OPEN ACCESS
The proposed work mainly deals with transformerbased threephase MultiLevel Inverter (MLI) where the switches in MLI are controlled by MultiCarrier PWM (MCPWM) techniques. In this article, 48V DC input voltage is fed to 48/400V HighVoltage DCDC boost converter which is controlled by Proportional–Integral (PI) and NeuralNetwork (NN) tuned PI controllers. 400V DC output from boost converter is connected to threephase MLI where the MLI switches are controlled by three different types of MCPWM techniques like Phase Disposition (PD), Phase Opposition and Disposition (POD) and Alternate Phase Opposition and Disposition (APOD). The overall system performance in terms of Total Harmonic Distortion (%THD) is analyzed and it is depicted that the proposed system under different modulation indices gives improved THD results with an NN tuned PI controllerbased boost converter. The whole analysis is performed on the MATLAB\Simulink platform.
multi carrier PWM, multilevel inverter, PD, POD, APOD, THD
In the present renewable energy world, usage and implementation of power electronic devices are high. In India, the utilization of renewable energy resources like solar, wind, tidal, etc., is increased drastically as compared to earlier years [1]. If AC load is to be fed from solar PV power, a boost converter is required to boost up the DC power from solar which is transformed to AC load by means of inverters or MultiLevel Inverters (MILs) [2]. In this article, a highvoltage DCDC boost converter was taken to boost up the 48V input voltage to 400V output voltage. Boost converter output voltage is given to the inverter to run ac loads. The inverter's main role is to change DC to AC power with the desired voltage magnitude and frequency [3]. MLIs are most suitable for standalone PV and gridconnected applications. The fundamental MLIs topologies are Cascaded HBridge, Flyingcapacitors, and Diodeclamped [4]. Out of these topologies, cascaded HBridge is more preferable in gridconnected applications. Earlier, researchers have proposed different MLI topologies for standalone PV and gridconnected applications. However, all those topologies have their own merits and demerits. MLI topologies are mainly of two types (i) Direct and (ii) Indirect, where in the direct type, generation system is directly connected to the load, but in the indirect type, generation system is connected to load through the DClink capacitors [5]. Direct type MLIs have poor efficiency hence indirect type topological MLIs are more preferable than direct type MLI [6]. Based on the input voltages, MLIs are mainly classified into two types in which one is Symmetrical and another is Asymmetrical [7]. In symmetrical MLIs, all DClinks (input voltage) is maintained at equal voltage level whereas in asymmetrical case DClinks do not have equal voltage levels [8]. Asymmetrical type MLIs result in high cost and more switching losses, hence it is preferable to consider symmetrical MLIs [9]. This article presented a threephase multilevel inverter with MCPWM techniques. The MLI topology comprises four diodes, five switches, and one threephase transformer to generate fivelevel output voltage.
Recently, many authors have reported their findings related to MLIs with SDC source instead of multiple sources for single and threephase applications [10]. A threephase inverter has been investigated using an SDC input with singlephase transformers for gridconnected applications. Three phase MLIs using 3phase transformers have been considered by the SDC source. This topology is suitable for gridconnected PV applications. All these topologies are used to enhance the output voltage level without increasing the number of levels, but, other hand utilization of transformers is also increased. From the discussion, it is found that all the inverter topologies are highly adaptable for standalone photovoltaic (PV) and grid connected applications [11]. A threephase topology has been designed with a reduced number of active switching devices, but it requires additional power diodes and DC bus capacitor. Recently, the inverter was designed for PV applications using the isolation transformer with three levels of the threephase inverter [12]. A three phase 4wire PV interface gridconnected inverter was investigated. That generated a 3level output voltage and required additional filtering circuits [13]. For controlling an inverter, a different multicarrier PWM technique has been applied [14].
For controlling the MLI different MultiCarrier PWM (MCPWM) techniques have been proposed, where for generating a fivelevel output voltage, M1 carriers are considered. In the MCPWM technique, M1 number of carrier signals are compared with a reference signal. Based on the positioning of the carrier signals the PWM techniques are again classified as Alternate Phase Opposition Disposition (APOD), Phase Opposition Disposition (POD) and Phase Disposition (PD) [11]. In this, a highvoltage DCDC boost converter was used in front of MLI, for controlling the boost converter Proportional Integral (PI) and Neural Network (NN) tuned PI controllers are considered.
Figure 1. Advantages of fivelevel inverter
The main objectives of the work are PI and NN tuned PI controllers are used to control the duty cycle of the boost converter similarly, for MLI controlling MCPWM techniques are proposed [15]. Then the overall performance of the system is observed in terms of Total Harmonic Distortion (%THD) analysis with the modulation index varying from 0.4 to 1. MLI has so many advantages that are clearly explained in Figure 1.
Isolation Sepic and fly back converter have come under Coupled Inductor (CI) DCDC boost converters. The CI based converters produce high output voltage as compared to conventional boost converters. Regarding that, a CIbased Highvoltage DCDC converter was used for the threephase MLI. The highvoltage dc/dc converter contains input dc voltage Vin, coupled inductor (CI), main switch S, one output capacitor C0, one output diode D0, one clamp capacitor C1, clamp diode D1, two diodes D2 and D3 and two capacitors C2 and C3. The CI contains leakage inductor (LI) Lk, magnetizing inductor (MI) Lm and an ideal transformer. Figure 2 represents the highvoltage DCDC converter. PI and NN tuned PI controllers are applied to the boost converter for controlled the duty cycle. In this, a 48V DC voltage is given to the input of the boost converter and it produces 400V DC voltage at the output side [13]. The operating stages of HighVoltage DCDC converter is explained in Figure 3.
Figure 2. Highvoltage DCDC converter
Figure 3. Stages of operation of the HighVoltage DCDC converter
In stage one diodes D0 and D1 both are in off state remaining diodes and switch are in on state. In this case Vin supplies the energy to the Lk and Lm, due to Lk secondary side of the CI current is decreases linearly. So, load (R) receives the energy from Co only. In stage two diode Do and switch both are in on state remaining are in off state. In this case Vin supplies the energy to the Lk, Lm and secondary side of the CI. The capacitors C1, C2 and C3 are in series the voltage is transfer to both Co and Load (R). In stage three diode Do and switch both are in on state remaining are in off state. In this case Vin supplies the energy to the parasitic capacitor (Cds) of the main switch through the Lk, Lm. The capacitors C1, C2 and C3 are still in series the voltage is transform to both Co as well as Load (R). In stage four diodes D2 and 3 are in on state remaining diodes and switch are in off state. Input voltage source (Vin) supplies energy to the C1 through the Lk and Lm. Load (R) and capacitor Co gets energy from the secondary side of the CI. In stage five diode Do and switch both are in off state remaining diodes are in on state. The energy in the Lk and Lm are transfer to the C1 similarly, C3 and C3 are charged by Lk and Lm in the secondary side of the CI. Load (R) gets energy from Co only.
2.1 Modelling of the highvoltage DCDC converter
Let, the energy in the L_{k} is transfer to the capacitor. Expression of the duty cycle D_{c1 }is
$D_{C 1}=\frac{t_{C 1}}{T_{S}}=\frac{2(1D)}{n+1}$ (1)
where, t_{c1} =Time interval
Stages I and III are knowingly small, the remaining three Stages are taken in CMS analysis. In Stage II, based on Figure 3 (Stage 2).
$V_{L k}^{2}=\frac{L_{k 1}}{L_{m}+L_{k 1}} V_{m}=(1k) V_{in}$ (2)
$V_{L 1}^{2}=\frac{L_{m}}{L_{m}+L_{k 1}} V_{i n}=k V_{in}$ (3)
$V_{L 2}^{2}=n V_{L 1}^{2}=n k V_{i n}$ (4)
$V_{0}=V_{i n}+V_{C 1}+V_{C 2}+V_{L 2}^{2}+V_{C 3}$ (5)
The subsequent equations are set, by utilizing the voltage second balance principle
$\int_{0}^{D T_{S}} V_{L k}^{2} d t+\int_{D T_{S}}^{T_{S}} V_{L k}^{5} d t=0$ (6)
$\int_{0}^{D T_{s}} V_{L 1}^{2} d t+\int_{D T_{S}}^{T_{S}} V_{L 1}^{5} d t=0$ (7)
$\int_{0}^{D T_{S}} V_{L 2}^{2} d t+\int_{D T_{S}}^{T_{S}} V_{L 2}^{5} d t=0$ (8)
Stage V voltage can be derived by substituting Eqns. (1), (2), (3) and (4) into Eqns. (6), (7) and (8). The voltages are expressed, according to the definition of voltage direction.
$V_{L k}^{5}=\frac{D(n+1)(1k)}{2(1D)} V_{in}$ (9)
$V_{L 1}^{5}=\frac{D k}{1D} V_{i n}$ (10)
$V_{L 2}^{5}=\frac{n D k}{1D} V_{i n}$ (11)
In Stage V, capacitors $\mathrm{C}_{1}, \mathrm{C}_{2}$ and $\mathrm{C}_{3}$ also charged. Based on Figure 3 (Stage 5) capacitor voltage can be denoted as.
$V_{C 1}=V_{L k}^{5}V_{L 1}^{5}$$=\frac{D}{1D} V_{i n} \frac{(1+k)+(1k) n}{2}$ (12)
$V_{C 2}=V_{C 3}=V_{L 2}^{5}=\frac{n D k}{1D} V_{i n}$ (13)
To obtain voltage gain by substituting equations 4, 12 and 13 into equation 5.
$M_{C C M}=\frac{V_{0}}{V_{in}}=\frac{1+n k}{1D}+\frac{D}{1D} \frac{(k1)+n(1+k)}{2}$ (14)
To obtain ideal voltage gain, at k=1
$M_{C C M}=\frac{1+n+n D}{1D}$ (15)
Voltage stresses on diodes $\mathrm{D}_{0}, \mathrm{D}_{1}, \mathrm{D}_{2}$ and $\mathrm{D}_{3}$ and switch are specified as
$V_{d s}=\frac{1}{1D} V_{in}=\frac{V_{0}+n V_{in}}{2 n+1}$ (16)
$V_{D 1}=\frac{1}{1D} V_{in}=\frac{V_{0}+n V_{in}}{2 n+1}$ (17)
$V_{D 2}=V_{D 3}=V_{D 0}=\frac{n}{1D} V_{i n}=\frac{n}{2 n+1}\left(V_{0}+n V_{in}\right)$ (18)
Eq. (16), (17) and (18) are in equal voltage ratio, by adjusting CI turns ratio the voltage stresses can be regulated.
MLIs are mostly used in gridconnected and various AC drive applications [5]. The fundamental topologies for MLIs are Cascaded HBridge, Flyingcapacitors, and Diodeclamped. All these fundamental topologies have their own merits and demerits, the fundamental topologies need a slightly huge number of switch count and occupy huge space. So, a new MLI topology was presented in this article.
The projected MLI topology has a very less number of switches as related to existing topologies and it produces minimum switching losses and probably less %THD. The projected topology is operated in five modes. Figure 4 indicates the block diagram of the transformerbased threephase multilevel inverter. MLI contains two capacitors, four diodes, and five switches. The switch 5 contains four diodes which are arranged as a bridge where switch 1 and 2 are in one leg and switch 3 and 4 are in another leg. Out of five switches, switch 1, 2 and 5 function at high frequency and the remaining two switches function at low frequency. Here low switching frequency and high switching frequency are referred to as fundamental frequency and carrier signal frequency respectively. Figure 5 indicates the four modes of operation of the inverter. In stage one, switches S_{a}_{4} and S_{a}_{5} are active and the remaining switches S_{a}_{1, }S_{a}_{2 }and S_{a}_{3} are inactive which results in the output voltage +V_{0}/2. In stage two, only switches S_{a}_{1 }and S_{a}_{4} are active, results+V_{0}. In the third stage, only switches S_{a}_{3 }and S_{a}_{5 }are active, which gives output voltageV_{0}/2. In stage four, only switches S_{a}_{2 }and S_{a}_{3 }are active, results V_{0} as output voltage. Finally, a fivelevel output voltage is obtained only by using five switches. Table 1 depicts the switching modes of fivelevel MLI. Figure 6 represents the inverter voltages and currents at various modulation index values.
Figure 4. Block diagram of the transformerbased threephase MLI
Figure 5. Four stages of operation of the inverter
Table 1. Switching modes of the inverter
S. No 
S_{1} 
S_{2} 
S_{3} 
S_{4} 
S_{5} 
Voltage (V_{0}) 
1 
X 
X 
X 
√ 
√ 
+V_{0}/2 
2 
√ 
X 
X 
√ 
X 
+V_{0} 
3 
√(or)X 
X(or)√ 
√(or)X 
X(or)√ 
X 
0 
4 
X 
X 
√ 
X 
√ 
V_{0}/2 
5 
X 
√ 
√ 
X 
X 
+V_{0} 
Figure 6. Inverter output voltages and currents with different MI values
Table 2 indicates the comparison among the existing MLI topologies and suggested MLI topology which represents that suggested topology has less number of switches compared to the existing topologies [15, 16].
Table 2. Comparison of components utilized in between the existing topologies and suggested topology
Reference 
Components utilization 

Active Switches 
Input DC Sources 
DC Bus Capacitors 
Transformers 
Total 

[R14] 
24 
01 
01 
06 
32 
[R15] 
24 
01 
01 
02 
28 
[R16] 
18 
03 
03 
01 
25 
Suggested 
15 
01 
02 
01 
19 
4.1 Multicarrier pulse width modulation
In the MCPWM techniques, four triangles are compared with a single sine wave to produce a pulse, in this M1 triangles are considered for an Mlevel inverter. Here, four triangles are considered for an MCPWM technique all triangles have equal magnitude and frequency. The Modulation Index (MI) of the inverter is given in (1)
$M_{inv}=\frac{A_{m}}{(m1) A_{C}}$ (19)
where, A_{m} represents the amplitude of the modulation signal and A_{c} represents the amplitude of the carrier signal. When A_{m}>A_{c} then only produces a positive pulse remaining condition no pulse is produced. The MCPWM techniques are categorized into three types such as Phase Disposition (PD), Phase Opposite Disposition (POD) and Alternative Phase Opposite Disposition (APOD). In this, carrier waves contain 2k Hz frequency and sine wave contains 50 Hz frequency. Hence, the Frequency Modulation ratio is $\mathrm{Mf}=\left(f_{s} / f_{r}\right)$. Each carrier is compared with a sine wave to produce pulses for MLI switches. The switching logic of the PD is presented in Figure 7. The similar way POD and APOD are also derived using logic gates.
Figure 7. Switching logic
Figure 8. Multicarrier PWM techniques (a) PD (b) POD (c) APOD
In PDPWM technique four carrier signals are used for generating switching pulses, all four carrier signals having equal amplitude and in phase with each other and the switching pattern also given in Figure 8 (a). In the PODPWM technique, four carrier signals are used for generating switching pulses, out of four carrier signals, two carrier waves are inphase with each other with above zero level and the other two carrier signals are maintained 1800 phase shift with below the zero points. All carrier signals are maintained the equal amplitude and frequency, but the difference in phase shift and the switching pattern also given in Figure 8 (b). In APODPWM, zero reference is placed in the middle of the carriers. Above the zero references, carriers are positive and below the zero references are negative carriers. In this, all carriers are maintaining 1800 shifted with each other and the switching pattern of AOPD is given in Figure 8 (c).
4.2 Proportional integral
The PI controller contains mainly two parameters those are Proportional and Integral. The transfer function of the PI controller is as follows:
$G_{C}(s)=K_{p}\left(1+\frac{1}{K_{i} S}\right)=K_{p}\left(\frac{K_{i} S+1}{K_{i} S}\right)$ (20)
where, K_{p} indicates the proportional constant and Ki indicates the integral constant. Figure 9 indicates the block diagram of the PI controller with unity feedback
Figure 9. Block diagram of the PI controller
From Figure 9 r*(t) indicates the reference value, r(t) indicates the actual value and e(t) indicates the error value. Most of the researchers prefer a PI controller for easier operation and implementation. The main purpose of the PI controller is to give fast response, to regulates the peak undershoot, settling time and peak overshoot of the system.
4.3 Neural network controller
NN algorithms are inspired by the structure of the human brain. NN takes data themselves and selftrain to get the desired output. NN is made up of several layers of neurons these neurons are core processing units of the network. NN controller has mainly three layers those are input, output, and hidden layers. The first input layer will seed the input, the output layer predicts the final output in between the existing hidden layers. Neurons of one layer are connected to the neurons of the next layer through the channels. Each of these channels is assigned a numerical value known as weights. From Figure 10 it is clearly observed that X1, X2 and X3 represents the input values, Y1 and Y2 represents the output values and W1 represents the weights of the network. Figure 10 indicates the neural network architecture.
Figure 10. The architecture of the neural network
In case if the output is not satisfied with the requirement by adjusting the weights of the network desired accurate output can be found. In the NN controller, so many learning algorithms are available those are supervised learning, Incremental, and Batch training [17]. Out of all these algorithms, an Incremental algorithm is stated as an ‘adaptive’ or ‘online’ training algorithm. The incremental algorithm is easily adaptable to the unknown situation, it can solve complex functions, easy to learn and train. Finally, it produces a better response of the system, with less settled time and a smooth response.
Figure 11. Closed loop control of the highvoltage DCDC converter
The simplified block diagram of NN tuned PI controller for Highvoltage DCDC converter to get the required output voltage is shown in the Figure 11. The actual output voltage is compared to the reference output voltage Vref. The error signal Er is obtained and is applied to the Neural Network tuned PI controller or PI controller. The output signal from the controller is applied to the power switch as gating signal. The neural network is trained with the best kp and ki values of the PI controller for various reference output voltages of HighVoltage DCDC converter.
PI and NN tuned PI control techniques are applied to the highvoltage DCDC boost converter and MCPWM control techniques are applied to the transformerbased threephase multilevel inverter. Figure 12 represents DCDC boost converter output voltage and, Figure 13, 14 and 15 represent the inverter voltages and currents with PI and NN tuned PI control techniques respectively.
Figure 12 (a) is illustrating the PIbased boost converter output voltage, it is settled at 0.15sec. Similarly, Figure 12 (b) is illustrating the NN tuned PI controllerbased boost converter output voltage is settled at 0.07sec. Settling time in case of NN tuned PI controller is less when compared to the PI controller along with less ripples and less overshoot with the NN tuned PI controller.
Figure 13 illustrate the PIbased inverter output voltages, it is settled at 0.065sec. Similarly, the NN tuned PI controllerbased inverter output voltages are settled at 0.045sec. Settling time in case of NN tuned PI controller is less when compared to the PI controller along with less ripples and less overshoot with the NN tuned PI controller.
Figure 14 is illustrating the PIbased inverter output currents, it is settled at 0.067sec. Similarly, the NN tuned PIbased inverter output currents are settled at 0.047sec. Settling time in case of NN tuned PI controller is less when compared to the PI controller along with less ripples and less overshoot with the NN tuned PI controller.
Figure 12. Boost converter output voltages (a) with PI controller (b) with NN tunes PI controller
Figure 13. Inverter output voltages (a) PIPD (b) PIPOD (c) PIAPOD (d) NN tuned PIPD (e) NN tuned PIPOD (f) NN tuned PIAPOD
Figure 14. Inverter output currents (a) PIPD (b) PIPOD (c) PIAPOD (d) NN tuned PIPD (e) NN tuned PIPOD (f) NN tuned PIAPOD
Figure 15. Threephase inverter output voltage and current (a) PIPD (b) PIPOD (c) PIAPOD (d) NN tuned PIPD (e) NN tuned PIPOD (f) NN tuned PIAPOD
Figure 15 is illustrating the PIbased threephase inverter output voltages and currents. it is settled at 0.067sec. Similarly, the NN tuned PIbased threephase inverter output voltages and currents are settled at 0.038sec. Settling time in case of NN tuned PI controller is less when compared to the PI controller along with less ripples and less overshoot with the NN tuned PI controller.
The settling time and settling point of the inverter output waveforms are represented in the Table 3. The settling time and settling point in case of NN tuned PI controller is less when compared to PI controller
Table 4 represents the THD analysis of the PD technique to MLI with PI and NN tuned PI controllers, to boost converter, by varying the modulation index from 0.4 to 1. The PI controllerbased system has produced 2.43% THD while the NN tuned PI controllerbased system has produced 2.21% THD at (Modulation Index) MI=1. Finally, the NN tuned PI controlled based system has produced less THD as compared to the PI controller.
Table 5 represents the THD analysis of the POD technique to MLI with PI and NN tuned PI controllers, to boost converter, by varying the modulation index from 0.4 to 1. The PI controllerbased system has produced 2.44% THD while the NN tuned PI controllerbased system has produced 2.22% THD at (Modulation Index) MI=1. Finally, the NN tuned PI controlled based system has produced less THD as compared to the PI controller.
Table 6 represents the THD analysis of the APOD technique to MLI with PI and NN tuned PI controllers, to boost converter, by varying the modulation index from 0.4 to 1. The PI controllerbased system has produced 2.48% THD while the NN tuned PI controllerbased system has produced 2.26% THD at (Modulation Index) MI=1. Finally, the NN tuned PI controlled based system has produced less THD as compared to the PI controller.
Figure 16 represents the THD analysis of the MLI system. Figure 16 (a) indicates the THD of PDMLI with PI controller has produces 2.43% THD, Figure 16 (b) indicates the THD of PODMLI with PI controller has produces 2.44% THD, Figure 16 (c) indicates THD of APODMLI with PI controller has produces 2.48% THD. Similarly, Figure 16 (d) indicates THD of PDMLI with NN tuned PI controller has produces 2.21% THD, Figure 16 (e) indicates THD of PODMLI with NN tuned PI controller has produces 2.22% THD finally Figure 16 (f) indicates THD of APODMLI with NN tuned PI controller has produces 2.26% THD.
Table 7 represents the THD analysis of the MLI system with PI and NN tuned PI controllers. From the Table 7 to clearly observe the PDPWM technique with the NN tuned PI controller significantly improve the performances of the inverter as compared to the other techniques and current THD is also within the IEEE standards.
Table 3. Summary of PI and NN tuned PI controllers
Parameters 
PI Controller 
NN tuned PI Controller 

Settling time 
Settling point 
Settling time 
Settling point 

DC output voltage 
0.0 to 0.15 
0.15 
0.0 to 0.07 
0.07 
Inverter voltages 
0.0 to 0.065 
0.065 
0.0 to 0.045 
0.045 
Inverter currents 
0.0 to 0.067 
0.067 
0.0 to 0.047 
0.047 
Threephase output 
0.0 to 0.067 
0.058 
0.0 to 0.038 
0.038 
Table 4. Summary of phase disposition
Modulation index 
MLI with PI controller 
MLI with NN tuned PI controller 

Voltage (V_{ph}) 
% THD 
Current (A) 
% THD 
Voltage (V_{ph}) 
% THD 
Current (A) 
% THD 

1 
215.3 
15.74 
1.87 
2.43 
215.6 
15.70 
1.87 
2.21 
0.9 
194.9 
16 
1.69 
2.46 
195.1 
15.91 
1.69 
2.20 
0.8 
175 
20.2 
1.51 
2.65 
175.1 
19.94 
1.52 
2.46 
0.7 
154.5 
22.33 
1.33 
2.88 
154.8 
22.25 
1.34 
2.61 
0.6 
133.1 
23.47 
1.14 
3.18 
132.8 
23.34 
1.13 
2.71 
0.5 
114.3 
32.62 
0.95 
3.86 
114.0 
32.45 
0.95 
2.79 
0.4 
94.96 
38.59 
0.78 
3.14 
94.89 
38.49 
0.77 
2.91 
Table 5. Summary of phase opposite disposition
Modulation index 
MLI with PI controller 
MLI with NN tuned PI controller 

Voltage (V_{ph}) 
% THD 
Current (A) 
% THD 
Voltage (V_{ph}) 
% THD 
Current (A) 
% THD 

1 
216.8 
19.87 
1.87 
2.44 
216.8 
19.80 
1.88 
2.22 
0.9 
199.6 
27.46 
1.69 
2.57 
199.7 
27.42 
1.69 
2.32 
0.8 
180.7 
32.71 
1.51 
2.79 
180.8 
32.68 
1.52 
2.60 
0.7 
160.2 
35.66 
1.33 
3.02 
160.0 
35.62 
1.34 
2.77 
0.6 
137.5 
35.33 
1.14 
3.28 
137.5 
35.25 
1.13 
2.84 
0.5 
115.9 
36.84 
0.95 
3.88 
115.8 
36.67 
0.94 
2.82 
0.4 
104.1 
61.81 
0.78 
3.56 
104.0 
61.73 
0.78 
3.36 
Table 6. Summary of alternate phase opposite disposition
Modulation index 
MLI with PI controller 
MLI with NN tuned PI controller 

Voltage (V_{ph}) 
% THD 
Current (A) 
% THD 
Voltage (V_{ph}) 
% THD 
Current (A) 
% THD 

1 
218.5 
23.42 
1.87 
2.48 
217.6 
23.35 
1.87 
2.26 
0.9 
199.0 
26.27 
1.69 
2.56 
199.2 
26.22 
1.69 
2.31 
0.8 
177.9 
27.27 
1.51 
2.71 
177.9 
27.22 
1.51 
2.52 
0.7 
155.8 
25.84 
1.33 
2.88 
160.1 
25.78 
1.34 
2.62 
0.6 
133.2 
23.60 
1.14 
3.16 
133.1 
23.49 
1.13 
2.69 
0.5 
115.9 
36.84 
0.95 
3.88 
115.2 
36.67 
0.94 
2.82 
0.4 
104.1 
61.81 
0.78 
3.56 
104.0 
61.73 
0.78 
3.36 
Table 7. Summary of % THD with PI and NN tuned PI controllers
Modulation Technique 
% THD with PI Controller 
% THD with NN tuned PI Controller 
PD 
2.43 
2.21 
POD 
2.44 
2.22 
APOD 
2.48 
2.26 
Figure 16. THD analysis of the MLI with PI and NN tuned PI controllers
This article mainly discussed a transformerbased threephase multilevel inverter with MCPWM techniques that comprises a minimum number of switches which is peculiar than existing topologies [1315]. This MLI topology is simulated using only 15 power semiconductor switches and it produces minimum switching losses. Multicarrier PWM techniques like PD, POD, and APOD are applied to control the MLI switches. The inverter topology is given an input supply of 400V AC which is derived from the highvoltage DCDC boost converter. The boost converter operates at 48V DC input voltage source and it is boosted to the general operating voltage of 400V DC. In that DCDC boost converter is controlled by two different controllers namely PI and NN tuned PI controllers. Comparative analysis in terms of THD in MLI's output voltage and output current is done on PIbased DCDC boost converter and NN tuned PI based DCDC boost converter. From Table 7 it can be depicted that NN tuned PI based DCDC boost converter with MLI resulted in less THD than PIbased DCDC boost converter with MLI at different modulation indices. Similarly, settling time in case of the NN tuned PI controller is less when compared to the PI controller along with less ripples and less overshoot with the NN tuned PI controller.
The authors are very thankful to the management of Vignan’s Foundation for Science, Technology, & Research for the successful completion of the work.
MLI 
Multilevel Inverter 
MCPWM 
Multi Carrier Pulse Width Modulation 
PD 
Phase Disposition 
POD 
Phase opposite Disposition 
APOD 
Alternate Phase opposite Disposition 
PI 
proportional–integral 
NN 
Neural Network 
MI 
Modulation Index 
[1] Gupta, K.K., Jain, S. (2014). A novel multilevel inverter based on switched DC sources. IEEE Transactions on Industrial Electronics, 61(7): 32693278. http://dx.doi.org/10.1109/TIE.2013.2282606
[2] Kavali, J., Mittal, A. (2016). Analysis of various control schemes for minimal total harmonic distortion in cascaded Hbridge multilevel inverter. Journal of Electrical Systems and Information Technology, 3: 428441. http://dx.doi.org/10.1016/j.jesit.2016.01.007
[3] Ali, J.S.M., Krishnaswamy, V. (2018). An assessment of recent multilevel inverter topologies with reduced power electronics components for renewable applications. Renewable and Sustainable Energy Reviews, 82(3): 33793399. https://doi.org/10.1016/j.rser.2017.10.052
[4] Pires, V.F., Cordeiro, A., Foito, D., Silva, J.F. (2018). Threephase multilevel inverter for gridconnected distributed photovoltaic systems based in three threephase twolevel inverters. Solar Energy, 174: 10261034. https://doi.org/10.1016/j.solener.2018.09.083
[5] Jana, J., Saha, H., Bhattacharya, K.D. (2017). A review of inverter topologies for singlephase gridconnected photovoltaic systems. Renewable and Sustainable Energy Reviews, 72: 12561270. https://doi.org/10.1016/j.rser.2016.10.049
[6] Kala, P., Arora, S. (2017). A comprehensive study of classical and hybrid multilevel inverter topologies for renewable energy applications. Renewable and Sustainable Energy Reviews, 76: 905931. https://doi.org/10.1016/j.rser.2017.02.008
[7] Prabaharan, N., Palanisamy, K. (2017). A comprehensive review on reduced switch multilevel inverter topologies, modulation techniques and applications. Renewable and Sustainable Energy Reviews, 76: 12481282. https://doi.org/10.1016/j.rser.2017.03.121
[8] Ramya, G., Ganapathy, V., Suresh, P. (2019). Comprehensive analysis of interleaved boost converter with simplified Hbridge multilevel inverter based static synchronous compensator system. Electric Power Systems Research, 176(105936): 114. https://doi.org/10.1016/j.epsr.2019.105936
[9] Suresh, Y., Panda, A.K. (2013). Investigation on hybrid cascaded multilevel inverter with reduced dc sources. Renewable and Sustainable Energy Reviews, 26: 4959. https://doi.org/10.1016/j.rser.2013.04.027
[10] Colak, I., Kabalci, E., Bayindir, R. (2011). Review of multilevel voltage source inverter topologies and control schemes. Energy Conversion and Management, 52(2): 11141128. https://doi.org/10.1016/j.enconman.2010.09.006
[11] Venkatesan, M., Rajeswari, R., Deverajan, N. (2015). A fuzzy logic based three phase inverter with single DC source for grid connected PV system employing three phase transformer. International Journal of Renewable Energy Research, 5(3): 739745. https://doi.org/10.3906/elk1404423
[12] Susheela, N., Kumar, P.S. (2017). Performance evaluation of carrier based PWM techniques for hybrid multilevel inverters with reduced number of components. Energy Procedia, 117: 635642. https://doi.org/10.1016/j.egypro.2017.05.164
[13] Hsieh, Y.P., Chen, J.F., Liang, T.J., Yang, L.S. (2013). Novel high stepup DC–DC converter for distributed generation system. IEEE Transactions on Industrial Electronics, 60(4): 14731482. http://dx.doi.org/10.1109/TIE.2011.2107721
[14] Singh, J., Dahiya, R., Saini, L.M. (2018). Recent research on transformer based single DC source multilevel inverter: A review. Renewable and Sustainable Energy Reviews, Renewable, 82(3): 32073224. https://doi.org/10.1016/j.rser.2017.10.023
[15] Song, S.G., Kang, F.S., Park, S.J. (2019). Cascaded multilevel inverter employing threephase transformers and single DC input. IEEE Transactions on Industrial Electronics, 56(6): 20052014. http://dx.doi.org/10.1109/TIE.2009.2013846
[16] Pires, V.F., Cordeiro, A., Foito, D., Silva, F. (2018). Threephase multilevel inverter for gridconnected distributed photovoltaic systems based in three threephase twolevel inverters. Solar Energy, 174: 10261034. https://doi.org/10.1016/j.solener.2018.09.083
[17] Fu, X., Li, S., Fairbank, M., Wunsch, D.C., Alonso, E. (2015). Training recurrent neural networks with the levenberg–marquardt algorithm for optimal control of a gridconnected converter. IEEE Transactions on Neural Networks and Learning Systems, 26(9): 19001912. http://dx.doi.org/10.1109/TNNLS.2014.2361267