POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 21 | NUMBER: 4 | 2023 | DECEMBER

Identification of Open-Circuit Faults in T-Type

Inverters Using Fuzzy Logic Approach

Amina MIMOUNI 1 , Souad LARIBI 1 , Tayeb ALLAOUI 1 , Morsli SEBAA 1 ,

Abdelkader Azzeddine BENGHARBI 1

1
Laboratory of Energy Engineering and Computer Engineering,Ibn Khaldoun University,
BP 78, Tiaret, Algeria

[email protected], [email protected], [email protected],

[email protected], [email protected]

DOI: 10.15598/aeee.v21i4.5218

Article history: Received May 14, 2023; Revised Aug 07, 2023; Accepted Sep 03, 2023; Published Dec 31, 2023.
This is an open access article under the BY-CC license.

Abstract. With the increasing adoption of solar Pho- 1. Introduction

tovoltaic (PV) systems in several applications, reliabil-
ity and service continuity are important challenges that
need to be addressed. Power converters are vital com- The demand for global electricity continues its per-
ponents in solar PV systems and inverters tend to be sistent growth, serving both domestic and industrial
the most likely devices of equipment to experience faults needs. However, this heightened demand relies pre-
which usually occur in the switching devices. It is there- dominantly on fossil fuels like coal, natural gas, and oil
fore critical to assess the functioning of the inverters for electricity generation. Unfortunately, the world’s
and identify these faults in order to lower risks and the reserves of these resources are rapidly depleting, com-
resulting financial losses. Open-circuit faults (OCF) pounded by their adverse environmental consequences
are among the most common. This paper suggests a including pollution and global warming [1, 2].
fuzzy-based fault detection approach for the T-Type in-
Analyzing the data from 2010 to 2019 regarding car-
verter in grid connected PV system based on the diag-
bon dioxide (CO2 ) emissions linked to global energy
nosis variables which are calculated using the average
production from the power sector reveals a significant
values of positive and the negative parts of the normal-
pattern. Starting at 30.4 Gt in 2010, these emissions in-
ized output currents data. After that, these variables
creased to 31.3 Gt in 2011 and continued to rise in sub-
are analyzed using a fuzzy logic technique, The single,
sequent years: 2012 (31.6 Gt), 2013 (32.2 Gt), and 2014
multiple power switch open circuit faults may all be de-
(32.3 Gt). Emissions then remained steady in 2015 and
tected and diagnosed utilizing this fuzzy-based fault di-
2016 at 32.2 Gt, followed by a climb to 32.7 Gt in 2017,
agnostic technique. The results of the simulation in
MATLAB show that the proposed method can accu- and ultimately reaching 33.3 Gt in both 2018 and 2019
[3]. This disconcerting pattern has prompted a gradual
rately identify and locate OCF the inverter switches.
shift towards adopting cleaner, sustainable, and eco-
nomically viable alternative energy sources. The driv-
ing force behind this shift is a significant reduction in
CO2 emissions, especially in advanced economies. No-
tably, within these advanced economies, the total CO2
emissions have decreased by over 3.2 %, equivalent to
a reduction of around 370 million metric tons of CO2 .
Keywords Importantly, the power sector has played a major role
in achieving this decline, accounting for a substantial
85% of this accomplishment. This achievement high-
Grid connected PV system, T-Type inverter, lights the profound impact of significant modifications
Fault diagnosis, Open Circuit Fault, Fuzzy logic in electricity generation methodologies on the substan-
approach. tial attenuation of CO2 emissions [4].

c 2023 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 282

POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 21 | NUMBER: 4 | 2023 | DECEMBER

Over the past decade, the deployment of solar Photo- rule base for the proposed fuzzy logic-based fault de-
voltaic (PV) around the world has massively increased. tection approach. In [15, 16], the authors presented a
The PV market expansion was essentially driven by a diagnosis method based on bridge voltages for a T-type
lower production costs and supportive policies from lo- 3L inverter. An OCF detection method for the T-type
cal governments. These incentives are making solar in- 3L inverter using the two separate fault signals that
stallations’ Return on Investment (ROI) appear more are acquired was presented in [16]. A fault detection
attractive [5, 6, 7]. Like all other industrial processes, method was presented in [17] to identify multiple tran-
a PV system is susceptible to a variety of faults and sistor OCF in a T-type 3L inverter. In the study, state
anomalies that can affect its performance or even make transitions under different failure modes were distin-
it completely inoperable. All these negative effects will guished using a Finite-State Machine (FSM) recording
undoubtedly lower the installation’s productivity and state transitions and Rough Set Theory (RST). The au-
consequently, its profit. Not to mention the mainte- thors in [18, 19, 20] developed a method for diagnosing
nance cost to get the system back to normal operation an OCF and proposed a fault-tolerant control approach
[8, 9]. The reliability, safety, and quality of the energy for T-type inverter systems. The method was based on
produced can be significantly impacted by inverters, the calculation of the average of the normalized phase
which are at the core of the PV system. Therefore, current and the change in neutral-point voltage. In
inverters must meet increased reliability requirements, [21], a unique fault detection technique of OCF for 3L
which remain crucial for highly efficient energy conver- T-type inverter based on Silicon Carbide (SiC) MOS-
sion [10]. FET is presented. The main purpose of this method
is to detect changes in the dc-bus neutral-point cur-
Recent energy challenges have stimulated the devel-
rent and compare them to the neutral-point current’s
opment of new power electronic converter topologies
expected value during normal operation. Several re-
and technological improvement of semiconductor de-
search studies focusing on the detection and location
vices to handle the increasing power demand. Mul-
of OCFs in a Neutral Point Converter (NPC) have been
tilevel inverters (MLIs) have proven superior perfor-
published. In [22, 23], the fault diagnosis of NPC in-
mance over the conventional two-level (2L) topolo-
verter is based on a Neural Network (NN) classification
gies for wide output power ranges, and they are be-
method. The authors in [22] proposed an OCF diagno-
ing widely in a variety of applications, including re-
sis technique for grid-connected NPC inverters which
newable energy systems. These topologies increase the
is based on the Independent Component Analysis fault
inverter’s power capacity and enhance the quality of
feature extraction and NN classifier. This method is
energy by reducing the level of harmonics. In addition,
effective and uses only on the phase currents. The au-
they allow low voltage ratings of the components and
thors in [24, 25] presented a fault diagnosis method
reduces the output grid filter [11, 12]. There are several
in PV systems. In [24], six possible fault scenarios in
MLI topologies used in grid-tied PV installations. The
grid connected PV system including the inverter were
T-type three-level (3L) inverter topology has received
studied. The approach used was based on the extrac-
the most attention among the existing topologies. The
tion of features from the output currents using Dis-
main benefits of T-type topology are the elimination of
crete Wavelet Transform (DWT) and the identification
clamping diodes and flying capacitors and the reduc-
is based on NN classification. In [25], Artificial Neu-
tion of conduction losses in the power switches [13, 14].
ral Network (ANN) were used to classify open circuit
With regard to continuity of service, inverters are failures in IGBT power converters for PV systems.
particularly sensitive to faults in their power switches.
This paper proposes an OCF diagnosis technique for
An (OCF) or short-circuit fault (SCF) in a power
the 3L T-Type inverter in grid-connected PV system.
switch can cause serious system malfunctions. Any
The proposed approach uses the three-phase currents,
fault that is not detected can quickly damage the en-
and the fault diagnosis variables are calculated using
tire power converter stage and auxiliary circuits. Once
the average positive and negative normalized currents.
a fault has occurred, it must be quickly detected to
The information on the OCF of the switches can then
prevent it from spreading to other system’s compo-
be obtained by analyzing these variables using fuzzy
nents. Therefore, the implementation of efficient and
logic. The proposed fuzzy-based fault diagnostic algo-
fast methods of fault detection is imperative [8, 15].
rithm provides identification and location of single and
The behavior of the inverters under an OCF of a multiple power switch OCF.
power semiconductor fault, have been the subject of
extensive research and several solutions have been pro-
posed to detect faults and overcome the impact of sub-
sequent failures on the overall system. In [8], the be-
havior of the measured output current and dc-link ca-
pacitor voltages of grid-tie T-Type inverter was ana-
lyzed in healthy and faulty operation to design a fuzzy

c 2023 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 283

POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 21 | NUMBER: 4 | 2023 | DECEMBER

Fig. 1: Schematic of the grid-connected PV system.

2. Modeling of the Studied

PV System
To effectively transfer energy from the PV system to
the grid, a two-stage grid-connected PV system topol-
ogy is proposed in this study. Fig. 1 depicts the struc-
ture of the system which consists of a PV generator
with a DC-DC converter (boost converter) to track
the Maximum Power Point (MPP) and a three-level
T-Type inverter connected to the grid. The T-type in-
verter topology offers high quality of energy and low
Total Harmonic Distortion (THD).
Fig. 2: Equivalent circuit of PV cell.

2.1. PV Cell Model

and Ish is the shunt resistance current. The Vt thermal
The PV cell allows the direct conversion of solar en- voltage is given by:
ergy’s radiation into electricity [26]. Several models AKT
have been proposed for the PV cell. The one-diode Vt = , (3)
q
model has often been the most common due to its ease
of use and effectiveness in simulating the characteristics where A is a diode ideality factor, K represents Boltz-
of PV cells. This model includes a current source Iph mann’s constant (1.38×10−23 (J/K)), and q is the Elec-
which represents the photogenerated current, A shunt tron charge (1.6 × 10−19 C).
resistance (Rsh ), and a series resistance (Rs ) that re-
flects the losses in the semiconductor material [26, 27].
Fig. 2 shows the equivalent circuit of the PV cell used
2.2. MPPT controller design for the
in this work. DC-DC converter
The output current of a PV cell is given by [27, 28]: The power output of the PV generator (GPV) depends
Ipv = Iph − Id − Ish , (1) on several climatic factors, such as irradiation and tem-
perature. However, real-time tracking of the Maximum
 Ipv ×Rs +Vpv  Ipv × Rs + Vpv
Ipv = Iph − I0 exp V t − , (2) Power Point (MPP) is required to optimize the system
Rsh performance [1, 2]. The MPPT proposed in this paper
where Iph represents the light generated current, Id is is based on Adaptive Neuro-Fuzzy Inference Systems
the diode current, I0 is the diode saturation current (ANFIS).

c 2023 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 284

POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 21 | NUMBER: 4 | 2023 | DECEMBER

Fig. 3 depicts PV system with the DC-DC boost

converter conversion and the ANFIS-based MPPT con-
trol. The PV array is coupled directly to a filtering ca-
pacitor.The MPPT algorithm uses only the PV voltage
and current measurements to regulate the duty cycle
of the boost converter [29, 30].
ANFIS technique combines ANN learning ability
with fuzzy inference system (FIS). The suggested AN-
FIS controller’s inputs are identical to those of a fuzzy
controller, which are the error, the variation of the er-
ror, and a single output called the variation of the duty
cycle. The two input variables are determined using
the error E and the change of error dE given by equa-
tions (4) and (5), respectively. The boost converter is
driven by the duty ratio at the output, which enables Fig. 4: PV system with ANFIS-based MPPT controller.
the PV to be operated at its peak power [31, 32].

There have several MLI topologies proposed in the

literature. MLIs can generate a waveform with more
than 2L will be a typical feature. This feature re-
duces THD and increases efficiency compared to a 2L
inverter. The most popular topologies are the Neu-
tral Point Clamped (NPC), the Flying Capacitor (FC)
and the Cascaded H-bridge (CHB). A T-Type inverter
is derived from the traditional NPC with less passive
components to offer a smaller size to implement and
to mitigate the drawbacks of 3L T-Type NPC MLI.
T-Type topologies also benefit from the advantages of
two converters, such as low power losses and simple op-
Fig. 3: PV system with ANFIS-based MPPT controller. eration, and 3L inverters, such as enhanced efficiency
and smoother output voltage waveforms [11, 13, 14].
Ppv (k) − Ppv (k − 1) The fundamental structure of a 3L T-Type inverter
E(k) = . (4) is shown in Fig. 5. The conventional 2L VSI design is
Vpv (k) − Vpv (k − 1)
extended with the addition of an active, bidirectional
dE(k) = E(k) − E(k − 1), (5) switch to the dc-link midpoint. The three switching
states [P], [N], and [O] that correspond to the three
where P (k) and P (k − 1) represent the current and inverter output voltages +V /2, -V /2, and 0 can be
dc dc
previous values of the measured PV output power re- used to describe the main concept of this inverter in
spectively, V (k) and V (k − 1)refer to the measured Tab. 1[11].
current and previous PV output voltage respectively.
The structure of ANFIS model is presented in Fig. 4 Tab. 1: Possible states of the 3L T-Type inverter.
The number of epochs used for training ANFIS is 2000
and the membership functions for input variables E and Commutation type T1 T2 T3 T4 VAO
P ON ON OFF OFF +Vdc /2
dE are 4 shown in Fig. 4, Hence, there are 16 rules all O OFF ON ON OFF 0
together. The results of this approach provide a very N OFF OFF ON ON -Vdc /2
accurate calculation of the GPV’s maximum power.

2.3. Three Phase T-Type inverter • Positive (P) when (T1, T2) are closed.

As the use of grid-connected PV system has become

more widespread due to environmental concerns, the • Zero (O) when (T2, T3) are closed.
significance of efficient energy conversion via voltage
source inverters (VSI) has increased. The superior per- • Negative (N) when (T3, T4) switches are closed.
formance and wider applicability of MLIs have made
them favorable over conventional 2L inverters [12].

c 2023 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 285

POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 21 | NUMBER: 4 | 2023 | DECEMBER

 2 1
id = ia − (ib + ic )

3 3
. (6)
iq = √1 (ib − ic )

3
q
is = (id )2 + (iq )2 . (7)
ix
ix,n = , x(A, B, C). (8)
|is |
(
ix.n ⇐ ix.n > 0
i+
x.n = . (9)
0 ⇐ ix.n > 0
Fig. 5: Structure of a three-phase 3L T-Type inverter. (
0 ⇐ ix.n > 0
i−
x.n = . (10)
ix.n ⇐ ix.n > 0

−
3. Proposed Fault Diagnosis The average values i+ x.n.av and ix.n.av are calculated
using equations below (11) and (12) over a moving av-
Method erage window [33].

1 t +
Z
i+
x.n.av (t) = i (t) dt. (11)
The use of inverters in grid PV systems has raised T t−T x.n
awareness to the urgent need to address reliability and
1 t −
Z
service continuity. The power switches are among the −
ix.n.av (t) = i (t) dt. (12)
most vulnerable devices in the inverter circuit. They T t−T x.n
play a crucial role in maintaining the reliability of T-
Type inverter systems. Therefore, T-Type inverter The fundamental period is denoted by T, and the
condition monitoring and fault diagnostics are neces- inverter leg index is x.
sary to reduce the risk of failures and economic losses The threshold values Iav th1+ th2+ th1−
, Iav ,Iav th2−
and Iav
caused by faults [8, 15]. in this diagnostic method are chosen based on several
There are two distinct types of inverter switch faults: simulation tests of different possible scenarios. The
SCF and OCF. Power device SCF can happen very purpose is to set thresholds that accurately differenti-
quickly and are destructive. Therefore, hardware pro- ate between normal and faulty behavior of the system
tection circuits should be able to detect SCFs and pro- being diagnosed. These thresholds values should be
vide system protection. OCFs, on the other hand, do chosen to minimize false flags. The accuracy of our
not cause the system’s breakdown, but lead to per- diagnostic method relies on carefully chosen threshold
th1+ th2+ th1−
formance degradation. However, if this type of failure values: Iav = 0.28, Iav = 0.1, Iav =-0.28, and
th2−
remains undetected for a long time, this might result in Iav =-0.1 . The following expressions can be used to
unexpected system damage because it can put healthy generate fault symptom variables.
transistors under too much stress [16, 17].  + th1+
LP, ix.n.av > Iav

For the proposed fault diagnosis technique, the aver-
µ+x =
th2+
M P, Iav < i+ +
x.n.av 6 ix.n.av . (13)
age values of the normalized currents are used as input 
 + th2+
variables, the models for calculating diagnostic vari- SP, ix.n.av 6 Iav
−
ables will be presented. These variables i+x.n and i x.n  − th1−
should be in the range 0.31 and -0.31 when there is LN, ix.n.av > Iav

no fault. However, under certain operating conditions, µ−x =
th2−
M N, Iav > i− −
x.n.av > ix.n.av . (14)
the variances are not always exactly 0.31 and - 0.31. 
SN, i− th2−

x.n.av > Iav
To use this diagnostic technique, a Park’s current
where LP = Large Positive, MP = Medium Positive,
vector magnitude is first determined from the three-
SP = Small Positive, LN = Large Negative, MN =
phase current using (6) and (7) .
Medium Negative, SN = Small Negative. The fault
− − −
The magnitude of the Park’s current vector is used to indicator variables µ+ + +
A , µA , µB , µB , µC and µC can
determine the normalized currents (8) .The two compo- combine to generate a fault signature that is associ-
nents of the normalized currents are then divided into ated with an OCF operating mode in the 3L inverter.
two parts: one with positive values calculated from (9) For example, in phase A, the OCF location and iden-
−
and (10) one with negative values. tification is determined using µ+ A and µA . When µA =
+

c 2023 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 286

POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 21 | NUMBER: 4 | 2023 | DECEMBER

LP and µ− A = LN, then there is no fault, when µA =

+
+
MP then T2 is the faulty switch, and for µA = SP,
the OCF has occurred in T1. Whereas when OCF has
occurred in T3, the fault will be detected when µ− A=
MN and T4 is detected when µ− A = SN. This fault diag-
nosis approach can be applied to single-power-switch
open-circuit faults as well as the double power switch
open-circuit faults as shown Tab. 2.

Tab. 2: Diagnostic variables identification.

Faulty Switch µ+
A µ−
A µ+
B µ−
B µ+
C µ−
C
Fig. 6: Membership functions of the inputs.
T1 SP LN LP LN LP LN
T2 MP LN LP LN LP LN
T3 LP MN LP LN LP LN
T4 LP SP LP LN LP LN
T5 LP LN SP LN LP LN
T6 LP LN MP LN LP LN
T7 LP LN LP MN LP LN
T8 LP LN LP SN LP LN
T9 LP LN LP LN SP LN
T10 LP LN LP LN MP LN
T11 LP LN LP LN LP MN
T12 LP LN LP LN LP SN
T1||T4 SP SN LP LN LP LN
T5||T8 LP LN SP SN LP LN Fig. 7: Membership functions of the outputs.
T9||T12 LP LN LP LN SP SN
T1||T8 SP LN LP SN LP LP
T1||T12 SP LN LP LN LP SN
T4||T5 LP SN SP LN LP LN 3.2. Fuzzy logic rules
T4||T9 LP SN LP LN SP LN
T5||T12 LP LN SP LN LP SN In general, the procedure for operating a fuzzy system
T8||T9 LP LN LP SN SP LN
is accomplished in three steps:
Healthy LP LN LP LN LP LN

• The fuzzification block which converts numeri-

cal values into linguistic values using membership
functions (MFs). Fuzzification provides a series of
3.1. Fuzzy logic fault diagnosis fuzzy variables, joined by a vector, which will be
approach introduced to the inference block [36, 37].

The operating principle of fuzzy logic-based fault di- • In the inference block, the values of the linguistic
agnosis is to extract and compute input signals, using variables are linked by several rules that charac-
fuzzy rules represented as membership functions, which terize the behavior of the system under normal
can contain all possible combinations that correspond operating conditions or in the presence of a fault.
to the fault mode considered [8, 34].
• In the defuzzification block, the fuzzy values are
This technique can diagnose and identify faults that
converted back to real values that allows to infer
occur in three phase inverters. The proposed technique
the state of the system [36, 37].
is based on the analysis of the output currents of the
inverter, under normal (healthy) and faulty operation
[35] .The Mamdani-type fuzzy inference system was The fuzzy rules are extracted using the relationships
− −
adopted with six input variables (µ+ +
A , µA , µB , µB shown in Tab. 2. In the fuzzy logic control system,
+ −
, µC , µC ) from the characteristic extraction functions fuzzification is used to create fuzzy variables from nor-
−
and an output variable representing the state of health malized inputs (µ+ x , µx ) using MFs. The varied output
of the inverter considered in the study. In this fuzzy levels are chosen from 0 to 12 for switch faults, which
logic-based diagnosis approach, the output is the faulty can accurately depict the number of each switch fault,
switch number (Ti = 0,. . .,12). with zero denoting a healthy condition.
The input quantities are normalized to a range [-1,1]. The proposed fault diagnosis method is illustrated by
The membership functions of the inputs and outputs Fig. 8. Which was based on using the measurement of
are depicted in Fig. 6 and Fig. 7 respectively. the currents.

c 2023 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 287

POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 21 | NUMBER: 4 | 2023 | DECEMBER

Fig. 8: Block diagram of the proposed diagnostic method. Fig. 10: Current generated by the GPV.

4. Simulation of the diagnosis

approach

This section presents the simulation results of OCF in

our PV system, as well as the performance of the pro-
posed fault diagnosis technique. The PV system under
study is simulated in MATLAB/ Simulink software. To
assess the effectiveness of the proposed fault detection
method, a series of simulation tests and scenarios are
performed. The parameters of the grid-connected PV
system are summarized in Appendix A.
Fig. 9 shows the time-varying irradiance used
as input to the grid-connected PV system. The
temperature was fixed at 25 0 C.

Fig. 11: Power generated by the GPV.

This section introduces the simulation results ob-

tained from the faulty operation of the grid-connected
3L inverter. These results were obtained by modeling
the system in Matlab.
Specifically, an open circuit fault was generated at
switch Ta1 at time t = 1 s. The fault detection method
used is presented in the previous section. Fig. 12 illus-
trates the three-phase currents and the output of the
fault detection approach.
Fig. 12 depicts the simulation results of the three
Fig. 9: Irradiance of the GPV. phase currents along with the diagnostic outcomes.
When an OCF is introduced at t= 1s in the IGBT
switch Ta1 is introduced. The positive half-cycle of
current Ia gets eliminated. As a result, the diagnostic
Fig. 10 and Fig. 11 present the simulation results
variable µ+A drops instantly, and converges to a value
that illustrate the evolution of the current and power
of approximately 0.5e-3, which is below the threshold
output of the photovoltaic panel, as obtained through th2+
value Iav .
the ANFIS method algorithm. As can be observed, the
power of the panel accurately tracks changes in irradi- For the fault diagnosis, the fuzzy fault diagnosis ap-
ation and remains stable, with only minor oscillations proach was employed using a two output fuzzy logic
around the optimal power points. These results demon- system to address scenarios involving multiple faults
strate the effectiveness of the MPPT ANFIS control. within the T-Type inverter, this enhanced methodol-

c 2023 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 288

POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 21 | NUMBER: 4 | 2023 | DECEMBER

ogy allows for the simultaneous detection of multiple As shown in Fig. 13, when the fault occurred at t =
fault cases. Based on Fig. 12(c) the output of the 1.2 s, it affected the negative part of the leg A current
fuzzy logic is the fault diagnosis signal with a magni- waveform, the diagnostic variable µ− A rises to -0.5e-3
tude of 1. the fuzzy logic output accurately reflected enabling effective identification and localization of the
the presence of the single OCF, indicating a value of 1. failure.
The successful identification of the single faulty switch
demonstrates the potential of the developed diagnos-
tic approach It can already be noted that the fault is
indeed detected within a half of period.

(a)

(a)

(b)
(b)

(c)
(c)
Fig. 12: Simulation results of the PV system under OCF in Ta1
(a) Three Phase Currents, (b) Fault detection (c) Fault
Fig. 13: Simulation results of the PV system under OCF in Ta4
Location
(a) Three Phase Currents, (b) Fault detection (c) Fault
. Location.

Fig. 13 shows the simulation results obtained under

the fault conditions as previously simulated, an open An OCF of switches Tb3 and Tb2 (inner switches) at
circuit fault is applied to switch Ta4 at time t=1.2s. t = 1.2s is simulated in Figs. 14 and 15 respectively.

c 2023 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 289

POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 21 | NUMBER: 4 | 2023 | DECEMBER

the open-circuit fault occur in one of the inner switch

of the T-Type inverter. The phase current is distorted
after the open fault occurs. The diagnostic variable
µ+
B decreases immediately, and converges to a value of
th1+
0.21, which is below the threshold value Iav in the
case of OCF at Tb2 .
The simulation results obtained showed the effective-
ness of the proposed fault detection method. All faults
were accurately identified and classified. The proposed
(a) fuzzy diagnostic algorithm is now tested on its ability
to locate faults. Previous figures show the location of
different single faulty switches.
Since the percentage of three or more OCF is very
low, this paper will focus also on double-switch failures.
Figs. 16-17 show the scenario of two IGBTs experienc-
ing OCFs at the same time.
The results presented here confirm that the proposed
fault detection of OCF can function well during simul-
taneous OCFs. First, Fig. 16 which correspond to
(b) faults occurring in switches Ta1 and Ta4 belonging to
the same inverter leg. The results show the behavior
Fig. 14: Simulation results of the PV system under OCF in Tb3 for simultaneous OCFs in the upper switch and second
(a) Three Phase Currents, (b) Fault location. bottom switch.

(a)

(a)

(b)

Fig. 15: Simulation results of the PV system under OCF in Tb2

(a) Three Phase Currents, (b) Fault location.

(b)
Fig. 14 and Fig. 15 show the wave-forms of the
simulation results when the switches Tb3 and Tb2 are Fig. 16: Simulation results of the PV system under OCF in Ta1
open-circuited, respectively. It can be seen that when and Ta4 (a) Three Phase Currents, (b) Fault location.

c 2023 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 290

POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 21 | NUMBER: 4 | 2023 | DECEMBER

At t = 0.7s, the OCF in IGBT Ta4 is applied to the Fig. 17 depicts the simulation results for the three
fault in Ta1 , the diagnostic variable µ+
A instantly falls phase currents, as well as the fuzzy logic output. At
th2+
to 0.5e-3 and falls below the threshold Iav . When the time t = 1.5 s, the gate signals of the IGBT’s Ta1
the second fault in Ta4 is added, the second diagnostic and Tb4 are open circuited, resulting in failure of an
variable µ−A , rises to -0.5e-3 which allow for effective open-circuit in the first phase A and in phase B. It is
identification and location of both failures. evident that the suggested fault diagnostic algorithm
has the ability to identify and localize the problem in
For our developed fuzzy logic-based diagnostic ap-
a single crucial period at a variety of fault types and
proach, we simulated a scenario involving the occur-
locations.
rence of two simultaneous faults within the same leg of
the T-Type inverter. Notably, our method successfully
detected both faults, as evidenced by the outcomes ob-
served in the diagnostic outputs. Specifically, the first 5. Conclusion
diagnostic output responded with a value of 1 indica-
tive of the detection of the initial fault, while the second
This paper focuses on the identification and diagnosis
diagnostic output exhibited a value of 4 corroborating of power switch open circuit fault in a 3L T-Type in-
the successful identification of the second concurrent verter in PV system. The proposed approach is based
fault. This outcome validates the robustness and effi- on the extraction and analysis of averaged positive and
cacy of our fuzzy logic approach in addressing multiple negative parts of the normalized currents. The per-
fault scenarios. formance of the extracted features is analyzed under
both healthy and defective conditions in order to es-
In this instance, both the top switch Ta1 in phase
tablish the fuzzy bases and derive the rules of the rec-
A and the bottom switch of phase B Tb4 are open cir-
ommended fuzzy logic detection approach. The results
cuited. The detection and localization algorithm of two
of the case study under consideration demonstrate to
simultaneous IGBTs open circuit fault is carried out.
the effectiveness of the suggested detection method in
The phases currents are shown in Fig. 17.
precisely locating the faulty switch and identifying the
numerous faulty combinations situations. The simula-
tion results show that the proposed fuzzy based fault
diagnosis approach is very effective in identifying single
and multiple open-circuit faults in the power switches.
This study suggests several possibilities for future re-
search. Using advanced machine learning for more
accurate and efficient method of diagnosis which can
be tested on different topologies of inverters. More-
over, transitioning from simulation-based verification
to real-hardware implementation would facilitate the
validation of the method’s reliability and practicality.

Author Contributions
The idea for this paper was conceived by A.M. and
S.L., with S.L. supervising the work. A.M. and A.A.B.
(a) collaborated on the design of the OCF diagnosis for
the inverter and the formulation of the research diag-
nosis method. A.M. contributed to the writing of the
manuscript, while S.L., T.A., and M.S. provided crucial
feedback and approved the final version for submission.

References
[1] SHEIKH AHMADI, S.H, M. KARAMI, M. GHO-
(b) LAMI and R. MIRZAEJ. Improving MPPT
Performance in PV Systems Based on Integrat-
Fig. 17: Simulation results of the PV system under OCF in Ta1 ing the Incremental Conductance and Particle
and Tb4 (a) Three Phase Currents, (b) Fault location. Swarm Optimization Methods. Iranian Journal

c 2023 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 291

POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 21 | NUMBER: 4 | 2023 | DECEMBER

of Science and Technology, Transactions of Elec- [10] MA, K., D, ZHOU and F, BLAABJERG. Evalua-
trical Engineering. 2022, vol. 46, pp. 27–39. tion and Design Tools for the Reliability of Wind
DOI: 10.1007/s40998-021-00459-0. Power Converter System. Journal of Power Elec-
tronics. 2015, vol. 15,no. 5,pp. 1149—1157. DOI:
[2] KOLLIMALA, S and M. MISHRA.A Novel
10.6113/JPE.2015.15.5.1149.
Adaptive P&O MPPT Algorithm Consid-
ering Sudden Changes in the Irradiance. [11] SALEM, A., M.A, ABIDO. T-Type Multi-
IEEE Transactions on Energy Conversion. level Converter Topologies: A Comprehensive Re-
2014, vol. 29,pp. 602–610. ISSN 1804-3119. view.Arab J Sci Eng.2019, vol. 44,pp. 1713—1735.
DOI: 10.1109/TEC.2014.2320930. DOI: 10.1007/s13369-018-3506-6
[3] KHALEEL, M.M, M.R. Adzman, and S. M. Zali.
[12] SATTI, A., A. HASSAN and M. AHMAD. A New
An integrated of hydrogen fuel cell to distribu-
Multilevel Inverter Topology for Grid-Connected
tion network system: Challenging and oppor-
Photovoltaic Systems. International Journal of
tunity for D-STATCOM. Energies.2021, vol. 14,
Photoenergy.2018, vol. 2018, Article ID 9704346,
no. 21, pp. 7073, 2021. DOI: 10.3390/en14217073.
9 pages. DOI: 10.1155/2018/9704346
[4] KHALEEL, M.M. T. MOHAMED GUENDOURI.
A. ALI AHMED. A. ALSHARIF and all. Im- [13] ALTIN, N., I, SEFA., H, Komurcugil
pact of mechanical storage system technologies: and S, OZDEMIR. Three-phase three-
A powerful combination to empowered the elec- level T-type grid-connected inverter with
trical grids application. 2022 IEEE 2nd In- reduced number of switches. 6th Inter-
ternational Maghreb Meeting of the Conference national Istanbul Smart Grids and Cities
on Sciences and Techniques of Automatic Con- Congress and Fair. Istanbul:IEEE,2018. DOI:
trol and Computer Engineering. Sabratha, Libya: 10.1109/SGCF.2018.8408942
IEEE, 2022,pp. 628–636. DOI: 10.1109/MI-
[14] ZORIG, A., M, BELKHEIRI., S, BARKAT.
STA54861.2022.9837670.
Control of three-level T-type inverter-based
[5] NASIRI, M., S. CHANDRA, M .TAHERKHANI grid connected PV system.13th International
and S.J. MCORMAK. Impact of Input Capacitors Multi-Conference in Systems, Signals and De-
in Boost Converters on Stability and Maximum vices. Leipzig, Germany:IEEE, 2016. DOI:
Power Point Tracking in PV systems. In: 48th 10.1109/SSD.2016.7473723.
Photovoltaic Specialists Conference. Fort Laud-
erdale, USA: IEEE, 2021. ISBN 0160-8371. DOI: [15] CHEN, D., Y, LIU., S, ZHANG. Open-circuit
10.1109/PVSC43889.2021.9518903. fault diagnosis method for the T-type inverter
based on analysis of the switched bridge voltage.
[6] MOHAMED, S and M. ABD ELSATTAR. A com- IET Power Elctronics.2019, vol. 12, pp. 295–302.
parative study of P&O and INC maximum power DOI: 10.1049/iet-pel.2018.5377.
point tracking techniques for grid-connected
PV systems. SN Appl. 2019, vol. 01,pp. 174. [16] DANJIANG, C., L, YUTIAN, Z,
DOI: 10.1007/s42452-018-0134-4. SHOAZHONG.Fault diagnosis of T-type three-
level inverter based on bridge voltages.Archives of
[7] NASIR, A., A. RASOOL., D. SIBTAIN and
Electrical Engineering. 2021, vol. 70, pp. 73–87.
all. Adaptive Fractional Order PID Controller
DOI:10.24425/aee.2021.136053.
Based MPPT for PV Connected Grid System
Under Changing Weather Conditions. J. Electr. [17] WANG, B., Z, LI, Z, BAI., P.T, KREIN
Eng. Technol. 2021, vol. 16,pp. 2599—2610. DOI: and H, MA.Real-time diagnosis of multi-
10.1007/s42835-021-00782-w. ple transistor open-circuit faults in a T-
[8] ALY, M and H. REZK. An Efficient Fuzzy Logic type inverter based on finite-state machine
Fault Detection and Identification Method of model.CPSS Transactions on Power Electron-
Photovoltaic Inverters. Computers Materials and ics and Applications. 2020, vol. 5, pp. 74–85.
Continua. 2021, vol. 64,pp. 1546—2226. DOI: DOI:10.24295/CPSSTPEA.2020.00007.
10.32604/cmc.2021.014786.
[18] CHOI, U.M., K.B, LEE and F, BLAAB-
[9] ALY, M., H, AHMED and M, SHOYAMA. JERG.Diagnosis method of an open-switch fault
Developing new lifetime prolongation SVM al- for a grid-connected T-type three-level in-
gorithm for multilevel inverters with thermally verter system.3rd IEEE International Symposium
aged power devices. IET Power Electronics. on Power Electronics for Distributed Genera-
2017, vol. 10,pp. 2248—2256. DOI: 10.1049/iet- tion Systems. 2012, Aalborg, Denmark:IEEE.
pel.2017.0327. DOI:10.1109/PEDG.2012.6254044.

c 2023 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 292

POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 21 | NUMBER: 4 | 2023 | DECEMBER

[19] CHOI, U.M., K.B, LEE and F, BLAAB- tem using random forests technique.Energy Explo-
JERG.Diagnosis and Tolerant Strategy of an ration and Exploitation. 2018, vol. 36, pp. 132–
Open-Switch Fault for T-Type Three-Level In- 148. DOI:10.1177/0144598717723648.
verter Systems.IEEE Transactions on Indus-
try Applications. 2014, vol. 50, pp. 495–508. [28] YAQOOB, S., A, SALEH, S, MOTAHHIR and
DOI:10.1109/TIA.2013.2269531. all. Comparative study with practical validation
of photovoltaic monocrystalline module for single
[20] CHOI, U.M and K.B, LEE.Detection method of and double diode models.Sci Rep. 2021, vol. 11,
an open-switch fault and fault-tolerant strategy pp. 21822. DOI:10.1038/s41598-021-01357-5.
for a grid-connected T-type three-level inverter
system.2012 IEEE Energy Conversion Congress [29] GHAYTH, A., Z, Yusupov and M, Khaleel.
and Exposition (ECCE). 2012, Raleigh, NC, Performance Enhancement of PV Array Utilizing
USA:IEEE. DOI:10.1109/ECCE.2012.6342254. Perturb and Observe Algorithm. International
Journal of Electrical Engineering and Sus-
[21] HE, J., N, DEMERDASH, N, WEISE., tainability (IJEES).2023, vol. 1, no. 2,pp. 29–37.
R, KATEBI.A Fast On-Line Diagnostic Method DOI:https://ijees.org/index.php/ijees/article/view/33.
for Open-Circuit Switch Faults in SiC-MOSFET-
Based T-Type Multilevel Inverters.IEEE Trans- [30] KHALEEL, M., Yusupov, Z., Yasser, N., and J.
actions on Industry Applications. 2017, vol. 53, El-Khozondar, H. Enhancing Microgrid Perfor-
pp. 2948–2958. DOI:10.1109/TIA.2016.2647720. mance through Hybrid Energy Storage System
Integration: ANFIS and GA Approaches. Interna-
[22] HU, H., F, FENG and T, WANG. Open- tional Journal of Electrical Engineering and Sus-
circuit fault diagnosis of NPC inverter IGBT tainability (IJEES).2023, vol. 1, no. 2, pp. 38–48.
based on independent component analysis and DOI:https://ijees.org/index.php/ijees/article/view/34.
neural network.2020 7th International Con-
ference on Power and Energy Systems En- [31] GUENOUNOU, O., B, DAHHOU and F,
gineering. 2020, Fukuoka, Japan:Energy Re- CHABOUR.Adaptive fuzzy controller based
ports.DOI:10.1016/j.egyr.2020.11.273. MPPT for photovoltaic systems.Energy Conver-
sion and Management. 2014, vol. 78, pp. 843–850.
[23] WAN, X., H, HU and Y, YU. Open-circuit Fault DOI:10.1016/j.enconman.2013.07.093.
Diagnosis for Grid-connected NPC Inverter based
on Independent Component Analysis and Neural [32] REZK, H., M, ALY, M, AL-DHAIFALLAH and
Network.TELEKMNIKA. 2017, vol. 15, pp. 36– M, SHOYAMA.Design and Hardware Imple-
47. DOI:10.12928/TELKOMNIKA.v15i1.3677. mentation of New Adaptive Fuzzy Logic-Based
MPPT Control Method for Photovoltaic Appli-
[24] BENGHARBI, A.A., S, LARIBI, T, ALLAOUI cations.IEEE Access. 2019, vol. 7, pp. 106427 –
and A, MIMOUNI. Photovoltaic system faults di- 106438. DOI:10.1109/ACCESS.2019.2932694.
agnosis using discrete wavelet transform based
artificial neural networks.Electrical Engineering [33] MTEPELE, K., D, CAMPOS DOLGADO,
and Electromechanics. 2022, vol. 6, pp. 42–47. A, VALDEZ FERNANDEZ and J, PECINA
DOI:10.20998/2074-272X.2022.6.07. SANCHEZ. Model-based strategy for open-circuit
faults diagnosis in n-level CHB multilevel con-
[25] MIMOUNI, A., S, LARIBI, M, SEBAA, T, AL- verters.IET Power Electronics. 2019,vol. 12,no. 4
LAOUI AND A.A. BENGHARBI. Fault diag- pp. 648-655. DOI:https://doi.org/10.1049/iet-
nosis of power converters in a grid connected pel.2018.5478.
photovoltaic system using Artificial Neural Net-
works.Electrical engineering and electromechan- [34] LARIBI, S., M, BENDIABELLAH, M, YOUCEF
ics. 2023, vol. 1, pp. 4–10. DOI:10.20998/2074- and S, MERADI.Use of neuro-fuzzy technique in
272X.2023.1.04. diagnosis of rotor faults of cage induction mo-
tor.5th International Conference on Electrical En-
[26] KHAMIS, A., M, RUDDIN, A, GHANI gineering. 2017,BOUMERDAS, ALGERIA:IEEE.
and all.Control Strategy for Distributed DOI:10.1109/ICEE-B.2017.8192148.
Integration of Photovoltaic and Bat-
tery Energy Storage System in Micro- [35] LARIBI, S., M, BENDIABELLAH and
grids.TELEKOMNIKA. 2018, vol. 5, pp. 2415– S, MERADI. Fault diagnosis of rolling el-
2427. DOI:10.12928/telkomnika.v16i5.10249. ement bearings using artificial neural net-
work.International Journal of Electrical and
[27] IBRAHIM, I.A., T, KHATIB, A, MOHAMED Computer Engineering. 2020, vol. 10, no. 5,
and W, ELMENREICH.Modeling of the out- p. 5288–5295. DOI:10.11591/ijece.v10i5.pp5288-
put current of a photovoltaic grid-connected sys- 5295.

c 2023 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 293

POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 21 | NUMBER: 4 | 2023 | DECEMBER

[36] SHETGAONKAR, S.Fault diagnosis in induc- Engineering (L2GEGI) at Ibn Khaldoun University of
tion motor using fuzzy logic.2017 Interna- Tiaret, Algeria. His major research interests are on
tional Conference on Computing Methodologies application of artificial intelligence and robust control
and Communication. 2017, Erode, India:IEEE. techniques of process.
DOI:10.1109/ICCMC.2017.8282693.
Abdelkader Azzeddine BENGHARBI was
[37] MEHTA, P., S, SAHOO and H, DHIMAN. Open born in Algeria, received his BSc degree in Elec-
Circuit Fault Diagnosis in Five-Level Cascaded trotechnical Engineering and his MSc in renewable
H-Bridge Inverter.International Transactions on energies on Electrotechnique Engineering from Djelfa
Electrical Energy Systems. 2022, vol. 2022, pp. 13. University, Algeria. He is currently a PhD student
DOI:10.1155/2022/8588215. at Tiaret University. He is also a member of the
Energetic Engineering and Computer Engineering
Laboratory (L2GEGI) at the same university. His
About Authors research interests include renewable energy systems,
electrical machines, power systems, diagnosis, control
and artificial intelligence.
Amina MIMOUNI (corresponding author) was
born in Algeria, in 2019 she received the Engineer
degree in electrical engineering from Higher School of
Applied Sciences–Tlemcen,Algeria. She is currently
working toward the PhD degree in L2GEGI laboratory
in Ibn khaldoun University, Algeria. Her research Appendix A Grid connected PV
interests include, Renewable Energy, Fault diagnosis system parameters.
in power converters.

Souad LARIBI was born in Algeria, received Quantity Value

her Ph.D. degrees in Electrical Engineering from Maximum power (W) 45032.175
Voltage at (Maximum power) (V) 394.5
the University of Sciences and Technology of Oran Current at (Maximum power) (A) 114.15
Mohamed Boudiaf (USTOMB), Oran, Algeria, in Open circuit voltage (V) 493.5
2016 she receives her Habilitation degrees from the Short circuit current (A) 123.15
University Ibn Khaldoun Tiaret, Algeria, in 2020 in Vdc (V) 800
Electrical Engineering. Currently she working Senior Cpv 100 x 10-6
L 5 x 10-3
Lecturer at Université Ibn Khaldoun Tiaret Algeria. Switching frequency of the boost (Hz) 5000
Her fields of interest are electrical machines associated Switching frequency of the inverter (Hz) 10000
with static converter, control, modeling, and diagnosis Grid frequency (Hz) 50
and renewable energy systems. She is a member in Grid line voltage (V) 400
Energetic Engineering

Tayeb ALLAOUI was born in Tiaret, Algeria.

He received the diploma of Electrotechnical Engineer-
ing degree from Ibn Khaldoune University of Tiaret,
Algeria. The Master degree, from the University of
Science and Technology of Oran, Algeria.in 2002.
The PhD degrees from the University of Science and
Technology of Oran, Algeria, in 2007. His research
activities are mostly concentrated in the study of
power systems, FACTS, Renewable and Sustainable
Energy, power management and research activities
focusing a smart grid system.

Morsli SEBAA was born in Algeria, he received the

magister degree in automatic control from National
Higher Technical School of Oran (Algeria), and PhD
degree in control engineering from University of
Science and Technology of Oran (Algeria). He is now
a teacher of electrical engineering at the University
Ibn Khaldoun of Tiaret (Algeria) and a member in the
Laboratory of Energetic Engineering and Computer

c 2023 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 294