See discussions, stats, and author profiles for this publication at: [Link]

net/publication/323547959

Losses Evaluation for a Two-Level Three-Phase Stand-Alone Voltage Source

Converter Using Model Predictive Control

Conference Paper · December 2017

DOI: 10.1109/SPEC.2017.8333624

CITATION READS

1 153

4 authors:

Mohamed Adel Alhasheem Tomislav Dragicevic

Aalborg University Aalborg University
7 PUBLICATIONS   14 CITATIONS    161 PUBLICATIONS   2,999 CITATIONS   

SEE PROFILE SEE PROFILE

Marco Rivera F. Blaabjerg

Universidad de Talca Aalborg University
266 PUBLICATIONS   3,567 CITATIONS    2,082 PUBLICATIONS   65,831 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Hybrid AC/DC Microgrids View project

Study on Harmonic Instability Mechanism and Mitigation Methods in Traction Power Supply System View project

All content following this page was uploaded by Mohamed Adel Alhasheem on 16 May 2018.

The user has requested enhancement of the downloaded file.

 Losses Evaluation for a Two-Level Three-Phase
Stand-Alone Voltage Source Converter Using Model
Predictive Control
M. Alhasheem1,2 , T. Dragicevic1 , M. Rivera3 and F. Blaabjerg1
1 Department of Energy Technology, Aalborg University, Aalborg, Denmark
2 Arab Acadmey for Science, Technology and Maritime Transport, Cairo, Egypt
3 Faculty of Engineering, University of Talca Curico, Chile

E-mail: mah@[Link], tdr@[Link], marcoriv@[Link], fbl@[Link]

Abstract— This paper discusses a Model Predictive as proposed in [8] is studied. The VSC, as shown in Fig.
Control (MPC) scheme for a Voltage Source Converter 2, consists of 6 IGBTs and diodes. The Finite Control Set
(VSC) where the aim is to minimize losses and Total Model Predictive Control (FCS-MPC) is applied to the VSC.
Harmonics distortion (THD) compared to the conventional A thermal description for a power module SKM 50GB 123D
MPC scheme. Different Cost Functions (CFs) are applied is used to calculate the switching and conduction losses. Also,
to the stand-alone VSC. A comparison between conven- the heat sink for the IGBTs is considered and controlled with
tional and improved controller is conducted in terms of a temperature circuit. Finally, a detailed comparison is done
losses. The switching and conduction losses for the VSC for the different CFs in terms of THD as discussed in [8]. This
are calculated using MATLAB/SIMULINK and PLECS paper includes the following sections; Section II presents the
software. A power level up to 1 kW is considered for the model of the system. Section III demonstrates how the losses
conventional and the improved schemes. All simulation and are calculated analytically. Section IV discusses the obtained
experimental results are conducted using the SEMIKRON results for one-step prediction and improved current control in
SKM 50GB 123D as the power module. terms of losses and THD. Section V gives the conclusion for
the paper.
I. I NTRODUCTION
Electrical power is indispensable for daily life. The conver-
sion from the primary form of energy to electrical energy has Vc *(k) LC filter
L1 ,f ,o
allowed to build a modern society. The VSC is one of the
most effective devices in the electrical power generation and Minimization S L2 Load
Vc(k)
conversion. It has effective capabilities such as, controlling of cost VSC (R)
,f (k) Predictive Vc(k+1) function
active and reactive power, the ability of connection to weak L3
control
,o (k)
AC networks, and it can also be influential during grid
restoration. Thus, VSC is used in many applications, such Vc
as Uninterruptible Power Supply (UPS) [1]. Researchers have C1 C2 C3
claimed that the different control structures such as, deadbeat
control [2], linear control [3, 4], and MPC [5–7] can be applied Fig. 1. Model predictive control scheme of a VSC.
to the VSC. The MPC can be used for the losses investigation,
which is the main focus of this paper. The MPC scheme,
as shown in Fig. 1, is used having linear loads. Generally,
different predictive controllers can be used with the VSC as S1 S2 S3
discussed in [1, 8, 9]. The conventional and improved current L ,f LC filter ,o
a
controllers using MPC will be considered in this work.
L Load
Different MPC controllers can be achieved by modifying b (R)
Vdc
the used CF. The objective of the CF is to minimize the error L
c
between the measured and reference values. One of the CF C C C
Vc
advantages is that it can track the variables with the different S4 S5 S6
physical nature. Two different CFs will be considered in this
paper for losses calculation. Firstly, a conventional CF that
evaluates the predicted and reference voltage using one-step Fig. 2. Power circuit schematic for a stand-alone VSC.
prediction. Secondly, an improved current control evaluation
 II. S YSTEM M ODEL
The power circuit of the three phase VSC, as shown in Vβ
Fig. 2, is considering where the two switches in each leg V3 V2
operate in a complementary mode. The switching states can S3(0,1,0) S2(1,1,0)
be represented by the switching signals Sa , Sb , and Sc , which
are defined as follows:

1 if S1 on and S4 off
Sa = (1)
0 if S1 off and S4 on V4 S4(0,1,1) S1(1,0,0) V1 Vα

1 if S2 on and S5 off
Sb = (2)
0 if S2 off and S5 on
S0(0,0,0)
 V0,7
1 if S3 on and S6 off S7(1,1,1)
Sc = (3)
0 if S3 off and S6 on
The filter inductance (L1, L2, L3) equation can be ex- S5(0,0,1) S6(1,0,1)
pressed in the vectorial form as: V6
V5
dIf
L = Vi − Vc (4)
dt Fig. 3. Voltage vectors generated by the inverter.
where L is the filter inductance. The equation that describes
the dynamic behaviour of the output voltage can be expressed
mathematically as:  Ts
dVc Bdq = expAτ B2 dτ (14)
C = I f − Io (5) 0
dt
This model is used to calculate the prediction of filter
where C is the filter capacitance (C1, C2, C3). These equa-
voltages and currents for every possible input voltage. The
tions can be rewritten in the state space model as:
selection of the optimal input voltage depends on the evalu-
dx ation of CF [9]. Consequently, the new switching states are
= Ax + B1 Vi + B2 Io (6) applied to the converter for the next sampling time.
dt
where,   III. E VALUATION OF POWER LOSSES
I
x= f (7)
Vc Power losses are important not only because of the effi-
  ciency reduction but also due to the generated heat from the
−R/L −1/L
A= (8) circulating currents. The heat must removed from the semi-
1/C 0
conductor devices and never be higher than a certain junction
 
1/L temperature (Tj) [10]. The junction temperature allowed in
B1 = (9) an IGBT module is usually up to 125 ◦ C. In this section,
0
  the equations used to calculate the relevant power losses are
0 presented in SEMIKRON - AN-8005 application note [11].
B2 = (10)
−1/C The PLECS toolbox is used for developing the thermal model
of the VSC. The conduction losses for SKM 50GB 123D
If and Vc are the filter current and voltage respectively. Io module, which is used in the simulations, is given by:
is the load current, which can be estimated or measured. In 
this paper, Io is measured and used in the evaluation of the 1 T
P = (VCE (t)I(t) + Rce I 2 (t))dt (15)
error. Vi is the inverter voltage of the system and it has 8 T 0
different voltage vectors as shown in Fig. 3. where VC E and Rce can be determined graphically from the
A discrete model is obtained from (6) and it can be converter data sheets. During the commutations, losses appear
expressed as follows: due to the switching pattern. The expression of the switching
        losses is given by:
x k + 1 = Aq x k + Bq Vi k + Bdq Io k (11) √
where, 2Irms
Pon+of f = fsw V K (16)
Aq = expATs (12) π
where K is the reference value as calculated in [11]. The losses
 Ts should be dissipated away in order to keep the semiconductor
Bq = expAτ B1 dτ (13) devices at a safe temperature. A thermal resistance will exist
0
due to the internal construction of the materials. Practically,
two factors should be considered and treated. Firstly, the
thermal resistance must be as small as possible. Secondly, the 
heat transformation from the conducting device to the heat

3RZHU:
sink should be more efficient. The heat sink is simulated in this
work through the PLECS toolbox and all equivalent thermal 
resistances are added.
IV. S IMULATION RESULTS AND DISCUSSION 
  
The performance of the VSC using MPC is simulated using
7LPH6HF
MATLAB/SIMULINK and PLECS software. The simulations
are done twice: a) conventional CF (one-step prediction) [9], Fig. 4. The conduction losses for one fundamental cycle of the conventional
b) improved current control methodology [8]. The simulations CF at 0.5 kW output power.
are conducted using the parameters of the VSC as listed in
Table I. The thermal model of VSC is simulated by using
PLECS. Moreover, the thermal characteristics of the SKM 
50GB 123D module are considered as a comparison of the 

3RZHU:
different predictive controllers.

TABLE I
PARAMETERS OF A THREE - PHASE VOLTAGE SOURCE CONVERTER (VSC). 


Quantity Value   
DC link voltage Vdc 520 V 7LPH6HF
Filter inductance L (L1,L2,L3) 2.4 mH
Fig. 5. The switching losses for one fundamental cycle of the conventional
Filter capacitor C (C1,C2,C3) 25 μF CF at 0.5 kW output power.
Resistive load R 60 - 400
Ω
Average switching frequency 6.6 kHz 


3RZHU:

A. Conventional CF - one-step prediction technique


The conventional CF used in the first set of analysis is given
by: 

∗ 
g = (Vcα − Vcαp
(k + 1))2 + (Vcβ∗ p
− Vcβ (k + 1))2 +   
(λsw npsw ) (17) 7LPH6HF
As shown in the Eqn. (17), the CF contains error terms for
Fig. 6. The switching and conduction losses for one fundamental cycle of
the voltage and the third term is for the switching frequency the conventional CF at 0.5 kW output power.
control. Generally, the absolute or the squared error terms give
similar results. The CF tracks the trajectory of the voltage
variable for the given sample time (Ts = 25 μS). It aims to B. Improved current control technique
minimize the error of the voltage to select a proper actuation
The second set of analysis is considering the following CF:
of the conducting devices.
∗
The conduction losses are calculated using Eqn. (15). As g = (Vcα − Vcα
p
(k + 1))2 + (Vcβ
∗ p
− Vcβ (k + 1))2 +
shown in Fig. 4, the conduction losses is fluctuating around (18)
3.5 W, as an average value, as the total value of the 6 IGBTs. (f actor ∗ gI ) + (λsw npsw )
∗ 2 ∗
On the other hand, the switching losses are shown in Fig. 5 gI = (If α − Ioα + Cωref (Vcβ )) + (If β − Ioβ + Cωref (Vcα ))2
and they are approximately 15.7 W, as an average value, for (19)
the 0.5 kW rated output power. Notably, the average time for where, the tracking of the voltage variables is already
the calculations of the losses is Ts . All the values of either accomplished by the first two terms in Eqn. (18). As proposed
the conduction or the switching losses are calculated for one in [8], the inability to control derivative of the capacitor
fundamental cycle (50 Hz). In Fig. 6, the total conduction and voltage causes high THD in the conventional scheme. The term
switching losses are deployed. gI is ensured that the voltage derivative tracks the reference
and its importance is controlled by f actor. This could be C. Constraints and limitations
obtained from the predicted currents, which are calculated
from the discrete model of the VSC. Hence, the considered Since the voltage in the case of the conventional CF is
CF, as described in (18), can predicts the voltage, control deteriorated and in order to have a fair comparison between the
the current, and control the switching frequency. Figs. 7 and two controllers, a compensation for the voltage has been ap-
8 show the conduction and the switching losses using the plied for the conventional CF. Thus, the three phase reference
improved CF respectively. It can be noticed that the losses voltages, applied to the VSC, are 200 V for the improved CF,
are reduced compared to the conventional CF. Fig. 9 shows while the reference voltages are 201.5 V for the conventional
the sum of both losses. CF. Moreover, in both CFs, a load variation was applied to
the VSC. The variation of the linear load is done in order
to have a gradual increment in the rated output power. The

output power is varied from 0 W until 1000 W with 100 W
intervals. In addition, the simulation is measuring the THD,
 the output voltage, and the switching frequency for both CFs
at different rated power. The average switching frequency for
1PXFS
8

all tests was 6.6 kHz ±100 Hz. The losses evaluation is done
 for linear loads and future work will include non-linear loads
for the both CFs. Furthermore, all the generated results were
captured for one fundamental cycle.
  
7LPH6HF 25
20
Power loss (W)

Fig. 7. The conduction losses for one fundamental cycle of the improved CF
at 0.5 kW load power. 15
10
5
0
0 100 200 300 400 500 600 700 800 900 1000
Load power (W)

3RZHU:

Conventional CF Improved CF

 Fig. 10. Total losses for the conventional and the improved CF at different
output power.


  
7LPH6HF D. Results discussion

Fig. 8. The switching losses for one fundamental cycle of the improved CF
Fig. 10 shows the total losses for the conventional and
at 0.5 kW load power. the improved CFs. The THD for each interval of the rated
output power is shown in Fig. 11. It can be seen that the
improved method has a better THD and lower losses below
1 kW. Fig. 12 illustrates the VSC efficiency when both CFs
 are applied. Similarly, the efficiency in the case of low loads
for the VSC is slightly better than the conventional CF. It is
worth to mention that both CFs have approximately the same

3RZHU:

efficiency at 1 kW but the THD with the improved method

is significantly better. Fig 13 proves that the reduction of the
 losses is relatively high at lower power and it continues to
decrease at higher rated output power. Fig 14 shows the total
losses in the experimental set-up. The power loss has been

   measured using the power analyser. Figs 15 and 16 depict the
7LPH6HF THD of one fundamental cycle using the conventional and the
improved CFs, respectively. The THD for the current control
Fig. 9. The switching and conduction losses for one fundamental cycle of is reduced at the same operating switching frequency. Finally,
the improved CF at 0.5 kW load power. Figs. 17 and 18 show the distribution of the conduction and
switching losses for both CFs.
 2 95

Efficiency (%)
1,5 90
THD (%)

1 85

0,5 80

0 75
0 200 400 600 800 1000 1200 380 580 780 980 1180
Load power (W) Load power (W)
THD for Conventional CF THD for Improved CF
Conventional CF Improved CF
Fig. 11. THD for the conventional and the improved CF at the same switching
Fig. 14. Total losses for the conventional and the improved CF at different
frequencey.
output power for the expermintal set-up.

97   !
96
Efficiency (%)

95 
94

93
92

2XWSXWYROWDJH9

91
90 
0 200 400 600 800 1000 1200
Load power (W) 

Conventional CF Improved CF 


Fig. 12. Efficiency for the conventional and the improved CF at different
intervals of rated power.

  
45%
7LPH6HF
40%
Loss reduction (%)

35%
30% Fig. 15. THD for output voltage Vα at 0.5 kW output power using the
25% conventional CF.
20%
15%   !
10%
5% 
0%
0 200 400 600 800 1000 1200 
Output power (W)
2XWSXWYROWDJH9



Fig. 13. Loss reduction when applying the improved CF. 


V. C ONCLUSION 
In this paper the VSC is controlled using the finite control

set model predictive control. The discrete form of the system
predicts the future value of the voltage for one sampling 
time. Two different CFs are applied and compared in terms of   
switching and conduction losses. This paper has demonstrated 7LPH6HF
the effect of additional variables in the CF to reduce the total
losses for the VSC below 1 kW. Reducing the percentage of
Fig. 16. THD for output voltage Vα at 0.5 kW output power using the
losses can improve the overall efficiency. Moreover, the THD improved CF.
for the improved CF is significantly reduced. All simulation
and experimental results are conducted for the SKM 50GB
123D power module.
 Conduction [3] P. C. Loh and D. G. Holmes, “Analysis of multiloop
losses control strategies for lc/cl/lcl-filtered voltage-source and
17%
current-source inverters,” IEEE Trans. Ind. Appl., vol. 41,
no. 2, pp. 644–654, March 2005.
[4] S. Buso, S. Fasolo, and P. Mattavelli, “Uninterruptible
power supply multiloop control employing digital pre-
dictive voltage and current regulators,” IEEE Trans. Ind.
Appl., vol. 37, no. 6, pp. 1846–1854, Nov 2001.
[5] J. Rodriguez, P. Cortes, R. Kennel, and M. P.
Kazrnierkowski, “Model predictive control – a simple
Switching losses
83% and powerful method to control power converters,” in
Proc. of 2009 IEEE 6th International Power Electronics
Conduction losses Switching losses
and Motion Control Conference, May 2009, pp. 41–49.
[6] J. Rodriguez, J. Pontt, C. A. Silva, P. Correa, P. Lezana,
Fig. 17. Distribution for conduction and switching losses using the conven- P. Cortes, and U. Ammann, “Predictive current control
tional CF at 0.5 kW load.
of a voltage source inverter,” IEEE Trans. Ind. Electron,
vol. 54, no. 1, pp. 495–503, Feb 2007.
Conduction
losses [7] S. Vazquez, J. Rodriguez, M. Rivera, L. G. Franquelo,
15% and M. Norambuena, “Model predictive control for
power converters and drives: Advances and trends,” IEEE
Trans. Ind. Electron, vol. 64, no. 2, pp. 935–947, Feb
2017.
[8] T. Dragicevic, M. Alhasheem, [Link], and F. Blaabjerg,
“Improved model predictive control for high voltage
quality in ups applications,” in proc. of 2017 IEEE
Energy Conversion Congress and Exposition (ECCE),
Switching losses
85% Cincinnati, OH.
[9] J. Rodriguez and P. C. Estay, Predictive control of power
Conduction losses Switching losses
converters and electrical drives. Chichester West Sussex
UK: Wiley-IEEE Press, 2012.
Fig. 18. Distribution for conduction and switching losses using the improved
CF at 0.5 kW load.
[10] B. Gadalla, E. Schaltz, Y. Siwakoti, and F. Blaabjerg,
“Thermal performance and efficiency investigation of
conventional boost, z-source and y-source converters,”
R EFERENCES in Proc. of 2016 IEEE 16th International Conference on
[1] P. Cortes, J. Rodriguez, S. Vazquez, and L. G. Franquelo, Environment and Electrical Engineering (EEEIC), June
“Predictive control of a three-phase ups inverter using 2016, pp. 1–6.
two steps prediction horizon,” in Proc. of 2010 IEEE In- [11] SEMIKRON, in IGBT Power Electronics Teaching Sytem
ternational Conference on Industrial Technology, March Principle for Sizing Power Converters, Application Note
2010, pp. 1283–1288. AN-8005.
[2] P. Mattavelli, “An improved deadbeat control for ups
using disturbance observers,” IEEE Trans. Ind. Electron,
vol. 52, no. 1, pp. 206–212, Feb 2005.

View publication stats