I. INTRODUCTION

Multi-level topologies have outstanding performance in terms of total harmonic distortion (THD) and efficiency [7]-[9]. Three-level topologies have been used in applications of wide-power range [1]-[6].

There are two types of three-level rectifier topologies. In the neutral-point clamped (NPC)-type, four switches are connected in series and two clamping diodes are connected to the neutral point. Therefore, the NPC-type rectifier can reduce the switching device’s collector-emitter voltage (VCE) in half. The other type of topology is the T-type rectifier as shown in Fig. 1. Although the switches used in the T-type rectifier have a VCE which is the same as that of the conventional two-level rectifier, the T-type rectifier has lower conduction and switching losses, and it does not need clamping diodes when compared to the NPC-type rectifier [7].

Fig. 1.Grid connected T-type rectifier.

Research on improvements in the reliability of systems using multi-level topologies has become an important issue. Therefore, systems using switching devices need to diagnose switching device faults and to maintain their operation by tolerance controls. Diagnosis methods for detecting the open-switch faults of NPC-type inverters and rectifiers are proposed in [8]-[11]. To detect open-switch faults, additional devices such as voltage sensors or current patterns are used.

In [12], an imperfect tolerance control for an NPC-type rectifier is proposed without additional devices. This tolerance control considers inner open-switch faults (Sx2 and Sx3) but not outer open-switch faults (Sx1 and Sx4) because the outer switches do not have an effect on rectifier operation with unity power factor [13]. These tolerance controls do not completely eliminate current distortion because of structural limitations of the NPC-type topology. In [14], there are two tolerance controls, which do not require additional devices, for the inner open-switch faults (Sx2 and Sx3) of an T-type rectifier with a high modulation index. These two methods are the replacement two-level switching (R2LS) and the maintenance three-level switching (M3LS) tolerance controls. These controls use a two-level switching method to completely restore distorted currents. They have opposite characteristic in terms of the current THD and the dc-link voltage ripple. The R2LS tolerance control maintains the two-level switching method in four of the six sectors. Therefore, the current THD in this control is higher than that of M3LS tolerance control. However, the M3LS tolerance control, which considers the neutral-point balance, has a larger dc-link voltage ripple than the R2LS tolerance control.

This paper proposes a new tolerance control based on the nearest space-vector PWM (SVM) method [15] for the Sx2 and Sx3 open-switch faults of a T-type rectifier. The proposed tolerance control guarantees operation in the whole modulation index and is advantageous in terms of efficiency when compared with other tolerance controls. Simulations and experiments are conducted to demonstrate the performance and characteristics of the proposed tolerance control.

II. T-TYPE RECTIFIER AND INNER-OPEN SWITCH FAULTS

A. Description of a Three-Level T-type Rectifier

In a three-level T-type rectifier, the pole voltage Vxz (x = a, b, c) with respect to the neutral-point Z can be either Vdc/2, 0, or -Vdc/2 depending on the operating state. The first operating state “P” has Vxz of Vdc/2. In “P”, the switches Sx1 and Sx2 are ON and the switches Sx3 and Sx4 are OFF. When Sx2 and Sx3 are ON and Sx1 and Sx4 are OFF, the operating state is “O” and Vxz is 0. When Sx1 and Sx2 are OFF and Sx3 and Sx4 are ON, Vxz is –Vdc/2 at the operation state “N”. The switch states and pole voltages depending on the operating states are shown in Table Ι.

TABLE IOPERATION STATE AND POLE VOLTAGE

According to the switching states of each phase, there are six large-, six medium-, twelve small- and three zero-vectors in the three-level space vector diagram of a three-level T-type rectifier shown in Fig. 2.

Fig. 2.Space vector diagram of three-level T-type rectifier.

B. Inner Open-Switch (Sx2 and Sx3) Faults

In a rectifier, according to the operating states and current direction, there are six current paths as shown in Fig. 3. The current of the rectifier is generated by the operating state “O” and the current continuously flows through a diode of the outer switches if the operating state is changed to “P” or “N”. Therefore, the impossibility of the “O” operating state only lead to a zero current like the current distortion.

Fig. 3.Current paths depending on operating state and current.

In the positive current part, most of the current of the rectifier, which has a unity power factor, flows through paths (a) and (b). Therefore, the range of path (c) can be ignored. This means that the “N” operating state does not appear in the positive current part. An Sx3 open-switch fault in the “O” operating state is fatal to rectifier operation. On the other hand, on a unity power factor, an Sx1 open-switch fault does not lead to current distortion in the “P” operating state because the current flows through a diode. Sx1 also does not affect the current in the negative current part. Based on the fact that Sx4 and Sx1 are not needed, a reduced part topology was proposed in [15]. Therefore, in this paper, Sa2 and Sa3 open-switch faults are considered to explain the effect of an open-switch fault.

Fig. 4 shows a rectifier’s input three phase currents, a-phase pole voltage, and dc-link voltage when Sa2 and Sa3 open-switch faults occur. Inner open-switch faults have a fatal effect on the a-phase input current. In the negative part of the a-phase current, as shown in Fig. 4(a), the valid operating state “O”, shown in Fig. 3(e), becomes infeasible. Therefore, the negative part of the a-phase current cannot be generated and the dc-link voltage has ripple. Like an Sa2 open-switch fault, an Sa3 open-switch fault has a fatal effect on the positive part of the a-phase current.

Fig. 4.Rectifier waveforms: (a) Sa2. (b) Sa3 open-switch fault.

III. PROPOSED TOLERANCE CONTROL FOR THE INNER OPEN-SWITCH FAULTS OF T-TYPE RECTIFIERS

The proposed tolerance control does not use the “O” operating vectors of a phase which contains an open-switch fault. Furthermore, in the proposed tolerance control, the two-level switching method is applied for an a phase containing an open-switch fault and two pole voltages (Vbz, and Vcz) are maintained to three levels (Vdc/0, 0, -Vdc/2) at a high modulation index. As a result, the proposed tolerance control is advantageous in terms of efficiency when compared with the other tolerance controls using the two-level switching method. In this section, an Sa2 open-switch fault is considered as the fault case. The proposed tolerance control is developed based on the nearest space-vector PWM (SVM) method. Fig. 5 shows the infeasible vectors when an Sa2 open-switch fault occurs.

Fig. 5.Space vector diagram of Sa2 open-switching fault.

A. Tolerance Operation at a High Modulation Index

An Sa2 open-switch fault leads two medium-vectors (OPN, ONP), five small-vectors (OPO, OPP, OOP, OON, ONO) and one zero-vector (OOO) being infeasible when the a-phase current is negative. At a high modulation index, which is shown in the white part of Fig. 5, the OPN, ONP, OPO, OPP, OPP, OON, and ONO vectors should be changed to other feasible vectors during negative current.

In sectors ΙΙΙ and ΙV, the infeasible P-type small-vectors OPO, OPP, and OOP are changed to the N-type small vectors NON, NOO, and NNO. The increased use of N-type small vectors in sectors ΙΙΙ and ΙV can cause a neutral-point unbalance. Therefore, in sectors Ι and VΙ, the N-type small-vectors OON, ONN, and ONO are changed to the P-type small-vectors PPO, POO, and POP. When an Sa2 open-switch fault occurs, the changing vector methods in sectors Ι, ΙΙΙ, ΙV, and VΙ are the same as those of the M3LS tolerance control in [14]. The M3LS tolerance control uses two-level switching in sectors ΙΙ and V. However, the two-level switching increases the switching loss of the three-phase switches.

Sector ΙΙ is divided into four parts as shown is Fig. 5. In sector ΙΙ-A, there are infeasible OPN and OPO vectors. The OPN vector can be changed to NPN and PPN vectors, and the PPO vector can be also changed to a PPO vector as shown in Fig. 6. As a result, the NPN, PPN, and PPO vectors retain the b- and c- phase pole voltages Vbz and Vcz. However, the a-phase pole voltage Vaz has Vdc/2 and -Vdc/2 without 0. The changing vector is easily performed by recalculating the switching time of the newly selected NPN, PPN, and PPO vectors, based on the SVM method as shown in Fig. 6. It has three on-times for each phase which are calculated from the reference voltage and are expressed as:

Fig. 6.Tolerance control in sector ΙΙ-A.

To use the NPN, PPN, and PPO vectors, average Vaz in Fig. 6(a) should be equal to average Vaz in Fig. 6(b) for a switching period Ts. This is satisfied by adding the offset value to the original on-time Ta,on of the a-phase and it is expressed as:

Fig. 7 shows the switching sequence of the proposed tolerance control in sector ΙΙ-B. Like sector ΙΙ-A, Vbz and Vcz are not changed. The on-time Ta,on,TC for the proposed tolerance control can be calculated by (2).

Fig. 7.Tolerance control in sector ΙΙ-B.

In sectors ΙΙ-C and II-D of Fig. 8, the on-time Ta,on,TC for the proposed tolerance control is calculated by using the off-time Ta,off. Ta,off is expressed as:

Fig. 8.Tolerance control in sector ΙΙ-C and -D.

To make the average Vaz of the proposed tolerance control equal to the original average Vaz, Ta,on,TC is defined as:

Likewise, in sector V, equations (2) and (4) are applied according to parts of sector V.

Consequently, the proposed tolerance control always keeps Vbz and Vcz as three levels and it makes the level of Vaz, which contains the faulty switch, two in sectors ΙΙ and V when an Sa2 open-switch fault occurs. The whole principle of the tolerance control method at a high modulation index is shown in Table ΙΙ.

TABLE IIPROPOSED TOLERANCE CONTROL AT HIGH MODULATION INDEX: SA2 OPEN-SWITCH FAULT

B. Tolerance Operation at a Low Modulation Index

Like the method mentioned in section ΙΙΙ-A, the proposed tolerance control at a low modulation index, which is shown in the blue part of Fig. 5, changes the infeasible vectors to other vectors. This principle is also different depending on the sectors.

In sectors ΙΙΙ and ΙV, the infeasible P-type small-vectors OPO, OPP, and OOP are changed to the N-type small-vectors NON, NOO, and NNO. Moreover, the zero-vector OOO cannot be used. Therefore, it is should be substituted with PPP and NNN.

Fig. 9(a) shows the switching sequence in sector ΙΙΙ-1 when the rectifier operates normally without any faults. This switching sequence consists of the NON, NOO, OOO and OPO vectors. The OPO and OOO vectors are infeasible. The P-type small-vector OPO can be replaced with NON, and the OOO zero-vector can be replaced with the NNN zero-vector as shown in Fig. 9(b). The switching sequence in Fig. 9(b) causes switching to occur two times in the b- and c- phases. Therefore, the switching sequence should be changed to NNN-NON-NOO-NON-NNN, as shown in Fig. 10, if the SVM method is used. Its switching sequence can be performed by adding the offset value as shown in Fig. 9(c). This is expressed as:

Fig. 9.Tolerance control in sector ΙΙΙ-1.

Fig. 10.Performance of tolerance control at high modulation index. (a) Sa2 open-switch fault. (b) Sa3 open-switch fault.

The positive voltage Vbz in Fig. 9(a), is changed to a negative voltage as shown in Fig. 9(c). Vaz and Vcz are maintained as negative voltages. Consequently, when the sign of Vxz is changed, –Ta,on is added to Tx,on for Vxz. Ts–Tx,on is added to Tx,on for Vxz when the sign of Vxz is not changed. This method is applied in sectors ΙΙΙ and ΙV.

For neutral-point balancing control, the N-type small vectors are changed to P-type small-vectors and the zero-vector OOO is changed to the zero-vector PPP in sectors Ι and VΙ. In sectors Ι and V, the two-level switching method is used. The whole principle of the tolerance control method at a low modulation index is shown Table ΙΙΙ.

TABLE IIIPROPOSED TOLERANCE CONTROL AT LOW MODULATION INDEX: SA2 OPEN-SWITCH FAULT

IV. SIMULATION RESULTS

Simulations were performed using PSIM. The simulation circuit is the same as the one shown in Fig. 1. The simulation parameters are shown in Table ΙV.

TABLE IVSIMULATION AND EXPERIMENT PARAMETERS

Fig. 10 shows the three-phase currents, dc-link voltage, and line-to-line voltage of the rectifier when the proposed tolerance control is applied at a high modulation index (Vline-to-line is 126 Vrms and the load is 33 Ω). Due to an inner open-switch fault, the current containing an open-switch fault becomes zero for a half period and the dc-link voltage has a large ripple. After 0.6 s, which means that the proposed tolerance control is applied, the three-phase currents are restarted as a sinusoidal waveform and the dc-link voltage ripple is eliminated. Additionally, the line-to-line voltage (Vab) between the a- and b- inputs of the proposed tolerance control is different from that of the conventional SVM method [15].

Fig. 11 shows the three-phase pole voltages and currents of the proposed tolerance control at a high modulation index. Vbz and Vcz maintain three levels. However, Vaz has only Vdc/2 in sectors Ι and VΙ or –Vdc/2 in sectors ΙΙΙ and ΙV. This is the same as that of the two-level switching method in sectors ΙΙ and V. Therefore, the a-phase current ripple increases in sectors ΙΙ and V due to the high dV/dt of the two-level switching method. The proposed tolerance control only uses P-type small vectors for the two sectors and N-type small vectors for the other two sectors. Therefore, the neutral-point voltages (Vtop and Vbottom) have a low frequency ripple which is the same as the fundamental frequency of the input current as shown in Fig. 11.

Fig. 11.The neutral-point and three-phase pole voltages of the proposed tolerance control at high modulation index.

Fig. 12 shows the three-phase currents, dc-link voltage, and line-to-line voltage of the rectifier when the proposed tolerance control is applied at a low modulation index (Vline-to-line is 63 Vrms and the load is 67 Ω). Similar the case of a high modulation index, the current containing an open-switch fault becomes zero for a half period and the dc-link voltage has a large ripple. After the tolerance control is applied, the three-phase currents are restarted as a sinusoidal waveform and the ripple of the dc-link voltage is eliminated. Additionally, the line-to-line voltage (Vab) between the a- and b- inputs of the rectifier has Vdc and 0 or –Vdc and 0 in sectors ΙΙ and V when the two-level switching method is applied.

Fig. 12.Performance of tolerance control at low modulation index: (a) Sa2 open-switch fault (b) Sa3 open-switch fault.

Fig. 13 shows the three-phase pole voltages and currents of the proposed tolerance control at a low modulation index. In sectors ΙΙ and V, the three-phase pole voltages have Vdc/2 or –Vdc/2. Therefore, the current ripple of the three-phase increases in sectors ΙΙ and V due to a high dV/dt. The switching operation of the remaining sectors is the same as that in the case of a high modulation index. At a low modulation index, the neutral-point voltages has a low frequency ripple which is the same as the fundamental frequency of the input current

Fig. 13.Neutral-point and three-phase pole voltages of proposed tolerance control at low modulation index.

V. EXPERIMENTAL RESULTS

Experiments are conducted to verify the validity of the proposed tolerance control. Fig. 14 shows the hardware setup which consists of a control board, sensors, gate drivers, and T-type IGBT modules. This module is 4MBI300VG-120R-50 from Fuji. The parameters for the experiments are the same as those used in the simulations.

Fig. 14.Experiment setup of T-type rectifier.

Fig. 15 shows the performance of the proposed tolerance control at a high modulation index (Vline-to-line is 126 Vrms and the load is 33 Ω) when an Sa2 open-switch fault occurs. Due to the Sa2 open-switch fault, the positive a-phase current (Ia) becomes zero and the dc-link voltage has a large ripple. After the proposed tolerance control for a high modulation index is applied, the a- and b-phase currents (Ia and Ib) are restored as the sinusoidal waveforms and the dc-link voltage ripple decreases as shown in Fig. 15. In the experimental results, the Vab of the proposed tolerance control, which is different from the Vab of the conventional SVM in two sectors, is the same as that in simulation.

Fig. 15.Experimental results of tolerance control at high modulation index when Sa2 open-switch fault occurs.

Fig. 16 shows the experimental results of the proposed tolerance control for a low modulation index when an Sa2 open-switch fault occurs. Like the results at a high modulation index, the current distortion is eliminated and the dc-link ripple decreases. The Vab of the proposed tolerance control for a low modulation index has Vdc and –Vdc in spite of the low modulation index. This is the same result as the simulation.

Fig. 16.Experimental results of tolerance control at low modulation index when Sa2 open-switch fault occurs.

VI. CONCLUSION

This paper proposes a tolerance control based on the nearest SVM methods for the Sx2 and Sx3 open-switch faults of a T-type rectifier. The proposed tolerance control does not need additional devices and restores the distorted input currents caused by Sx2 and Sx3 open-switch faults. In addition, it guarantees the operation at the whole modulation index with inner open-switch faults. In comparison with the R2LS and M3LS tolerance controls, the switching loss is decreased at a high modulation index because only the input voltage of the phase containing an open-switch fault has two levels in two sectors. The performance of the proposed tolerance control is verified by the simulation and experimental results.