2014 IEEE International Conference on Power Electronics, Drives and Energy Systems (PEDES)

Comparison of Fuzzy and MPC based Buck

Converter
M.Praveen Kumar P.Ponnambalam S.Sreejith
School of Electrical Engineering School of Electrical Engineering School of Electrical Engineering
VIT University, Vellore, India, VIT University, Vellore, India, VIT University, Vellore, India,
Vellore, India, [email protected] [email protected] sreejith.s@vit.

J.BelwinEdward K.Krishnamurthy
School of Electrical Engineering Dean, SBMS Kongu Engineering College
VIT University, Vellore, India, [email protected] Perundurai, Erode 638052, [email protected]

Abstract- This article presents the design and operation [4]. The usual requirement of a control system for
construction of Model Predictive Controller (MPC) for a this DC/DC converter is to maintain the output voltage
Pulse Width Modulation (PWM) based buck converter, constant irrespective of variations in the input voltage and
working in Continuous Conduction Mode (CCM). The source current [3]. Conventional solutions for controller
converter operates at a switching frequency of 100 KHz. requirements for these DC/DC converters were based on
The buck converter is mathematically modeled and classical control theory or modern control theory.
implemented in matlab simulink. The open loop Controllers designed on the basis of classical control theory
response of the buck converter is used to obtain the require precise linear mathematical models of the plants.
transfer function (second order) of the model, which in These controllers failed to perform satisfactorily under
turn is used to design the MPC controller. This MPC is parameter variation, non-linearity, load disturbance, etc [5].
used to control the output voltage of the buck converter The stability of non-linear systems can be established only
and it has been found that the output voltage is on system-by-system basis and hence the design procedures
maintained constant accurately for dynamic load, and for a controller that meets the stability, robustness and good
the peak overshoot is reduced drastically. The output dynamical response are not available for large classes of
voltage obtained with MPC controller is then compared such systems [6]. One of the major solutions for this kind of
with the output voltage obtained from fuzzy based buck problem is using a fuzzy controller, but fixing the input and
converter and it has been observed that the MPC gives output membership functions for controlling the output
better output than the fuzzy controller voltage with minimum peak overshoot is arduous.
Model Predictive Control (MPC) allows us to
Keywords - Buck Converter, Fuzzy Inference System, address problems like feasibility, stability and performance
Fuzzy Logic, MPC Controller, Prediction Horizon, in a rigorous manner [7].Model Predictive Control (MPC)
Control Horizon, Pulse Width Modulation will serve as a better alternative for controlling the output
voltage of the buck converter with minimum peak
overshoot. The main advantage of Model Predictive Control
I. INTRODUCTION (MPC) is the fact that it allows the current time slot to be
The switched mode DC/DC converters convert one level optimized while keeping future time slots in account, this is
of electrical voltage to another level by switching action [1]. achieved by optimizing a finite time horizon but only
They have gained popularity these days because of their implementing the current timeslot.
high efficiency and smaller size [2]. DC/DC converters can Fig.1 shows the functional block diagram of a
provide a controllable and theoretically lossless DC voltage Model Predictive Controller (MPC) controlled buck
transformation, in addition to that it can also provide voltage converter. From the figure it can be observed that the
isolation by incorporating small high frequency transformer reference voltage and the measured voltage is given to the
[3]. The switching nature of the DC/DC converter circuit MPC. The MPC will take care of finding out the error signal
makes the simulation difficult to converge unless some and controlling the duty ratio of the buck converter. The
default parameters are suitably changed, and hence load resistor is shown as variable resistor, which specifies
MATLAB/SIMULINK is quite suitable for simulating such that the load of the buck converter is dynamic. In this paper
kind of converters, where the parameters modification is the performance of the MPC controller is verified for
possible [2]. variations in reference (set point) voltage and variations in
In general, power electronic DC/DC converters are the load.
periodic time variant systems due to their inherent switching

978-1-4799-6373-7/14/$31.00 2014 IEEE

 II. MODELING OF BUCK CONVERTER 1
In this section the mathematical modeling of buck .
converter is described. Fig. 1 shows the basic Buck 4
converter circuit diagram for which the mathematical Where
modeling is to be derived. The table. 1 shows the parameters
used for the buck converter taken into consideration. 5
1 6
Supply
Voltage
Reference

7
Voltage
MPC Controlled d Buck Converter R
Measurement 1 8
And d is the duty cycle whose value varies
between 0 and 1 and is defined as [9]:
Measured value
9
Fig. 1 Functional block diagram of MPC controller Buck
Converter The above derived state space equations are used to
The basic operation of the buck converter can be develop the Simulink model for the buck converter, the
explained in two steps, first the inductor is charged through output voltage of which will be controlled by the MPC for
the switch S, second the switch S is opened and the inductor different input voltages and different load.
discharges through the load [8]. Practical switching
converters take into account the series resistance of both the
inductor and the capacitor in the feedback compensation. III. MODEL PREDICTIVE CONTROLLER (MPC)
The capacitor ESR, RC, introduces a zero frequency to the Traditionally, power electronics circuits and systems have
transfer function [5]. been controlled in industry using linear controllers
S L combined with non-linear procedures like Pulse Width
Vg Modulation (PWM). The models used for controller design
Rs Rl are a result of simplifications that include averaging the
Vin Rg Rc behavior of the system over time (to avoid modelling the
R Vo
d switching) and linearizing around a specific operating point
D C disregarding all constraints. As a result, the derived
controller usually performs well only in a neighborhood
around the operating point [10].
The Model Predictive Controller (MPC) also known as
the receding horizon control technique is one of the most
Fig. 2 Buck converter. successful controller for practical applications [11] and also
is one of the emerging research field for control
Table. 1 Buck Converter parameters requirements. Model Predictive Control (MPC) is a standard
Symbol Values Symbol Values
R 1 C 305 F
and widely accepted tool to control multivariable systems in
R1 80 m RC 5m the presence of constraints [12]. This control method deals
L 5 H fsw 100 KHz in general with constrained regulation problems where the
control objectives are defined through an objective function
Systems equations are given by [4] and constraints imposed over the systems inputs, outputs
1 and states. Each time step, the current state is used to solve
.
an optimization problem over a finite horizon which yields
1 an optimal control sequence that minimizes the given
1 objective function. According to the receding horizon
strategy, only the first element of the sequence is kept and
2 applied to the system in order to obtain a new state. At the
next time-step, the computation is repeated with the updated
3 state over a shifted horizon. The solution of each
optimization problem respects the constraints imposed and
The transfer function of the output filter is the controller is performant with respect to the chosen
obtained by rearranging the above objective function [13].
equations [5]
 The core MPC Toolbox algorithm is based on a model of At the next sampling instant, k+1, the M-step
the system to be controlled, a performance index driving the control policy is re-calculated for the next M sampling
selection of the decision variables, a set of constraints to be instants, k+1 to k+M, and implement the first control move,
fulfilled, and a state estimator to reconstruct the model's u(k+1).
internal states [14]. The first two steps are repeated for subsequent sampling
Fig.3 shows the basic block diagram of Model Predictive instants.
Controller (MPC). The block diagram also specifies the
basic elements necessary for designing a MPC. The control calculation for the single input single output
At the k-th sampling instant, the values of the case without loss of generality consists of the following
manipulated variables, u, at the next M sampling instants, optimization problem [15].
{u(k), u(k+1), , u(k+M -1)} are calculated.
min
Set-point ..
calculations

Set Points

Predicted Control
Predictions Inputs Process Process outputs
outputs calculations

Inputs Model Model outputs

10
Where P is the prediction horizon, M is the number of
moves to be optimized, ysp is the set point, is the model
Residuals prediction, and are weighting factors. u is the change in
manipulated variables defined as u(k) = u(k) - u(k-1). In
Fig. 3 MPC Block Diagram general, M P and therefore u(k + M 1) = u(k + M) = =
Past Future u(k + P 1). The model prediction can be written as

Set Point (Target) 1

1,2, . . 11
Past Output
Predicted Future Output Where y*(k + j) is the contribution to the future values
Past Control Action due to past input moves (up to time k-1), Smi are the step
Control Horizon, M Future Control Action response coefficients for the manipulated variables and
w(k+j) captures all unmodelled effects, y*(k+j) can be
written as

Prediction Horizon, P

k-1 k k+1 k+2 k+M-1

Sampling Instant
k+P

1,2, . . 12
Fig. 4 Basic Concept for Model Predictive Control
The term is the contribution of the most recent
o This set of M control moves is calculated so as to measured disturbance change. To predict the output values
minimize the predicted deviations from the reference using equation (11) w(k+j) must be estimated. This is
trajectory over the next P sampling instants while carried out as follows:
satisfying the constraints.
o Typically, an LP or QP problem is solved at each
1,2, . . 13
sampling instant.
Where y(k) is the current measurement. The solution to
o Terminology: M = control horizon, P = prediction
equation (10) subject to equation (11), reduces to a least
horizon
square solution [15].
Then the first control move, u(k), is implemented.
The MPC controller is analyzed with MATLAB Simulink
Then the first control move, u(k), is model. To use the MPC controller first the Transfer
implemented. Function of the model is estimated by black box modeling
by obtaining the input and output data from the open loop
buck converter and the obtained Transfer Function is given Fig. 7 it can be observed that there is minimum change in
by voltage for change in load current
The platform power delivery efficiency for a CPU of a
. .
14 Digital Signal Processors (DSP) Integrated Circuits (IC) in
. . portable computers and handheld devices at all load levels is
significant for energy saving and battery life [16], and hence
The Transfer Function obtained shows that the system is the closed loop converter designed has to be checked with
a second order system, and the fitness to the estimation data different load conditions
is 99.3%. This Transfer Function is used to design the MPC Fig.8 shows the output voltage and corresponding
controller in MATLAB Simulink model, the MPC controller inductor current for changes in load current, from the figure
designed is with 2 poles. The designed MPC controller is it can be understood that there is minimum change in
then checked with the buck converter in closed loop inductor current for changes in load current. Fig. 9 shows
configuration and it has been found that the output obtained the output voltage when both the reference voltage and load
is with almost zero oscillation and zero peak overshoot, current is varied with a same time interval of 0.001 seconds,
which gives us an idea that MPC controller can be a better the voltage variations are as same as the previous
controller than any other controller for controlling the experimentation and the variation in the load current is 1,
output voltage of buck converter for variations in input 0.75, 0.85, 1.15, 1.5, 2, 1.25, 1.75, 2.25 A for every 0.001
voltage and load current. seconds.
12

IV. SIMULATION RESULTS

Output Voltage
Reference Voltage
10
In most practical problems, states of the system are not

Output Voltage (V)

accessible directly and must be estimated. The quality of 8

state estimates has important bearings on the overall 6

performance of a model predictive controller, especially of

one based on a non-linear model. Unlike the linear case,
4

however, there is no established method for non-linear state 2

estimation. The most popular method is the extended 0

Kalman filter, which simply relinearizes the non-linear 0 0.001 0.002 0.003 0.004 0.005
Time (s)
0.006 0.007 0.008 0.009 0.01

model at each time step and updates the gain matrix and the Fig. 6 Output voltage for various reference voltages
co-variance matrix on the basis of linear filtering theory [7]. 12

This method is used for estimating the model of the given 10

system and giving out the proper control signal to get exact
Output Volttage (V)

error less output. 8

The simulation results for the closed loop control of

Output Voltage
6 Reference Voltage

Model Predictive Control (MPC) is shown in the figures 5

to . Fig. 5 shows the output voltage for constant reference
4

voltage maintained at a voltage of 10 V. The figure shows 2

that there is very less peak overshoot 0

0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.01
Time (s)
12
Fig. 7 Output voltage for variations in load
10
From the Fig. 10 it can be observed that the output
voltage is following the reference voltage even with the
variations in the load current. It can be observed that the
O utput V oltage (V )

6
output voltage is having very minimum overshoot and it is
almost following the reference voltage and the inductor
4 current also does not vary very widely with changes in load
current, from which we can come to a conclusion that the
ModeL Predictive Controller (MPC) works effectively in
2

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
most of the working conditions.
Time (s) x 10
-3
The fuzzy PID controller for the buck converter is
Fig.5 Output voltage for constant reference voltage performed and compared with the MPC controller in matlab
Fig.6 shows the output and reference voltage with Simulink environment. For the fuzzy PID controller design
time axis when the reference voltage is changed as 0, 5, 2.5, the Proportional (P), Integral (I) and Derivative (D) action
7, 11, 8, 9, 6, 1, 3, Volts. The results shows that the output of the PID controller is described in the form of rule base
voltage almost follows the reference voltage with almost nil [5]
overshoot, Fig.7 shows the output voltage for different load
currents, keeping the reference voltage constant. From the
 The structure of the fuzzy PID controller is shown in Fig
In d u c to r C u rre n t (A ) O u tp u t V o lta g e (V )

15
11.The inputs and the outputs of the three FISs are related
10
by these fuzzy rules which are given in the form of Table. 2
5
[5].
0
0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.01
Fig. 12 & 13 show the member ship ship function
100
Time (s) considered for fuzzy proportional controller. Fig. 14 shows
the performance comparison of fuzzy controller with the
50
MPC controller. The controllers are given with an input
reference voltage of 5 V and load current of 1 A, for which
0 the performance of the buck converter has been checked and
from which it can be inferred that MPC controller gives
0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.01
Time (s)
Fig. 8 Output voltage and inductor current for variations better performance in terms of settling time and peak
in load current overshoot
12
Output Voltage Error Fuzzy
Reference Voltage Proportional
10
Controller
O utput V oltage (V )

To Buck
6 Fuzzy Converter
e Integral
4
Controller
2

0
0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.01
Time (s) Fuzzy
de/
Fig 9.output voltage for change in reference voltage and dt
Differential
Controller
change in load current
Fig.11 Structure of Fuzzy PID controller used for the buck
converter
In d u c to r C u rre n t ( A ) O u tp u t V o lta g e (V )

15

10 Table. 2 Fuzzy Rules

5
Np Ni Nd
0
If e(k)=N If i e(k)=N If d e(k)=N
0 0.001 0.002 0.003 0.004 0.005
Time (s)
0.006 0.007 0.008 0.009 0.01
then y=N then y=N then y=N
50
If e(k)=Z If i e(k)=Z If d e(k)=Z
0
then y=Z then y=Z then y=Z
If e(k)=P If i e(k)=P If d e(k)=P
-50
0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.01 then y=P then y=P then y=P
Time (s)
Fig. 10 Output voltage and inductor current for variations
in reference voltage and load current

The advantages of fuzzy controllers are that (i) it does not

need accurate mathematical model; (ii) it can work with
imprecise inputs; (iii) it can handle nonlinearity, and (iv) it
is more robust than conventional nonlinear controllers [17].
The fuzzy control algorithm has a set of fuzzy control rules
which are related by the concepts of fuzzy implication and
the compositional rule of inference. These inference rules
are determined depending on the behavior of the converter.
A fast action is required if the output is away from the target
[18]. During defuzzification process we get only a single
output point from the whole output curve. Center of Area
(COA) method defuzzification is used for defuzzification Fig12:Member ship functions for error in fuzzy
and the output is described as [19]. proportional fuzzy controller


15
 [3] A. J. Forsyth and S. V. Mollov, "Modeling and Control of DC-DC
Converters", PoweEngineering Journal, Vol. 12, No. 5, pp. 229236,
1998.
[4] J. Mahdavi, A. Emadi, H. A. Toliyat, "Application of State Space
Averaging Method to Sliding Mode Control of PWM DC/DC
Converters", presented at conf on IEEE Industry Applications Society,
Oct. 1997.
[5] G. Abbas, N. Abouchi, A. Sani, and C. Condemine, Design and
analysis of fuzzy logic based robust PID controller for pwm-based
switching converter, IEEE International Symposium on circuits and
systems, pp 777780, 15-18, May 2011
[6] K. S. Narendra and K. Parthasarathy, Identification and control of
dynamical systems using neural networks, IEEE Trans. on Neural
Networks, vol. 1, no. 1, pp. 427, 1990.
[7] M. Morari and J. H. Lee, Model predictive control: past, present and
future, Comput. Chem. Eng., vol. 23, no. 4/5, pp. 667682, May
1999.
Fig13:Member ship functions for output in fuzzy [8] K. Lian, J. Liou C. Huang LMI Based Integral Fuzzy Control of
proportional controller DC-DC Converters, IEEE Transactions on Fuzzy Systems, Vol. 14,
No. 1, February 2006.
7
[9] Simon S. Ang and Oliva Alejandro, "Power Switching Connverters",
MPC Controller Output
Second Edition, Taylor & Francis, Inc. March 2005.
6 Fuzzy PID Controller Output [10] T. Geyer, G. Papafotiou, and M. Morari Model predictive control in
5
power electronics: A hybrid systems approach, Proc. CDC-ECC05,
pp. 56065611, 2005.
O utput V oltage (V)

4 [11] P. Corts, M. P. Kazmierkowski, R. M. Kennel, D. E. Quevedo, and J.

3
Rodrguez, Predictive control in power electronics and drives, IEEE
Trans. Ind. Electron., vol. 55, no. 12, pp. 43124324, Dec. 2008.
2 [12] S.J. Qin, T.A. Badgwell, An overview of industrial model predictive
control technology, in: J.C. Kantor, C.E. Garcia, B. Carnahan (Eds.),
1
Fifth International Conference on Chemical Process Control, AIChE
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Symposium Series 316, 93, 1997, pp. 232256.
Time (s) x 10
-3
[13] C. Vlad, P. Rodriguez-Ayerbe, E. Godoy, P. Lefranc Explicit Model
Fig14. Comparision between Fuzzy and MPC controler For Predictive Control of Buck Converter 15th International Power
Electronics and Motion Control Conference, EPE-PEMC 2012 ECCE
a reference voltage Europe, Novi Sad, Serbia.
[14] A. Bemporad, N. Ricker, and J. Owen, Model predictive control
New tools for design and evaluation, in American Control
V. CONCLUSION & FUTURE SCOPE Conference, Boston, MA, 2004, pp. 56225627.
[15] K. O. Temeng, P. D. Schnelle, and T. J. McAvoy, Model predictive
The Model Predictive Controller (MPC) designed for the control of an industrial packed bed reactor using neural networks, J.
buck converter is much more robust and simple compared to Process Cotrol 5 (1995), pp. 1927.
the design of fuzzy based buck converter and PI controller [16] J.A.A. Qahouq, On load adaptive control of voltage regulators for
controlled buck converter. The results obtained using power managed loads: control schemes to improve converter
efficiency and performance, IEEE Transactions on Power
MATLAB Simulink shows that the peak overshoot, ripple Electronics, Vol. 22, 2007, pp. 1806 - 1819.
factor of the voltage, change in inductor current, rise time, [17] V.S.C Raviraj and P.C Sen., Comparative study of proportional
settling time, etc are much more better than the other integral, sliding mode, and Fuzzy Logic Controllers for power
controller used for controlling the buck converter. Since Converters, IEEE Transactions on Industry Applications, vol. 33, no.
2, March/April 1997.
buck converter is basically used as control for voltage [18] Jan Jantzen, "Design Of Fuzzy Controllers". Tech report no 98-E 864
source in switched mode power supply, which involves (design), 19 Aug 1998.
variation in input voltage and load current, the simulation is [19] Mamun Rabbani, H. M. Mesbah Maruf, Tanvir Ahmed, Md.
checked with various reference voltages and various load Ashfanoor Kabir and Upal Mahbub Fuzzy Logic Driven Adaptive
PID Controller for PWM Based Buck Converter IEEE/OSA/IAPR
currents and it has been observed that the MPC controller International Conference on Informatics, Electronics & Vision.
has good control performance compared with other
controllers. The MPC controlled buck converters
performance is compared with the performance of fuzzy
PID controlled buck converter and it has been observed that
the MPC based buck converter gives good performance
compared with the fuzzy PID controlled converter.

REFERENCES

[1] S. S. Ang, A practice-oriented course in switching converters, IEEE

Trans. Educ., vol. 39, pp. 1418, Feb. 1996.
[2] J-H. Su, J-J. Chen, and D-S. Wu, Learning Feedback Controller
Design of Switching Converters Via MATLAB/Simulink, IEEE
Trans. on Education, Vol. 45, No. 4, pp.307315, 2002.