IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 59, NO.
2, FEBRUARY 2012
Letters
Delay Compensation in Model Predictive Current Control
of a Three-Phase Inverter
Patricio Cortes, Jose Rodriguez,
Cesar Silva, and Alexis Flores
Abstract—When control schemes based on finite control set model
predictive control are experimentally implemented, a large amount of
calculations is required, introducing a considerable time delay in the
actuation. This delay can deteriorate the performance of the system if
not considered in the design of the controller. In this paper, the problem
is described, and the solution to this issue is clearly explained using a
three-phase inverter as an example. Experimental results to validate this
solution are shown.
Index Terms—Digital control, digital signal processing, inverters, non-
linear control systems, optimal control, predictive control, static power
converters.
I. I NTRODUCTION
In recent years, finite control set model predictive control (FCS-
MPC) has been proposed as an interesting alternative for the control
of power converters and drives. This control technique uses a model
of the system to calculate predictions of the future behavior of the
system for a given set of possible actuations for a predefined time
horizon. The sequence of actuations that minimizes a cost function,
which synthesizes the control objectives, is selected, and the first
element of the sequence is applied. This optimization is repeated
every sampling time considering the new measured data. Several
applications of this control principle are summarized in [1]. FCS-
MPC has been used for the current control of three-phase inverters [2],
[3], multilevel inverters [4]–[8], multiphase inverters [9], and matrix
converters [10], [11]. Other interesting applications include torque
and flux control in electrical machines [12]–[14], power control in
pulsewidth-modulation rectifiers [15], uninterruptible power supplies
[16], and direct-current (dc)–dc converters [17]. One of the advantages
of FCS-MPC is its flexibility to control different types of variables
and to include constraints and other requirements in a simple way,
as can be observed from the wide variety of applications. However,
this kind of control algorithms needs a high amount of calculations, as
compared with a classical control scheme. This makes the calculation
time considerable, and the delay between the measurements and the
actuation can cause some problems if not considered.
The compensation of the calculation delay has been considered
in several studies published to date [14]–[16], [18], [19]. Similar
compensation methods have been also proposed for other predictive
control schemes such as deadbeat control [20], [21]. However, the
problem and its solution have not been addressed in detail in any of
Manuscript received September 21, 2010; revised January 15, 2011; accepted
March 3, 2011. Date of publication May 19, 2011; date of current version
October 18, 2011. This work was supported in part by Universidad Tecnica Fed-
erico Santa Maria, by the Chilean National Fund of Scientific and Technological
Development (FONDECYT) under Grant 1101011, by research grants from
CCTVal, Basal Project FB021 “Valparaiso Center for Science and Technology,”
and by Proyecto Anillo ACT-119.
The authors are with the Department of Electronics Engineering, Uni-
versidad Técnica Federico Santa María, 2390123 Valparaiso, Chile (e-mail:
patricio.cortes@usm.cl).
Digital Object Identifier 10.1109/TIE.2011.2157284
0278-0046/$26.00 © 2011 IEEE
1323
Fig. 1. Predictive control scheme for a three-phase inverter.
these studies. This paper provides a clear and detailed explanation
of the problem, which is not available in the previous studies, and a
simple solution is presented. Simulation and experimental results show
the effect of this delay and the performance improvement when the
compensation is considered.
II. M ODEL P REDICTIVE C URRENT C ONTROL
In this paper, the control of a three-phase inverter [2] is used as an
example application for the delay compensation method.
The predictive current control scheme using MPC is shown in Fig. 1,
and it consists of the following steps:
1) measurement of the load currents;
2) prediction of the load currents for the next sampling instant for
all possible switching states;
3) evaluation of the cost function for each prediction;
4) selection of the switching state that minimizes the cost function;
5) application of the new switching state.
In the case of current control, the cost function is defined as the
error between the reference current and the predicted currents for a
given switching state, and it is expressed as
 
g = |i∗α − ip ∗ p
α | + iβ − iβ (1)
where i∗α and i∗β are the real and imaginary parts of the reference
p
current vector, respectively, and ip
α and iβ are the real and imaginary
parts of the predicted load current vector ip (k + 1), respectively. The
predicted load current vector is calculated using a discrete-time model
of the load, which is a function of measured currents i(k), inverter
voltage (the actuation) v(k), and load back electromotive force (EMF)
e, and it is expressed as
 RTs
 Ts
ip (k + 1) = 1 − i(k) + (v(k) − ê(k)) (2)
L L
where R and L are the load resistance and inductance, respectively, Ts
is the sampling time, and ê is the estimated back EMF.
To graphically illustrate the predictive current control, only iβ is
shown in Fig. 2. This simplifies the example as the seven different
voltage vectors produce only three different values for their β compo-
nent, and hence, there are only three possible trajectories for iβ . In this
figure, the dashed lines represent the predictions for iβ , as given by
(2), and the solid line is the actual trajectory given by application of
the optimum voltages obtained by minimization of cost function (1).
In the ideal case, the time needed for calculations is negligible, and
the predictive control operates, as shown in Fig. 2(a). This ideal case
is shown for comparison. The currents are measured at time tk , and
the optimal switching state is instantly calculated. The switching state
that minimizes the error at time tk+1 is selected and applied at time tk .
Then, the load current reaches the predicted value at tk+1 .
1324 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 59, NO. 2, FEBRUARY 2012
Fig. 2. Operation of the predictive current control. (a) Without delay: calcu-
lation time is zero (ideal case). (b) With delay and without compensation: long
calculation time (real case). (c) With delay and compensation: long calculation
time (real case).
As the three-phase inverter has seven different voltage vectors,
predicted current (2) and cost function (1) are calculated seven times.
This way, depending on the sampling frequency and the speed of the
microprocessor used for the control, the time between the measure-
ment of the load currents and the application of the new switching
state can be considerable.
If the calculation time is significant compared with the sampling
time, there will be a delay between the instant in which the currents
are measured and the instant of application of the new switching
state, as shown in Fig. 2(b). During the interval between these two
instants, the previous switching state will continue to be applied. As
it can be observed in the figure, the voltage vector selected using
measurements at tk will continue being applied after tk+1 , making
the load current go away of the reference. The next actuation will be
selected considering the measurements in tk+1 and will be applied
near tk+2 . As a consequence of this delay, the load current will
oscillate around its reference, increasing the current ripple.
III. D ELAY C OMPENSATION
A simple solution to compensate this delay is to take into account
the calculation time and apply the selected switching state after the
next sampling instant. This way, the control algorithm is modified as
follows:
1) measurement of the load currents;
2) application of the switching state (calculated in the previous
interval);
3) estimation of the value of the currents at time tk+1 , considering
the applied switching state;
4) prediction of the load currents for the next sampling instant tk+2
for all possible switching states;
5) evaluation of the cost function for each prediction;
6) selection of the switching state that minimizes the cost
function.
The operation of the predictive control with a compensation delay
is shown in Fig. 2(c). Here, the measured currents and the applied
switching state at time tk are used in (2) to estimate the value of the
load currents at time tk+1 . Then, this current is used as a starting
point for the predictions for all switching states. These predictions
are calculated using the load model shifted one step forward in
time, i.e.,
 RTs
 Ts
ip (k + 2) = 1 − î(k + 1)+ (v(k + 1) − ê(k + 1)) (3)
L L
where î(k + 1) and ê(k + 1) are the estimated current and
back-EMF vectors, and v(k + 1) is the actuation to be
evaluated. Then, the cost function is evaluated for predicted currents
ip (k + 2), and the switching state that minimizes this cost function
is selected and stored to be applied at the next sampling instant. The
estimation of î(k + 1) and ê(k + 1) increases the calculation times
but only marginally, as this calculation is performed only once.
IV. R ESULTS
Simulation and experimental results are presented, illustrating the
effect of the calculation delay and the effectiveness of the delay
compensation method. MATLAB/Simulink was used for simulations,
considering the following system parameters: Vdc = 520 V, R =
10 Ω, L = 7 mH, and Ts = 50 μs. As shown in Fig. 3, when there
is a delay in the actuation, the ripple of the load current is considerably
increased. These problems are solved by including a compensation of
the control delay, as explained in the previous section. Results with the
delay compensation present a reduced load current ripple.
Experimental results were carried out using a modified 5.5-kW
industrial inverter, which is externally controlled by a digital signal
processor (DSP). The predictive control algorithm was implemented
in a TMS320C6713 DSP, which is a field-programmable gate-array-
based daughter card that handles simple tasks such as digital input
and outputs and analog-to-digital converters. The system parameters
are the same as those in the simulations. The calculation time of
the control algorithm is 17 μs. The effect of the delay introduced
by this calculation time is shown in Fig. 4. It can be observed in
this figure that delay compensation improves the performance of the
predictive controller by reducing the ripple of the load currents. The
total harmonic distortion of the load currents is reduced from 8.5% to
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 59, NO. 2, FEBRUARY 2012
Fig. 3. Simulation results of the predictive current control in steady-state
operation without and with delay compensation (Ts = 50 µs).
Fig. 4. Experimental results of the predictive current control without and with
delay compensation (Ts = 50 µs).
3.6%. Although the sampling frequency is the same for compensated
and uncompensated delays, the average switching frequency increases
when the compensation method is used. The switching frequency
without compensation is 2.72 kHz and increases to 4.62 kHz with
delay compensation.
V. C ONCLUSION
The problem of the delay due to the calculation time in MPC
schemes has been explained in this paper, and a solution has been
presented. The compensation method is simple and can be applied to
any MPC application. Simulation and experimental results illustrate
the problem and its solution.
The delay compensation method is verified with experimental
results. Using this method, the load current ripple is considerably
reduced, and the switching frequency is increased. No significant effect
in the dynamic behavior has been observed.
1325
R EFERENCES
[1] S. Kouro, P. Cortes, R. Vargas, U. Ammann, and J. Rodriguez, “Model
predictive control—A simple and powerful method to control power
converters,” IEEE Trans. Ind. Electron., vol. 56, no. 6, pp. 1826–1838,
Jun. 2009.
[2] J. Rodríguez, J. Pontt, C. Silva, P. Correa, P. Lezana, P. Cortés, and
U. Ammann, “Predictive current control of a voltage source inverter,”
IEEE Trans. Ind. Electron., vol. 54, no. 1, pp. 495–503, Feb. 2007.
[3] A. Linder and R. Kennel, “Model predictive control for electrical
drives,” in Proc. IEEE Power Electron. Spec. Conf., Jun. 12–16, 2005,
pp. 1793–1799.
[4] P. Lezana, R. Aguilera, and D. E. Quevedo, “Model predictive control
of an asymmetric flying capacitor converter,” IEEE Trans. Ind. Electron.,
vol. 56, no. 6, pp. 1839–1846, Jun. 2009.
[5] M. A. Perez, P. Cortes, and J. Rodriguez, “Predictive control algorithm
technique for multilevel asymmetric cascaded H-bridge inverters,” IEEE
Trans. Ind. Electron., vol. 55, no. 12, pp. 4354–4361, Dec. 2008.
[6] R. Vargas, P. Cortes, U. Ammann, J. Rodriguez, and J. Pontt, “Predictive
control of a three-phase neutral-point-clamped inverter,” IEEE Trans. Ind.
Electron., vol. 54, no. 5, pp. 2697–2705, Oct. 2007.
[7] J. Barros and J. Silva, “Multilevel optimal predictive dynamic voltage
restorer,” IEEE Trans. Ind. Electron., vol. 57, no. 8, pp. 2747–2760,
Aug. 2010.
[8] P. Cortes, A. Wilson, S. Kouro, J. Rodriguez, and H. Abu-Rub,
“Model predictive control of multilevel cascaded H-bridge invert-
ers,” IEEE Trans. Ind. Electron., vol. 57, no. 8, pp. 2691–2699,
Aug. 2010.
[9] F. Barrero, M. R. Arahal, R. Gregor, S. Toral, and M. J. Duran, “One-
step modulation predictive current control method for the asymmetrical
dual three-phase induction machine,” IEEE Trans. Ind. Electron., vol. 56,
no. 6, pp. 1974–1983, Jun. 2009.
[10] P. Correa, J. Rodriguez, M. Rivera, J. R. Espinoza, and J. W. Kolar,
“Predictive control of an indirect matrix converter,” IEEE Trans. Ind.
Electron., vol. 56, no. 6, pp. 1847–1853, Jun. 2009.
[11] R. Vargas, U. Ammann, B. Hudoffsky, J. Rodriguez, and P. Wheeler, “Pre-
dictive torque control of an induction machine fed by a matrix converter
with reactive input power control,” IEEE Trans. Power Electron., vol. 25,
no. 6, pp. 1426–1438, Jun. 2010.
[12] T. Geyer, G. Papafotiou, and M. Morari, “Model predictive direct torque
control—Part I: Concept, algorithm, and analysis,” IEEE Trans. Ind. Elec-
tron., vol. 56, no. 6, pp. 1894–1905, Jun. 2009.
[13] G. Papafotiou, J. Kley, K. G. Papadopoulos, P. Bohren, and M. Morari,
“Model predictive direct torque control—Part II: Implementation and
experimental evaluation,” IEEE Trans. Ind. Electron., vol. 56, no. 6,
pp. 1906–1915, Jun. 2009.
[14] H. Miranda, P. Cortes, J. I. Yuz, and J. Rodriguez, “Predictive torque
control of induction machines based on state-space models,” IEEE Trans.
Ind. Electron., vol. 56, no. 6, pp. 1916–1924, Jun. 2009.
[15] P. Cortes, J. Rodriguez, P. Antoniewicz, and M. Kazmierkowski, “Direct
power control of an AFE using predictive control,” IEEE Trans. Power
Electron., vol. 23, no. 5, pp. 2516–2523, Sep. 2008.
[16] P. Cortes, G. Ortiz, J. I. Yuz, J. Rodriguez, S. Vazquez, and
L. G. Franquelo, “Model predictive control of an inverter with output
LC filter for ups applications,” IEEE Trans. Ind. Electron., vol. 56, no. 6,
pp. 1875–1883, Jun. 2009.
[17] A. G. Beccuti, S. Mariethoz, S. Cliquennois, S. Wang, and M. Morari,
“Explicit model predictive control of dc–dc switched-mode power sup-
plies with extended Kalman filtering,” IEEE Trans. Ind. Electron., vol. 56,
no. 6, pp. 1864–1874, Jun. 2009.
[18] P. Cortes, J. Rodriguez, D. E. Quevedo, and C. Silva, “Predictive current
control strategy with imposed load current spectrum,” IEEE Trans. Power
Electron., vol. 23, no. 2, pp. 612–618, Mar. 2008.
[19] M. Arahal, F. Barrero, S. Toral, M. Duran, and R. Gregor, “Multi-phase
current control using finite-state model-predictive control,” Control Eng.
Pract., vol. 17, no. 5, pp. 579–587, May 2009.
[20] H. Abu-Rub, J. Guzinski, Z. Krzeminski, and H. Toliyat, “Predictive
current control of voltage-source inverters,” IEEE Trans. Ind. Electron.,
vol. 51, no. 3, pp. 585–593, Jun. 2004.
[21] J. Moreno, J. Huerta, R. Gil, and S. Gonzalez, “A robust predictive cur-
rent control for three-phase grid-connected inverters,” IEEE Trans. Ind.
Electron., vol. 56, no. 6, pp. 1993–2004, Jun. 2009.