Two-level inverter and three-level neutral point diode clamped inverter for traction applications: a comparative analysis study

Asmaa A. Abdelsalam, Mohamed Adel Esmaeel, Salama Abo Zaid Tabbin Institute for Metallurgical Studies, Cairo, Egypt Department of Electrical Power and Machines Engineering, Faculty of Engineering, Helwan University, Cairo, Egypt Faculty of Engineering and Technology, Badr University in Cairo (BUC), Cairo, Egypt Department of Electrical Engineering, Faculty of Engineering, Al-Azhar University, Cairo, Egypt


INTRODUCTION
The motivation for the diminution of fuel consumption and harmful emissions in transportations increased the interest in developing the electrified power trains in the automotive industry, besides the advantages of low maintenance, cost-effectiveness, safety drive, popularity, and no noise pollution. Since breakdown is not accepted on the road, reliability is in high demand. Also, the allowed weight of the vehicle is a challengeable topic as mass production at low cost is a requirement. These matchless demands and tradeoffs in traction applications pose significant challenges for the electrical propulsion system of EVs. The EV propulsion system mainly consists of a motor, power inverter, and an educated control system. EV manufacturers majority utilize a two-level inverter for machine drives that is a well-established technique. Since the reduction of passive components in the traction drive, advanced control of the electric machine, and a rise of the inverter power density is a demand, the switching frequency increase of the converter to several kHz is a must. At high switching frequencies, two-level inverters have high power losses that reduce the inverter efficiency [1].
The recent direction for high power high voltage electric drives are multi-level inverters since they prove the production of an enhanced output voltage waveform, less total harmonic distortion (%THD), lower switching losses at the high switching frequency, reduced electromagnetic interference (EMI) emission, and lower dv/dt [2]- [6]. The three-level Neutral Point Clamped Inverter (NPCI) has been introduced in 1981 by Akira Nabae and Isao Takahashi [7], then it has been taken into account in many medium and high-power industries. Above 6.5 kV, high voltage locomotives with high-speed, (NPCI) has found acceptance because of their various advantages like more minor switching losses, lower distortion, and the improved voltage waveform. Recently, many research papers mentioned that (NPCI) exhibits excellent accomplishment in terms of reliability, efficiency, and better frugal advantages, especially during high switching frequency operation. The switching losses of the NPCI are less than that for a two-level inverter for the same DC bus voltage as NPCI has a lower voltage rating for switches so when the switching frequency comes up, the switching losses go down [5], [8]. Hence, nowadays, manufacturers are directed to three-level NPCI for electric machine drives since it provides an increase of the dc voltage level despite the fact of its drawback; the capacitors' DC-link voltages unbalance.
For an EV, the most significant features of the electric motor are high reliability, low maintenance requirements, high efficiency, reliable drive control, besides the ability to handle voltage fluctuations of the source. Motors represented for EVs are Induction machines (IM), DC machines, permanent magnet synchronous machines (PMSMs), and switched reluctance machines (SRMs). Among the mentioned machines, the Induction motor has the advantages of elevated max. speed, a wide range of field weakening, higher power density, low no-load current, sturdy design, and low production costs. Nowadays, the semiconductors and power electronics uprising make the IM a strong candidate with better chances on the market for vehicle propulsion. Recently, multiple research has paid attention to an analytical comparison between two-level, Figure 1 (a), and three-level NPC inverters, Figure 1 (b) [1], [3], [5]. Several works of literature have mentioned the switching losses and harmonics comparisons however, steady-state and transient operation at different speeds is not discussed.
This research states a comparison between a two-level voltage source inverter and a three-level neutral point diode clamped inverter NPCI used for IM traction drives in terms of the switching losses, quality of the output voltage waveform, total harmonic distortion (THD%), steady-state and transient response taking into account the dynamic PF of the IM operation by MATLAB/Simulink. Figure 2 shows a simplified block diagram of the scheme. The most popular pulse width modulation (PWM) techniques are multi-carrier sine pulse width modulation (SPWM) [9]- [11], Space Vector Pulse width Modulation (SVPWM) [12]- [16], and selective harmonic elimination PWM (SHE) [17]- [19]. The comparison between different techniques is made based on the output line voltage THD, the common-mode voltage, and the utilization of DC-link voltage [20], [21]. Figure 3 presents the classification of various PWM techniques. The most commonly used PWM control strategies are carrier-based PWM and space-vector-based PWM (SV-PWM) techniques due to the advanced PWM quality. In this work carrier-based, continuous PWM is applied to both Two-level and three-level inverters.

OPERATING PRINCIPLE OF THREE-LEVEL NPCI
A Neutral point clamped three-phase inverter scheme is shown in Figure 1 (b). Each phase contains four IGBT switches and two clamping diodes (D1 and D2 for leg a), the midpoint of the two diodes is linked to the DC-link capacitors midpoint (O) [22]. Figure 4 illustrates phase/leg (a) as an example. Three statues of operation are shown. The neutral point (O) represents the zero-potential point.
First status, Figure 4 (a): the switches (Sa1 and Sa2) are on, allowing the current ia1 to flow from the DC link to the load while switches (Sa3 and Sa4) are off, the output voltage is Vdc/2. The current ia2 flows from the load to the DC power supply through the diodes of Sa1 and Sa2 that also provides an output voltage equal to Vdc/2. Second status, Figure 4 (b): the switches (Sa2 and Sa3) are on, allowing the current ia1 to flow from the DC link to the load through D1 and Sa2 while switches (Sa1and Sa4) are off, producing an output voltage equal to zero. The current ia2 flows from the load to the DC power supply through D2 and Sa3 that also provides an output voltage equal to zero. Third status, Figure 4 (c): the switches (Sa3 and Sa4) are on, allowing the current ia1 to flow from the DC link to the load through the clamping diodes of Sa3 and Sa4 while switches (Sa1and Sa2) are off, producing an output voltage equal to -Vdc/2. The current ia2 flows from the load to the DC power supply through Sa3 and Sa4 that also provides an output voltage equal to -Vdc/2.

PERFORMANCE ANALYSIS
The performance evaluation of the introduced work is modeled by MATLAB/Simulink. A closedloop scalar control V/Hz based induction motor drive is presented to obtain a transient response with different speeds, the quality of the waveform, the power loss analysis for both switching and conduction periods for full range distribution of the modulation index and load PF angle is introduced. The simulation parameters are: A DC-link voltage of 600 V for both two-level and three-level NPCI. The NPCI capacitors are 500 μF each, and the switching frequency is 5 kHz. The induction motor rating is 400 V, 5.4 HP, 4-pole, 50 Hz,1430 rpm, which is loaded by a dc-generator coupling.

Steady-state operation
This section inspects the steady-state performance of the applied scheme for both inverters.

Transient operation
The scheme attitude is examined during transients by changing the speed for the broad zone of operation. Figure 7(a)-(d), and Figure 8 (a)-(d), represent the performance for the two inverters for constant load torque of 27 N.m; while the speed varies, in the beginning, the induction motor operates with 500 rpm for 0.4 sec, then is raised to reach the rated speed, 1430 rpm for another 0.4 sec. Finally, the speed was declined to 1000 rpm. Figure 8 shows the transient performance of a three-level neutral point inverter over a range of different speeds. It is observed that the rise of the modulation index is meet by an increase of the output voltage steps to be closer to a sinusoidal wave-shape distribution if compared with the two-level inverter; as a result, the dv/dt stress across the power switches goes down. In Figure 8 (c), It is noticeable that the behavior of the line output voltage of the NPCI acts as that of the two-level inverter at 500 rpm due to the modulation technique used.

CONDUCTION POWER LOSS CALCULATIONS 4.1. Two-level inverter
The two-level inverter conduction losses have been investigated for the inverter switches, IGBT and its anti-parallel diode over a full range distribution of the modulation index (m) (0 < m < 1) and a full range of the phase angle (φ) (0 < φ < π/2). The conduction losses of a single IGBT switch and its anti-parallel diode could be described as given in (1) Where is the average conduction losses in power switches and is the average conduction losses in the anti-parallel diodes.
is the saturation voltage of the anti-parallel diodes while is the saturation voltage of the IGBT switch. The maximum phase peak current is denoted as . Figure 9 shows the conduction losses of the two-level inverter at different modulation indexes and different phase angles. The IGBT losses are shown in Figure 9 (a), while the ani-parallel diode losses are depicted in Figure 9 (b). It is evident that with the modulation index rise, the conduction losses of the IGBT come up while the opposite acquires with their anti-parallel diodes; the conduction losses decrease with the modulation index increase. In contrast, the rise in the power factor angle reduces the IGBT conduction losses and shift up the anti-parallel diode losses.

Three-level NPCI
For a 3-level NPCI, the IGBT switches Sa1 and Sa4 have the same conduction period; thus, both switches have the same conduction losses ) 1 , 4 . Similarly, with switches Sa2 and Sa3. The average IGBT switches conduction losses ) 2 , 3 are expressed in (3)  While the conduction losses of the clamping diodes D1 and D2, ) 1 , 2 are given in (5) and for the anti-parallel diodes Da, in (6).  Figures 10 (c) and 10 (d), respectively. The rise of the modulation index is met by an increase in the conduction loss while it declines with the increment of the power factor angle. Unlike the twolevel inverter, Figure 9, the 3-level NPCI switches do not have equal conduction power losses distribution, which leads to inhomogeneous heat distribution over the different switches. The anti-parallel diode of the NPCI IGBT switch has almost half of the maximum conduction loss of the two-level inverter's one.

SWITCHING POWER LOSS CALCULATIONS 5.1. Two-level inverter
The voltage across the switch and the load current are defining its power-switching losses. Turn-on "Eon "and turn-off "Eoff" losses are the observed loss for an IGBT. Commonly both of them are defined by the manufacturer. The diode has only reverse recovery losses during turn-off periods Err. Hence the average switching losses for an IGBT is and for the anti-parallel diode is can be expressed as in (7) and (8) Considering = + is the total switching losses as a function of the load current, the and can be expressed as shown in (9) and (10): Replacing the values of the switching losses from (9) and (10) into (7) and (8) From the (11) and (12), it is perspicuous that the switching losses are mainly related to the switching frequency. The losses are directly proportional to the switching frequency, and it does not depend on the modulation index.

Three-level inverter
The average switching power losses of the NPCI switches can be calculated as [2]: From expressions, it is clear that the four IGBT switches for the three-level NPC inverter have the same switching power loss distribution, same for the anti-parallel diodes and the two clamped diodes. The switching losses depend on the switching frequency and do not rely on the modulation index. Since the number of switches/leg is greater than that of the two-level inverter then the NPCI switching losses for the single switch are lower.

THE THREE-LEVEL NPCI CAPACITOR VOLTAGES UNBALANCE
In an EV that runs with an NPCI induction motor drive, a periodic dc-link voltage unbalance may exist due to the dynamic operating conditions. The major reason for this unbalance is the flow of average 1617 current in the neutral point because of the unregular switching of the nonideal components; the modulation topology applied is also a reason. This non-zero neutral current generates accelerated fluctuations that cause the voltage two unbalance the two dc-link capacitors. For steady, reliable operation of an NPCI, these voltage fluctuations, shown in Figure 11 are unwanted for balanced EV traction inverter drive. The modulation technique applied should be able to reduce this unbalance issue in case no external circuit is added. Many researchers have reported solutions for the unbalance of the neutral current [23]- [26]. In [26], the neutral point current was reduced by adding a measured zero-sequence voltage signal to the reference voltages, which improved the converter performance that is highly recommended for EVs applications. Figure 11. The NPCI capacitors voltage unbalance

TOTAL HARMONIC DISTORTION (THD%)
The output line voltage THD% of the two-level inverter is illustrated in Figure 12 (a) while, the NPCI output voltage THD% is presented in Figure 12 (b). Form Figure 12 (a) and Figure 12 (b), it is obvious that the NPC 3-level inverter THD% is almost 50% less than that of the two-level inverter at the same modulation index. The THD% remarkable reduction leads to the deduction of the size of the passive components such as the inductor of the interference filter on the load side and the electromagnetic filter that existed at the source side, which leads to the reduction of the system power losses existed at the source side.

CONCLUSION
This paper compares the operation of a two-level inverter and a three-level NPCI traction IM drive using MATLAB/Simulink. The comparative performance analysis has been classified into different attributes, which are the steady-state operation, the transient operation at a wide range of speed variation, the power loss, including both conduction and switching losses calculations for a full range of modulation index distribution and PF angle variation, waveform quality, and the total harmonic distortion of the output line voltage for both inverters. Also, the dc-link capacitors' unbalance voltage of the NPCI is analyzed. The output results of the applied scheme show its applicability with both inverter drives. The total harmonic distortion of the NPCI output line voltage is 50% lower than that for the two-level inverter at the same 1618 modulation index using the control topology introduced. Hence, Using the three-level neutral point diode clamped as an induction motor drive-based Electric Vehicle is considered an attractive candidate.