This paper proposes a maximum efficiency operation strategy for threelevel Ttype inverter in entire operation areas. The threelevel Ttype inverter has higher and lower efficiency areas compared with twolevel inverter. The proposed strategy aims to operate in the maximum efficiency point for the lowvoltage and lowpower home appliances. The threelevel Ttype inverter is analyzed in detail, and the two operation mode selection strategy is developed. The proposed algorithm is verified by theoretical analysis and experimental results.
1. Introduction
Voltage source inverters are widely used in many applications, such as industrial systems and home appliances, to achieve energy conservation and to improve high motion control quality. Especially, the twolevel inverter to drive a multipole permanent magnet synchronous motor is a general choice for modern home appliances and industrial systems
[1
,
2]
. In recent years, the lowvoltage applications with 270V~600V DC link voltage for home appliances is continually increased and for higher efficient energy conversion and low cost, the microcontrol unit and the power electronics technology are advanced
[3]
.
A twolevel voltage source inverter is commonly used for home appliances and industrial systems since its configuration is simple and the reliability is sufficiently ensured. However, there is a limit to improve the twolevel inverter efficiency and performance because twolevel inverter output voltage is decided by ±V
_{DC}
/2. Furthermore, the inverter efficiency and performance are varied according to the switching method. In the case of the sixstep modulation, the switching loss can be reduced, but the copper loss and torque ripple are increased because of the low frequency harmonics. The pulse width modulation method (PWM) is commonly used to solve the harmonic problem of this sixstep method in many applications. PWM methods can be reduced the copper loss and the torque ripple compared with sixstep method since PWM method switching frequency is higher than sixstep method, but the switching loss becomes increased. The switching loss is more prominent when the switching frequency is increased and the DC link voltage becomes higher
[4]
. Therefore, in order to overcome these limitations of the twolevel inverter efficiency and performance, a threelevel inverter is being researched in various applications for further improvement of energy efficiency, reliability, power and density
[5]
.
The multilevel topology such as the threelevel inverter has been developed to drive mediumvoltage level (DC link voltage: 600V~2500V) and highvoltage level (DC link voltage: 2500V~) since a semiconductor voltage blocking capability in conjunction is limited. Furthermore, it is verified that the multilevel topology offered a superior harmonic spectrum, lower overvoltage stress at cable and end windings of motors, lower commonmode voltage and lower switching loss
[6]
. Among various wellestablished multilevel topologies, the neutralpoint clamped inverter (NPCI), flying capacitor inverter (FCI), and cascaded Hbridge inverter (CHBI) have been widely used and investigated as shown in
Fig. 1
[7

14]
. For these topologies, the switching loss in each switch is half and the conduction losses become double of the counterpart of the twolevel inverter due to the two switch series connection. Therefore, these threelevel inverter topologies are not suitable for lowvoltage and lowpower applications. To overcome these characteristics in the multilevel topology, the threelevel Ttype inverter (3LTI) has been proposed for the high efficiency and performance in lowvoltage applications
[15]
.
Conventional multilevel inverter topologies
Circuit configuration of threelevel Ttype inverter
Even though the 3LTI is generally evaluated and investigated to apply in lowvoltage industrial systems
[16
,
17]
, it has difficulties to apply in lowvoltage and lowpower home appliances such as refrigerators and air conditioners because the increased efficiency is not sufficient due to its complexity and cost problem. More accurately, since the utilization rate of switches is varied according to the modulation index (MI), the 3LTI has together higher and lower efficiency points. Therefore, in this paper, the maximum efficiency operation strategy for the 3LTI in entire operation areas is proposed. With the proposed strategy, the 3LTI is always operated in the maximum efficiency point in lowvoltage and lowpower applications. The proposed strategy is theoretically explained in detail and its validity is verified by experiment results.
2. General Characteristics of ThreeLevel Ttype Inverters
 2.1 Operational principle
Basically, the output bridge can be connected one of three states:

1) SX, Highis turnon It is connected to positive DC link voltage level The output voltage is VDC/2

2) SX, N1or SX, N2is turnon It is connected to neutral voltage level The output voltage is 0

3) SX, Lowis turnon It is connected to negative DC link voltage level The output voltage is VDC/2
The 3LTI is implemented by comparing the reference voltage with two triangular waves which are corresponding to the highside capacitor voltage (V
_{DC1}
) and lowside capacitor voltage (V
_{DC2}
). During the period of the positive reference, S
_{X, High}
and S
_{X, N1}
are operated complementary through the relationship between the reference voltage and the triangular wave corresponding to V
_{DC1}
, and S
_{X, N2}
is kept onstate in order to guarantee the current path through the antiparallel diode of S
_{X, N1}
. Therefore, it is possible to control the phase voltage in the period of the positive reference. On the contrary, during the period of the negative reference, S
_{X, Low}
and S
_{X, N2}
are operated complementary through the relationship between the reference voltage and the triangular wave corresponding to V
_{DC2}
, and S
_{X, N1}
is kept onstate. Then, since improper voltage sharing on DC link capacitors leads to an overvoltage on switches and, a failure of the 3LTI, the capacitor voltage change should be reflected in the triangular wave to maintain the DC link voltage balancing.
 2.2 Analysis of operating modes
One phase of the Ttype inverter has six operating modes according to the direction of the output current (i
_{X}
) and reference voltage (V
_{X, RER}
) as shown in
Fig. 3
. Modes 1, 3, 4, and 6 are same as twolevel inverter operation modes. Mode 2 and 5 construct bidirectional current path between the neutral point and the output bridge. Therefore, the operating mode is selected one of twolevel inverter operation modes (Mode 1, Mode 3, Mode 4, Mode 6) and one of bidirectional current path modes (Mode 2, Mode 5) as shown in
Fig. 4
. For instance, in period 1, since the output bridge is connected to the positive voltage level for the negative output current, mode 1 and mode 2 are selected. Therefore, mode 4 and mode 5 are selected in period 2, mode 5 and mode 6 are selected in period 3, and mode 2 and mode 3 are selected in period 4.
Operation mode for one phase of Ttype inverter
Operating mode classification according to the relationship between the reference voltage and the load current
The proportion of each period is changed according to the phase delay. If the phase delay is increased, the proportion of period 1 and period 3 is increased, but the proportion of period 2 and period 4 is decreased. Furthermore, the utilization rate of S
_{X, High}
and S
_{X, Low}
is changed according to the magnitude of V
_{X, REF}
. If the size of V
_{X, REF}
is bigger than the present value, S
_{X, High}
and S
_{X, Low}
are more frequently used. As a result, in the high MI areas, the 3LFI and the twolevel inverter are similar in the utilization rate of switches. Therefore, the 3LFI efficiency is influenced by the phase delay and the MI.
3. Proposed Maximum Efficiency Operation Algorithm
 3.1 Utilization rate of switch for 3LTI
The utilization rate of switches for the 3LTI is influenced by the MI, switching method and reference voltage angle. Especially, the utilization rate of switches is instantaneously varied according to the angle of the reference voltage in each cycle. For example, when the angle of the reference voltage is 30°, the utilization rate of the Vphase switch is always two regardless of the MI and switching method since the Vphase reference voltage is zero voltage. However, in this case, the utilization rate of the Uphase switch and Wphase switch is varied according to the MI and switching method.
Fig. 5
shows the variation of the reference voltage in space vector pulse width modulation (SVPWM) and classification of areas according to the reference voltage. As a result, controllable switches are changed according to the reference voltage. In case 1, case 2 and case 6, V
_{U, Ref}
is the positive value. Therefore, in these cases, V
_{U, Ref}
determines S
_{U, High}
and S
_{U, N1}
switching patterns and the utilization rate of S
_{U, High}
and S
_{U, N1}
. However, in case 3, case 4 and case 5, V
_{U, Ref}
is the negative value. Therefore, in contrast with case 1, case 2 and case 6, V
_{U, Ref}
determines S
_{U, Low}
and S
_{U, N2}
switching patterns and the utilization rate of S
_{U, Low}
and S
_{U, N2}
. Therefore, the instantaneous utilization rate of the switch (S
_{θ}
) for the SVPWM 3LTI can be calculated as
where V
_{SN}
is the offset voltage according to the switching method and θ is the angle of the reference voltage.
Classification of areas according to reference voltage
Basically, regardless of the MI and switching method, the average utilization rate of switches for the twolevel inverter is always three. The average utilization rate of switches for the threelevel NPCI and FCI is always six. However, the average utilization rate of switches for the 3LTI is changed based on the MI and switching method as shown in
Table 1
. Neutral point switches are more used compared with highside and lowside switches in the low MI. Since if the MI is increased, the utilization rate of highside and lowside switches is also increased, the average utilization rate of switches is decreased. Therefore, if the MI is low, conduction losses of the 3LTI are similar to the threelevel NPCI and FCI. However, if the MI is sufficiently high, conduction losses of the 3LTI are similar to the twolevel inverter. Furthermore, in the case of using a discrete pulse width modulation, the average utilization rate of switches is much more decreased.
Average utilization rate of switches according to the MI and the inverter topology
Average utilization rate of switches according to the MI and the inverter topology
 3.2 Efficiency calculation method and proposed maximum efficiency operation algorithm
Commonly, switching losses can be calculated by using the switching energy and switching frequency. However, since all switching actions during a fundamental period have to be considered, two assumptions are required. First, the switching frequency (f
_{SW}
) is higher than the fundamental frequency. Second, the switching actions evenly distributed over the fundamental period. Therefore, the average switching energy losses (E
_{SWITCH}
) and the switching losses (P
_{SWITCH}
) can be expressed as
where V
_{SWITCH}
is the switching voltage, and V
_{BASE}
is the reference switching voltage, used in the datasheet. Using these two value, the switching energy is linearly scaled. θ represents the current angle, I
_{PEAK}
is the maximum current value, and α
_{1}
and α
_{2}
represent the switching intervals within one fundamental period. A and B are the curvefitting constants for IGBTs (A
_{S}
, B
_{S}
) and diodes (A
_{D}
, B
_{D}
) switching energy.
In order to calculate the 3LTI switching losses, operation switches are classified depending on the switching intervals as shown in
Table 2
. V
_{SWITCH}
is determined according to the connection position of the operation switch. The highside and lowside switches (S
_{X, High}
, S
_{X, Low}
) have to block the full DC link voltage. However, since the full DC link voltage is blocked by using two IGBTs in the neutral point switches (S
_{X, N1}
, S
_{X, N2}
), V
_{SWITCH}
of the neutral point switches is half of the full DC link voltage. Therefore, based on
Table 2
, the average switching losses for each switch can be calculated as
Operation switches and operation modes according to the switching intervals for the 3LTI
Operation switches and operation modes according to the switching intervals for the 3LTI
where ø represents the phase delay angle.
Conduction losses can be calculated using the conduction IV characteristics of the IGBT and antiparallel diode. These characteristics can be approximated by using two curvefitting constants with instantaneous current (i), onstate zerocurrent saturation voltage (V
_{0}
) and onstate resistance (R)
Then, the average conduction losses (P
_{CON}
) can be expressed as
where D is a duty cycle, and β
_{1}
and β
_{2}
represent conduction angle intervals within one fundamental period.
The conduction interval and conduction switch can be classified as shown in
Table 3
. For example, in period 2, mode 4 and mode 5 are operated complementary by using S
_{X, High}
, D
_{X, N1}
and S
_{X, N2}
. Therefore, the conduction interval can be determined as 0~(π – ø). The duty cycle is assumed according to the output bridge state change. Based on
Table 3
, average conduction losses for each switch can be calculated as
Conduction switches and duty cycle according to the conduction intervals for the 3LTI
Conduction switches and duty cycle according to the conduction intervals for the 3LTI
where V
_{0,S}
and R
_{S}
is the IGBT onstate zerocurrent saturation voltage and onstate resistance, V
_{0,D}
and R
_{D}
is the diode onstate zerocurrent saturation voltage and onstate resistance, respectively. Therefore, total conduction losses for the 3LTI can be calculated as
Table 4
shows calculation results of switching and conduction losses according to the MI in the half and full load condition. Conduction and switching losses are calculated based on the IKP10B60T (600V, 10A, Infineon), which was selected to implement the lowvoltage and lowpower 3LTI prototype. The percent value represents the ratio of the 3LTI calculation result according to the MI compared with the twolevel inverter. 3LTI switching losses are smaller than twolevel inverter, and conduction losses are varied from 193% to 133% according to the MI. Consequentially, the 3LTI has together higher and lower efficiency areas compared with the twolevel inverter. Especially, in the high MI, the 3LTI efficiency is higher than the twolevel inverter. The magnitude and ratio of the calculated loss might be changed according to the electrical characteristics of the switching device, the overall trend of loss analysis could be similar to the evaluated results in
Table 4
.
Calculation results of switching and conduction losses according to the MI
Calculation results of switching and conduction losses according to the MI
If the 3LTI is applied in the lowvoltage and lowpower home appliance such as refrigerators and air conditioners, compared with the twolevel inverter, it has low efficiency in the low speed region, but it has high efficiency in the high speed region. Especially, due to the low power factor, the efficiency is much lower in the lowload and lowspeed region since the proportion of conduction losses is further increased.
The 3LTI consists of the general twolevel inverter and six IGBTs. Six IGBTs are connected between the DC link midpoint and each output bridge in order to generate the zero voltage in the twolevel inverter. Therefore, unlike the other wellknown threelevel inverter such as NPCI, FCI and CHBI, the 3LTI is possible the selective operation as shown in
Fig. 6
, such as the twolevel operation mode and the threelevel operation mode. If the operation mode is selected to drive maximum efficiency according to the load current, MI and power factor as shown in
Fig. 7
, the 3LTI is possible to always obtain the maximum efficiency. Based on each mode efficiency, the 3LTI is operated twolevel operation mode in lowspeed region, and it is operated threelevel operation mode in highspeed region.
Threelevel and twolevel operation mode
Operation mode selection according to each mode efficiency
4. Experimental Results
A prototype of the 3LTI system was built to verify the validity of the proposed maximum efficiency operation strategy.
Fig. 8
shows the entire block diagram of the 3LTI testbed, along with the prototype. The strategy was implemented base on a digital signal process 28335, and 600V 10A Infineon IKP10B60T single IGBT is used for the 3LTI. Also, an interior permanent magnet synchronous motor (IPMSM) for a refrigerator driver system is applied to the experiment testbed and the proposed strategy is verified under the operation condition of refrigerator. The conditions and parameters of the experiments are shown in
Table 5
. The output waveform of the threelevel operation mode and twolevel operation mode are shown in
Fig. 9
.
Experimental testbed
Experimental conditions and motor parameter
Experimental conditions and motor parameter
Output voltage and current waveforms according to operation modes (V_{UV}: 50V/div., I_{U}: 1A/div., I_{V}: 1A/div., V_{U}^{*}: 1V/div.)
Fig. 10
shows the 3LTI efficiency without and with the proposed strategy under half and full loads. In the low MI region such as 1000RPM and 1800RPM, the 3LTI efficiency is improved about 2% compared with the twolevel mode by using the proposed strategy. Similarly, in the half load condition, the 3LTI efficiency with the proposed strategy is improved about 2%. In the motor speed higher than 2000RPM, since the efficiency of the threelevel operation mode is higher than the efficiency of the twolevel operation mode, the 3LTI with the proposed strategy is operated as the threelevel operation mode.
3LTI efficiency comparison between the threelevel mode operation and the proposed strategy operation
Fig. 11
shows 3LTI efficiency with the proposed strategy and the twolevel inverter efficiency. In the low MI region, since the 3LTI is operated the twolevel operation mode, the twolevel inverter and the 3LTI efficiency are similar. However, in the middle and high MI region, the 3LTI efficiency is higher than the twolevel inverter since the 3LTI is operated as the threelevel operation mode. As a result, the 3LTI with the proposed strategy can be always obtained the maximum efficiency in all operation areas.
Efficiency comparison between the twolevel inverter and the proposed strategy operation
5. Conclusion
This paper has analyzed the maximum efficiency operation strategy for the 3LTI in lowvoltage and lowpower home appliance. Unlike the other wellknown threelevel inverter such as the threelevel NPCI and FCI, the efficiency of the 3LTI is effected by the utilization rate of switches. In the low MI region, the efficiency of the 3LTI is similar to the general threelevel inverters since the utilization rate of switches is increased. However, in the high MI region, the efficiency of the 3LTI is similar to the twolevel inverter since the utilization rate of switches is decreased. Therefore, compared with the twolevel inverter, the 3LTI has together higher and lower efficiency areas according to the MI. To overcome this characteristic, two operation mode selection strategy is proposed for maximum efficiency operation. Based on each mode efficiency, the 3LTI is operated twolevel operation mode in low MI, and it is operated threelevel operation mode in high MI. As a result, the 3LTI can always obtain the maximum efficiency in all operation areas. The validity of the proposed strategy has been verified by experimental results.
BIO
SeungMin Shin He received his B.S. and M.S. degrees in Electrical Engineering from Sungkyunkwan University, Suwon, Korea, in 2009 and 2011, respectively. Since 2009, he has studied for Ph.D. degree in electrical Engineering at Sungkyunkwan University. His research interests are electric vehicle drives, power electronics and advanced motor drive systems.
JungHoon Ahn He received the B.S. and the M.S. degrees from Sungkyunkwan University, Suwon, Korea, in 2011 and 2013, respectively. Since 2013, he has worked for his Ph.D in Electrical Engineering at Sungkyunkwan University. His research interests include DC home appliance like future home system, battery management system (BMS), high / low power DCDC converter for PHEV/EV and advanced motor drive systems.
ByoungKuk Lee He received the B.S. and the M.S. degrees from Hanyang University, Seoul, Korea, in 1994 and 1996, respectively and the Ph.D. degree from Texas A&M University, College Station, TX, in 2001, all in electrical engineering. From 2003 to 2005, he has been a Senior Researcher at Power Electronics Group, Korea Electrotechnology Research Institute (KERI), Changwon, Korea. From 2006 Dr. Lee joins at School of Information and Communication Engineering, Sungkyunkwan University, Suwon, Korea. His research interests include charger for electric vehicles, hybrid renewable energy systems, dc distribution systems for home appliances, power conditioning systems for fuel cells and photovoltaic, modeling and simulation, and power electronics. Prof. Lee is a recipient of Outstanding Scientists of the 21^{st} Century from IBC and listed on 2008 Ed. of Who’s Who in America. Prof. Lee is an Associate Editor in the IEEE Transactions on Industrial Electronics and Power Electronics. He was the General Chair for IEEE Vehicular Power and Propulsion Conference (VPPC) in 2012.
ün Emre
,
Hava Ahmet M.
2009
“A NearState PWM Method with Reduced Switching Losses and Reduced CommonMode Voltage for ThreePhase Voltage Source Inverters,”
IEEE Trans. Industry Applications
45
(2)
782 
793
DOI : 10.1109/TIA.2009.2013580
Choudhury Abhijit
,
Pillay Pragasen
,
Amar M
,
Williamson Sheldon S
2014
“Performance Comparison study of Two and ThreeLevel Inverter for Electric Vehicle Application,”
Transportation Electrification Conference and Expo (ITEC), 2014 IEEE
Dearborn, USA
Shin Hojoon
,
Ha JungIk
2014
“Phase Current Reconstructions from DCLink Currents in ThreePhase ThreeLevel PWM Inverters,”
IEEE Trans. Power Electronics
29
(2)
582 
593
DOI : 10.1109/TPEL.2013.2257866
Itoh Junichi
,
Ogura Takumi
2010
“Evaluation of Total Loss for An Inverter and Motor by Applying Modulation Strategies,”
Proceedings of IEEE EPE, 2010
Ohrid, Macedonia
S1221 
S1228
Schweizer M.
,
Lizama I.
,
Friedli T.
,
Kolar J. W.
2012
“Comparison of the chip area usage of 2level and 3level voltage source converter Topologies,”
Proc. 36th Annu. Conf. IEEE Ind. Electron.
391 
396
Teichmann R.
,
Bernet S.
2005
“A comparison of threelevel converters versus twolevel converters for lowvoltage drives, tractions, and utility applications,”
IEEE Trans. Ind. Appl.
391396, Nov. 2012
41
(3)
391 
396
Dordevic O
,
Jones M
,
Levi E
2009
“A Comparison of CarrierBased and Space Vector PWM Technique for the ThreeLevel FivePhase Voltage Source Inverter,”
IEEE Trans. Industrial Informatics
9
(2)
609 
619
Schweiser M
,
Kolar W
2013
“Comparative Evaluation of Advanced ThreeLevel Inverter / Converter Topologies against TwoLevel Systems,”
IEEE Trans. Industrial Electronics
60
(12)
5515 
5527
DOI : 10.1109/TIE.2012.2233698
Padilha Felipe J. C.
,
Suemitsu Walter Issamu
,
Bellar Maria Dias
,
Lourenco Plutarcho Maravilha
2009
“LowCost Gate Drive Circuit for Threelevel NeutralPointClamped VoltageSource Inverter,”
IEEE Trans. Ind. Electron
56
(4)
1196 
1204
DOI : 10.1109/TIE.2008.2007554
Felipe J.
,
Padilha C.
2011
“A Hybrid Cascade Converter Topology with SeriesConnected Symmetrical and Asymmetrical DiodeClamped HBridge Cells,”
IEEE Trans. Power Electron.
26
(1)
51 
65
DOI : 10.1109/TPEL.2009.2031115
Padilha Felipe J. C.
,
Suemitsu Walter Issamu
,
Bellar Maria Dias
,
Lourenco Plutarcho Maravilha
2013
“Modified DCLink Voltage Balancing Algorithm for a 3Level Neutral Point Clamped (NPC) Traction Inverter Based Electric Vehicle PMSM Drive,”
Proc. Annual Conf. of the IEEE Industrial Electronics Society
Vienna, Austria
4660 
4665
AbuRub H.
,
Holtz J.
,
Rodriguez J.
,
Baoming G.
2010
“Medium Voltage Multilevel Converters: State of the art, Challenges and Requirements in Industrial Applications,”
IEEE Trans. Ind. Electron.
57
(8)
2581 
2596
DOI : 10.1109/TIE.2010.2043039
Rodriguez J.
,
Bernet S.
,
Steimer P. K.
,
Lizama I. E.
2009
“A Survey on NeutralPointClamped Inverters,”
IEEE Trans. Ind. Electron.
57
(7)
2219 
2230
Sharkh Suleiman M.
,
Abusara Mohammad A.
,
Orfanoudakis Georgios I.
,
Hussain Babar
2014
`Loss Comparison of Two and ThreeLevel Inverter Topologies: Power Electronic Converter for Microgrids
John Wiley & Sons
Schweizer Mario
,
Kolar Johann W.
2013
“Design and Implementation of a Highly Efficiency ThreeLevel TType Converter for LowVoltage Applications,”
IEEE Trans. Ind. Electron.
28
(2)
889 
907
Choi UiMin
,
Lee KyoBeum
,
Blaabjerg Frede
2014
“Diagnosis and Tolerant Strategy of an OpenSwitch Fault for TType ThreeLevel Inverter Systems,”
IEEE Trans. Ind. Appl.
50
(1)
495 
508
DOI : 10.1109/TIA.2013.2269531
Komatsu K.
,
Yatsu M.
,
Miyashita S.
,
Okita S.
,
Nakazawa H.
,
Igarashi S.
,
Takahashi Y.
,
Okuma Y.
,
Seki Y.
,
Fujihira T.
2010
“New IGBT Modules for Advanced NeutralPointClamped 3Level Power Converters,”
Conf. Rec. IPEC
523 
527