2022 IEEE International Conference on Power Electronics, Smart Grid, and Renewable Energy (PESGRE)

Open Circuit Fault Detection for a Grid-tied Hybrid

Multi-level Inverter
Dheeraj Etta Richa Goel Anubrata Dey
2022 IEEE International Conference on Power Electronics, Smart Grid, and Renewable Energy (PESGRE) | 978-1-6654-4837-6/22/$31.00 ©2022 IEEE | DOI: 10.1109/PESGRE52268.2022.9715961

Department of Electrical Engineering Department of Electrical Engineering Department of Electrical Engineering

Indian Institute of Technology Indian Institute of Technology Indian Institute of Technology
Roorkee, India Roorkee, India Roorkee, India
[email protected] [email protected] [email protected]

Abstract—This paper proposes a single switch open circuit inverters, as the high value of current generated due to
fault detection strategy for a grid-tied seven-level hybrid shorting of dc supply may damage the inverter within
inverter that is first of its kind for an unconventional multi-level microseconds. As a result, the majority of short circuit fault
inverter (MLI) topology. The faulty switch can be detected and detection techniques are hardware-based that are identified
located properly by analyzing few error variables associated to
locally mostly using a state-of-the-art gate driver. There are
absolute phase currents which undergo average value changes
in comparison to those of healthy conditions. The proposed very few strategies reported till date which are algorithmic
algorithm for the considered MLI topology also observes the based [6]. As fault detection time for open circuit faults can
change in neutral point voltage and flying capacitor voltages for be accommodated in few milliseconds, algorithmic based
the fault detection purpose. But it does not require additional strategies can be employed to work for open circuit fault
circuitry, sensors, and complex offline/online calculations. It is detection in most applications. While protection against short
observed that this strategy is able to detect faults within two circuit or overcurrent has become a common feature in the
fundamental periods. Simulation results shows the feasibility IGBTs used in industry, open circuit faults still have not
and effectiveness of the proposed fault detection strategy which received the necessary attention [5]. So, development of
can well be extended to other MLI topologies using similar
algorithm-based strategies for open circuit failure detection
logics.
has become an important research field and over the years
Keywords—fault detection, open circuit fault, Park’s vector, researchers have proposed various schemes for two/three-
error variables, hybrid inverter, neutral-point, flying capacitor. level inverters. Furthermore, fault detection is also essential
for any fault tolerant operation strategy which can only be
I. INTRODUCTION implemented after the faulty switch location is detected
For over three decades now, multilevel inverters (MLIs) reliably.
have been under research, and they are proving to be a useful In [5] & [6], a survey of open and short circuit fault
technology in industries for drive, grid, and renewable energy detection strategies is presented. One of the initial open
applications [1], [2]. Floating capacitor, cascaded H-bridge circuit fault detection strategy is based on Park’s vector
and neutral point clamped are the conventional MLI approach [7]. Here, Park’s transformation is applied on
topologies. The output current of MLI tends to perfectly average phase currents and magnitude and phase angle of
become sinusoid as the number of voltage levels increases, Park’s vector are analyzed to detect faults. However, the main
but this will increase the number of power devices and may disadvantage of this approach is that it is dependent on the
add complexity in the inverter control scheme. There are also magnitudes of load currents. In [8] & [9], phase currents are
unconventional topologies available in the literature [3], [4] normalized using phase current Park’s modulus before
which are basically mixture of conventional topologies. carrying out Park’s transformation on average phase currents
Hybrid inverter topology shown in [4] is one such and deviations in neutral point voltage are also monitored to

&
unconventional topology and it has several benefits over exactly locate the faulty switch. In [10], phase current
conventional MLI topologies like reduced device count, patterns are plotted using values and Concordia
better fault-tolerant response, and performance under current pattern radius is analyzed to detect faults. Apart from
different load conditions. These hybrid topologies can be these current based techniques, in [11], line voltages and
effective in industrial applications both in terms of neutral point voltages deviations from the healthy operation
performance and cost. of the three level T-type converter are analyzed to detect open
As components count increases in the inverters, the circuit faults. Average current Park’s vector-based technique
reliability of these inverters decreases, and this is one of the is implemented in three-level T-type inverters in [12] & [13]
main issues with any type of multilevel inverters. Around and error variables, average phase currents and phase angle
38% of the faults in voltage source inverters used in ac drive are used to detect single as well as multiple phase faults.
systems is due to the failure of the power devices [5]. Now- Faults in MLIs may also be detected using neural network
a-days, IGBTs are used as power devices in most of the MLIs and fuzzy logic-based methods as discussed in [14]-[16], but
because of their high current and voltage ratings. Although this will require significant computational effort and
IGBTs are robust and capable of handling short circuit application specific data for training the algorithm. In [17], a
currents for few µs, they are prone to failures because of voltage-based detection strategy is presented for CHB MLIs
electrical and thermal stress and load side faults created in to detect open circuit faults based on voltage deviation and
many applications. Broadly, IGBT faults can be of two types current flow direction and fault verification process is carried
- open circuit switch and short circuit switch faults. Open out after isolating the possible fault.
circuit failures are not fatal to inverters in most cases, but they Above discussed fault detection methods are either for
can induce secondary problems in the inverter over time. On two or three-level inverters and a very few are oriented
the other hand, short-circuit failures are more critical to towards higher level inverters. In the literature also, there are

978-1-6654-4837-6/22/$31.00 ©2022 IEEE

Authorized licensed use limited to: Consortium - Algeria (CERIST). Downloaded on August 02,2022 at 23:57:54 UTC from IEEE Xplore. Restrictions apply.
 2022 IEEE International Conference on Power Electronics, Smart Grid, and Renewable Energy (PESGRE)

not many fault detection strategies focusing on MLIs,

especially on the unconventional topologies. Generally, a
short circuit switch fault induces stress and rupture in IGBTs,
and this might cause lifting of bond wires which lead to an
open circuit switch fault [6]. Also, it is very rare that open
circuit failures occur in more than one switch at the same
time. Considering all these, a single switch open circuit fault
detection strategy for the inverter depicted in Fig-1 is
described in this article. To test the proposed algorithm, grid
tied seven-level inverter is created in Simulink environment
and different open circuit faults are introduced. Simulation
results presented in the end proves the reliable performance
of the strategy and its ability to detect and locate the faults.
II. SEVEN-LEVEL HYBRID INVERTER
Fig. 1. Seven-level hybrid inverter
A. Operation of seven-level hybrid inverter
TABLE I.
The primary unit of the inverter shown in Fig. 1 is a three- SWITCHING TABLE FOR THE HYBRID SEVEN-LEVEL INVERTER
level T-type structure supplied by a single DC source, .

/4, is
Pole Switching States State of FCA
To each front-end T-type leg, an H-bridge cell with flying Voltage

3 /4
capacitor (FC) which is maintained at a voltage VAO S1 S2 SA1 SA2 SA3 SA4 SA5 SA6
connected in series to increase the number of voltage levels 1 0 0 0 1 1 1 0 ↓ ↑

3 /4, /2, /4, 0, /4, /2and 3 /4. /2

[4]. This inverter can achieve at a maximum of seven levels 1 0 1 0 1 0 1 0

/2, 0, and /2, FC voltage

─ ─
1 0 0 1 0 1 1 0

/4
For pole voltage levels 1 0 1 1 0 0 1 0 ↑ ↓
( ) is not affected whereas for the remaining voltage 0 0 0 0 1 1 1 1 ↓ ↑
levels FC charges or discharges depending on the load current

/4 and /4 pole voltage levels, FC voltage can

0 0 1 0 1 0 1 1
direction through the flying capacitor (as shown in Table-I). 0 ─ ─
0 0 0 1 0 1 1 1
For
/4
3 /4 and 3 /4 voltage levels,
be controlled with the help of switching state redundancies. 0 0 1 1 0 0 1 1 ↑ ↓
Whereas for 0 1 0 0 1 1 0 1 ↓ ↑

/2
redundant states are not available and hence for these levels 0 1 1 0 1 0 0 1
FC voltages are not controllable and this poses a restriction ─ ─

3 /4
0 1 0 1 0 1 0 1
either on modulation index range or load power factor for 0 1 1 1 0 0 0 1 ↑ ↓
seven level operation. Unity power factor (upf) condition is
the worst possible situation for FC balancing in seven level ↓’ - Discharging; ‘↑’ - Charging; ‘─’ - No Change, ‘1’ – IGBT on condition,
‘0’ - IGBT off condition.
operation of above-mentioned topology. Load power factor is
not controllable always, so modulation index can be limited,
and the limiting value is found out to be 0.8 under upf switches S1, SA1 and SA3. However, for phase current > 0,
conditions [18]. In seven level operation, for modulation if an open circuit fault occurs in switch S1 then the current
index less than 0.8, FC voltages can be balanced properly and flows through the SA5, SA3, body diode of SA1 and body diode

0 instead of /2. For < 0 during the pole voltage level

maintained within a certain tolerance band under all possible of SA6 as shown in Fig. 2(a). Hence the output pole voltage is

/2, the current flows through the body diode of S1 (as

linear load conditions. NP voltage can also be balanced under
all load conditions using the balancing algorithm specified in
[19]. Zero sequence voltage (ZSV) component is calculated body diode remains operational even in open circuit fault
at every instant by forcing average neutral current in each condition) in both healthy and fault cases as shown in Fig.
carrier period to zero and this ZSV component is added to all 2(b). As a result, no distortion is observed in the phase current
phase modulating signals to suppress NP voltage variations. and pole/line voltage when < 0 which can be clearly seen

In case of SA5 fault, for the desired pole voltage level 0

in waveforms shown in later part (Fig. 8(a)).
III. ANALYSIS OF SEVEL-LEVEL INVERTER DURING OPEN
CIRCUIT-SWITCH FAULT and > 0, the current flows via switch SA3, body diode of
If an open circuit fault occurs, the current path will deviate SA1 and body diode of S2 instead of SA5, SA3, body diode of

/2 . Under normal
from its normal path and due to this deviation, intended SA1 and body diode of SA6 as illustrated in Fig. 3(a). Instead
voltage level will not be achieved. This will produce of 0, the pole voltage becomes
distortions in output pole/line voltages, output phase currents, circumstances, for < 0, the phase current passes via
NP voltage and FC voltages. switches SA6, SA1, the body diode SA3 and body diode of SA5.
A. Effect on Output Pole/line Voltages and Phase Currents For an open circuit switch fault in SA5, current path remains
same as that of healthy conditions as illustrated in Fig. 3(b).

′0′, the current path is formed through SA5, SA3, body diode
Fig. 2–4 show the current paths for various switch faults
As shown in Fig. 4(a), for > 0 and desired pole voltage
under healthy and fault conditions. Current flowing out from

/2 then
the inverter to the grid is defined as positive current. Under
healthy conditions, if the desired pole voltage is of SA6 and body diode of SA1 regardless of switch SA1 fault.
both positive and negative phase current flows through the For < 0, if the open circuit fault arises in SA1, the current

Authorized licensed use limited to: Consortium - Algeria (CERIST). Downloaded on August 02,2022 at 23:57:54 UTC from IEEE Xplore. Restrictions apply.
 2022 IEEE International Conference on Power Electronics, Smart Grid, and Renewable Energy (PESGRE)

flows through switch SA6, FCA, body diode of SA5, body diode

/4 instead of 0.
of SA4 and body diode of SA3, and the output pole voltage
becomes
Because of these open circuit switch faults, undesirable
output pole voltages are produced, and this will distort the
phase currents. For some open-switch faults, it also causes
unbalance in neutral-point voltage and flying capacitor
voltages and sustained operation under such conditions may
be not acceptable.
B. Effect on Neutral-Point and Flying Capacitor Voltages
In the event of fault occurrences, due to change in current
paths, phase current waveforms will also change. The NP (a) (b)
voltage and FC voltage balancing schemes are phase current Fig. 2. Phase a current path under healthy and switch S1 open circuit fault
conditions (a) > 0, (b) < 0.
dependent, so these voltages will be affected accordingly. In

/2 state in the T-
case of S1 fault, as observed in Fig. 8(b) average phase a

type end will not be achieved instead of which 0 state is

current would be highly negative and

achieved. Therefore, in this way T-type leg will not be

responding to NP balancing algorithm properly and capacitor
C1 voltage becomes larger than capacitor C2 voltage thus
reducing the NP voltage. Similarly, as phase average current
values changes by large values, FC voltages of not only phase
a but also of other phases show variations. (As clearly seen
in Fig. 8(a)).

achieving 3 /4 & 3 /4 levels and any fault in

Switches S1, S2, SA2 & SA3 play an important role in

these switches, affects both the FC and NP voltage balancing

schemes severely. So, for open circuit faults in these (a) (b)
switches, the FC and NP voltages will deviate a lot from its Fig. 3. Phase a current path under healthy and switch SA5 open circuit fault
conditions (a) > 0, (b) < 0.
respective nominal values. For the remaining single switch
faults, there will not be much deviation in the NP and FC
voltages as phase currents will not deviate significantly from
its healthy condition waveforms.
IV. OPEN-CIRCUIT SWITCH FAULT DETECTION METHOD
As discussed in the earlier sections, whenever an open-
circuit fault arises, the phase current waveforms are distorted,
and the neutral-point voltage and flying capacitor voltages are
unbalanced. Sensing such abnormalities, the faulty switch is
identified in conventional two/three-level inverters [12]-[13].
For the considered hybrid inverter, the phase/line voltages,
capacitor voltages and phase currents also carry the fault
signatures and thus they are investigated below to detect and (a) (b)
localize the fault. Fig. 4. Phase a current path under healthy and switch SA1 open circuit fault
The proposed detection method for the hybrid inverter conditions (a) > 0, (b) < 0.
adapts the basic average current Park’s vector-based
technique presented in [12] and uses new variables (apart To make fault detection method independent of
from the phase current based variables) which depicts magnitudes of phase currents and robust against false alarms,
capacitor voltages behavior to analyze and detect various the measured grid phase currents are normalized by dividing

= (2 )
switch fault conditions in the seven-level topology shown in each phase current with Park’s vector modulus.
√
(1)
= ( )
Fig. 1. A signal flow block diagram summarizes the proposed
algorithm as shown in Fig. 5. The three phase grid currents,
√"
(2)
neutral-point voltages and flying capacitor voltages are the
unique inputs required for this strategy. As inverter levels where and are d-axis and q-axis currents respectively,

vector modulus | $ | is given by

increases, complexity of the fault detection algorithm may and , , and are the grid phase currents. The Park’s

| $| = %
also increase because more fault diagnostic variables are
required to properly detect a fault. But the proposed fault " " (3)
detection strategy utilizes the variables that are currently
&'

&' =
(
employed by the primary inverter control under healthy The normalized phase current is calculated as follows
|() |
condition. As a result, it avoids the need of additional sensors (4)
and the resulting increase in hardware/software complexity.

Authorized licensed use limited to: Consortium - Algeria (CERIST). Downloaded on August 02,2022 at 23:57:54 UTC from IEEE Xplore. Restrictions apply.
 2022 IEEE International Conference on Power Electronics, Smart Grid, and Renewable Energy (PESGRE)

= +, sin(0$ 1 ø)
Grid phase currents can be represented as follows

= +, sin(0$ 1 ø)
"3
& = * 4
= +, sin(0$ 1 ø)
"3
(5)

4
where +, is the amplitude of phase currents, 0$ is the grid
frequency and ø is initial phase angle.
Now error variable 5& (6 = 7, 9, :) are evaluated using the

5& = ; 〈| &' |〉
below equation.

〈| &' |〉 denotes the average absolute values of the three

(6) Fig. 5. Block diagram for fault detection algorithm.

normalized grid phase currents. ; is the average absolute

value of the normalized perfect sinusoid grid phase current
which signifies healthy condition value.
;= %
>
3 4
(7)
Phase angle ?$@A is a useful variable for exactly locating the
faulty switch. Average Park’s vector phase angle settles at
some value during faulty conditions and upon accurate

?$@A can be calculated as shown in (8) and (9). + @A and + @A

determination of this angle, specific faults can be identified.

average phase current values @A , @A & @A .

can be found out by applying Park’s transformation on

+$@A = + @A B+ @A = |+$@A | ?$@A

?$@A = tanE (+ @A ⁄+ @A )
(8)
(9)
Apart from these variables, average neutral point voltage
( 'G@A ) and flying capacitor average voltage values
( @A , 6 = 7, 9, :) are also used to diagnose open-switch

In healthy conditions, 〈| &' |〉 for each phase is almost

fault in this seven-level hybrid inverter.

equal to ; , so error variables 5& (6 = 7, 9, :) are almost

Fig. 6. Flowchart depicting the fault detection and localization algorithm.

V. SIMULATION RESULTS
equal to zero and phase angle is not a constant value. As
proper capacitor voltage balancing algorithms are Seven-level hybrid inverter is modelled and simulated

inverter, 'G@A &

implemented during healthy seven-level operation of the with grid load conditions as well as the proposed fault
detection strategy is implemented using the Simulink
@A
/2 and /4
remain within their tolerance limits
environment. To control the power injected from the inverter
and these average values are close to input source to grid and to synchronize inverter output with
respectively. Whenever an open-switch fault occurs, first the grid, the required modulating signals for the inverter are
faulty phase is detected by the sign of error variables. generated using a DQ current controller. The carrier wave
Faulty phase error variable drifts towards a sufficiently based PWM technique is used to produce switching signals
positive value whereas other two phases error variables will for the inverter.
be negative or almost zero. After finding out the faulty phase,

i.e., ?$@A , 'G@A &

exact faulty switch can be detected using the other variables TABLE II.
@A
. Fault detection and localization PARAMETER OF THE SIMULATION
logic for single switch open circuit faults is explained using a Parameter Value
self-explanatory flowchart shown in the Fig. 6. Load 3-Phase Grid
Here, ranges of K1, K2 and the phase angle, the NP and FC
voltage values which are used for differentiating the fault Grid Frequency 50 Hz
type, can be easily evaluated by carrying out extensive DC Link Voltage 250 V
simulation studies of the considered system for different fault
Carrier Frequency 2 kHz
conditions. The values evaluated for phase a will also work
similarly for phases b and c. In simulations, conditions like Grid Line-Line rms Voltage 180 V
deadtime effect, code execution time delay, parasitic drops NP Voltage Tolerance Limits [124 V, 126 V]
and sensor sampling time are considered to evaluate these
values. They are also properly tuned, or their range is FC Voltage Tolerance Limits [60.5 V, 64.5 V]
properly set to avoid false alarms and this way robustness of Active Power Reference 1.5 kW
the algorithm can be ensured.
Reactive Power Reference 0 Var

Kp & Ki value for DQ PI controller 6 & 90

Deadtime Value & Sensor
4 µs & 50 µs
Sampling Time

Authorized licensed use limited to: Consortium - Algeria (CERIST). Downloaded on August 02,2022 at 23:57:54 UTC from IEEE Xplore. Restrictions apply.
 2022 IEEE International Conference on Power Electronics, Smart Grid, and Renewable Energy (PESGRE)

(a) (b) (a) (b)

(c) (d) (c) (d)

Fig. 7. Healthy condition simulation results with grid load conditions (a) Pole
voltage H , (b) Line voltage , (c) Phase current , (d) Capacitor
voltages & 'G .

Both neutral point voltage and flying capacitor voltage

balancing schemes as mentioned in [18] & [19] are employed.
Fig. 7 demonstrates the performance of a healthy seven-level
grid connected hybrid inverter. Thereafter, few results are (e) (f)
presented and discussed below to evaluate the performance
of the detection method under various fault conditions. In this
section, two distinct fault conditions are considered, and, in
every simulation, open circuit fault is introduced at 0.4s (to
maintain the uniformity) by removing the respective gating
signals and keeping the body diode connected.
A. Open Circuit Switch Fault in S1: (g)
Fig. 8. Fault diagnostic variables under S1 open circuit fault conditions (fault
occurred at 0.4s) (a) Pole voltage H , (b) Line voltage
, (d) error variables 5& , (e) Average phase current 〈 &' 〉, (f) Average phase
Fig. 8 shows the waveforms for the inverter pole/line , (c) Phase current

angle ?$@A , (g) Average capacitor voltages

voltage, grid phase currents and various diagnostic variables
for open circuit fault in S1. At 0.4s, gating pulse to S1 is @A
& 'G@A .

variables are almost zero, when the fault occurs, variable 5

removed thus introducing open switch fault. Until 0.4s, error
much deviation in FC and NP voltages. After the fault event,

the other two variables 5 & 5 settles in ranges [-0.13, -0.09]

the phase angle settles at constant value i.e., 21 degrees. With
immediately increases and settles in a range [0.19, 0.25] and
the help of this value, SA1 fault can be easily distinguished
from other faults which have almost similar error variable
& [-0.08, -0.03] respectively showing that fault is in phase a.
ranges.
Although the error variables settle in a similar fashion for
In similar manner, the other switch faults can be detected.
other single open circuit faults (like SA2 or SA3 or S2 fault), S1
Further simulations are also done for various fault conditions
fault can be clearly distinguished from those faults by using
in b and c phases. It is proved that for b and c phases, fault
the behavior of additional parameters i.e., neutral point
detection and localization logic is similar to that of shown in
average voltage and flying capacitor average voltages. As T-
Fig. 6. This method however gives reliable results for power
leg switch failed, deviations can be observed in neutral point
factors ranging from 1 to 0.4 (both leading and lagging). At
voltage. Also, because of absence of positive phase current in
low power factors, it is observed that the diagnostic variables
phase a, average current value in phase a will be highly
and their pre-set detection ranges may overlap under different
negative and in other two phases it will be positive, this will
fault cases and thus it will be difficult to distinguish the faulty
severely affect the FC voltage balancing and as a result these
switch locations. However, in almost all practical cases, the
voltages also deviate from the normal values. After the fault
grid tied inverter for renewable power extracting purpose the
event, phase angle also settles around -140 degrees. From the
identified stable operating power factor range can be
results it can be clearly seen that fault can be safely detected
sufficient, and the proposed fault detection strategy can be
in 40ms i.e., within two fundamental periods.
employed there reliably.
B. Open Circuit Switch Fault in SA1:
VI. CONCLUSION
fault. When SA1 fault occurs at 0.4s, error variable 5 settles
Fig. 9 shows the fault diagnostic parameters in case of SA1
In this paper, a single switch open circuit fault detection

variable 5 settles at a negative value which is in the range [-

at a positive value which is in the range [0.01, 0.06] and strategy based on Park’s vector technique is presented for a

0.04, -0.01]. Whereas 5 settles at almost zero value. Under variables 5& (6 = 7, 9, : Iℎ7K5K) (which are evaluated from
seven-level hybrid inverter. The diagnostic variables are error

from its normal path and as a result 3 /4 level is not

this fault condition, for negative the current will deviate the three phase grid currents), average phase currents,

/2 level is achieved as
average neutral point and flying capacitor voltages. This
achievable and instead of this, method is independent of load current magnitude because
current flows through flying capacitor. Here, there is not normalization of phase currents is carried out before using

Authorized licensed use limited to: Consortium - Algeria (CERIST). Downloaded on August 02,2022 at 23:57:54 UTC from IEEE Xplore. Restrictions apply.
 2022 IEEE International Conference on Power Electronics, Smart Grid, and Renewable Energy (PESGRE)

[3] P. Roshankumar, P. P. Rajeevan, K. Mathew, K. Gopakumar, J. I.

Leon and L. G. Franquelo, "A Five-Level Inverter Topology with
Single-DC Supply by Cascading a Flying Capacitor Inverter and an
H-Bridge," in IEEE Transactions on Power Electronics, vol. 27, no. 8,
pp. 3505-3512, Aug. 2012.
[4] T. T. Davis and A. Dey, "Enhanced Floating Capacitor Voltage
Balancing Schemes for Single-Source Seven-Level Inverters With
Capacitor Fed H-Bridge Units," in IEEE Transactions on Industrial
(a) (b) Electronics, vol. 67, no. 8, pp. 6227-6236, Aug. 2020.
[5] F. W. Fuchs, "Some diagnosis methods for voltage source inverters in
variable speed drives with induction machines - a survey," IECON'03.
29th Annual Conference of the IEEE Industrial Electronics Society
(IEEE Cat. No.03CH37468), 2003, pp. 1378-1385 Vol.2.
[6] B. Lu and S. K. Sharma, "A Literature Review of IGBT Fault
Diagnostic and Protection Methods for Power Inverters," in IEEE
Transactions on Industry Applications, vol. 45, no. 5, pp. 1770-1777,
Sept.-oct. 2009.
(c) (d) [7] A. M. S. Mendes and A. J. Marques Cardoso, "Voltage source inverter
fault diagnosis in variable speed AC drives, by the average current
Park's vector approach," IEEE International Electric Machines and
Drives Conference. IEMDC'99. Proceedings (Cat. No.99EX272),
1999, pp. 704-706.
[8] U. Choi, K. Lee and F. Blaabjerg, "Diagnosis and Tolerant Strategy
of an Open-Switch Fault for T-Type Three-Level Inverter Systems,"
in IEEE Transactions on Industry Applications, vol. 50, no. 1, pp. 495-
508, Jan.-Feb. 2014.
(e) (f) [9] U. Choi, J. Lee, F. Blaabjerg and K. Lee, "Open-Circuit Fault
Diagnosis and Fault-Tolerant Control for a Grid-Connected NPC
Inverter," in IEEE Transactions on Power Electronics, vol. 31, no. 10,
pp. 7234-7247, Oct. 2016, doi: 10.1109/TPEL.2015.2510224.
[10] U. Choi, H. Jeong, K. Lee and F. Blaabjerg, "Method for Detecting
an Open-Switch Fault in a Grid-Connected NPC Inverter System," in
IEEE Transactions on Power Electronics, vol. 27, no. 6, pp. 2726-
2739, June 2012.
[11] S. Xu, J. Zhang and J. Hang, "Investigation of a Fault-Tolerant Three-
(g) Level T-Type Inverter System," in IEEE Transactions on Industry
Fig. 9. Fault diagnostic variables under SA1 open circuit fault conditions Applications, vol. 53, no. 5, pp. 4613-4623, Sept.-Oct. 2017.
(fault occurred at 0.4s) (a) Pole voltage H , (b) Line voltage
, (d) error variables 5& , (e) Average phase current 〈 &' 〉, (f)
, (c) Phase [12] J. O. Estima and A. J. Marques Cardoso, "A New Approach for Real-

Average phase angle ?$@A , (g) Average capacitor voltages

current Time Multiple Open-Circuit Fault Diagnosis in Voltage-Source
@A
& 'G@A . Inverters," in IEEE Transactions on Industry Applications, vol. 47, no.
6, pp. 2487-2494, Nov.-Dec. 2011.
them for evaluation of diagnostic variables. Based on the [13] A. Topcu and Y. Sozer, "Multiple Device Open Circuit Fault
Diagnosis for T-Type Multilevel Inverters," 2018 IEEE Energy
diagnostic variables, a simple flowchart is presented which is Conversion Congress and Exposition (ECCE), 2018, pp. 4056-4061.
helpful in locating the faults. Simulation results for a grid tied [14] M. R. Mamat, M. Rizon, and M. S. Khanniche, “Fault detection of 3-
system are shown to prove the effectiveness of the proposed phase VSI using wavelet-fuzzy algorithm,” Amer. J. Appl. Sci., vol.
method to detect and locate the single switch open circuit 3, no. 1, pp. 1642–1648, 2006.
[15] F. Charfi, F. Sellami, and K. Al-Haddad, “Fault diagnosis in power
faults. Fault detection time of less than 40ms (i.e., two system using wavelet transforms and neural networks,” in Proc. IEEE
fundamental periods) is observed in simulations. All Int. Symp. Ind. Electron., 2006, pp. 1143–1148.
thresholds and ranges for fault diagnostic variables are [16] M. A. Awadallah and M. M. Morcos, “Diagnosis of switch open-
evaluated properly through simulations by considering few circuit fault in PM brushless dc motor drives,” in Proc. Large Eng.
Syst. Conf. Power Eng., May 2003, pp. 69–73.
realistic conditions such as dead-time effect, code execution [17] J. Lamb and B. Mirafzal, "Open-Circuit IGBT Fault Detection and
time delay, parasitic drops, sensor sampling and thus it makes Location Isolation for Cascaded Multilevel Converters," in IEEE
the proposed strategy robust against false alarms. Authors Transactions on Industrial Electronics, vol. 64, no. 6, pp. 4846-4856,
are presently working towards the hardware validation of the June 2017.
[18] T. T. Davis and A. Dey, "Investigation on Extending the DC Bus
strategy. Utilization of a Single-Source Five-Level Inverter With Single
Capacitor-Fed H-Bridge Per Phase," in IEEE Transactions on Power
REFERENCES Electronics, vol. 34, no. 3, pp. 2914-2922, March 2019.
[1] X. Zhang, T. Zhao, W. Mao, D. Tan and L. Chang, "Multilevel [19] T. T. Davis and A. Dey, "A Neutral Point Voltage Balancing Scheme
Inverters for Grid-Connected Photovoltaic Applications: Examining With Improved Transient Performance for 5-Level ANPC and TNPC
Emerging Trends," in IEEE Power Electronics Magazine, vol. 5, no. Inverters," in IEEE Transactions on Power Electronics, vol. 34, no.
4, pp. 32-41, Dec. 2018. 12, pp. 12513-12523, Dec. 2019.
[2] S. Kouro et al., "Recent Advances and Industrial Applications of
Multilevel Converters," in IEEE Transactions on Industrial
Electronics, vol. 57, no. 8, pp. 2553-2580, Aug. 2010.

Authorized licensed use limited to: Consortium - Algeria (CERIST). Downloaded on August 02,2022 at 23:57:54 UTC from IEEE Xplore. Restrictions apply.