference Qrefk+2 is kept to 0 VAR to At each sampling interval, the 8 possible switching
obtain unity power factor operation. Thus, without using any combinations (6 active vectors and 2 zero vectors) are used
Phase-Locked-Loop (PLL), as is the case in current control to evaluate the cost function. The switching combination that
where the grid phase angle estimation is necessary for gives the minimum cost function value is selected and hence
current reference generation, the supply voltage and current applied. Similarly, in the next sampling interval the process
are synchronized by maintaining a close to unity power repeats. Using this approach, the use of a modulator to
factor by directly controlling active and reactive powers. generate the converter switching states is no longer required
Moreover, under low SCR scenarios, re-tuning of PLL as in [7], [8] or [14]. It is possible, however, to achieve the
algorithms may have computational constraints, adding more same switching combination to be applied at two or even
complexities to the control structure. Therefore, assuming three consecutive sampling intervals, thus resulting in a
unity power factor, the active power reference is calculated variable switching frequency for the converter. The cost
using the system power balance equation [15] as shown in function, G, used for the proposed MP-DPC, is shown in
(11) where, Ps is the supply active power, Pr is the power (18).

S&
loss on the input filter resistance and Pload is the active power
R U $%: $%:
U
across the capacitor and resistor at the DC side.
=G < =
MT=
? ? ) ?KLM 11
S:
) U? =G $%: ?< $%:
U
For a balanced and undistorted 3-phase system with a unity ? MT= =
power factor, it can be assumed that
SV
3 ) UE $%:
E< $%:
U 18
? ' $%& $%&
12 ? MT= =G =
2
3 where, S& , S: and SV are the weighting factors for the 3
:
? @ $%&
C 13
2
terms in the cost function. Vdc-rated and Prated are used as
normalizing factors where Vdc-rated is the rated dc-link voltage
where $%& and ' $%& and Prated is the apparent power. In [11], the authors
are the predicted supply current emphasize that special attention must be paid while
and estimated grid voltage obtained from (3) and (36), designing the weighting factors; however, there is not a
respectively. Therefore, (11) can be expressed in a straightforward approach through which an accurate value
discretized form in terms of grid voltage and supply current, for these parameters can be selected. Since the weighting
with regards to Figure 1, as shown in (14). factors selection is an on-going research topic, their values

' $%&
2?KLM $%&
are mostly designed based on empirical procedures.
$%& : $%&
) 0 14
3
III. DEAD-TIME COMPENSATION

Solving (14) for isk+1 and multiplying by '

In order to adapt the proposed inductance estimation
$%& method to practical converter implementation, the presence
, the
active supply power reference is hence obtained in (17), of dead-times in the devices switching needs to be taken into
where Pload is the load power and idck+1 is the rectified current account. In fact, as described in section IV, the proposed
defined, respectively, by (15) and (16). method considers the switching signals for each leg of the
converter and the dc-link voltage. However, switching dead-
?KLM $%& $%&
∙ =G
$%&
15 times (Td region in Figure 3), considered equal to 2µs in this
work, results in a voltage drop which affects the estimation
$%& $ $
method. Therefore, this voltage drop needs to be
!∙ ) 16
$%& =G
8 "
compensated. A dead-time compensation method has been
included in the model predictive control by incorporating an
additive term representing the voltage drop due to the dead
$%& :
3 ' 8 ?KLM $%&
∙
time Td. The voltage loss in the ‘Td’ region for interval Sk for
? =G $%: ∙ ∙ O1 P1 Q 17
4 3 ' $%& :
leg A is expressed as

1 8
WX , ∙ W∙ ∙ 2YX$ YZ$ YW$ 19
As it can be noted, the expression of the active power 3 8 & & &

reference requires an appropriate grid voltage estimation

algorithm, which is described in section V and takes In the same way the voltage in the ‘Ts – Td’ region for
advantage of the inductance estimation algorithm described interval Sk is expressed as follows.
in section IV.
 1 8 8
WX , ∙ W∙ ∙ 2YX$ YZ$ YW$ 20
seem to be significantly affected by high frequency
,
3 8
components of grid impedance, thanks to its inherent
robustness. Therefore, it is reasonable to assume a low
frequency grid impedance model [26]. Referring to Figure 5,
Therefore, the total converter voltage applied during the the estimation approach works on the principle of assuming a
sampling interval Sk is now expressed as constant grid voltage magnitude between two consecutive

WX[$ WX , ) WX , 21
sampling instants. Although there will be a phase shift of
, ‘ω·ts,’ the magnitude of the grid voltage vector will,
however, not vary significantly.
Similarly, (19)–(21) are applied for the other legs with
reference to the direction of the supply currents in the
respective legs [23]. The converter voltages are hence
transformed to their respective α and β components to be
used in the current predictions for (3) and (6) and for the
inductance estimation method, (24) and (25).

Figure 5. Grid voltage representation in α-β.

In α-β reference frame, the square of the grid voltage

magnitude at the instant k is expressed as
Figure 3. Dead-time in top switch leg-A
$ : $ : $ :
IV. INDUCTANCE ESTIMATION METHOD U U @ A C )@ B C 22
The estimation method proposed in this paper estimates
According to Figure 2 and Figure 5, the grid voltage
the total inductance, i.e. supply inductance plus converter
magnitude at the time instant k can be expressed using the
inductance, and feeds the estimated value as an update to the
value of vgridk extracted from the system model in α-β:
model based predictive controller, as shown in Figure 4.
Using the total estimated inductance value, we can easily :
$ :
find the variation in supply inductance, LS, and make an @ C ∙ ) ∙ $
) $
23
estimation of the supply/grid voltage inside the controller.

Hence, the square of the grid voltage magnitude at instant

k is expressed as:

:
$ :
U U ∙ ) ∙ )
A $ $
A A
:
)\ ∙ ) ∙ ) ] 24
B $ $
B B

Similarly, the square of the grid voltage at the previous

instant, k – 1, is equated as:

:
$ & :
U U ∙ ) ∙ )
A $ & $ &
A A
:
)\ ∙ ) ∙ ) ] 25
Figure 4. Block scheme of proposed MPC with online inductance and B $ & $ &
grid voltage estimation B B

According to [26], since the voltage harmonic magnitude

decreases as the frequency increases, the effect of high
frequency harmonics on the grid impedance are often limited vc-α and vc-β in the equations above are calculated using
in power systems. Moreover, the predictive control does not the optimized switching signals from (18) and the measured
 $%& $
f[ ∙
dc-link voltage. After discretizing the current derivatives,
' ) 36
$ A,B A,B $
A,B
8a < A,B
(24) and (25) are solved for the value of the total inductance
L such that:
ijk i
$ : $ & :
U U U U 0 26 -g,h -g,h
,
where, is the discretized supply current

at PCC. Using the ' l

derivative and vpccαβk is the actual measurement of the voltage
$
A,B and [ in (3) and (6) for current
This resulted in the quadratic equation of (27), with the
terms A, B and C expressed, respectively, by equations (28)- predictions, it is possible to apply the previously described
(30) with the single terms defined in (31)- (33). MP-DPC in a way that the resultant control action takes into
:
∙^) ∙_)! 0 27
account supply impedance variations and the system
performance is not affected. In cases when the grid has a low
: :
^ ∆^A& :
) @∆^B& C ∆^A: :
@∆^B: C 28
SCR, i.e. very high grid inductance, a phase shift between
the supply current and vpcc is present in addition to the

_ 2 ∙ @∆_A& ∙ ∆^A& ) ∆_B& ∙ ∆^B&

distorted vpcc. Moreover, the fundamental component of vpcc
does not accurately approximate vgrid because of the variable
∆_A: ∙ ∆^A: ∆_B: ∙ ∆^B: C 29 grid inductance, thus resulting in a phase and magnitude
error with respect to vgrid. Therefore, estimating the grid
: :
! ∆_A& :
) @∆_B& C ∆_A: :
@∆_B: C 30
voltage is much more efficient and suitable, particularly in a
low SCR scenario.

$%& $ $%& $
∆^A& , ∆^B& ,
A A B B
VI. SIMULATION RESULTS
8a 8a The proposed MP-DPC is applied to the three-phase two
level AFE of Figure 1 with the system parameters reported in
$ $ & $ $ &
∆^A: , ∆^B: 31
A A B B
table I. Simulation results obtained in Matlab-Simulink
8a 8a
environment are shown and discussed in this section.

∆_A& $
) , ∆_B& $
) 32
SIMULATION PARAMETERS
$ $
TABLE I.
A 2A B 2B
Parameter Symbol Value

∆_A: A
$ &
) 2A
$ &
, ∆_B: B
$ &
) 2B
$ &
33
PCC voltage vPCC 100Vrms
Sample time Ts 50µs
Converter filter inductance Lc 4.5mH
After substituting (28)-(30) in (27), the total inductance is Converter filter resistance r 0.4Ω
Active Power P 2400W
evaluated as
S1 , S2 , S3
DC-link capacitance C 2200µF

FSW
Weighting factors 1.5, 1, 1
1 _ 4∙!∙^
∙ ∙ d 1 ) P1 e 34
Average switching frequency 9.7 kHz

= T bMT Lc
2 ^ _: Although the converter has an average switching
frequency, [11], of ~9.7 kHz, the sample time used for the
predictive control is fixed at 20 kHz. Figure 6 shows the
The estimated value of inductance in (34), once excluded estimated inductance and its reference value, where the
the negative root, is used as an update to the inductance method has been tested in the extreme condition of a
value in the current predictions of (3) and (6). frequent variation in LS. The estimation algorithm evaluates
the total inductance quite accurately even when the supply
V. GRID VOLTAGE ESTIMATION inductance LS has been varied in order to reach a value
If the voltage at the point of common coupling vpcc greater than 200% of LC.
presents substantial distortion and it is used in the model, it 16
Reference Estimated
will in turn induce distortion on the current predictions. We 15

need therefore an estimation of the grid voltage. Once the 14

total inductance has been estimated the supply inductance LS 13

12
is estimated as:
Inductance (mH)

f[ 35
11

= T bMT Lc 2
10

where, the value of LC is usually known since it is the 7

converter input inductance. Referring to the block diagram in 6

Figure 4, the supply voltage in αβ reference frame used for 5

0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
the predictive controller is corrected using the estimation in Time (s)

(35): Figure 6. Estimated Total Inductance with a variation in LS , LC =4.5mH

 In terms of system SCR, results are shown in table II, When the LS is 0.5mH, the current THD is ~5.48%. Due
where the SCR has been defined as [27]: to the large mismatch between the real and the model value

Yℎq W Wr sqtu (Y!?)

of LS at 0.09s because of the step change, the control and
Y!"
estimator performance gets affected, boosting the THD to
(37) 22.35%. However, once the estimator has reached a steady
_Xau ?qtu (Yvw[x )
state within a fundamental period, the corrected values of LS
The estimation method proposed in this paper works and grid voltage largely improve the control performance
effectively for SCR values larger than 2, at which the system achieving a current THD of ~3.66%.
starts showing an unstable behavior. Table III shows the THD of the supply current for
different variations in LS when the grid voltage has also been
TABLE II. SCR VALUES FOR DIFFERENT LS VALUES estimated. Once the LS is correctly estimated, the increase in
LS LC ZS (Ω) SCP SBASE SCR the total inductance on the AC side obviously improves the
(mH) (mH) (kVA) (kVA) quality of the current waveform, as it can be seen from the
1.0 4.5 0.314 95.47 2.4 39.8 Stiff grid reduction of THD values from 5.48% (LS = 0.5mH) to 3.18%
2.0 4.5 0.628 47.77 2.4 19.93 (LS = 5.0mH) in the case of online estimation. Thus,
4.0 4.5 1.257 23.86 2.4 9.954 presenting highly improved performance.
6.0 4.5 1.885 15.91 2.4 6.637
8.0 4.5 2.513 11.94 2.4 4.98 TABLE III. CURRENT THDS : ESTIMATION TO VARIATION OF LS
10.0 4.5 3.142 9.55 2.4 3.98 Weak grid
12.0 4.5 3.770 7.96 2.4 3.32 LS (mH) LC (mH) THD (%)
0.5 4.5 5.48
Figure 7 shows total inductance and grid voltage 1.0 4.5 4.93
estimation results for the case when LS is 3.0mH and the 2.0 4.5 4.29
3.0 4.5 3.76
converter input filter inductance is 4.5mH. The same value 4.0 4.5 3.39
of LS has been used in the experimental results section. The 5.0 4.5 3.18
supply voltage used for the predictive controller is corrected
by using the estimated value of the supply inductance where VII. EXPERIMENTAL RESULTS
the PCC voltage and grid voltage estimation have a THD of Experimental tests have been carried out using the
~18% and 1.16% respectively. laboratory setup as shown in Figure 9. The experimental
implementation uses circuit configuration and data as in
Voltage (V), Current*3 (A)

150 Vpcc V-grid estimation Supply Current

100
simulation.
50
0
Interface PC Resistive Load
-50
-100
-150
0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0.11 PCC
10
Lreference Lestimation
‘Grid Inductor’
Inductance (mH)

5
0.02 0.03 0.04 0.05 0.06 0.07
Time (s)
0.08 0.09 0.1 0.11 0.12
DSP/FPGA
and AFE Power Supply
Figure 7. Actual vPCC, vgrid-estimation, current, L-est. LS = 3.0mH Protection
Figure 8 shows the supply current and estimated grid Converter Input
voltage for a step variation of 3.0mH in LS from 0.5mH to Inductance
3.5mH at 0.09s. VARIAC
Vpcc V-grid estimation Supply Current
150

100
Voltage (V), Current*3 (A)

50
Figure 9. 3-Phase 2-Level AFE laboratory setup

0
The control system is composed of a main board
featuring a TMS320C6713 digital signal processor (DSP)
-50
with 225MHz clock frequency and an auxiliary board
-100
equipped with a field programmable gate array (FPGA)
THDi = 5.48% THDi = 22.35% THDi = 3.66% ActelProASIC3A3P400 used for data acquisition and PWM
-150
0.06 0.07 0.08 0.09 0.1 0.11 0.12 0.13 0.14
generation with 50MHz clock frequency. The experimental
Time (s)
tests have been carried out using a controlled pure sinusoidal
Figure 8. vGRID estimation and Current– LS= 0.5mH 3.5mH
voltage power supply manufactured by Chroma
(Programmable AC Source 61511) and also using a 10(e) where the amplitude of current increases due to an
VARIAC (Variable Voltage Auto-Transformer CMV20E-3) increase in dc-link voltage. Despite this large variation in dc-
connected with the mains. In figure 9 the experimental rig is link voltage reference, the grid voltage shows a good
powered through a VARIAC autotransformer, which is estimation even though the PCC voltage remains distorted.
connected to a power supply protection circuit in case of The supply current shows a small variation in THD from
over-current or short-circuits in the system. The power 5.03% before the step to 5.96% after the step in dc-link
supply protection is then connected to a three phase inductor voltage reference. Figure 10(f) shows the oscilloscope results
emulating the unknown grid inductance, shown as LS in for CHROMA supply voltage, supply current and the dc-link
figure 4, with 3mH per phase. Measurements of supply voltage to a variation in dc-link voltage reference. The
voltages for data processing are taken at the point of increase in the supply current amplitude and also the dc-link
common coupling, PCC, after the grid inductance. A voltage can be seen in figure 10(f). When figure 10 is
converter inductance of 4.5mH per phase, LC as shown in compared with the simulation results of figure 7 an increase
figure 4, is connected between the PCC and the AFE input of ~1.5% in the supply current THD value can be noted. This
and is known to the converter control. The AFE is controlled is caused by parameter uncertainties like, for example, power
using the above mentioned DSP/FPGA structure connected switching device voltage drop that have not been taken into
to an interface PC, as shown in figure 9. The estimation account in the simulation. The vpcc and vgrid-est have ~17% and
algorithm is executed in c-code in the DSP and run following ~2.7% THD values, respectively. Figure 11 shows the results
the procedure presented in (28)-(33) where, after calculation for the case when the system is supplied by the mains
of each of the terms ∆^A& , ∆^A: , ∆^B& , ∆^B: , ∆_A& , ∆_A: , through a VARIAC. At time 0.55s, a step in input voltage
∆_B& and ∆_B: in (31)-(33), they are substituted in (28)-(30) from 33V to 100V was introduced using the VARIAC. This
for calculation of the terms A, B and C. The terms A, B and operation changes the VARIAC built-in inductance
C, are then substituted in (34) to obtain the estimated value generating an increase in the total system input inductance
of total inductance. The total computation time required by from 7.5mH to approximately 8.3mH as estimated by the
the algorithm, including control and estimation routines, is algorithm and shown in Figure 11.
~43µs for an interrupt time of 50µs. This means that the mismatch between the supply
The experiments for the proposed inductance estimation inductance and the converter input inductance is increased
have been carried out both in steady state and transient from 66.67% to 85%. Moreover, it has been noted that,
conditions. That includes a 500VAR variation in reactive considering the small sampling time of the
power reference, 35V variation in DC-Link voltage reference control/estimation algorithm with respect to the dynamics of
and also a 67V variation in the supply voltage when the eventual variations of Ls or supply voltage in a real
VARIAC is used. The estimation results show a good scenario, imposing constant magnitude of vgrid and
behavior even in the presence of these variations. Figure 10 considering Ls constant between two sampling intervals,
shows an experimental test where the Chroma voltage supply does not cause any uncertainty in the estimation process.
was used. In Figure 10 (a), the reactive power was varied
from 0VAR to 500VAR and the system is required to Due to laboratory safety reasons, the tests were not
estimate a supply inductance of LS=3.0mH for a total line conducted for higher supply inductance values; however, the
inductance of L=7.5mH. The reactive power follows the simulation and practical results shown support the proposed
requested change and the inductance estimation responds methodology. The effect of capacitance on the transmission
very well to this condition too. However, a very small steady line has not been taken into consideration in the proposed
state error of ~50VAR can be seen on the reactive power. estimation algorithm. Since the effect of capacitance appears
This is due to the presence of ripple on the estimated at high frequency, its presence will not have a significant
inductance, effect of mutual inductance on the transmission effect on the estimation algorithm since the fundamental
line, other parameter uncertainties and model discretization frequency of the power system is 50Hz.
errors that have not been compensated. The corresponding VIII. CONCLUSION
results in Figure 10(b) show the distorted measured PCC
voltage, the estimated grid voltage and the quasi-sinusoidal In recent years, grid-connected converters have gained
supply current with a THD value of 5.31%. Figure 10(c) popularity particularly for renewable energy systems
shows the oscilloscope results for the CHROMA supply integration, where in some cases the grid is not stiff, like
voltage, supply current and the dc-link voltage when a weak micro grids. Small grid parameter variations in these
reactive power variation of 500VAR occurs. Figure 10(d) systems can substantially affect the performance of the
shows the test results and estimation results in the case of a converter control. This grid parameter variation can be
35V step in the dc-link voltage reference. The dc-link regarded as a variation in the grid impedance, which is
voltage responds to this reference variation and so does the mostly dominated by variation in grid inductance. Therefore,
estimated inductance for the same LS and LC values as for this paper presents a novel MP-DPC algorithm where
reactive power variation. Since in an active front end the variations in grid inductance are estimated online and used to
exchange of power is between the AC side and DC side, update the inductance value in the control system. The
increasing the dc-link voltage reference essentially means the increase in vpcc ripple along with the variation in supply
supply current is increased as well. This is shown in Figure inductance can greatly affect the predictive control
performance. Estimating the variation in grid inductance adaptable in the case of any topology of grid connected
allows an estimation of the grid voltage inside the controller converters, PV applications, variable load on the distribution
so that the vpcc can be updated at each sampling interval for a network or in the case of a different grid connection like for
better quality operation. Though the system in Figure 1 has example the use of LCL filters. The estimation approach has
been tested considering vgrid is from a low voltage substation, been integrated within the model based predictive control,
the proposed algorithm can be easily adapted and modified thus making the estimation and control effective and
to be used for general supply impedance estimation, medium efficient. Simulation and experimental results support the
or high voltage applications. The algorithm is also easily proposed methodology.
15 15
Inductance (mH)

Inductance (mH)
Estimated Inductance Reference Inductance Estimated Inductance Reference Inductance
10 10

5 5

0 0
0 0.5 1 1.5 2 2.5 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 4
1.5 240
Reactive Power Reactive Power Reference DC-Link Voltage DC-Link Reference

Voltage (V)
1 220
Q(kVar)

0.5
200
0
180
-0.5
0 0.5 1 1.5 2 2.5 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 4
Time (s) Time (s)

(a) (d)
150
Voltage (V), Current*3 (A)

150 Vpcc-a Vg-est-a is-a VDC* variation Vpcc-a Vg-est-a is-a

Q* variation
Voltage (V), Current*3 (A)

100
100

50 50

0 0
-50
-50
-100 THDi = 5.31% THDi = 5.62% THDi = 5.03% THDi = 5.96%
-100
-150
1.31 1.32 1.33 1.34 1.35 1.36 1.37 1.38 1.39 1.4 3.08 3.1 3.12 3.14 3.16 3.18
Time (s) Time (s)

(b) (e)
Q* variation VDC* variation

Effect of a positive variation in Q* Effect of a variation in VDC*

variation

(c) (f)
Figure 10. (a) Estimated Total Inductance and reactive power with a variation in Q*, (b) PCC Voltage, Grid Voltage estimation, Supply Current with a
variation in Q*. LC = 4.5mH, LS = 3.0mH, (c) Scope result for CHROMA voltage (yellow), supply current (pink) and dc-link voltage (blue) with a variation
in Q*, (d) Estimated Total Inductance and VDC with a variation in VDC*, (e) PCC Voltage, Grid Voltage estimation, Supply Current with a variation in VDC*.
LC = 4.5mH, LS = 3.0mH, (f) Scope result for CHROMA voltage (yellow), supply current (pink) and dc-link voltage (blue) with a variation in VDC*
15 Vpcc-a Vg-est-a is-a
Estimated Inductance
Voltage (V), Current*3 (A)

40
Inductance (mH)

10 20
A
0
5
-20 THDi = 8.3%
0 -40
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0.29 0.3 0.31 0.32 0.33 0.34
Time (s)
Voltage (V), Current*3 (A)

Vpcc-a Vg-est-a is-a

100 B
Vpcc-a Vg-est-a is-a
50 100
A
Voltage (V), Current*3 (A)

0 50 B
0
-50
-50
-100 THDi = 6.3%
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 -100
Time (s) 1.17 1.18 1.19 1.2 1.21 1.22
Time (s)

(a) (b)
Figure 11. (a) Estimated Total Inductance, PCC voltage, grid voltage estimation and supply current with a variation in VARIAC, (b) Zoomed in PCC
Voltage, Grid Voltage Estimation, Supply Current
 REFERENCES [19] P. Antoniewicz, M. P. Kazmeirkowski. “Virtual-flux-based
predictive direct power control of ac/dc converters with
[1] J. R. Rodriguez, J. W. Dixon, J. R. Espinoza, J. Pontt, P. online inductance estimation”, IEEE Trans. Electron.,
Lezana. “PWM regenerative rectifiers: state of the art”, IEEE volume 55, no. 12, pp. 4381-4390, Dec. 2008.
Trans. Ind. Electr., volume 52, no. 1, pp. 5-22, (Feb. 2005).
[20] J. G. Norniella, J. M. Cano, G. A. Orcajo, C. Garcia, J. F.
[2] J. N. Reddy, M. K. Moorthy, D. V. A. Kumar. “Control of Pedrayes, M. F. Cabanas, M. G. Melero. “Analytic and
grid connected PV cell distributed generation systems”, IEEE iterative algorithms for online estimation of coupling
TENCON 2008 Conference, pp. 1-6, (Nov. 2000). inductance in direct power control three-phase active
[3] V. S. Tejwani, P. V. Sandesara, H. B. Kapadiya, A. M. Patel. rectifiers”, IEEE Trans. Electron., volume 26, no. 11, pp.
“Power electronic converter for wind power plant”, ICCEET 3298-3307, Nov. 2011.
2012 Conference, pp. 413-423, (March. 2012). [21] J. G. Norniella, J. M. Cano, G. A. Orcajo, J. F. Pedrayes, M.
[4] S. Kwak, H. A. Toliyat. “Design and rating comparisons of F. Cabanas, M. G. Melero. “New strategies for estimating the
PWM voltage source rectifiers and active power filters for coupling inductance in grid-connected direct power control-
AC drives with unity power factor”, IEEE Trans. Power based three-phase active rectifiers”, IEEE PES., pp. 1-5, July.
Electr., volume 20, no. 5, pp. 1133-1142, (Sept. 2005). 2013.
[5] P. Zanchetta, P. Cortes, M. Perez, J. Rodriguez, C. Silva. [22] M. Ciobotaru, R. Teodorescu, F. Blaabjerg. “A New Single-
“Finite state model predictive control for shunt active filters”, Phase PLL Structure Based on Second Order Generalized
IECON 2011- 37th Annual Conference on IEEE Industrial Integrator”, PESC ’06, pp. 1-6, June. 2006.
Electronics Society, pp. 581-586, (Nov. 2011). [23] Seon-Hwan Hwang, Jang-Mok Kim. “Dead-time
[6] K. Ilango, P. V. Manitha, M. G. Nair. “An enhanced Compensation method for Voltage-Fed PWM inverter”,
controller for shunt active filter interfacing renewable energy Energy Converter, IEEE Trans., volume 25, no.1, pp. 1-10,
source and grid”, IEEE Third Intl. Conf. on ICSET, pp. 305- March. 2010.
310, (Sept. 2012). [24] B. Arif, L. Tarisciotti, P. Zanchetta, J. Clare, M. Degano,
[7] M. P. Kazmeirkowski, F. Blaabjerg, R. Krishnan. “Control in “Integrated grid inductance estimation technique for Finite
power electronics, selected problems”, Elsevier science, pp. Set Model Predictive Control in grid-connected converters”,
444, ISBN : 0-12-402772-5. USA 2002. ECCE 2014, IEEE, pp. 5797-5804, Sept. 2014.
[8] M. Malinowski, M. P. Kazmeirkowski, A. M. [25] M. M. Canteli, A. O. Fernandez, L. I. Enguiluz, C. R.
Trzynadlowski. “A comparative study of control techniques Estebanez, “Three-phase adaptive frequency measurement
for PWM rectifiers in AC adjustable speed drives”, IEEE based on Clarke’s transformation”, Power Delivery, IEEE
Trans. Power Elect., volume 20, no. 5, pp. 1133-1142, Sept. Trans., volume 21, no. 3, pp. 1101-1105, July 2006
2005. [26] Gu Herong, Guo Xiaoqiang, Wang Deyu, Wu Weiyang,
[9] S. Vazques, J. I. Leon, L. G. Franquelo, J. Rodriguez, H. A. “Real-time grid impedance technique for grid-connected
Young, A. Marquez, P. Zanchetta, “Model Predictive power converters”, Ind.Electr., ISIE 2012, pp 1621-1626,
Control: A review of Its Applications in Power Eectronics”, May 2012
IEEE Magazine, vol. 8, no. 1, pp. 16-31, March 2014 [27] Soenke Grunau and Friedrich W. Fuchs, “Effect of Wind-
[10] P. Zanchetta, J. Rodriguez, M. P. Kazmeirkowski, J. R. Energy Power Injection into Weak Grids”, EWEA 2012
Espinoza, H. A. Abu-Rub, H. A. Young, C. A. Rojas. “State
of the art of finite control set model predictive control in
power electronics”, IEEE Trans. Informatics, volume 9, no.2,
pp. 1003-1016, May. 2013.
[11] J. Rodriguez, P. Cortes. “Predictive control of power
converters and electrical drives”, John Wiley & Sons, Ltd,
ISBN : 978-1-119-96398-1, 2012.
[12] J. Rodriguez, J. Pontt, C. A. Silva, P. Correa, P. Lezana, P.
Cortes, U. Amman. “Predictive current control of a voltage
source inverter”, IEEE Trans. Elect., volume 54, no.1, pp.
495-503, Feb. 2007.
[13] P. Correa, J. Rodriguez, I. Lizama, D. Andler. “A predictive
control scheme for current-source rectifiers”, IEEE Trans.
Elect., volume 56, no.5, pp. 1813-1815, 2009.
[14] M. P. Kazmeirkowski, L. Malesani. “Current control
techniques for three-phase voltage-source pwm converters: A
survey”, IEEE Trans. Elect., volume 45, no.5, pp. 691-
703,Oct. 1998.
[15] D. E. Quevedo, R. P. Aguilera, M. A. Perez, P. Cortes.
“Finite control set MPC of an AFE rectifier with dynamic
references”, IEEE Technology, pp. 1265-1270, March. 2010.
[16] P. Cortes, J. Rodriguez, P. Antoniewicz, M. P.
Kazmeirkowski. “Direct power control of an afe using
predictive control”, IEEE Trans. Elect., volume 23, no.5, pp.
2516-2523, Sept. 2008.
[17] M. Sumner, B. Palethorpe, D. W. P. Thomas, P. Zanchetta,
M. C. Piazza. “A technique for power supply harmonic
impedance estimation using a controlled voltage
disturbance”, IEEE Trans. Electron. volume 17, no. 2, pp.
207-215, March 2002.
[18] A. V. Timbus, P. Rodriguez, R. Teodorescu, M. Ciobotaru.
“Line impedance estimation using active and reactive power
variations”, IEEE Power Electron., pp. 1273-1279, June
2007.

9
 Bilal Arif completed his undergradute
degree (BEng (hons.)) from the University
of Nottingham in the field of Electrical
and Electronics engineering in July of
2011. In September 2011, Bilal started his
Ph.D. study with the Power Electronics
Machines and Controls (PEMC) group at
the University of Nottingham, UK.
Currently he is in the final stages of his
Ph.D. and his research interests include
control strategies for grid connected
converters with special emphasis on grid impedance estimation
techniques for cases when power converters are connected to weak
grids that have low short circuit ratio values.
Luca Tarisciotti received the Master’s
degree in electronic engineering from The
University of Rome "Tor Vergata" in
2009 and his Ph.D. degree in electrical
and electronic engineering in the PEMC
group, University of Nottingham in 2015.
He is currently working as Research
Fellow at the University of Nottingham,
UK. His research interests include
multilevel converters, advanced
modulation schemes, and advanced power converter control.
Professor Pericle Zanchetta received his
degree in Electronic Engineering and his
Ph.D. in Electrical Engineering from the
Technical University of Bari (Italy) in
1993 and 1997 respectively. In 1998 he
became Assistant Professor of Power
Electronics at the same University and in
2001 he joined the PEMC research group
at the University of Nottingham – UK,
where he is now Professor in Control of
Power Electronics systems. He is Vice-
chair of the IAS Industrial Power Converters Committee (IPCC)
and Associate Editor of the IEEE Transactions on Industry
Applications and IEEE Transactions on Industrial Informatics.
Jon C. Clare (M’90–SM’04) was born in
Bristol, U.K. He received the B.Sc. and
Ph.D. degrees in electrical engineering
from the University of Bristol, U.K. From
1984 to 1990, he was a Research Assistant
and Lecturer at the University of Bristol
involved in teaching and research in power
electronic systems. Since 1990 he has been
with the Power Electronics, Machines and
Control Group at the University of Nottingham, U.K., and is
currently Professor in Power Electronics and Head of Research
Group. His research interests are power electronic converters and
modulation strategies, variable speed drive systems, and
electromagnetic compatibility.
Marco Degano received the 5 years Laurea
Degree in Electronic Engineering from the
Universita' degli studi di Udine (Italy) in
April 2004. In February 2008 he joined the
Power Electronics Machines and Control
(PEMC) research group at the University
of Nottingham and received a PhD degree
in Electrical Engineering in 2012. He now
works as a research fellow in the same
group developing a power converter for
aerospace application. His main interests
are power electronics and energy storage systems.

10