1

MPC Based Phase Angle Adjustment for Optimal

Power Transfer in Neutral Point Clamped Dual
Active Bridge Converter
Somnath Meikap Ram Gopal Chandan Kumar Jose Rodriguez
Student Member, IEEE Student Member, IEEE Senior Member, IEEE Life Fellow, IEEE
EEE Dept., IIT Guwahati EEE Dept., IIT Guwahati EEE Dept., IIT Guwahati Universidad San Sebastian
Guwahati, India Guwahati, India Guwahati, India Santiago, Chile
somnathmeikap@gmail.com r.gopal@iitg.ac.in chandank@iitg.ac.in jose.rodriguezp@uss.cl

Abstract
For high-voltage applications, neutral point clamped-based dual active bridge (NPC-DAB) converters are commonly
used. In literature, different fixed phase shift control methods like double phase shift control (DPS), triple phase shift
control (TPS), etc., are used to reduce losses in NPC-DAB converter. In this paper, a variable phase shift method is
proposed to reduce the losses in the converter. This variable phase shift angle is obtained by utilizing a power loss
minimization function in a model predictive controller (MPC). The simulation result shows that MPC based variable
phase shift control reduces power loss and improves transient performance in the system in comparison to PI control
based fixed phase shift control.
Index Terms
Neutral point clamped, model predictive control, dual active bridge.
I. I NTRODUCTION
Recently multilevel dual active bridge (DAB) converters have been explored for medium-voltage dc (MVDC) systems, high
voltage dc transmission, high step-up dc-dc converters, etc. [1]. These DAB converters improves conversion efficiency by
achieving higher voltage blocking capability, higher degree of freedom, and wider zero-voltage switching (ZVS) operation
range.
In [2], it is shown that fixed phase shift control techniques like triple phase shift (TPS) control gives higher power losses in
comparison to hybrid control. The hybrid control is designed to regulate the transferred power and its direction by one phase
shift angle and other phase shift angle controls the inductor current peak. In [3], a nondominated sorting genetic algorithm II
(NSGA-II) based rms current and peak current optimization control is designed for three-level ANPC-DAB system for better
efficiency. But, genetic algorithm based controls requires large number of iterations to reach optimal point. This increases
computational burden for controller.
This paper proposes a variable phase shift angle control technique to reduce power loss in the NPC-DAB operation. The
controller technique consists of two parts, one is a searcher and another is optimisation controller. A binary searcher is designed
to find all the possible phase shift angles for the required output power from a dataset containing all possible output power.
A model predictive controller (MPC) is designed to optimize the power loss and current ripple to get the optimal phase shift
angles.

II. NPC-DAB S YSTEM C ONFIGURATION

The circuit configuration of NPC-DAB converter is shown in Fig 1. A 3/2-level NPC-DAB converter consists of a three-level
NPC bridge at the primary side and a two-level H-bridge at the secondary side. The converter system is designed for battery
charging application. Vin is the input dc link voltage, Vo is the output dc link voltage, Vpri is the input port voltage, Vsec is
the output port voltage and VBat is the battery terminal voltage in the NPC DAB converter.
The NPC-DAB system uses a four-level operation at the MVDC side, maintaining capacitor voltage balance. The voltage of
the NPC bridge to other H bridges is shown in Fig. 2, with angle D1 representing phase shift and D2 representing half turn
off period of upper switch.
D12 D22
   
nVin Vo nVin Vo
Po = D1 − − D1 D2 , 0 < D1 < D2 , Po = D1 − D12 − , D2 < D1 < 0.5 (1)
2fs L 2 2fs L 2

III. O PERATION AND MPC D ESIGN FOR NPC-DAB

The NPC-DAB converter is designed to operate over a wide range of input and output voltages. This is particularly helpful
for a battery charger for various terminal voltages. The predictive controller is designed to select the phase shift angles to
reduce power loss.
 2

Ii
D2
S1 S3 Vin
+ Io Vin/2
LBat
VIn D1 D3 -Vin/2 t
S3l L 1:n Q1 Q3
2 S1l VBat - Vin
Vo (a)
iL Vo D1
Vpri Vsec
D2 D4
VIn t
2 S2l S4l - Vo
Q4 (b)
Q2

S2 S4
Fig. 2. (a) ac output voltage of the NPC bridge (bridge 1) and (b) ac phase
shift voltage of H bridge in the proposed control technique.
Fig. 1. NPC-DAB converter circuit.

A. Operation of the NPC-DAB

The operation of NPC-DAB can be divided into three cases depending upon the value of phase shift angles D1 and D2 .
The three cases are D1 = D2 , D1 < D2 and D1 > D2 . The inductor current’s rms for the case D1 = D2 is as follows
Vin D1 Ts p
IL = (1 − D1 ). (2)
L
Likewise, when D1 < D2 , the inductor current’s rms value is as follows.
r
Vin Ts n (D1 +D2 )(3D22 +3D12 −D1 D2 )
IL = 18 + (1 − 2D2 )D12 . (3)
L
Similarly for the case of D1 > D2 , the rms value of inductor current is as follows
r
Vin Ts (5D13 + 29D12 D2 + 44D1 D22 + 5D23 ) 2
IL = + (1 − D2 − D1 )D12 . (4)
L 36 3
Utilizing (2)-(4), the inductor current rms value can be predicted for a particular set of D1 , D2 and Vin . Here, Ts is half time
period of DAB converter and L is the series inductor value in the DAB system.

B. Model Predictive Controller Design

The model predictive controller (MPC) is divided into two parts. One is a binary search algorithm and other is the optimization
of power loss using inductor current model of NPC DAB (as shown in Fig. 3).
First, a dataset is created containing output power values for all possible D1 and D2 values using (1). After that depending
upon the required output power, all possible sets are searched using binary searching technique and sent to MPC module for
optimization (in this case power loss minimization). Next the MPC module finds the optimal values of phase shifts D1o and
D2o depending upon the required output power, the state of charge (SoC) of battery, and output current. The SoC is used to
select the mode of charging, i.e. constant current (CC) or constant voltage (CV) operation. The MPC selects the cost function
depending on CC or CV operation and sends the optimal phase shift angles directly to the logic gate or the PI controller. The
optimal phase shift angles are used to reduce the operation region of PI controller for better transient performance. Finally,
the logic gate selects the optimal phase shift angles from the MPC module or PI controller depending upon SoC of battery
and sends it to the modulator for switching pulse generation for NPC-DAB.

C. Power Loss Optimization

The optimization problem is to explore the optimum combinations of the control variables (D1 and D2 ) to reduce the power
loss. To optimize the power losses, it is needed to reduce both conduction loss and switching losses. The conduction loss and
switching losses are given as follows [4]
2 2
2
Vin /2 · IL,rms (Vin /2) · |IL,rms |
F1 = IL,rms req and F2 = + (5)
2 · si 2 · sv
where IL,rms is the rms value of inductor current at the primary side, req is the equivalent series resistance, si and sv are the
current and voltage transition slopes, and Vin is the input voltage. Considering Vin , req , si and sv as constant throughout the
operation. The optimization expression to reduce losses is directly dependent on the magnitude of inductor current rms.
 3

For D1 < D2 For D1 > D2

For D1 = D2 Considering that Vin = nVo Considering that Vin = nVo
Considering that Vin = nVo Utilizing the inductor equation. Utilizing the inductor equation.
Utilizing the inductor equation.
dIL dIL
dIL L = VL (8) L = VL (10)
L = VL (6) dt dt
dt
We get the value of I1 , I2 and I3 as We get the value of I1 , I2 and I3 as
We get the value of I1 and I2 as shown in Fig. ??(d). shown in Fig. ??(d).
shown in Fig. ??(d).
Vin (D1 + D2 )Ts Vin D1 Ts Vin D1 Ts Vin (2D1 − D2 )Ts
Vin D1 Ts Vin D1 Ts I1 = , I2 = , I1 = , I2 = ,
I1 = , and I2 = 2L L L 2L
L 2L
(7) Vin (2D1 −D2 )Ts Vin (2D2 −D1 )Ts
and I3 = . (9) and I3 = 2L . (11)
2L

Start

Po Io Vo IL Po Io SoC CC controller
D1o Get the discrete value of output power, input voltage, output voltage
Binary and inductor.
MPC D2o
searcher Modulator
D1o D2o Logic /pulse Search all possible set of D1 and D2 for the required power from a dataset
D1 and D2 set SoC case generator made from (1)

Vbsat PI All possible datasets of D1 and D2 sent to MPC.

CV controller
Vb IF D1= D2
No
Yes
IF D1 > D2
Fig. 3. NPC-DAB converter controller block diagram. Yes
Calculate predicted
Calculate predicted Calculate predicted inductor current as
TABLE I inductor current as per (3) inductor current as per (4) per (2)
NPC-DAB C ONVERTER S YSTEM D ETAILS .

Parameters Values No CV mode Yes

IF CC
Active bridge 1 inductance 15 µH mode
NPC bridge 1 voltage 500 V
Calculate cost function (J) as per (8) Calculate cost function (J) as per (9)
Active bridge 2 voltage 120 V
Switching frequency 10 kHz Find minimum (J) Find minimum (J)
Battery nominal voltage 100 V
Battery nominal capacity 25 Ahr Send the optimal value of D1 and D2 Send the optimal value of D1 and D2
of Jmin to the PI controller of Jmin to the pulse generator
Battery active bridge dc link capacitor 4.7 mF
Battery active bridge output inductor 0.12 mH
Battery current charging rating 30 A Stop
LVDC active bridge dc link capacitor 4.7 mF
Fig. 4. NPC-DAB MPC flow chart.

F = k|IL | or F = k|IL2 |. (12)

But in case of CC operation, the equivalent primary side ac current (ILr ) for the output reference current is calculated. Then
the optimization equation to minimize the inductor current rms changes as follows.
2
F = k|ILr − IL | or F = k|ILr − IL2 |. (13)

For CV operation, the cost function for the optimization operation is as follows
J = a1 |(Por − Po (k + 1)| + a2 |IL (k + 1)| (14)

where Por is the reference output power. IL (k + 1) is the calculated rms value of inductor current from all the possible sets
of D1 and D2 . The first part checks that the output power of the predicted variable matches the reference power. The second
part tries to find the set with minimum inductor current rms value as this leads to lower conduction losses in switches and
transformers. a1 and a2 are constants in cost function used as weighting factor. They are tuned to reduce power loss without
affecting the output power. Similarly, in case of CC operation the cost function for the operation is as follows
2
J = a1 |(ILr − IL2 (k + 1)| + a2 |(Por − Po (k + 1)| (15)

where ILr is the ideal inductor current rms value with no harmonics. The first part of (9) tries to bring the inductor current
to ideal region so that the harmonics in the inductor current is reduced. This will ultimately reduce the conduction loss and
switching loss by minimizing the inductor current to the required current value. Second part tries to keep the output power
 4

matches required value. a1 and a2 are constants in cost function used as weighting factor. Fig. 4 explains the whole MPC
operation in form of a flowchart. In case of CV operation, the optimal values of D1 and D2 are given to PI controller from
MPC module for faster transient operation with minimal power loss in comparison to conventional PI control operation, as the
phase shift angles are restricted near optimal values (D1o and D2o ).

500 Input port voltage 500 Input port voltage

Voltage (V)

Voltage (V)
0 0

-500 -500
(a) (a)
Output port voltage Output port voltage
100 100

Voltage (V)
Voltage (V)

0
0
-100
-100 (b)
(b) 80
80 Input Inductor current 40 Input Inductor Current

Current (A)
Current (A)

40
0 -40
-40
-80
-80 1.13772 1.13776 1.13780 1.13784 1.13788
1.13772 1.13776 1.13780 1.13784 1.13788 (c) Time (sec)
(c) Time (sec)
Power Loss (W)
100
Power loss (W)

110
88
108
86
104
0 0.1 0.2 0.3 0.4 0.5 0.6
0.1 0.2 0.3 0.4 0.5 (d) Time (sec)
0
(d) Time (sec)
Fig. 6. Simulation results of NPC-DAB converter during proposed control
technique with MPC based optimal power control and PI control based CV
Fig. 5. Simulation results of NPC-DAB converter during PI control based operation of battery charging. (a) Voltage of active bridge 1. (b) Voltage of
CV operation of battery charging. (a) Voltage of active bridge 1. (b) Voltage active bridge 2. (c) Active bridge 1 inductor current. (d) NPC-DAB power
of active bridge 2. (c) Active bridge 1 inductor current. (d) NPC-DAB loss during proposed control technique with MPC based optimal power
power loss during PI control based CV operation of battery charging. control and PI control based CV operation of battery charging.

IV. S IMULATION R ESULTS

Table I shows the DAB converter design details used for simulation. Fig. 5 and 6 show the simulated results during the
conventional PI control in DAB, and the output results when proposed method is used. Fig. 5(a)-(d) shows voltage of active
bridge 1, voltage of active bridge 2, active bridge 1 inductor current, and power loss during conventional PI control operation.
Fig. 6(a)-(d) shows voltage of active bridge 1, voltage of active bridge 2, active bridge 1 inductor current, active bridge 1
inductor voltage, and power loss during proposed control operation. Fig. 5(c) and 6(c) shows the variation in the slope of
inductor current due to the application of proposed control technique. Lastly, Fig. 5(d) and 6(d) shows that there is significant
change in power loss during conventional PI control to proposed control technique. Fig. 5(d) shows the oscillation due to
inherent nature of PI control to move near its reference value. This leads to oscillation at the output power. Fig. 6(d) shows
lower oscillations in power loss because the MPC restricts the operating region of PI control around optimal power loss region.
The power loss is calculated by subtracting input power from output power during both control. The results show that the
proposed control strategy lowers power loss by 12 percent compared to PI control.

V. C ONCLUSION
A phase angle adjustment technique is proposed for an NPC-DAB based battery charger to reduce power loss by selecting
the optimal phase shift angle. The proposed control system consists of a binary search algorithm and a model predictive control
(MPC) based power loss optimization technique. It provides a direct reference for constant current operation and an optimal
phase shift angle range to the voltage controller for constant voltage operation. The proposed controller provides better transient
response and reduced inductor current ripple. It also reduces transients in the output power. In the final paper, detailed analysis
and experimental results with the proposed control technique will be shown.
 5

R EFERENCES
[1] A. Filba-Martinez, S. Busquets-Monge, and J. Bordonau, “Modulation and capacitor voltage balancing control of multilevel npc dual active bridge dc–dc
converters,” IEEE Trans. Ind. Electron., vol. 67, no. 4, pp. 2499–2510, 2020.
[2] J. Dong, J. Pou, X. Li, Y. Zeng, J. Kotturu, M. Cupelli, and A. K. Gupta, “Switching characteristics analysis for hybrid neutral point clamped dual active
bridge with hybrid duty ratio phase-shift modulation,” in IECON 2023- 49th Ann. Conf. IEEE Ind. Electron. Soc., 2023, pp. 1–4.
[3] J. Tian, C. Zhuo, F. Wang, and H. Deng, “Full-power-range multiobjective optimization scheme for three-level anpc-dab-based distributed generation
system,” IEEE J. Emerg. Sel. Top. Power Electron., vol. 12, no. 2, pp. 1754–1768, 2024.
[4] A. Filba-Martinez, S. Busquets-Monge, J. Nicolas-Apruzzese, and J. Bordonau, “Operating principle and performance optimization of a three-level npc
dual-active-bridge dc–dc converter,” IEEE Trans. Ind. Electron., vol. 63, no. 2, pp. 678–690, 2016.