Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Open Circuit Fault Diagnosis in Five-Level Cascaded H-Bridge Inverter

Open Circuit Fault Diagnosis in Five-Level Cascaded H-Bridge Inverter Hindawi International Transactions on Electrical Energy Systems Volume 2022, Article ID 8588215, 13 pages https://doi.org/10.1155/2022/8588215 Research Article Open Circuit Fault Diagnosis in Five-Level Cascaded H-Bridge Inverter 1 2 2 Pavan Mehta , Subhanarayan Sahoo, and Harsh Dhiman Gujarat Technological University, Ahmedabad, Gujarat, India Adani Institute of Infrastructure Engineering, Adalaj, Gujarat, India Correspondence should be addressed to Pavan Mehta; pvnmehta55@gmail.com Received 3 January 2022; Revised 23 March 2022; Accepted 25 March 2022; Published 20 April 2022 Academic Editor: Jesus Valdez-Resendiz Copyright © 2022 Pavan Mehta et al. 'is is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. 'e development of power electronic converter, especially multilevel converter, is remarkable for several decades. 'e complex switching and increased power of semiconductor devices are prime reasons for faults in multilevel inverters and have raised question about reliability. To improve the reliability, a cost-effective solution in terms of fault diagnosis is essential. In this context, this study proposed an open circuit fault (OCF) diagnosis technique for a switching device in a five-level cascaded H-bridge multilevel inverter using fuzzy logic control. 'e OCF features like output voltage total harmonic distortion (THD) and normalized average output voltage are fuzzed as input variables of the fuzzy logic controller. 'ese input variables are divided into various triangular antecedent membership function (MF). 'e output produced by the fuzzy controller as consequent MFs is divided into different levels to identify the faulty switch. In order to make a complete fault-tolerant structure, a reduced modulation index-based postfault control is suggested to get a balanced output voltage. 'e MATLAB/Simulink results and prototype results are the evidence to support the proposed fault diagnosis technique. point clamped (ANPC) [6] and DC link converter [7] are 1. Introduction deduced from basic structures to overcome their limitations. Over the last several decades, the application area of mul- In the ML inverter, the probabilities of failure of power tilevel (ML) inverters is grown very fast. 'e different ap- semiconductor switching devices are higher due to inter- plications of ML inverters are mine hoists, gas turbine action between the numbers of other switching devices. It starters, hydro-pumped storage, high-voltage DC (HVDC) reduces the reliability of the ML inverter system. In a broad, transmission, reactive power compensation, wind energy the causes of fault in a power semiconductor device can be conversion, power generation using PV cells, railway trac- categorized as exterior and interior. 'e exterior possibilities tion, AC drives, and marine propulsion [1]. Traditionally, for switch failure may be swelled or dipped from the external the two-level three-phase voltage source inverters (VSIs) supply connected to the terminals of the ML inverter, and were used to operate the drives. For the high-power ap- high dynamic changes at the load side and short circuit of the plications, the limitations with two-level VSI arise due to the load. On the other side, the interior possibilities for switch limited voltage withstand capacity of power semiconductor failures may be thermomechanical fatigue, gate misfiring, switching devices. 'erefore, multilevel (ML) VSI was and saturation of semiconductor materials [8]. Due to the adopted, which has many advantages such as low total abovementioned causes, the power switches may be either harmonic distortion (THD), low electromagnetic interfer- permanently opened or closed. Depending upon the cause of ence (EMI), reduced filter size, and low dv/dt [2] due to its failure, if the switching device is permanently open then it is staircase output waveform quality. 'e neutral point called an open circuit fault (OCF) and if the switch is clamped (NPC) [3], flying capacitor (FC) [4], and cascaded permanently closed then it is called a short circuit fault (SCF). 'e SCF in any switching device must be detected H-bridge (CHB) [5] are the basic structures of ML con- verters. 'e other popular structures such as active neutral within 10 µsec (depending on the semiconductor chip). 2 International Transactions on Electrical Energy Systems After the fault detection of SCF, the faulty switch must be where E is the output voltage of any one cell of the th cell X isolated, and a complete shutdown is mandatory [9]. On the phase, Sf and Sf are the switching functions (0 or 1) 1 or 5 2 or 6 other side, during the OCF the ML inverter peruses to of switch S or S and S or S of a cell, and E is the X1 X5 X2 X6 DC operate in faulty condition with reduced output quality. DC source of a cell. 'e total output phase voltage E is the OX However, this may lead to increased voltage stresses across summation of the individual cell output and can be given as the other healthy switches. Hence, the faulty switch location follows: must be identified. E � E + E . (2) OX cell1 cell2 'e different fault diagnosis techniques are reported in the literature such as time voltage criterion [10], switching 'eoutputvoltagedependsonitsoccupiedswitchingstates, time-domain OCF detection [11], asymmetric zero voltage i.e., 2p, p, o, n, and 2n. For an illustration, as shown in the switching [12], neural network and artificial intelligence (AI) waveformofFigure2,duringthehealthycondition,iftheoutput [13, 14], histogram [15], harmonic frequency analysis [16], voltagelevelsofanylegare+2E –+E –0–−E –−2E then DC DC DC DC and output voltage or current analysis [17, 18]. In general, allswitchingstatesareoccupiedandnoneoftheswitchingstate the fault diagnosis process is to perform different signal is empty. Figures 3(a) and 3(b) show the output phase voltage, systems or mathematical operations on the sensed output empty and occupied switching states, and current flow during quantity and to extract the unusual features of the sensed healthy and faulty conditions for switches S and S , re- X1 X6 output quantity like voltage, current, and power. 'e fault spectively, during their switching operations. diagnosis techniques are classified into three basic cate- As shown in Figure 3, if we apply Kirchhoff’s voltage law gories: waveform analysis, AI-based techniques, and har- (KVL) in the circuit, then the green line will be the actual monic frequency analysis as shown in Figure 1. path of current before the fault at the respective switching 'e abovementioned techniques require extra sensors, state of the switch. When there is the condition of OCF veryfastcontrollers,andmore computationaleffortsforfault occurs, then according to KVL the current path will change diagnosis.'erefore,thisstudyproposesacost-effectiveOCF and flow as per the red dotted line at respective switching diagnosistechniqueforaswitchoffive-levelCHBMLinverter states. For an illustration, during the healthy condition of S X1 using fuzzy logic control merged with waveform analysis. the current path is E -S -a-d-S -E -S -b-S -E , DC X1 X7 DC X5 X3 DC Also,thefaultdiagnosistimetakenbytheproposedtechnique which is the cause for generating +2E voltage level. While DC is very less is ensured. 'e different fuzzy logic control lit- during the faulty condition of S , if we apply KVL then the X1 erature studies are available for the following: switching of direction of current will be as follows: a-d-S -E -S -b- X7 DC X5 CHB ML inverter [19], nine-phase IM drive fault-tolerant S -D -a, which will be the cause for missing of +2E X3 X4 DC operation[20],andMLinverterwithphotovoltaiccell[21].In voltage level. 'erefore, the switching states p, o, n, and 2n this study, fuzzy control theory is used to fuzz the fault are occupied and 2p is considered as empty as mentioned in symptom variables and related fuzzy output levels generated Table 1. In this manner, we can find the current path and toidentifythefaultyswitch.Forimplementingthistechnique, output voltage during healthy and faulty conditions for all werequireonlyonevoltagesensorperphase,whichisalready switches. 'e expected current path, actual current path, available with main control of any closed-loop operation and expected output voltage level, and the missing output voltage no requirement of any extra hardware circuitry. It can be level at the time of OCF are summarized in Table 2. implemented in any existing system without making major 'e other switching state’s occupancy, emptiness, and changes in the control scheme. outputvoltagerelatedtothefaultyswitchinleg xwhere xϵ{A, In Section 2, the overview of CHB ML inverter for B, C} are shown in Table 1. From the analysis of Table 1, the healthy and faulty conditions and the principle of fault output voltage THD of the xth phase depends on the empty diagnosis technique are presented. In Section 3, the fuzzy states and related possible output voltage level. 'is is because logic control is highlighted for OCF diagnosis. 'e THD and thenonlinearityoftheoutputvoltagewillincreaseastheempty normalized average output phase voltage threshold esti- outputvoltagestateincreases.'eTHDindex(T )forOCFin mation is discussed in Section 4. 'e simulation results are S and S is comparatively low but higher than the healthy X1 X6 discussed in Section 5. 'e postfault control strategy is condition T . Similarly, the T of S and S are almost the X X X2 X5 suggested in Section 6. 'e hardware prototype of the five- same as they have only one empty state been n and p, re- level CHB ML inverter and its results are presented in spectively. 'e lower switches in each bridge of all the three Section 7. phasesareutilizedforzerooutputvoltageswitching.Hence,the OCFintheswitchesS ,S ,S ,andS givesthehighestT as X3 X4 X7 X8 X theycreatethetwoemptystates,namely,(o, p),(o,n),(o,p),and 2. CHB ML Inverter and OCF (o, n), respectively. In this contrast, THD of the output phase voltage can be a fault symptom variable. 'e identification of 'e generalized structure of the single-phase five-level CHB the different range of T is achieved by selecting appropriate ML inverter and healthy condition output phase voltage E X OX threshold values. 'e T of output phase voltage is given by is shown in Figure 2. 'e output voltage of any phase X equation (3). However, to get confidence in the correct fault depends on the individual cell output (E ). 'e output cell diagnosisandexactfaultlocationanotherthresholdisrequired voltage of a cell can be expressed as follows: for two-stage verification. 'erefore, the normalized average E � Sf − Sf 􏼁, (1) cell 1or5 2or6 output phase voltage (E ) is computed and compared with SavN International Transactions on Electrical Energy Systems 3 Fault detection in CHB ML inverter Waveform AI based Frequency Analysis Techniques Analysis • Time Voltage Criterion • Harmonic Frequency • Neural Network • Asymmetric Zero Voltage • AI & Neural Network Analysis Switching • Histogram & Neural • THD & Normalized Current Network Factor Figure 1: Classi“cation of fault diagnosis methods for CHB ML inverter. S S x1 x2 x1 x2 b E E cell1 DC 2p +2E DC S S x4 x3 D D x4 x3 p +E DC o 0 n -E S S DC x5 x6 x5 x6 2n -2E DC cell2 DC S S x8 x7 D D x8 x7 Figure 2: Structure of “ve-level CHB ML inverter and output phase voltage. their threshold values. To compute the value of E , the SavN E [k] E [k]+ j × E [k], (5) sav dav qav average values of all the phase voltages (E , E , and E ) A[k] B[k] C[k] arecalculatedasperequation(4),wherenisthetotalnumberof Where samples and k is sampling index. For better resolution, at least E [k] E × sin ω × T × k ,   A m s 50 samples should be collected from each fundamental cycle.  �e fundamental output phase voltages are expressed as in equation (6), where E is the maximum voltage and ω is the  2π E [k] E × sinω × T × k − , B m s (6) fundamental frequency. �e average park’s transformation on voltage vector is applied and represented as per equation (5).  During the normal operating conditions or healthy operation,   2π  E [k] E × sin ω × T × k + , the value of E is almost zero. When the fault occurs, then the C m s Xav value of E deviates and crosses the threshold limit. Xav Where h 2 max h2 OXh (3) 2 1 %THD T  × 100,   X  E [k] E [k]−  E [k]+ E [k],  dav Aav Bav Cav OX1  3 3 (7) 1 1 E [k]   E [k],Xε{} A, B, C , (4) Xav √ X  E [k] E [k]− E [k]. qav Bav Cav 3 3 k1 Output Voltage States 4 International Transactions on Electrical Energy Systems x1 x2 x2 x1 DC S S x4 x3 D D +E x4 x3 DC -E S DC x5 x6 x5 x6 2n -2E DC DC S S x8 x7 D D x8 x7 (a) x1 x2 x2 x1 DC +2E DC 2p S S x4 x3 +E DC D D x4 x3 -2E DC x5 x6 x5 2n x6 DC S S x8 x7 D D x8 x7 (b) Figure 3: Operation of “ve-level CHB ML inverter under OCF condition for (a) S and (b) S . X1 X2 However, this deviation needs to be a normalized value On the other side, when the fault occurs in any of the called E that is calculated and expressed as equation (8) switches, which is cause for 2n or n empty states, the E XN SavN and the average of normalized value E is calculated as per goes to a diœerent range in positive values. SavN equation (10). �e normalized average output phase voltage varies depending upon the faulty switch. For illustration, E [k] E  ,Xε{} A, B, C , (8) XN during the normal operating conditions the E is almost SavN E [k] zero and šoats between +0.1 and −0.1. When the fault occurs in any of the switches, which is the cause for 2p or p empty E [k] E [k]+ jE [k], (9) s sd sq states, the E goes to a diœerent range in negative values. SavN Output Voltage State Output Voltage State International Transactions on Electrical Energy Systems 5 Table 1: Occupied states, empty states, and membership functions for OCF in switch. Antecedent Condition of switch in leg x Consequent MFs MFs Occupied states Empty state Possible output voltage level S S S S S S S S T E S X1 X2 X3 X4 X5 X6 X7 X8 X savN Xi Ok Ok Ok Ok Ok Ok Ok Ok 2p, p, o, n, 2n null +2E , +E , 0, −E , −2E t �S e �ZR 0 DC DC DC DC 0 0 f Ok Ok Ok Ok Ok Ok Ok p, o, n, 2n 2p +E , 0, −E , −2E t �M e �NS I DC DC DC 1 2 Ok f Ok Ok Ok Ok Ok Ok 2p, p, o, 2n n +2E , 0, −E , −2E t �M e �PM II DC DC DC 1 3 Ok Ok f Ok Ok Ok Ok Ok 2p, n, 2n p, o +2E , −E , −2E t �L e �NL III DC DC DC 2 6 Ok Ok Ok f Ok Ok Ok Ok 2p, p, 2n n, o +2E , +E , −2E t �L e �PL IV DC DC DC 2 5 Ok Ok Ok Ok f Ok Ok Ok 2p, o, n, 2n p +2E , 0, −E , −2E t �M e �NM V DC DC DC 1 4 Ok Ok Ok Ok Ok f Ok Ok 2p, p, o, n 2n +2E , +E , 0, −E t �M e �PS VI DC DC DC 1 1 Ok Ok Ok Ok Ok Ok f Ok 2p, n, 2n p, o +2E , −E , −2E t �L e �NL VII DC DC DC 2 6 Ok Ok Ok Ok Ok Ok Ok f 2p, p, 2n n, o +2E , +E , −2E t �L e �PL VIII DC DC DC 2 5 Table 2: Current path for healthy and faulty condition for all switches. OCF switch Current path for healthy condition Current path for faulty condition Missing output voltage level S E -S -a-d-S -E -S -c-b-S -E a-d-S -E -S -c-b-S -D -E +2E X1 DC X1 X7 DC X5 X3 DC X7 DC X5 X3 X4 DC DC S E -S -b-c-S -D -d-b-S -E b-c-S -D -d-b-S -D -a -E X2 DC X2 X8 X7 X3 DC X8 X7 X3 X4 DC S E -S -a-d-S -D -c-b-S a-d-S -D -c-b-D -S +E X3 DC X1 X7 X8 X3 X7 X8 X2 X1 DC S E -S -b-c-S -D -d-b-S b-c-S -D -d-a-D -S -E X4 DC X2 X8 X7 X4 X8 X7 X1 X2 DC S E -S -c-b-S -D -a-d-S c-b-S -D -a-d-S -D +E X5 DC X5 X3 X4 X7 X3 X4 X7 X8 DC S E -S -b-c-S -E -S -d-a-S -E E -S -b-c-S -D -d-a-S -E -2E X6 DC X2 X8 DC X6 X4 DC DC X2 X8 X7 X4 DC DC S c-b-S -D -a-d-D -S E -S -c-b-S -D -a-d-S +E X7 X3 X4 X6 X5 DC X5 X3 X4 X7 DC S E -S -d-a-S -D -b-c-S d-a-S -D -b-c-D S -E X8 DC X6 X4 X3 X8 X4 X3 X5- X8 DC 'e (δ –δ ) and (λ –λ ) are different thresholds for T 1 3 1 6 X E [k] sav E [k] � . (10) savN and E for different empty states. 'e antecedent SavN E [k] membership functions are designed based on the input variables. 'e distribution of T is performed as t , t , and t X 0 1 2 present small T (S), medium T (M), and large T (L), X X X 3. OCF Diagnosis Using Fuzzy Logic Control respectively. Similarly, the distribution of E is performed savN as e , e , e , e , e , e , and e present different normalized From the above discussion, the T and E are the fault 0 1 2 3 4 5 6 X SavN output voltages, namely, zero (ZR), positive small (PS), symptom variables and combinedly used to accurately di- positive medium (PM), positive large (PL), negative small agnose the faulty switch. 'e detailed algorithms of the (NS), negative medium (NM), and negative large (NL), proposed method are as follows. respectively. 'e consequent MFs are designed in such a way to represent the faulty switch identification number from I 3.1. Fuzzy Logic Fault Diagnosis Reasoning. 'e diagnosis (S ) to VIII (S ), and 0 represents the healthy condition. X1 X8 architecture is created based on the analytical and heuristic Figures 4(a) and 4(b) show the graphical representation of knowledge of symptoms of the CHB ML inverter. Heuristic the antecedent MFs and consequent MFs, respectively. 'e knowledge in the form of qualitative process models can be output of fuzzy logic will be designated as 0 or 1 for healthy expressed as if-then rules. 'e nine input MFs and variables and faulty conditions, respectively, for each switch. T and E are fuzzified as equation (11). X SavN t , 0≤ T ≤ δ , ⎧ ⎪ 0 X 1 3.2. Extraction of Fuzzy Rules. Based on the relationships in Table 1, the fuzzy rules are extracted. In general, the fuzzy T � t , δ ≤ T ≤ δ , X 1 1 X 2 control system consists of fuzzification, fuzzy inference, and t , δ ≤ T ≤ δ , 2 2 X 3 defuzzification illustrated in Figure 5. e , E < λ , λ , ⎧ ⎪ ⎪ 0 SavN 2 1 (i) Fuzzification: Fuzzification is the process of defining e , λ < E < λ , 1 1 SavN 3 the fuzzy variables from input variables using MFs. (11) ⎪ As explained in Section 3.1, the T and E are the ⎪ X SavN e , λ < E < λ , ⎪ 2 2 SavN 4 input variables. 'e (S, M, and L) and (ZR, PS, PM, E � e , λ < E < λ , SavN 3 3 SavN 5 ⎪ PL, NS, NM, and NL) are different MFs of T and ⎪ E , respectively. e , λ < E < λ , ⎪ SavN 4 4 SavN 6 ⎪ (ii) Fuzzy interface: It expresses the relation between ⎪ e , λ < E , ⎪ 5 5 SavN ⎪ input fuzzy variables and output using if-then rules. e , λ < E . 6 6 SavN 'e various fuzzy rules can be expressed in Table 1. 6 International Transactions on Electrical Energy Systems δ S i λ Xi 0 Ι ΙΙ ΙΙΙ ΙV V VΙ VΙ VΙΙΙ t t t e e e e e e 0 1 2 6 4 2 1 3 5 T E X SavN (a) (b) Figure 4: (a) Antecedent MF distribution. (b) Consequent MF distribution. Fuzzy Membership Input Variable Defuzzification Interface Functions of Input Membership Output Functions of Membership Output Input Variable Functions of Input Fuzzy Rules Figure 5: Fuzzy interfacing process. �ere will be total of nine rules associated per phase conditions and modulation index. On the other side, the (x) in the fuzzy interface. For illustration, if T  M normalized average phase output voltage has diœerent levels and E  NS then output S  I represents S of value in positive and negative regions depending upon the SavN Xi X1 OCF. If T  M and E  PM then output S  II faulty switch. Once the faulty phase is identi“ed using T , the X SavN Xi X represents S OCF. If T  M and E  PS, then E will be calculated. �e actual value is compared with the X2 X SavN SavN output S  VI represents S OCF. �e fault fea- thresholds, and the faulty switch is declared. Xi X6 tures created by switches S –S and S –S are the Figures 6(a) and 6(b) show the T and E variation X3 X7 X4 X8 X SavN same. Hence, identi“cation of the faulty switch is with respect to modulation index and their thresholds for achieved on a priority basis. If the OCF occurred at various fault conditions. �e (δ , δ , and δ ) are the 1 2 3 , then S will be initially appeared as a faulty thresholds selected for THD for no-fault condition, fault in X7 X3 switch as per priority. However, in the next funda- S /S , and fault in S /S , respectively. �e values beyond X1 X6 X2 X5 mental cycle the correct faulty switch will be diag- δ represent the fault in S /S /S /S . �e selection of the 3 X3 X4 X7 X8 nosed, i.e., S . On theother side, ifthe OCFoccurred THD threshold gives a fair fault signature up to a 0.4 X7 at S , then S will be initially appeared as a faulty modulation index. In practical conditions, it is not an ad- X8 X4 switch as per priority and in the next fundamental visable modulation index below 0.5 because the output cycle the correct faulty switch will be diagnosed, i.e., voltage will be halved. �e value of E šoats between SavN S . �e time taken for fault diagnosis will be higher +0.01 and −0.01 in healthy condition. When the fault occurs X8 as much as one fundamental cycle for S and S in the switch, which is cause for the 2p or p empty states, the X7 X8. (iii) Defuzzi“cation: �e defuzzi“cation can be de“ned value of E goes to negative. While the OCF is in the SavN as the procedure of converting the fuzzy output set switch, which causes 2n or n empty states, the value of E SavN into the crisp set. �e max-min composition and goes to positive. �e (λ , λ , λ , and λ ) are the thresholds 1 2 5 6 centroid of area method are selected for defuzzi“- selected for E for faults in S ,S ,S , and S re- SavN X1 X6 X5 X2 cation in the proposed method. �e output of spectively. �e thresholds beyond λ and λ represent the 3 4 consequent MFs as shown in Figure 4(b) will decide faults in S /S and S /S , respectively. X3 X7 X4 X8 the output of the fuzzy logic controller and so faulty switch. �e diœerent levels of output are selected 5. Simulation Results and Discussion from 0 to VIII for switch faults S to S , which can X1 X8 directly rešect the number of respective switch fault �e proposed fault diagnosis method is tested using as shown in the simulation results section, where MATLAB/Simulink. For testing the robustness of the fault zero indicates the healthy condition. diagnosis technique, the simulation is performed with a 0.8 power factor with a three-phase IM drive. �e level-shifted pulse width modulation technique (LS-PWM) is used for 4. Estimation of Thresholds switching. �e threshold estimation process plays a key role in the se- Figure 7 shows the THD plot for all three phases for lection of fuzzy rules. Hence, the acceptable threshold au- healthy and S faulty conditions. When there is a healthy A1 thentication is required to avoid false fault diagnosis. We condition, the THD of all three phases are the same and so know the fact that THD may change with the diœerent loading T will not cross the threshold limit. As soon as the OCF X International Transactions on Electrical Energy Systems 7 +0.07 +0.05 +0.01 -0.01 -0.05 δ λ 3 3 -0.07 0 0.5 1 0 0.5 1 Modulation Index Modulation Index E at S /S Faults E at S Faults SavN X4 X8 SavN X1 %THD at S /S /S /S Faults X3 X4 X7 X8 E at S Faults E at S Faults SavN X2 SavN X5 %THD at S /S Faults X2 X5 E at S Faults E at S /S Faults SavN X6 SavN X3 X7 %THD at S /S Faults X1 X6 %THD at No Faults (a) (b) Figure 6: (a) THD threshold estimation. (b) E threshold estimation. SavN Phase-A Phase-B Phase-C Healthy Condition THD=30.01% THD=30.10% THD=29.99% S OCF Condition THD=42.50% THD=30.00% THD=30.05% A1 0 1 2 3 4 5 01 2 3 4 5 0 1 2 3 4 5 Frequency (kHz) Figure 7: THD plot for healthy and S faulty conditions. A1 occurs in switch S , the T will cross the threshold limit δ both will have the same fault features. �e value of E is A1 A 1 SavN as discussed in Section 4. �e THD of the remaining two +0.05, and fuzzy output settles at 4 V. �e fault diagnosis of phases will be the same. �erefore, the faulty phase is now S is achieved by approximately 0.038 sec. X4 identi“ed and E will be calculated. �e value of T for Figure 9(a) shows the OCF results for switch S . �e SavN X X6 diœerent switch faults will be diœerent depending upon the value of E is +0.035, and fuzzy output represents the 6 V. SavN absence of output voltage level as discussed in Section 2. Figure 9(b) shows the results for OCF in switch S . It can be X7 By analyzing Figure 8(a), it can be noted that the fault is observed that the S and S provide the same empty states, X3 X7 created at 0.14 sec in switch S , yet the eœect of fault in the i.e., o and p. As per priority, S will be diagnosed as the X1 X3 output phase voltage waveform and output current cannot faulty switch and S remains undiagnosed but in the next X7 be seen. �e fault diagnosis algorithm will only be active fundamental cycle S is detected as the faulty switch. �e X7 when the faulty switch’s switching state arrives. fault is detected in two and half fundamental cycles or Hence, till the switching state of the respective faulty 0.045 sec. switch, the OCF remains undiagnosed. �e value of E at SavN the time of fault is −0.035 as S is cause for missing in 2p X1 6. Postfault Control empty state. �e fuzzy logic output is considered as a fault diagnosis signal, and its magnitude is 1 V as per the fuzzy In Section 2, we have seen that the CHB ML inverter will algorithm. �e OCF for switch S is diagnosed within pursue to operate with reduced output quality under the X1 0.025 sec. Figure 8(b) shows the simulation results for OCF OCF in a switch, which prevents the complete shutdown of in switch S . As per discussion in Table 1, the fault in S load. However, this will create unbalancing in the three- X4 X4 and S both will cause for n and o empty states. �erefore, phase output voltage. �e unbalanced three-phase output X8 %THD (Output phase voltage) Mag (% of fundamental) SavN 8 International Transactions on Electrical Energy Systems -200 -200 -10 +0.05 -0.035 OCF OCF Created OCF Created Diagnosed OCF Diagnosed 0.1 0.12 0.14 0.16 0.18 0.2 0.22 0.1 0.12 0.14 0.16 0.18 0.2 0.22 Time (sec) (a) (b) Figure 8: Simulated OCF diagnosis for (a) S and (b) S . X1 X4 will create the undesired harmonics, torque pulsations, and we will get the balanced output voltage by reducing the MI. �e vibrations in the load. �erefore, after the fault diagnosis the time taken to recon“gure the OCF depends on the faulty faulty switch must get isolated and a balanced three-phase switch, its switching state, sample time, and controller speed. supply must be ensured to the load. �ere are several fault recon“guration techniques reported in literature such as 7. Hardware Prototype Results and Discussion zero or neutral shifting, redundant cascaded H-bridge cell, and redundant switching states [22]. �e proposed fuzzy logic-based fault diagnosis technique is In this study, we have implemented a reduced modulation also validated through hardware prototype results. �e index (MI) strategy as postfault control. �is will bypass the single-phase 0.5 HP induction motor load is considered for faulty switch cell, and one cell forms the remaining two phases analysis. �e 20N10 IGBT is utilized as a switching device, and gives the balanced three-phase three-level output. �e MI and 1 kHz switching frequency pulses are given through the can be de“ned as equation (12), and Figure 10 shows the STM32F407VG controller and TLP 250 gate driver and principle of reduced MI, where A is the amplitude of the isolator. �e LV25P voltage sensor is used for the feedback of modulation wave, A is the modulation of the carrier wave, output phase voltage. �e OCF is created by removing the and n de“nes the number of levels for output. respective gate signal from the controller. �e complete hardware setup is shown in Figure 12. �e output of fuzzy MI  . (12) logic control is also shown to validate the fault diagnosis (n − 1)× A technique. �e fault diagnosis time in prototype results is also very similar to the simulation results. If we consider the OCF in switch S , then after fault X1 diagnosis the switching cell of the faulty switch and switching Figure 13(a) shows the results for output voltage with S X1 cell from the other two phases will get bypassed by making S fault. �e x-axis is being set at 10 ms/div, and the y-axis is X3 and S permanently turned on for the faulty cell. �erefore, being set at 1,000 V/div and 5 V/div for channels 1 and 2, X4 the updated switching strategy will be as shown in Table 3. respectively. �e fuzzy output is high after 26 ms of fault Figure 11 shows the simulated output phase voltage and creation with 1V magnitude, which shows the fault diagnosis current pro“le for postfault control, when the OCF occurs in time for switch S . In the simulation result, the fault di- X1 switch S . �e other switch OCF and its updated switching agnosis time was 25 ms, which is very close. �e T is X1 X strategy can be de“ned in the same manner. It is now clear that calculated from the output phase voltage fast Fourier Fuzzy Output ESavN Motor Line Current Phase A Voltage ree Phase Voltage International Transactions on Electrical Energy Systems 9 -200 -200 -10 +0.035 -0.085 OCF Created 6 OCF Created OCF Diagnosed OCF Diagnosed 0.1 0.12 0.14 0.16 0.18 0.2 0.22 0.1 0.12 0.14 0.16 0.18 0.2 0.22 Time (sec) (a) (b) Figure 9: Simulated OCF diagnosis for (a) S and (b) S . X6 X7 Before Fault Aer Fault +0.5 -0.5 -1 0 0.02 0.04 0.06 0.08 Time (sec) Figure 10: Fault recon“guration control. Table 3: Updated switching strategy after OCF in S . X1 S S S S S S S S X1 X2 X3 X4 X5 X6 X7 X8 +2E — — — — — — — — DC +E — — On On On Oœ On Oœ DC 0 — — On On Oœ Oœ On On −E — — On On Oœ On Oœ On DC −2E — — — — — — — — DC transform (FFT). �e mapping of T is shown for the healthy indicate the faulty condition. Figure 13(b) shows the output condition in Figure 14. �e 32% of THD is mapped as phase voltage for OCF in S , which is taking 30 ms for fault X4 320 mV. �e deviation of T more than the threshold will diagnosis. �e reason behind it is the fault will be diagnosed Fuzzy Output ESavN Motor Line Current Phase A Voltage ree Phase Voltage Magnitude 10 International Transactions on Electrical Energy Systems Output with reduced MI -100 -200 Output with reduced MI -5 -10 0 0.02 0.04 0.06 0.08 0.1 0.12 Time (Sec) Figure 11: Output voltage and current with fault recon“guration for OCF in S . X1 Clamper Circuit MOSFET Driver Supply Transformers STM32F407VG 5L CHB Inverter DC Supply Motor Load LV25P Figure 12: Complete hardware prototype setup. Fault Created Fault Created THD = 45.5 THD = 60.1 x-axis: 10ms/div x-axis: 10ms/div Fuzzy Output y-axis: 100V/div (ch-1) y-axis: 100V/div (ch-1) Fuzzy Output y-axis: 5V/div (ch-2) y-axis: 5V/div (ch-2) (a) (b) Figure 13: Continued. Phase current (A) Phase voltage (V) International Transactions on Electrical Energy Systems 11 Fault Created Fault Created THD = 45.3 THD = 61.3 x-axis: 10ms/div Fuzzy Output x-axis: 10ms/div y-axis: 100V/div (ch-1) y-axis: 100V/div (ch-1) y-axis: 5V/div (ch-2) Fuzzy Output y-axis: 5V/div (ch-2) (c) (d) Figure 13: Hardware prototype results for OCF in (a) S (b) S (c) S and (d) S . X1 X4 X6 X7 Mapped THD reshold Figure 14: Mapping of output voltage THD for healthy condition. Fault Detected Fault Post Fault Created Control Output Voltage Switching Pulse Fault Detection Signal x-axis: 10ms/div, y-axis: 100V/div (ch-1), y-axis: 5V/div (ch-2), y-axis: 1V/div (ch-3) Figure 15: Fault recon“guration for S OCF. X1 12 International Transactions on Electrical Energy Systems only when its switching state will arrive. 'e fault remains References undiagnosed till its switching state is not present, and it is [1] S. Kouro, M. Malinowski, K. Gopakumar et al., “Recent very clear that it is the cause for missing in o and n empty advances and industrial applications of multilevel converters,” states. Figures 13(c) and 13(d) show the fault diagnosis IEEE Transactions on Industrial Electronics, vol. 57, no. 8, results for OCF in S and S , respectively. 'e time taken X6 X7 pp. 2553–2580, 2010. for fault diagnosis for each OCF will vary from one fun- [2] H. Akagi, “Multilevel converters: fundamental circuits and damental cycle to two and half fundamental cycles systems,” Proceedings of the IEEE, vol. 105, no. 11, depending upon the empty state. pp. 2048–2065, 2017. Figure 15 shows the output voltage waveform, switching [3] J. Rodriguez, S. Bernet, P. K. Steimer, and I. E. Lizama, “A pulse, and fault-diagnosed signal for OCF in switch S with X1 survey on neutral-point-clamped inverters,” IEEE Transac- the postfault control technique. After the fault, the output tions on Industrial Electronics, vol. 57, no. 7, pp. 2219–2230, voltage will get unbalancedfor all three phases. If we reduce the [4] S. Dargahi, E. Babaei, S. Eskandari, V. Dargahi, and M. Sabahi, MI and change the switching strategy as mentioned in Table 3 “Flying-capacitor stacked multicell multilevel voltage source after fault diagnosis, then the five-level ML inverter will inverters: analysis and modelling,” IET Power Electronics, produce the balanced three-level output as shown in Figure 15. vol. 7, no. 12, pp. 2969–2987, 2014. From the above results, the OCF of a switch can be [5] J. Rodr´ıguez, S. Bernet, B. Wu, J. O. Pontt, and S. Kouro, diagnosed in one switching period to two fundamental “Multilevelvoltage-source-convertertopologiesfor.pdf,” IEEE cycles of 50Hz (5ms to 40ms) depending upon the faulty Transactions on Industrial Electronics, vol. 54, pp. 2930–2945, switch and switching state at the time of fault. However, the fault diagnosis time may depend upon the controller, sensor, [6] P. Mehta and M. Kumar, “Capacitor voltage balancing and switching frequency, and the sample time chosen. In the THD analysis in ANPC multilevel inverter,” in Proceedings of literature mentioned in Section 1, the fault diagnosis time the 2018 IEEMA Engineer Infinite Conference, eTechNxT, claimed by different techniques mentioned in Section 1 are pp. 1–5, New Delhi, India, March 2018. [7] G.-J. Su, “Multilevel DC-link inverter,” IEEE Transactions on as follows: AI-based techniques take about six fundamental Industry Applications, vol. 41, no. 3, pp. 848–854, 2005. cycles of 60Hz (up to 130ms), for asymmetric zero voltage [8] V. Smet, F. Forest, J.-J. Huselstein et al., “Ageing and failure switching, the diagnostic time is about two fundamental modes of IGBTmodules in high-temperature power cycling,” cycles of 60Hz (up to 32ms), and for harmonic switching IEEE Transactions on Industrial Electronics, vol. 58, no. 10, frequency analysis, the fault diagnostic time will vary from pp. 4931–4941, 2011. one switching frequency to one fundamental cycle. [9] Y. Shaoyong, D. Xiang, A. Bryant, P. Mawby, Li. Ran, and P. Tavner, “Condition monitoring for device reliability in power electronic converters: a review,” IEEE Transactions on 8. Conclusions Power Electronics, vol. 25, pp. 2734–2752, 2010. 'is study proposed a fault diagnosis technique for the five- [10] S. Ouni, J. R. Rodr´ıguez, M. Shahbazi et al., “A fast and simple method to detect short circuit fault in cascaded H-Bridge level ML inverter using fuzzy logic control. 'e proposed multilevel inverter,” in Proceedings of the IEEE International method is very easy to implement and cost-effective com- Conference on Industrial Technology (ICIT), pp. 866–871, pared to other methods in terms of sensor use and other Seville, Spain, March 2015. control requirements. Moreover, the computational efforts [11] M. Kumar, “Characterization and detection of open switch and conceptual complexity are also less compared to AI- faults for h-bridge inverter,” in Proceedings of the 9th IEEE based and frequency analysis-based fault diagnosis tech- International Conference on Power Electronics, Drives and niques. 'e fault diagnosis is achieved by output phase Energy Systems, PEDES 2020, Jaipur, India, December 2020. voltage fault symptom variables. 'ese variables are fuzzed [12] H.-W. Sim, J.-S. Lee, and K.-B. Lee, “Detecting open-switch in the fuzzy logic controller, and the fault detection is faults: using asymmetric zero-voltage switching states,” IEEE achieved using fuzzy rules. 'e simulation and prototype Industry Applications Magazine, vol. 22, no. 2, pp. 27–37, results support the algorithm. It is also applicable with [13] S. Khomfoi and L. M. Tolbert, “Fault diagnostic system for a different modulation indices and power factors. 'e post- multilevel inverter using a neural network,” IEEE Transactions fault control makes the system completely fault tolerant. As on Power Electronics, vol. 22, no. 3, pp. 1062–1069, 2007. the number of levels increases, the probable faulty switches [14] T. Wang, H. Xu, J. Han, E. Elbouchikhi, and M. E. H. Benbouzid, also increase; hence, the proposed fault diagnosis technique “Cascaded H-bridge multilevel inverter system fault diagnosis can only limit for five-level ML inverter OCF. using a PCA and multiclass relevance vector machine approach,” IEEE Transactions on Power Electronics, vol. 30, no. 12, Data Availability pp. 7006–7018, 2015. [15] S. Sedghi, A. Dastfan, and A. Ahmadyfard, “Fault detection No data were used to support this study. and reconfiguration of a modular multilevel inverter using histogram analysis and neural network,” International Review on Modelling and Simulations, vol. 4, pp. 2057–2065, 2011. [16] P. Lezana, R. Aguilera, and J. Rodriguez, “fault detection on Conflicts of Interest multicell converter based on output voltage frequency anal- 'e authors declare that they have no known conflicts of ysis,” IEEE Transactions on Industrial Electronics, vol. 56, interest. no. 6, pp. 2275–2283, 2009. International Transactions on Electrical Energy Systems 13 [17] P. Mehta, S. Sahoo, and M. Kumar, “A fault-diagnosis and tolerant control technique for five-level cascaded H-bridge inverters,” IET Circuits, Devices and Systems, vol.15, pp.1–11, [18] P. Mehta, M. Kumar, and S. Sahoo, “fault diagnosis in five- level CHB inverter using normalization factor and THD analysis,” in Proceedings of the IEEE India International Conference on Power Electronics, pp. 1–5, Jaipur, India, De- cember 2019. [19] H. Azeem, S. Yellasiri, V. Jammala, B. S. Naik, and A. K. Panda, “A fuzzy logic based switching methodology for a cascaded H-bridge multi-level inverter,” IEEE Transactions on Power Electronics, vol. 34, no. 10, pp. 9360–9364, 2019. [20] Z. Liu, Z. Zheng, and Y. Li, “Enhancing fault-tolerant ability of a nine-phase induction motor drive system using fuzzy logic current controllers,” IEEE Transactions on Energy Conversion, vol. 32, no. 2, pp. 759–769, 2017. [21] C. Cecati, F. Ciancetta, and P. Siano, “A multilevel inverter for photovoltaic systems with fuzzy logic control,” IEEE Trans- actions on Industrial Electronics, vol. 57, no. 12, pp. 4115– 4125, 2010. [22] P. Lezana, J. Pou, T. A. Meynard, J. Rodriguez, S. Ceballos, and F. Richardeau, “Survey on fault operation on multilevel inverters,” IEEE Transactions on Industrial Electronics, vol. 57, no. 7, pp. 2207–2218, 2010. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png International Transactions on Electrical Energy Systems Hindawi Publishing Corporation

Open Circuit Fault Diagnosis in Five-Level Cascaded H-Bridge Inverter

Loading next page...
 
/lp/hindawi-publishing-corporation/open-circuit-fault-diagnosis-in-five-level-cascaded-h-bridge-inverter-ENfZ09qnuY
Publisher
Hindawi Publishing Corporation
Copyright
Copyright © 2022 Pavan Mehta et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
eISSN
2050-7038
DOI
10.1155/2022/8588215
Publisher site
See Article on Publisher Site

Abstract

Hindawi International Transactions on Electrical Energy Systems Volume 2022, Article ID 8588215, 13 pages https://doi.org/10.1155/2022/8588215 Research Article Open Circuit Fault Diagnosis in Five-Level Cascaded H-Bridge Inverter 1 2 2 Pavan Mehta , Subhanarayan Sahoo, and Harsh Dhiman Gujarat Technological University, Ahmedabad, Gujarat, India Adani Institute of Infrastructure Engineering, Adalaj, Gujarat, India Correspondence should be addressed to Pavan Mehta; pvnmehta55@gmail.com Received 3 January 2022; Revised 23 March 2022; Accepted 25 March 2022; Published 20 April 2022 Academic Editor: Jesus Valdez-Resendiz Copyright © 2022 Pavan Mehta et al. 'is is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. 'e development of power electronic converter, especially multilevel converter, is remarkable for several decades. 'e complex switching and increased power of semiconductor devices are prime reasons for faults in multilevel inverters and have raised question about reliability. To improve the reliability, a cost-effective solution in terms of fault diagnosis is essential. In this context, this study proposed an open circuit fault (OCF) diagnosis technique for a switching device in a five-level cascaded H-bridge multilevel inverter using fuzzy logic control. 'e OCF features like output voltage total harmonic distortion (THD) and normalized average output voltage are fuzzed as input variables of the fuzzy logic controller. 'ese input variables are divided into various triangular antecedent membership function (MF). 'e output produced by the fuzzy controller as consequent MFs is divided into different levels to identify the faulty switch. In order to make a complete fault-tolerant structure, a reduced modulation index-based postfault control is suggested to get a balanced output voltage. 'e MATLAB/Simulink results and prototype results are the evidence to support the proposed fault diagnosis technique. point clamped (ANPC) [6] and DC link converter [7] are 1. Introduction deduced from basic structures to overcome their limitations. Over the last several decades, the application area of mul- In the ML inverter, the probabilities of failure of power tilevel (ML) inverters is grown very fast. 'e different ap- semiconductor switching devices are higher due to inter- plications of ML inverters are mine hoists, gas turbine action between the numbers of other switching devices. It starters, hydro-pumped storage, high-voltage DC (HVDC) reduces the reliability of the ML inverter system. In a broad, transmission, reactive power compensation, wind energy the causes of fault in a power semiconductor device can be conversion, power generation using PV cells, railway trac- categorized as exterior and interior. 'e exterior possibilities tion, AC drives, and marine propulsion [1]. Traditionally, for switch failure may be swelled or dipped from the external the two-level three-phase voltage source inverters (VSIs) supply connected to the terminals of the ML inverter, and were used to operate the drives. For the high-power ap- high dynamic changes at the load side and short circuit of the plications, the limitations with two-level VSI arise due to the load. On the other side, the interior possibilities for switch limited voltage withstand capacity of power semiconductor failures may be thermomechanical fatigue, gate misfiring, switching devices. 'erefore, multilevel (ML) VSI was and saturation of semiconductor materials [8]. Due to the adopted, which has many advantages such as low total abovementioned causes, the power switches may be either harmonic distortion (THD), low electromagnetic interfer- permanently opened or closed. Depending upon the cause of ence (EMI), reduced filter size, and low dv/dt [2] due to its failure, if the switching device is permanently open then it is staircase output waveform quality. 'e neutral point called an open circuit fault (OCF) and if the switch is clamped (NPC) [3], flying capacitor (FC) [4], and cascaded permanently closed then it is called a short circuit fault (SCF). 'e SCF in any switching device must be detected H-bridge (CHB) [5] are the basic structures of ML con- verters. 'e other popular structures such as active neutral within 10 µsec (depending on the semiconductor chip). 2 International Transactions on Electrical Energy Systems After the fault detection of SCF, the faulty switch must be where E is the output voltage of any one cell of the th cell X isolated, and a complete shutdown is mandatory [9]. On the phase, Sf and Sf are the switching functions (0 or 1) 1 or 5 2 or 6 other side, during the OCF the ML inverter peruses to of switch S or S and S or S of a cell, and E is the X1 X5 X2 X6 DC operate in faulty condition with reduced output quality. DC source of a cell. 'e total output phase voltage E is the OX However, this may lead to increased voltage stresses across summation of the individual cell output and can be given as the other healthy switches. Hence, the faulty switch location follows: must be identified. E � E + E . (2) OX cell1 cell2 'e different fault diagnosis techniques are reported in the literature such as time voltage criterion [10], switching 'eoutputvoltagedependsonitsoccupiedswitchingstates, time-domain OCF detection [11], asymmetric zero voltage i.e., 2p, p, o, n, and 2n. For an illustration, as shown in the switching [12], neural network and artificial intelligence (AI) waveformofFigure2,duringthehealthycondition,iftheoutput [13, 14], histogram [15], harmonic frequency analysis [16], voltagelevelsofanylegare+2E –+E –0–−E –−2E then DC DC DC DC and output voltage or current analysis [17, 18]. In general, allswitchingstatesareoccupiedandnoneoftheswitchingstate the fault diagnosis process is to perform different signal is empty. Figures 3(a) and 3(b) show the output phase voltage, systems or mathematical operations on the sensed output empty and occupied switching states, and current flow during quantity and to extract the unusual features of the sensed healthy and faulty conditions for switches S and S , re- X1 X6 output quantity like voltage, current, and power. 'e fault spectively, during their switching operations. diagnosis techniques are classified into three basic cate- As shown in Figure 3, if we apply Kirchhoff’s voltage law gories: waveform analysis, AI-based techniques, and har- (KVL) in the circuit, then the green line will be the actual monic frequency analysis as shown in Figure 1. path of current before the fault at the respective switching 'e abovementioned techniques require extra sensors, state of the switch. When there is the condition of OCF veryfastcontrollers,andmore computationaleffortsforfault occurs, then according to KVL the current path will change diagnosis.'erefore,thisstudyproposesacost-effectiveOCF and flow as per the red dotted line at respective switching diagnosistechniqueforaswitchoffive-levelCHBMLinverter states. For an illustration, during the healthy condition of S X1 using fuzzy logic control merged with waveform analysis. the current path is E -S -a-d-S -E -S -b-S -E , DC X1 X7 DC X5 X3 DC Also,thefaultdiagnosistimetakenbytheproposedtechnique which is the cause for generating +2E voltage level. While DC is very less is ensured. 'e different fuzzy logic control lit- during the faulty condition of S , if we apply KVL then the X1 erature studies are available for the following: switching of direction of current will be as follows: a-d-S -E -S -b- X7 DC X5 CHB ML inverter [19], nine-phase IM drive fault-tolerant S -D -a, which will be the cause for missing of +2E X3 X4 DC operation[20],andMLinverterwithphotovoltaiccell[21].In voltage level. 'erefore, the switching states p, o, n, and 2n this study, fuzzy control theory is used to fuzz the fault are occupied and 2p is considered as empty as mentioned in symptom variables and related fuzzy output levels generated Table 1. In this manner, we can find the current path and toidentifythefaultyswitch.Forimplementingthistechnique, output voltage during healthy and faulty conditions for all werequireonlyonevoltagesensorperphase,whichisalready switches. 'e expected current path, actual current path, available with main control of any closed-loop operation and expected output voltage level, and the missing output voltage no requirement of any extra hardware circuitry. It can be level at the time of OCF are summarized in Table 2. implemented in any existing system without making major 'e other switching state’s occupancy, emptiness, and changes in the control scheme. outputvoltagerelatedtothefaultyswitchinleg xwhere xϵ{A, In Section 2, the overview of CHB ML inverter for B, C} are shown in Table 1. From the analysis of Table 1, the healthy and faulty conditions and the principle of fault output voltage THD of the xth phase depends on the empty diagnosis technique are presented. In Section 3, the fuzzy states and related possible output voltage level. 'is is because logic control is highlighted for OCF diagnosis. 'e THD and thenonlinearityoftheoutputvoltagewillincreaseastheempty normalized average output phase voltage threshold esti- outputvoltagestateincreases.'eTHDindex(T )forOCFin mation is discussed in Section 4. 'e simulation results are S and S is comparatively low but higher than the healthy X1 X6 discussed in Section 5. 'e postfault control strategy is condition T . Similarly, the T of S and S are almost the X X X2 X5 suggested in Section 6. 'e hardware prototype of the five- same as they have only one empty state been n and p, re- level CHB ML inverter and its results are presented in spectively. 'e lower switches in each bridge of all the three Section 7. phasesareutilizedforzerooutputvoltageswitching.Hence,the OCFintheswitchesS ,S ,S ,andS givesthehighestT as X3 X4 X7 X8 X theycreatethetwoemptystates,namely,(o, p),(o,n),(o,p),and 2. CHB ML Inverter and OCF (o, n), respectively. In this contrast, THD of the output phase voltage can be a fault symptom variable. 'e identification of 'e generalized structure of the single-phase five-level CHB the different range of T is achieved by selecting appropriate ML inverter and healthy condition output phase voltage E X OX threshold values. 'e T of output phase voltage is given by is shown in Figure 2. 'e output voltage of any phase X equation (3). However, to get confidence in the correct fault depends on the individual cell output (E ). 'e output cell diagnosisandexactfaultlocationanotherthresholdisrequired voltage of a cell can be expressed as follows: for two-stage verification. 'erefore, the normalized average E � Sf − Sf 􏼁, (1) cell 1or5 2or6 output phase voltage (E ) is computed and compared with SavN International Transactions on Electrical Energy Systems 3 Fault detection in CHB ML inverter Waveform AI based Frequency Analysis Techniques Analysis • Time Voltage Criterion • Harmonic Frequency • Neural Network • Asymmetric Zero Voltage • AI & Neural Network Analysis Switching • Histogram & Neural • THD & Normalized Current Network Factor Figure 1: Classi“cation of fault diagnosis methods for CHB ML inverter. S S x1 x2 x1 x2 b E E cell1 DC 2p +2E DC S S x4 x3 D D x4 x3 p +E DC o 0 n -E S S DC x5 x6 x5 x6 2n -2E DC cell2 DC S S x8 x7 D D x8 x7 Figure 2: Structure of “ve-level CHB ML inverter and output phase voltage. their threshold values. To compute the value of E , the SavN E [k] E [k]+ j × E [k], (5) sav dav qav average values of all the phase voltages (E , E , and E ) A[k] B[k] C[k] arecalculatedasperequation(4),wherenisthetotalnumberof Where samples and k is sampling index. For better resolution, at least E [k] E × sin ω × T × k ,   A m s 50 samples should be collected from each fundamental cycle.  �e fundamental output phase voltages are expressed as in equation (6), where E is the maximum voltage and ω is the  2π E [k] E × sinω × T × k − , B m s (6) fundamental frequency. �e average park’s transformation on voltage vector is applied and represented as per equation (5).  During the normal operating conditions or healthy operation,   2π  E [k] E × sin ω × T × k + , the value of E is almost zero. When the fault occurs, then the C m s Xav value of E deviates and crosses the threshold limit. Xav Where h 2 max h2 OXh (3) 2 1 %THD T  × 100,   X  E [k] E [k]−  E [k]+ E [k],  dav Aav Bav Cav OX1  3 3 (7) 1 1 E [k]   E [k],Xε{} A, B, C , (4) Xav √ X  E [k] E [k]− E [k]. qav Bav Cav 3 3 k1 Output Voltage States 4 International Transactions on Electrical Energy Systems x1 x2 x2 x1 DC S S x4 x3 D D +E x4 x3 DC -E S DC x5 x6 x5 x6 2n -2E DC DC S S x8 x7 D D x8 x7 (a) x1 x2 x2 x1 DC +2E DC 2p S S x4 x3 +E DC D D x4 x3 -2E DC x5 x6 x5 2n x6 DC S S x8 x7 D D x8 x7 (b) Figure 3: Operation of “ve-level CHB ML inverter under OCF condition for (a) S and (b) S . X1 X2 However, this deviation needs to be a normalized value On the other side, when the fault occurs in any of the called E that is calculated and expressed as equation (8) switches, which is cause for 2n or n empty states, the E XN SavN and the average of normalized value E is calculated as per goes to a diœerent range in positive values. SavN equation (10). �e normalized average output phase voltage varies depending upon the faulty switch. For illustration, E [k] E  ,Xε{} A, B, C , (8) XN during the normal operating conditions the E is almost SavN E [k] zero and šoats between +0.1 and −0.1. When the fault occurs in any of the switches, which is the cause for 2p or p empty E [k] E [k]+ jE [k], (9) s sd sq states, the E goes to a diœerent range in negative values. SavN Output Voltage State Output Voltage State International Transactions on Electrical Energy Systems 5 Table 1: Occupied states, empty states, and membership functions for OCF in switch. Antecedent Condition of switch in leg x Consequent MFs MFs Occupied states Empty state Possible output voltage level S S S S S S S S T E S X1 X2 X3 X4 X5 X6 X7 X8 X savN Xi Ok Ok Ok Ok Ok Ok Ok Ok 2p, p, o, n, 2n null +2E , +E , 0, −E , −2E t �S e �ZR 0 DC DC DC DC 0 0 f Ok Ok Ok Ok Ok Ok Ok p, o, n, 2n 2p +E , 0, −E , −2E t �M e �NS I DC DC DC 1 2 Ok f Ok Ok Ok Ok Ok Ok 2p, p, o, 2n n +2E , 0, −E , −2E t �M e �PM II DC DC DC 1 3 Ok Ok f Ok Ok Ok Ok Ok 2p, n, 2n p, o +2E , −E , −2E t �L e �NL III DC DC DC 2 6 Ok Ok Ok f Ok Ok Ok Ok 2p, p, 2n n, o +2E , +E , −2E t �L e �PL IV DC DC DC 2 5 Ok Ok Ok Ok f Ok Ok Ok 2p, o, n, 2n p +2E , 0, −E , −2E t �M e �NM V DC DC DC 1 4 Ok Ok Ok Ok Ok f Ok Ok 2p, p, o, n 2n +2E , +E , 0, −E t �M e �PS VI DC DC DC 1 1 Ok Ok Ok Ok Ok Ok f Ok 2p, n, 2n p, o +2E , −E , −2E t �L e �NL VII DC DC DC 2 6 Ok Ok Ok Ok Ok Ok Ok f 2p, p, 2n n, o +2E , +E , −2E t �L e �PL VIII DC DC DC 2 5 Table 2: Current path for healthy and faulty condition for all switches. OCF switch Current path for healthy condition Current path for faulty condition Missing output voltage level S E -S -a-d-S -E -S -c-b-S -E a-d-S -E -S -c-b-S -D -E +2E X1 DC X1 X7 DC X5 X3 DC X7 DC X5 X3 X4 DC DC S E -S -b-c-S -D -d-b-S -E b-c-S -D -d-b-S -D -a -E X2 DC X2 X8 X7 X3 DC X8 X7 X3 X4 DC S E -S -a-d-S -D -c-b-S a-d-S -D -c-b-D -S +E X3 DC X1 X7 X8 X3 X7 X8 X2 X1 DC S E -S -b-c-S -D -d-b-S b-c-S -D -d-a-D -S -E X4 DC X2 X8 X7 X4 X8 X7 X1 X2 DC S E -S -c-b-S -D -a-d-S c-b-S -D -a-d-S -D +E X5 DC X5 X3 X4 X7 X3 X4 X7 X8 DC S E -S -b-c-S -E -S -d-a-S -E E -S -b-c-S -D -d-a-S -E -2E X6 DC X2 X8 DC X6 X4 DC DC X2 X8 X7 X4 DC DC S c-b-S -D -a-d-D -S E -S -c-b-S -D -a-d-S +E X7 X3 X4 X6 X5 DC X5 X3 X4 X7 DC S E -S -d-a-S -D -b-c-S d-a-S -D -b-c-D S -E X8 DC X6 X4 X3 X8 X4 X3 X5- X8 DC 'e (δ –δ ) and (λ –λ ) are different thresholds for T 1 3 1 6 X E [k] sav E [k] � . (10) savN and E for different empty states. 'e antecedent SavN E [k] membership functions are designed based on the input variables. 'e distribution of T is performed as t , t , and t X 0 1 2 present small T (S), medium T (M), and large T (L), X X X 3. OCF Diagnosis Using Fuzzy Logic Control respectively. Similarly, the distribution of E is performed savN as e , e , e , e , e , e , and e present different normalized From the above discussion, the T and E are the fault 0 1 2 3 4 5 6 X SavN output voltages, namely, zero (ZR), positive small (PS), symptom variables and combinedly used to accurately di- positive medium (PM), positive large (PL), negative small agnose the faulty switch. 'e detailed algorithms of the (NS), negative medium (NM), and negative large (NL), proposed method are as follows. respectively. 'e consequent MFs are designed in such a way to represent the faulty switch identification number from I 3.1. Fuzzy Logic Fault Diagnosis Reasoning. 'e diagnosis (S ) to VIII (S ), and 0 represents the healthy condition. X1 X8 architecture is created based on the analytical and heuristic Figures 4(a) and 4(b) show the graphical representation of knowledge of symptoms of the CHB ML inverter. Heuristic the antecedent MFs and consequent MFs, respectively. 'e knowledge in the form of qualitative process models can be output of fuzzy logic will be designated as 0 or 1 for healthy expressed as if-then rules. 'e nine input MFs and variables and faulty conditions, respectively, for each switch. T and E are fuzzified as equation (11). X SavN t , 0≤ T ≤ δ , ⎧ ⎪ 0 X 1 3.2. Extraction of Fuzzy Rules. Based on the relationships in Table 1, the fuzzy rules are extracted. In general, the fuzzy T � t , δ ≤ T ≤ δ , X 1 1 X 2 control system consists of fuzzification, fuzzy inference, and t , δ ≤ T ≤ δ , 2 2 X 3 defuzzification illustrated in Figure 5. e , E < λ , λ , ⎧ ⎪ ⎪ 0 SavN 2 1 (i) Fuzzification: Fuzzification is the process of defining e , λ < E < λ , 1 1 SavN 3 the fuzzy variables from input variables using MFs. (11) ⎪ As explained in Section 3.1, the T and E are the ⎪ X SavN e , λ < E < λ , ⎪ 2 2 SavN 4 input variables. 'e (S, M, and L) and (ZR, PS, PM, E � e , λ < E < λ , SavN 3 3 SavN 5 ⎪ PL, NS, NM, and NL) are different MFs of T and ⎪ E , respectively. e , λ < E < λ , ⎪ SavN 4 4 SavN 6 ⎪ (ii) Fuzzy interface: It expresses the relation between ⎪ e , λ < E , ⎪ 5 5 SavN ⎪ input fuzzy variables and output using if-then rules. e , λ < E . 6 6 SavN 'e various fuzzy rules can be expressed in Table 1. 6 International Transactions on Electrical Energy Systems δ S i λ Xi 0 Ι ΙΙ ΙΙΙ ΙV V VΙ VΙ VΙΙΙ t t t e e e e e e 0 1 2 6 4 2 1 3 5 T E X SavN (a) (b) Figure 4: (a) Antecedent MF distribution. (b) Consequent MF distribution. Fuzzy Membership Input Variable Defuzzification Interface Functions of Input Membership Output Functions of Membership Output Input Variable Functions of Input Fuzzy Rules Figure 5: Fuzzy interfacing process. �ere will be total of nine rules associated per phase conditions and modulation index. On the other side, the (x) in the fuzzy interface. For illustration, if T  M normalized average phase output voltage has diœerent levels and E  NS then output S  I represents S of value in positive and negative regions depending upon the SavN Xi X1 OCF. If T  M and E  PM then output S  II faulty switch. Once the faulty phase is identi“ed using T , the X SavN Xi X represents S OCF. If T  M and E  PS, then E will be calculated. �e actual value is compared with the X2 X SavN SavN output S  VI represents S OCF. �e fault fea- thresholds, and the faulty switch is declared. Xi X6 tures created by switches S –S and S –S are the Figures 6(a) and 6(b) show the T and E variation X3 X7 X4 X8 X SavN same. Hence, identi“cation of the faulty switch is with respect to modulation index and their thresholds for achieved on a priority basis. If the OCF occurred at various fault conditions. �e (δ , δ , and δ ) are the 1 2 3 , then S will be initially appeared as a faulty thresholds selected for THD for no-fault condition, fault in X7 X3 switch as per priority. However, in the next funda- S /S , and fault in S /S , respectively. �e values beyond X1 X6 X2 X5 mental cycle the correct faulty switch will be diag- δ represent the fault in S /S /S /S . �e selection of the 3 X3 X4 X7 X8 nosed, i.e., S . On theother side, ifthe OCFoccurred THD threshold gives a fair fault signature up to a 0.4 X7 at S , then S will be initially appeared as a faulty modulation index. In practical conditions, it is not an ad- X8 X4 switch as per priority and in the next fundamental visable modulation index below 0.5 because the output cycle the correct faulty switch will be diagnosed, i.e., voltage will be halved. �e value of E šoats between SavN S . �e time taken for fault diagnosis will be higher +0.01 and −0.01 in healthy condition. When the fault occurs X8 as much as one fundamental cycle for S and S in the switch, which is cause for the 2p or p empty states, the X7 X8. (iii) Defuzzi“cation: �e defuzzi“cation can be de“ned value of E goes to negative. While the OCF is in the SavN as the procedure of converting the fuzzy output set switch, which causes 2n or n empty states, the value of E SavN into the crisp set. �e max-min composition and goes to positive. �e (λ , λ , λ , and λ ) are the thresholds 1 2 5 6 centroid of area method are selected for defuzzi“- selected for E for faults in S ,S ,S , and S re- SavN X1 X6 X5 X2 cation in the proposed method. �e output of spectively. �e thresholds beyond λ and λ represent the 3 4 consequent MFs as shown in Figure 4(b) will decide faults in S /S and S /S , respectively. X3 X7 X4 X8 the output of the fuzzy logic controller and so faulty switch. �e diœerent levels of output are selected 5. Simulation Results and Discussion from 0 to VIII for switch faults S to S , which can X1 X8 directly rešect the number of respective switch fault �e proposed fault diagnosis method is tested using as shown in the simulation results section, where MATLAB/Simulink. For testing the robustness of the fault zero indicates the healthy condition. diagnosis technique, the simulation is performed with a 0.8 power factor with a three-phase IM drive. �e level-shifted pulse width modulation technique (LS-PWM) is used for 4. Estimation of Thresholds switching. �e threshold estimation process plays a key role in the se- Figure 7 shows the THD plot for all three phases for lection of fuzzy rules. Hence, the acceptable threshold au- healthy and S faulty conditions. When there is a healthy A1 thentication is required to avoid false fault diagnosis. We condition, the THD of all three phases are the same and so know the fact that THD may change with the diœerent loading T will not cross the threshold limit. As soon as the OCF X International Transactions on Electrical Energy Systems 7 +0.07 +0.05 +0.01 -0.01 -0.05 δ λ 3 3 -0.07 0 0.5 1 0 0.5 1 Modulation Index Modulation Index E at S /S Faults E at S Faults SavN X4 X8 SavN X1 %THD at S /S /S /S Faults X3 X4 X7 X8 E at S Faults E at S Faults SavN X2 SavN X5 %THD at S /S Faults X2 X5 E at S Faults E at S /S Faults SavN X6 SavN X3 X7 %THD at S /S Faults X1 X6 %THD at No Faults (a) (b) Figure 6: (a) THD threshold estimation. (b) E threshold estimation. SavN Phase-A Phase-B Phase-C Healthy Condition THD=30.01% THD=30.10% THD=29.99% S OCF Condition THD=42.50% THD=30.00% THD=30.05% A1 0 1 2 3 4 5 01 2 3 4 5 0 1 2 3 4 5 Frequency (kHz) Figure 7: THD plot for healthy and S faulty conditions. A1 occurs in switch S , the T will cross the threshold limit δ both will have the same fault features. �e value of E is A1 A 1 SavN as discussed in Section 4. �e THD of the remaining two +0.05, and fuzzy output settles at 4 V. �e fault diagnosis of phases will be the same. �erefore, the faulty phase is now S is achieved by approximately 0.038 sec. X4 identi“ed and E will be calculated. �e value of T for Figure 9(a) shows the OCF results for switch S . �e SavN X X6 diœerent switch faults will be diœerent depending upon the value of E is +0.035, and fuzzy output represents the 6 V. SavN absence of output voltage level as discussed in Section 2. Figure 9(b) shows the results for OCF in switch S . It can be X7 By analyzing Figure 8(a), it can be noted that the fault is observed that the S and S provide the same empty states, X3 X7 created at 0.14 sec in switch S , yet the eœect of fault in the i.e., o and p. As per priority, S will be diagnosed as the X1 X3 output phase voltage waveform and output current cannot faulty switch and S remains undiagnosed but in the next X7 be seen. �e fault diagnosis algorithm will only be active fundamental cycle S is detected as the faulty switch. �e X7 when the faulty switch’s switching state arrives. fault is detected in two and half fundamental cycles or Hence, till the switching state of the respective faulty 0.045 sec. switch, the OCF remains undiagnosed. �e value of E at SavN the time of fault is −0.035 as S is cause for missing in 2p X1 6. Postfault Control empty state. �e fuzzy logic output is considered as a fault diagnosis signal, and its magnitude is 1 V as per the fuzzy In Section 2, we have seen that the CHB ML inverter will algorithm. �e OCF for switch S is diagnosed within pursue to operate with reduced output quality under the X1 0.025 sec. Figure 8(b) shows the simulation results for OCF OCF in a switch, which prevents the complete shutdown of in switch S . As per discussion in Table 1, the fault in S load. However, this will create unbalancing in the three- X4 X4 and S both will cause for n and o empty states. �erefore, phase output voltage. �e unbalanced three-phase output X8 %THD (Output phase voltage) Mag (% of fundamental) SavN 8 International Transactions on Electrical Energy Systems -200 -200 -10 +0.05 -0.035 OCF OCF Created OCF Created Diagnosed OCF Diagnosed 0.1 0.12 0.14 0.16 0.18 0.2 0.22 0.1 0.12 0.14 0.16 0.18 0.2 0.22 Time (sec) (a) (b) Figure 8: Simulated OCF diagnosis for (a) S and (b) S . X1 X4 will create the undesired harmonics, torque pulsations, and we will get the balanced output voltage by reducing the MI. �e vibrations in the load. �erefore, after the fault diagnosis the time taken to recon“gure the OCF depends on the faulty faulty switch must get isolated and a balanced three-phase switch, its switching state, sample time, and controller speed. supply must be ensured to the load. �ere are several fault recon“guration techniques reported in literature such as 7. Hardware Prototype Results and Discussion zero or neutral shifting, redundant cascaded H-bridge cell, and redundant switching states [22]. �e proposed fuzzy logic-based fault diagnosis technique is In this study, we have implemented a reduced modulation also validated through hardware prototype results. �e index (MI) strategy as postfault control. �is will bypass the single-phase 0.5 HP induction motor load is considered for faulty switch cell, and one cell forms the remaining two phases analysis. �e 20N10 IGBT is utilized as a switching device, and gives the balanced three-phase three-level output. �e MI and 1 kHz switching frequency pulses are given through the can be de“ned as equation (12), and Figure 10 shows the STM32F407VG controller and TLP 250 gate driver and principle of reduced MI, where A is the amplitude of the isolator. �e LV25P voltage sensor is used for the feedback of modulation wave, A is the modulation of the carrier wave, output phase voltage. �e OCF is created by removing the and n de“nes the number of levels for output. respective gate signal from the controller. �e complete hardware setup is shown in Figure 12. �e output of fuzzy MI  . (12) logic control is also shown to validate the fault diagnosis (n − 1)× A technique. �e fault diagnosis time in prototype results is also very similar to the simulation results. If we consider the OCF in switch S , then after fault X1 diagnosis the switching cell of the faulty switch and switching Figure 13(a) shows the results for output voltage with S X1 cell from the other two phases will get bypassed by making S fault. �e x-axis is being set at 10 ms/div, and the y-axis is X3 and S permanently turned on for the faulty cell. �erefore, being set at 1,000 V/div and 5 V/div for channels 1 and 2, X4 the updated switching strategy will be as shown in Table 3. respectively. �e fuzzy output is high after 26 ms of fault Figure 11 shows the simulated output phase voltage and creation with 1V magnitude, which shows the fault diagnosis current pro“le for postfault control, when the OCF occurs in time for switch S . In the simulation result, the fault di- X1 switch S . �e other switch OCF and its updated switching agnosis time was 25 ms, which is very close. �e T is X1 X strategy can be de“ned in the same manner. It is now clear that calculated from the output phase voltage fast Fourier Fuzzy Output ESavN Motor Line Current Phase A Voltage ree Phase Voltage International Transactions on Electrical Energy Systems 9 -200 -200 -10 +0.035 -0.085 OCF Created 6 OCF Created OCF Diagnosed OCF Diagnosed 0.1 0.12 0.14 0.16 0.18 0.2 0.22 0.1 0.12 0.14 0.16 0.18 0.2 0.22 Time (sec) (a) (b) Figure 9: Simulated OCF diagnosis for (a) S and (b) S . X6 X7 Before Fault Aer Fault +0.5 -0.5 -1 0 0.02 0.04 0.06 0.08 Time (sec) Figure 10: Fault recon“guration control. Table 3: Updated switching strategy after OCF in S . X1 S S S S S S S S X1 X2 X3 X4 X5 X6 X7 X8 +2E — — — — — — — — DC +E — — On On On Oœ On Oœ DC 0 — — On On Oœ Oœ On On −E — — On On Oœ On Oœ On DC −2E — — — — — — — — DC transform (FFT). �e mapping of T is shown for the healthy indicate the faulty condition. Figure 13(b) shows the output condition in Figure 14. �e 32% of THD is mapped as phase voltage for OCF in S , which is taking 30 ms for fault X4 320 mV. �e deviation of T more than the threshold will diagnosis. �e reason behind it is the fault will be diagnosed Fuzzy Output ESavN Motor Line Current Phase A Voltage ree Phase Voltage Magnitude 10 International Transactions on Electrical Energy Systems Output with reduced MI -100 -200 Output with reduced MI -5 -10 0 0.02 0.04 0.06 0.08 0.1 0.12 Time (Sec) Figure 11: Output voltage and current with fault recon“guration for OCF in S . X1 Clamper Circuit MOSFET Driver Supply Transformers STM32F407VG 5L CHB Inverter DC Supply Motor Load LV25P Figure 12: Complete hardware prototype setup. Fault Created Fault Created THD = 45.5 THD = 60.1 x-axis: 10ms/div x-axis: 10ms/div Fuzzy Output y-axis: 100V/div (ch-1) y-axis: 100V/div (ch-1) Fuzzy Output y-axis: 5V/div (ch-2) y-axis: 5V/div (ch-2) (a) (b) Figure 13: Continued. Phase current (A) Phase voltage (V) International Transactions on Electrical Energy Systems 11 Fault Created Fault Created THD = 45.3 THD = 61.3 x-axis: 10ms/div Fuzzy Output x-axis: 10ms/div y-axis: 100V/div (ch-1) y-axis: 100V/div (ch-1) y-axis: 5V/div (ch-2) Fuzzy Output y-axis: 5V/div (ch-2) (c) (d) Figure 13: Hardware prototype results for OCF in (a) S (b) S (c) S and (d) S . X1 X4 X6 X7 Mapped THD reshold Figure 14: Mapping of output voltage THD for healthy condition. Fault Detected Fault Post Fault Created Control Output Voltage Switching Pulse Fault Detection Signal x-axis: 10ms/div, y-axis: 100V/div (ch-1), y-axis: 5V/div (ch-2), y-axis: 1V/div (ch-3) Figure 15: Fault recon“guration for S OCF. X1 12 International Transactions on Electrical Energy Systems only when its switching state will arrive. 'e fault remains References undiagnosed till its switching state is not present, and it is [1] S. Kouro, M. Malinowski, K. Gopakumar et al., “Recent very clear that it is the cause for missing in o and n empty advances and industrial applications of multilevel converters,” states. Figures 13(c) and 13(d) show the fault diagnosis IEEE Transactions on Industrial Electronics, vol. 57, no. 8, results for OCF in S and S , respectively. 'e time taken X6 X7 pp. 2553–2580, 2010. for fault diagnosis for each OCF will vary from one fun- [2] H. Akagi, “Multilevel converters: fundamental circuits and damental cycle to two and half fundamental cycles systems,” Proceedings of the IEEE, vol. 105, no. 11, depending upon the empty state. pp. 2048–2065, 2017. Figure 15 shows the output voltage waveform, switching [3] J. Rodriguez, S. Bernet, P. K. Steimer, and I. E. Lizama, “A pulse, and fault-diagnosed signal for OCF in switch S with X1 survey on neutral-point-clamped inverters,” IEEE Transac- the postfault control technique. After the fault, the output tions on Industrial Electronics, vol. 57, no. 7, pp. 2219–2230, voltage will get unbalancedfor all three phases. If we reduce the [4] S. Dargahi, E. Babaei, S. Eskandari, V. Dargahi, and M. Sabahi, MI and change the switching strategy as mentioned in Table 3 “Flying-capacitor stacked multicell multilevel voltage source after fault diagnosis, then the five-level ML inverter will inverters: analysis and modelling,” IET Power Electronics, produce the balanced three-level output as shown in Figure 15. vol. 7, no. 12, pp. 2969–2987, 2014. From the above results, the OCF of a switch can be [5] J. Rodr´ıguez, S. Bernet, B. Wu, J. O. Pontt, and S. Kouro, diagnosed in one switching period to two fundamental “Multilevelvoltage-source-convertertopologiesfor.pdf,” IEEE cycles of 50Hz (5ms to 40ms) depending upon the faulty Transactions on Industrial Electronics, vol. 54, pp. 2930–2945, switch and switching state at the time of fault. However, the fault diagnosis time may depend upon the controller, sensor, [6] P. Mehta and M. Kumar, “Capacitor voltage balancing and switching frequency, and the sample time chosen. In the THD analysis in ANPC multilevel inverter,” in Proceedings of literature mentioned in Section 1, the fault diagnosis time the 2018 IEEMA Engineer Infinite Conference, eTechNxT, claimed by different techniques mentioned in Section 1 are pp. 1–5, New Delhi, India, March 2018. [7] G.-J. Su, “Multilevel DC-link inverter,” IEEE Transactions on as follows: AI-based techniques take about six fundamental Industry Applications, vol. 41, no. 3, pp. 848–854, 2005. cycles of 60Hz (up to 130ms), for asymmetric zero voltage [8] V. Smet, F. Forest, J.-J. Huselstein et al., “Ageing and failure switching, the diagnostic time is about two fundamental modes of IGBTmodules in high-temperature power cycling,” cycles of 60Hz (up to 32ms), and for harmonic switching IEEE Transactions on Industrial Electronics, vol. 58, no. 10, frequency analysis, the fault diagnostic time will vary from pp. 4931–4941, 2011. one switching frequency to one fundamental cycle. [9] Y. Shaoyong, D. Xiang, A. Bryant, P. Mawby, Li. Ran, and P. Tavner, “Condition monitoring for device reliability in power electronic converters: a review,” IEEE Transactions on 8. Conclusions Power Electronics, vol. 25, pp. 2734–2752, 2010. 'is study proposed a fault diagnosis technique for the five- [10] S. Ouni, J. R. Rodr´ıguez, M. Shahbazi et al., “A fast and simple method to detect short circuit fault in cascaded H-Bridge level ML inverter using fuzzy logic control. 'e proposed multilevel inverter,” in Proceedings of the IEEE International method is very easy to implement and cost-effective com- Conference on Industrial Technology (ICIT), pp. 866–871, pared to other methods in terms of sensor use and other Seville, Spain, March 2015. control requirements. Moreover, the computational efforts [11] M. Kumar, “Characterization and detection of open switch and conceptual complexity are also less compared to AI- faults for h-bridge inverter,” in Proceedings of the 9th IEEE based and frequency analysis-based fault diagnosis tech- International Conference on Power Electronics, Drives and niques. 'e fault diagnosis is achieved by output phase Energy Systems, PEDES 2020, Jaipur, India, December 2020. voltage fault symptom variables. 'ese variables are fuzzed [12] H.-W. Sim, J.-S. Lee, and K.-B. Lee, “Detecting open-switch in the fuzzy logic controller, and the fault detection is faults: using asymmetric zero-voltage switching states,” IEEE achieved using fuzzy rules. 'e simulation and prototype Industry Applications Magazine, vol. 22, no. 2, pp. 27–37, results support the algorithm. It is also applicable with [13] S. Khomfoi and L. M. Tolbert, “Fault diagnostic system for a different modulation indices and power factors. 'e post- multilevel inverter using a neural network,” IEEE Transactions fault control makes the system completely fault tolerant. As on Power Electronics, vol. 22, no. 3, pp. 1062–1069, 2007. the number of levels increases, the probable faulty switches [14] T. Wang, H. Xu, J. Han, E. Elbouchikhi, and M. E. H. Benbouzid, also increase; hence, the proposed fault diagnosis technique “Cascaded H-bridge multilevel inverter system fault diagnosis can only limit for five-level ML inverter OCF. using a PCA and multiclass relevance vector machine approach,” IEEE Transactions on Power Electronics, vol. 30, no. 12, Data Availability pp. 7006–7018, 2015. [15] S. Sedghi, A. Dastfan, and A. Ahmadyfard, “Fault detection No data were used to support this study. and reconfiguration of a modular multilevel inverter using histogram analysis and neural network,” International Review on Modelling and Simulations, vol. 4, pp. 2057–2065, 2011. [16] P. Lezana, R. Aguilera, and J. Rodriguez, “fault detection on Conflicts of Interest multicell converter based on output voltage frequency anal- 'e authors declare that they have no known conflicts of ysis,” IEEE Transactions on Industrial Electronics, vol. 56, interest. no. 6, pp. 2275–2283, 2009. International Transactions on Electrical Energy Systems 13 [17] P. Mehta, S. Sahoo, and M. Kumar, “A fault-diagnosis and tolerant control technique for five-level cascaded H-bridge inverters,” IET Circuits, Devices and Systems, vol.15, pp.1–11, [18] P. Mehta, M. Kumar, and S. Sahoo, “fault diagnosis in five- level CHB inverter using normalization factor and THD analysis,” in Proceedings of the IEEE India International Conference on Power Electronics, pp. 1–5, Jaipur, India, De- cember 2019. [19] H. Azeem, S. Yellasiri, V. Jammala, B. S. Naik, and A. K. Panda, “A fuzzy logic based switching methodology for a cascaded H-bridge multi-level inverter,” IEEE Transactions on Power Electronics, vol. 34, no. 10, pp. 9360–9364, 2019. [20] Z. Liu, Z. Zheng, and Y. Li, “Enhancing fault-tolerant ability of a nine-phase induction motor drive system using fuzzy logic current controllers,” IEEE Transactions on Energy Conversion, vol. 32, no. 2, pp. 759–769, 2017. [21] C. Cecati, F. Ciancetta, and P. Siano, “A multilevel inverter for photovoltaic systems with fuzzy logic control,” IEEE Trans- actions on Industrial Electronics, vol. 57, no. 12, pp. 4115– 4125, 2010. [22] P. Lezana, J. Pou, T. A. Meynard, J. Rodriguez, S. Ceballos, and F. Richardeau, “Survey on fault operation on multilevel inverters,” IEEE Transactions on Industrial Electronics, vol. 57, no. 7, pp. 2207–2218, 2010.

Journal

International Transactions on Electrical Energy SystemsHindawi Publishing Corporation

Published: Apr 20, 2022

References