Integration of a cascaded multilevel

inverter (CMLI) with AC mains

By

Muhammad Arslan Fida

Muhammad Muzammil Nawaz

Noman Khan

Thesis submitted to the faculty of Engineering in partial

fulfillment of requirements for the Degree of BS Electrical

Engineering.

Department of Electrical Engineering,

Pakistan Institute of Engineering and Applied Sciences,

Nilore, Islamabad 45650, Pakistan.

January 06, 2017.iii

Department of Electrical Engineering

Pakistan Institute of Engineering and Applied Sciences (PIEAS)

Nilore, Islamabad 45650, Pakistan

Declaration of Originality

I hereby declare that the work contained in this thesis and the intellectual content

of this thesis are the product of my own research. This thesis has not been

previously published in any form nor does it contain any verbatim of the published

resources which could be treated as infringement of the international copyright law.

I also declare that I do understand the terms copyright and plagiarism, and that

in case of any copyright violation or plagiarism found in this work, I will be held

fully responsible of the consequences of any such violation.

Signature:

Name: Muhammad Arslan Fida

Muhammad Muzammil Nawaz

Noman Khan

Date: January 06, 2017.

Place: PIEASiv

Certificate of Approval

This is to certify that the work contained in this thesis entitled

\Integration of a cascaded multilevel inverter (CMLI) with AC mains"

was carried out by

Muhammad Arslan Fida

Muhammad Muzammil Nawaz

Noman Khan

under my supervision and that in my opinion, it is fully adequate,

in scope and quality, for the degree of BS Electrical Engineering

from Pakistan Institute of Engineering and Applied Sciences

(PIEAS).

Approved By:

Signature:

Supervisor: Mr. Tanveer Abbas

Verified By:

Signature:

Head, Department of Electrical Engineering

Stamp:v

'' Copyright by Muhammad Arslan Fida

Muhammad Muzammil Nawaz

Noman Khan, 2017.

All rights reserved.vi

Table of Contents

List of Figures viii

List of Tables viii

1 Introduction 1

1.1 Goals and Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.2 Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.3 Approach and Methodology . . . . . . . . . . . . . . . . . . . . . . 3

1.4 Overview of remainder of the report . . . . . . . . . . . . . . . . . . 4

2 Literature Review 5

2.1 Synchronization Techniques . . . . . . . . . . . . . . . . . . . . . . 5

2.1.1 Zero Crossing detection . . . . . . . . . . . . . . . . . . . . . 5

2.1.2 Kalman Filter . . . . . . . . . . . . . . . . . . . . . . . . . . 6

2.1.3 Discrete Fourier transform (DFT) . . . . . . . . . . . . . . . 6

2.1.4 Phase Locked Loop (PLL) . . . . . . . . . . . . . . . . . . . 6

2.1.5 Adaptive Notch Filter . . . . . . . . . . . . . . . . . . . . . 6

2.1.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

3 Adaptive Notch Filter (ANF) 8

3.1 Simulations: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

3.1.1 Results: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

3.1.2 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11vii

4 Accomplishments and Future Plan 12

Appendices 13

Appendix A Code for ANF Simulations 14

References 16viii

List of Figures

1.1 Block Diagram of System . . . . . . . . . . . . . . . . . . . . . . . . 3

3.1 Variation in '' at '' = 0:85 . . . . . . . . . . . . . . . . . . . . . . . 10

3.2 Variation in '' at '' = p2=2 . . . . . . . . . . . . . . . . . . . . . . . 10

A.1 ANF Differential Equations . . . . . . . . . . . . . . . . . . . . . . 14

A.2 Simulation code using ode45 . . . . . . . . . . . . . . . . . . . . . . 15

List of Tables

2.1 Comparison of Synchronization Methods . . . . . . . . . . . . . . . 7

4.1 Accomplishments Summary . . . . . . . . . . . . . . . . . . . . . . 121

1. Introduction

Renewable energy resources will play an important role in the world's future. The

energy resources have been split into three categories: fossil fuels, renewable resources and nuclear resources [1]. Renewable energy resources include solar energy,

wind energy, biomass energy, geothermal energy. The ever-increasing demand of

energy cannot be supplied by the conventional grid alone. Thus, renewable energy resources are of prime importance and generation of energy from renewable

resources will be an integral part of the future grid.

In present day, one of the major sources of renewable energy is sun and this energy

is harvested in DC form and hence is not readily available to be used as majority

of our appliances work with AC. Hence, a conversion of this solar energy to AC is

needed and this is done by an inverter. On the basis of overall system structure,

inverters could be divided into two categories: (1) Standalone Inverters, (2) Gridconnected Inverters.

Stand-alone inverters produce power from a single independent source. These are

more suitable for remotest locations where the grid cannot penetrate and there is

no other source of energy [2].

Grid-connected inverter, however, functions alongside with the utility grid. It not

only converts DC into AC but it also injects that energy into the utility grid. The

injection of the solar energy into the grid is impossible without synchronization

of the inverter output with the grid. This is done by a grid-side controller which

can have the following tasks [3]:2

1. Grid synchronization;

2. Control of active power delivered to the grid;

3. Control of reactive power transfer between the inverter and utility grid;

4. Ensure high quality of the injected power.

This project aims to addresses the problem of grid synchronization by designing

a controller.

1.1 Goals and Objectives

The goal of this project is to develop a controller which should be able to synchronize a cascaded multi-level inverter (CMLI) with the utility grid.

This goal is set to be achieved through the following objectives:

' Literature review;

' Selection of a suitable technique;

' Simulations of selected technique;

' Hardware Implementation of the controller;

' Testing of the controller;

' Thesis writing.

1.2 Scope

This project deals with the complex problem of synchronizing a CMLI with the

utility grid. Designing a CMLI is not in the scope of this project and it will be

provided during the testing phase. A block diagram of the whole grid-connected

system is shown below in Figure 1.1.3

Figure 1.1 Block Diagram of System

1.3 Approach and Methodology

A literature review of the present grid synchronization techniques is to be carried out and a suitable technique is to be selected to design the controller. After

selection of the technique, simulations are to be carried out to assess the performance and working of the technique.The simulation results will enable us to

further enhance the performance of the selected technique. After that, hardware

implementation is to be carried out. This will complete the design of the controller. This controller is then to be tested by synchronizing a CMLI with the

utility grid. After this, thesis writing is to be done.4

1.4 Overview of remainder of the report

Chapter 2 describes the literature review carried out. Chapter 3 details our selected technique. Chapter 4 deals with current accomplishments and future work

plan.5

2. Literature Review

There are different practical methods used for synchronization. Some of them are

native and can only be used for some specific applications, while other techniques

are comparatively more complex and are used for the applications where the loss of

synchronization cannot be compromised. The remainder of this chapter presents

a comprehensive discussion on these techniques.

2.1 Synchronization Techniques

Following are some of the techniques used for synchronization of grid-connected

inverters.

2.1.1 Zero Crossing detection

Zero crossing detection is one of the most popular methods to determine frequency

and time period of a signal. When determining the frequency is our priority, we

usually calculate the frequency of a number of cycles of the signal because by

doing so the error will be reduced to a very small magnitude. Zero crossing is the

point of choice for measuring phase and frequency. The reference is usually easy

to establish. However, phase synchronized triggering requires placing additional

constraints on zero crossing detection [4].

Although the system looks very easy and simple but it has a disadvantage that it

is very welcoming to noise. That noise alters the results and the synchronization

can be compromised.6

2.1.2 Kalman Filter

The Kalman gain can be evaluated off-line, so that the filter can be implemented

using fixed gains, simplifying its implementation. This is capable of generating

the synchronization signals even when the input signal contains harmonics and

measurement noise. The computational complexity of this technique is more and

difficult to handle and covariance selection is also a problem.

2.1.3 Discrete Fourier transform (DFT)

In this case, the proposed algorithm takes samples in a window synchronous with

the exact rate as that of the nominal frequency of the network. If the frequency is

exactly equal to nominal value i.e., 50Hz the system will produce synchronization.

This method, however, produces a phase shift when the DFT sampling is asynchronous with power grid.

2.1.4 Phase Locked Loop (PLL)

The difference between phase angle of the input and that of the output signal

is measured by the phase detection and passed through the loop filter .The loop

filter output signal drives the voltage-controlled oscillator (VCO) to generate the

output signal. The method is very simple and easy to implement. It gives most

accurate results as compared to other techniques.

However, using PLL may have a negative impact on the performance of HVDC in

weak ac system and during unsymmetrical voltage faults, 2nd harmonic produced

by negative sequence will propagates through PLL system and will be reflected in

the '' _.

2.1.5 Adaptive Notch Filter

It is a notch filter with adaption mechanism that tracks the frequency and phase

of the input signal. This information can be used for the purpose of the synchronization. A second-order notch filter is used to update the frequency and the7

voltage-controlled oscillator (VCO) is replaced by two extra integrators.

In addition to a simple structure (no VCO) ANF offers higher immunity to noise as

compared with other conventional methods. The transient response is also faster

than the PLL-based method.

This technique has been chosen for this project and detailed discussion of this

technique will be carried out in Chapter 3.

Comparison table of different techniques:

A comparison table is given below with the purpose to give an overview of different

techniques:

Table 2.1 Comparison of Synchronization Methods

Synchronization method Single phase Three phase Harmonic detection Implementation Complexity

Zero crossing detection Yes Yes No Simple

Kalman filter Yes Yes No Complex

Discrete Fourier transform No Yes Yes Medium

Phase-locked loop Yes Yes Yes Medium

Adaptive notch filter Yes Yes Yes Simple

2.1.6 Summary

Several techniques are available for synchronization of grid connected inverters

among which Zero crossing is robust to the frequency variations but the results

are distorted by noise. In case of Kalman filter, the computational complexity is an

issue. Discrete Fourier transform produces a phase shift when the DFT sampling

is asynchronous with power grid. PLL is immune to noise as compared to other

techniques but ANF provides much higher noise immunity as compared to PLL. It

includes a second-order notch filter to update the frequency and replaces the VCO

of PLL with two integrators. Also, its response time is faster than that of PLL.

Faster response rate and simplicity in design makes ANF our preferred technique

to synchronize a CMLI with AC Mains. A detailed discussion of ANF is given in

the Chapter 3.8

3. Adaptive Notch Filter (ANF)

ANF is preferred over other techniques because of its simplicity and faster response. A Notch filter (NF) is basically band-Stop filter with very narrow stopband corresponding to high Quality Factor. NF basically gives a notch at particular frequency. For example if it is set to give a notch at 50 Hz frequency; whenever

a signal is applied with a 50 Hz component in it, it would give a notch. An adaptive filters work on feedback algorithm by adjusting the variable parameters of its

transfer function.

So, an adaptive filter combined with a notch filter can track down a change in frequency and phase. ANF is the most recent technique being used in power systems

for frequency estimation and phase detection.

A modified version of HSU's ANF has been proposed in [5] and [6].

The response of ANF proposed in [5] and [6] is characterized by following set of

differential equations:

x '' + 2''x _ + ''2x = 2''2y (3.1)

'' _ = '''x(''2y ' ''x _) (3.2)

Where y is the input signal, x is the estimated phase and '' is the estimated

frequency in radians per second.

As compared to PLL, ANF doesn't use a VCO which makes it simpler to instead

it uses two extra integrators which makes it simpler to implement. But as far

as digital implementation is concerned, ANF needs a higher sampling rate as

compared to PLL as integrators must operate on signals that are changing faster.9

The design parameters are '' and '' . The parameter '' determines the depth

of notch and hence the sensitivity of ANF. While '' defines the adaption speed

and hence the frequency tracking capability of ANF [7]. A trade-off between the

accuracy and speed of the estimation is done by adjusting the ANF parameters

'' and ''. By increasing the value of '', one can achieve a faster estimation of

frequency but at the same time '' should be increased to avoid oscillatory behaviors.

Following are the constraints for a pure sinusoidal input signal y(t) = A1sin(!1t +

''1) and for '' > 1:

A2

1''

2

< 1 (3.3)

For 0 < '' < 1, the stability condition is no longer independent from ''. However,

one can use the more restricted condition that:-

A2

1''

2

< 0:8 (3.4)

Here A2

1=2 is the power of the sinusoidal signal with amplitude A. So, from equation

3.4 it is clear that knowledge of input signal power is adequate for selecting ''

independently from ''. Choosing '' = p2=2 and '' in the range [0.8, 1] results in

three to four cycles of transient response for a unity amplitude, sinusoidal signal

of frequency of 50 or 60 Hz [7].

3.1 Simulations:

The differential equations presented in Equation 3.1 and 3.2 are simulated in Matlab for a unity amplitude sinusoidal signal. Matlab codes are given in Appendix A

for reference. The results of the simulations are shown below.

3.1.1 Results:

After setting the initial conditions, we '' and '' to obtain some optimum value.

Figure 3.1 shows the response observed after varying the '' parameter while keeping

'' = 0:85.10

Figure 3.1 Variation in '' at '' = 0:85

Figure 3.2 shows the response observed after varying the '' parameter while

keeping '' = p2=2.

Figure 3.2 Variation in '' at '' = p2=211

3.1.2 Summary

ANF has been employed for the estimation of system Frequency and Phase. ANF

gives faster response by increasing the value of '' but this would result in overshoots

and undershoots. These overshoots and undershoots can be reduced by increasing

'' but this would increase the settling time. Whenever there is change in system

frequency the ANF will take some time to adapt to new frequency and the output

of CMLI will be disconnected during that time. So, we don't need to worry about

'' and hence the overshoots and undershoots. We increased '' to a certain limit

(0.95) to decrease the tracking time. After this limit the response of ANF becomes

unstable.12

4. Accomplishments and Future

Plan

Following table summarizes the accomplishments so far and future work plan for

the next semester.

Table 4.1 Accomplishments Summary

Objectives Deliverable Schedule Status

Literature Review Chapter 2 September-October 2016 Completed

Technique Explanation Chapter 3 October 2016 Completed

Simulations Simulation Results November-December 2016 Completed

Hardware Implementation Controller Hardware February-March 2017 Not Completed

Testing Practical Results April 2017 Not Completed

Thesis Writing Thesis 7th and 8th Semester Partially Complete13

Appendices14

A. Code for ANF Simulations

Following codes were used for the simulation of ANF in Matlab.

1 function dy = rigid(t,y,f)

2 3

clc

4 Zeta=sqrt(2)/2;

5 gamma=0.85;

6 dy = zeros(3,1);

7 if (t>3)

8 f=60.5;

9 end

10 if (t>5)

11 f=61;

12 end

13 if (t>7)

14 f=61.5;

15 end

16 dy(1) = y(2);

17 dy(2) = '2*Zeta*y(3)*y(2)'(y(3))''2*y(1)+2*Zeta*(y(3))''2* ...

(1*sin(f*t));

18 dy(3) = ('gamma * y(1)) *( (y(3))''2 * 1*sin(f*t) ' (y(3)*y(2)));

Figure A.1 ANF Differential Equations15

1 y0 = [0 0 59.5];

2 tspan = [0 10];

3 f = 60;

4 options = odeset('RelTol',1e'4,'AbsTol',[1e'4 1e'4 ...

1e'5],'OutputFcn',@odeplot);

5 [T,Y] = ode45(@(t,y) rigid(t,y,f),tspan,y0);

6 7

p = plot(T,Y(:,3),''');

8 hold on;

9 p.LineWidth = 1;

Figure A.2 Simulation code using ode4516

**...(download the rest of the essay above)**